There seems to be a growing field of what you could call “auxiliary AI” for safety/interpretability/control. Auxiliary in that we don’t know how the main ML model functions, so we tack on an extra ML model to help interpret the main ML model. The monosemanticy example was a good case, but there’s tons of these now, in a bunch of different fields. Seems a bit silly to me, because you’re ostensibly just moving the goalposts: if we don’t know how the auxiliary AI is doing what it’s doing, then how do we know it’s actually performing the interpretability/control function we want it to? You can make all sorts of arguments about the structure of a network and how it enforces certain desirable behaviors/patterns/structures in the working of the model, but at the end of the day, if its still just a giant inscrutable tensor of weights, are we really getting anywhere? Seems to me we should either just earnestly abandon our efforts at creating thinking digital structures we can understand, or abandon the neural network paradigm but it doesn’t seem like you can do both. Maybe I’m just not as deep in the lit as a I should be tho.
Hi for those interested in Scott's piece on AI monosemanticity I wrote a piece about the implications of AI interpretability for biology and drug discovery.
I am a big fan of the "Thrive-Survive Theory of Politics" by Scott Alexander (https://slatestarcodex.com/2013/03/04/a-thrivesurvive-theory-of-the-political-spectrum/). One thing I always found interesting/incongruous with this theory is that it claims that (material) security leads people to feel secure, and thus become more likely to identify as leftist and adopt their outlooks on life (i.e. post-materialism or "globalism"), while those who are more concerned with "surviving" will be more likely to be right-wing and adopt their outlook (i.e. being more traditional and more harsh towards human failure). I feel that this theory certainly makes sense on a global level - people in poorer countries are more likely to be in a "survive" mindset and thus less likely to be progressive politically than those in wealthier countries. Even inside the US it makes sense, with right-wing "Red States" tending to be poorer than left-wing "Blue States". But there is a major flaw in this theory IMO. Namely that the USA, which is the richest major economy in the world (https://en.wikipedia.org/wiki/List_of_countries_by_GDP_(PPP)_per_capita) is definitely (on average) more towards the "survive" Spectrum than most other rich Western countries, including all other Anglo countries (as can be seen in the WVS map: https://www.worldvaluessurvey.org/WVSContents.jsp?CMSID=Findings). But since the US is richer than almost all other Western countries, wouldn't one expect the US to be more towards the right side ("self-expression) than the other Anglosphere countries and/or Western European countries?
I guess one argument is that Western European countries have a better social safety net than the US and thus allowing people to focus more on "thrive" attitudes politically...but even then, why didn't the US develop a social safety net like those Western European countries in the first place? Is it just because of racial politics? Also, might this change if the US becomes more progressive politically overall? Especially if Western European countries become more focused on "survival" politically because of the energy/geopolitical situation in Europe? Anyway, I would like to know others' view on this here...
1. I read that post a couple of years ago and disagreed with it strongly. I'm glad you've given me the opportunity to comment on it without requiring the ability to go back in time to 2013. Thanks.
2. One major problem (and honestly I find that Scott makes this mistake amazingly often, though many others do it even more) is basically focusing entirely on 2010s US politics, and even then mostly on the social side only. For much of the last two centuries economic issues have dominated, and poorer voters have been more progressive. This is *still* the case in a number of western countries. And even within the modern US, where both parties have been hijacked by social and cultural activists to an unusual degree, when actually voting directly on things like minimum wage and medicaid expansion, the people of poorer red states are, unsurprisingly, quite liberal on those things. And also, Trump does better than more ideologically right-wing candidates among poorer republicans, again unsurprisingly. When your theory of the political spectrum largely ignores most of the time the spectrum has existed, most of the world even now, and much of the economic side of it even "here and now", I don't think it's unreasonable to say it's not a great theory.
3. Even focusing entirely on "2010s-20s US social issues" this theory is pretty bad. It doesn't explain basically anything to do with feminism: believing and punishing rape accusations without question is what you do in the zombie apocalypse when any threat to group safety and cohesion is potentially fatal, while due process is a luxury for the prosperous utopia. Similarly, protecting unborn life is, like protecting the environment, a function of a moral concern born of prosperity; "you're inconveniencing me, you die!" is a feature of the brutal zombie world. It doesn't explain attitudes to climate change: the leftists are responding fearfully to an existential threat, the rightists are saying "nah we'll be fine, our society can handle anything". Scott admits it doesn't explain school choice, and I'd say it also doesn't explain political correctness or hate speech laws: non-conforming ideas that may create bad feelings are tolerared in a stable utopia, and suppressed in a zombie situation. Overall it probably doesn't explain about half of all social issues alone. So, not a great theory at all.
4. I'm not at all surprised it's a bad theory, like every other theory about what "left" and "right" mean. What I am surprised at is that anyone as smart as Scott would think they can come up with such a theory, when every other attempt has been such a disaster. No one knows what left and right really mean. No one! I have literally never seen an explanation that doesn't completely break down for at least one major issue. "The right wants to return to the past", except for drug laws. "The right wants to ban destructive behaviours", except those involving the environment. "The left wants humane treatment for the most vulnerable", unless you happen to be in the womb, in which case sucks to be you I guess! "The left wants freedom in your personal life", except for hate speech laws, and gun control, and affirmative action... I think left and right mean *something*, for people to organise themselves so consistently around it, but whatever that something is isn't remotely obvious or clear, is a collection of historical accidents and unconscious social instincts that we don't understand, and only makes recognisable sense in a very particular kind of modern, democratic, western society.
5. And just about Europe being less polarised, more tolerant, more liberal, and so on. Both the fact that this is absurdly, ridiculously false, and the reason they're making an enormous effort to appear that way right now, are so blindingly obvious that it shouldn't need to be explicitly stated. I find it truly breathtaking how many people make the above claims as though the world was created ex nihilo in 1995, or (if we're just talking about western Europe) in 1950. I can only assume the blame lies with an overbloated media ecosystem that "reacts" to such a degree to every single political and social development, year after year after year, that people can become so confused that they actually *forget* utterly unforgetable things that happened within living memory.
It's a "survival" mindset that doesn't have anything to do with levels of wealth.
It seems like you have similar views to Bryan Caplan on this (https://www.econlib.org/archives/2017/09/whats_wrong_wit_23.html ). But IMO that view is clouded by his (libertarian) biases to some extent...now I agree that the terms "left" and "right" change with time and place, but I still believe that there's some truth to Scott Alexander's theory (namely that left-wing values tend to be correlated with post-materialistic values, and post-materialism tends to be correlated with wealth).
Also, I think that generally everyone would agree that Western European countries (besides Italy maybe) are more "left-wing" than the US as it is commonly understood in the modern Western political discourse. I am not sure when this first became the common wisdom among most people, but it certainly has been ever since I started partaking in political discussions on the internet (mid-2000s).
I think the actual laws of the USA *are* more "self-expression" than the EU, on average - For most cultural things, like gay rights, or accepting transgender as a thing, legal recognition came relatively early in at least some states of the USA. Even after the effective overturning of Roe vs Wade, half of the USA still has very liberal abortion laws. Overall, though ,for a lot of these cultural issues I think religiosity is far more predictive than wealth, it's just that religiosity happens to inversely correlate with wealth in general.
Re social safety nets - they are arguably a "survive" policy, at least from the perspective of any given voter. Lower risk of disaster, at the cost of higher taxes reducing the potential maximum 'thriving'
I am not sure if I agree with your argument...it's true that the US allows more "freedom of expression". But in most Western European countries, laws tend to protect workers and consumers better than in the US. Also, certain "vices" are legal in Western Europe while being illegal in the US. In terms of gay rights, it's true that the US introduced Gay marriage earlier than many countries in Western Europe (though some introduced it earlier than the US) at a federal level, but I see that more as an artifact of the supreme court being "liberal" in 2015 rather than true progressive/thrive sentiment in the US. Most opinion polls show that in the US agreement with gay marriage is lower than anywhere in Western Europe (i.e. https://www.pewresearch.org/short-reads/2023/11/27/how-people-around-the-world-view-same-sex-marriage/). Thus, the US is arguable less "thrive" (live and let live) than Western Europe here. Now, this might be explained by the higher prevalence of highly religious people in the US than in Western Europe. But if you look at the WVS map (https://www.worldvaluessurvey.org/images/Map2023NEW.png ) then you can see that more religious countries tend to be on the left side - the "survive" side of the map - and that the US is more towards the "survive" side than most other Western countries (not surprising considering the higher religiosity in the US). Thus, I don't think that the US is arguably more if a "thrive" mindset than most of Western Europe - in fact, the WVS map and opinion polling shows the opposite.
As for social safety nets, I disagree with your take here too. If a welfare state is an expression of a "survive" mindset, then why do conservatives throughout the Western World (people who tend to be more of a "survival" mindset as per Scott Alexander) tend to oppose them more than leftists (who are more of a "thrive" mindset as per Scott Alexander). Thus, my argument stands in my opinion: that there is an incongruence with Scott Alexander's theory of "Thrive vs Survive", namely that Western European countries which are poorer than the US per capita are more "thrive" politically and socially than the US...
Walk & Talk: We usually have an hour-long walk and talk after the meeting starts. Two mini-malls with hot takeout food are readily accessible nearby. Search for Gelson's or Pavilions in the zip code 92660.
Share a Surprise: Tell the group about something unexpected that changed your perspective on the universe.
Future Direction Ideas: Contribute ideas for the group's future direction, including topics, meeting types, activities, etc.
Recommend quotes around article title, or at least move the question mark, for the 3rd conversation starter, which currently reads, “Is Enlightment Compatiable with sex scandals by Scott Alexander?”
How good was the average life of the Middle-Ages peasant compared to the average life of the middle-class American today? I'd guess 1/2 as good but with less variance. Probably about as many miserable people as a percentage.
From whose viewpoint? An average middle-ages peasant was probably averagely happy/unhappy with his/her life, just like an average American is today. If you transport an average American to M-A peasantry he'll quicky succumb to disease/exposure/etc. If you transport an average M-A peasant to modern day America he'll probably go mad.
I'm not sure how we're defining good, but I'd say 20% as good. You'd be living in a shed in the cold, wearing two uncomfortable shirts all the time, having half your kids die when they're little, eating bread and not much else, getting conquered and perhaps slaughtered periodically, having no actual entertainment apart from church and sitting around and talking when you're not doing hard physical labor. It's all pretty bad.
I keep finding the phrase "Family, friends, health and wealth" to be a really useful way of classifying human happiness. Like, you can add or subtract something but it's hard to imagine a good life without these things.
In our favor:
Wealth, obvious, and I think it's hard to conceive of how much further we are ahead.
Health, and it's not like we're healthy today but a lifetime of physical labor, especially peasant agriculture, is just brutal. Your back, knees, everything just shot. I'm pretty sure a ton of 19th century drug and alcohol abuse is just pain management.
In their favor:
Family, it's not close, they had entire extended family and kinship groups we've totally lost.
Mixed:
Friends. I mean, in general I think the middle-ages peasant had more friends and seeing people in person is better but...this whole internet acquaintance thing...I mean it has advantages. Let's give a point to each
So, we're scoring 3/4 and they're scoring 2/4, that means they 2/3 as happy as us or we're 50% happier than them.
I'd say the worst things are more like casinos and opioid addiction, and perhaps junk food - these are things that appeal to the short-term interests that determine what a person does in the moment, even though they go against the person's long-term interests (as you can tell based on asking someone what they would prefer to do one month or one year from today, rather than asking what they would prefer to do this very instant).
It's possible that junk movies and junk music are like these things in the ways that junk food tends to be, but I think none of them are as extreme as the traditional "vices".
> Still other texts say the Messiah will come in a generation that is both the most righteous and the most wicked. I don’t even know what to think of that one.
I do! Because I came across this while reading the perennialists:
> The traditional world was essentially good but contained much evil, while the modern world is essentially evil but contains much good.
You can definitely get away with claiming the modern world is essentially evil if you judge by the standards of the traditional world, but then, much good is done! So we truly are the most righteous and most wicked generation already.
I thought of looking at what the Arabic Wikipedia says about the events of October 7 and I'm deeply impressed. The page is https://ar.wikipedia.org/wiki/عملية_طوفان_الأقصى and then use auto-translation if you don't know Arabic (I don't).
There's no mention at all of any civilian deaths in Israel. Really, none whatsoever. Apparently what happened was, Hamas fighters broke through the fence, entered various Israeli cities, army bases and "settlements" near Gaza and clashed with the Israeli army there. They captured lots of Israeli soldiers and officers. Nothing is said about any civilian hostages. The music festival is not mentioned.
Is this really that surprising given all the ink that has been spilled on the economics and the incentives of Wikipedia participation and their governing structure ? read any Anglo-USA culture war topic on English wikipedia (e.g. GamerGate) and you will witness the same one-sidedness, the selection of which facts to emphasize in the official and authoritative article voice and which to put between quotes and attribute to external parties, and selective wording of facts.
As far as I can see, no mention of the 14000+ dead, no mention of White Phosphorus being used in one of the densest civilian areas, no mention of Amnesty International or the resignation of human rights officials in the UN or the international condemnation of Israel from Belgium and Spain. As a matter of fact the entirety of post-10th october or post-13th october is extremely hazy in the article, almost dream-like, almost non-existent. Which is coincidentally also the implied general position of the Israeli propaganda : There was the 7th of October and a few days after where Israelis were victims, and then a huge black hole after that in the Calendar where whatever happens can't really be blamed or even traced to individual deliberate decisions.
Truth is one, but everybody wants to take a different piece. Nobody is immune to this.
No, that's wrong. The Hebrew page you linked to does not talk about the 14000+ dead because it's devoted to Oct 7 only, and even there only to the Hamas attack. It links prominently to another page devoted to the entire war, as Israel defines the current conflict: https://he.wikipedia.org/wiki/מלחמת_חרבות_ברזל
You'll find the 14000+ number there.
There are separate pages for the West Bank in this conflict
It's possibly not as detailed as you might want. I don't see a breakdown of the number of children killed (according to the Hamas-run Gaza Healthy Ministry) or allegations about white phosphorus. But there's plenty of accusations against Israel presented on those pages, and I'm reasonably sure these too could be added.
The Arabic page is of a completely different kind. Such total denial of reality of civilians being killed - which reverberated around the world and was shown in mass media everywhere - is chilling, and no, it's not business as usual in Wikipedia as it's run in English, French, or Hebrew for that matter. Interestingly, even the Farsi wikipedia mentions the civilian slaughter prominently (I went on to check if Wikipedia is blocked in Iran, which would mean the Farsi wikipedia being diaspora-run, but I believe it's not). But not the Arabic wikipedia. This level of disconnect from reality is really something.
I stand corrected. I should have followed the links in the Hebrew Wikipedia more before I commented. Thanks for your effort in tracing down where I'm wrong.
I generally agree that Arab sources are more radicalized and polarized against the Israelis significantly more than Israeli sources are radicalized and polarized against Arabs.
Since you link to other pages, however, wouldn't it be fair to also link to this other Arabic page about the Nova festival ? It was tricky to find because neither it nor its Hebrew counterpart which I found it through call the festival by the ubiquitous media name "Nova Festival '', instead calling it the name رعيم (R'aaim) festival, which is apparently the location of the festival or something like that :
In that page, everything that I have seen said against Hamas in international media is reiterated, including the horrific videos of the Chinese-Israeli young woman being taken on a motorcycle and the video of the German woman Shani.
I'm sure that with some effort that I'm too lazy to expend right now (but will happily do so if you asked me), I can find Arabic equivalents to pages like the ones you posted. The Hebrew Wikipedia "wins" this battle of neutrality in my view by a slim margin since their October 7th page at least links to the page about the Nova massacre, whereas the Arabic one doesn't as far as I can see.
Still, the Hebrew Wikipedia makes some interesting word usage choices, like the fact that they call Hamas "Terrorists" in the authoritative article voice without quoting, and indeed likes to use the phrase "Palestinian Terrorists" a lot. Meanwhile, the English Wikipedia page never describes **ISIS** as terrorists except in the words of someone else, "Designated as Terrorists' ', "Considered by such and such <country name> as terrorist organization' ', etc.... I don't disagree that Hamas has many of the major hallmarks of a terrorist organization, it's just that Wikipedia English still doesn't want to use the emotionally charged word with a much worse group.
In one of the pages you linked, probably the one about the West Bank and Arab Israelis, they put "Human Rights" between quotes when describing human rights organizations. Putting things between quotes is an interesting motte-and-bailey, the Motte is that they're just being accurate, everything a journalist or a second hand source like Wikipedia says is technically reported speech that the strict rules of grammar say should be put between quotes. But the Bailey, of course, is that putting things between quotes has the widespread double meaning that the reporter doesn't trust or agree with the things being reported, so "The 'Human Rights' organizations" is really "Those organizations that call themselves Human Rights organizations in a way that doesn't convince me". Which at least primes the reader or the listener to look for reasons to disbelieve all what follows.
I think that what I got out of this is not (only) that sources outright lie by omission (although they often do), it's that the mere act of organizing information, the act of linking 2 pieces of information, the very statement that they're related and that a reader of one should read the other, is itself a source of bias.
(Needless to say, my username doesn't extend to the citizens of Israel, who I wish the peaceful among them all good and peaceful things)
Yeah, you're right, the Re'emi page does describe all the accusations. Which makes it even weirder that it's not linked off the main Oct7 page and that that page doesn't even mention civilian victims. My guess is that one of the two pages is an outlier. Perhaps the ar wiki is actually in habit of correctly identifying the main facts, and the Oct7 page is a huge outlier, protected by some zealot editors. Or maybe the Oct7 page is business as usual and the Re'emi page was not noticed by its creators.
Note however that the music festival was but one of the sites of the massacre,
No, I still find it difficult to recognize Hebrew letters by name. Reading is even more difficult because each Hebrew letter can be pronounced in at least 3 different-but-related ways (some has even more), those different ways are indicated by diacritics, but those diacritics are omitted from the vast majority of Hebrew writing I have seen : Hebrew on Twitter and YouTube comments, Hebrew on street signs and government IDs, news sites, even some Hebrew videos on Youtube with subtitles aimed at making it easier for learners. Diacritics are only present in the Hebrew bible and a tiny minority of other writings.
I feel that speaking is a bit easier because Hebrew apparently borrows so much from Arabic and English, and the sentence structure looks Arabic-like, but I'm still at the phase where spoken Hebrew appears as an unintelligible stream of sounds to my ears, except the very few words I know (pronouns, "Man", "news", "war", "occupation", ...). I plan to watch a couple of TV series to train my ears and pick a few words, this is the same strategy I'm currently following with German.
There is one Hebrew speaker I'm currently emailing, we write in English about a variety of (non-Israel-related) topics and Hebrew did came up, they assured me that Hebrew is much easier than Arabic and mentioned something about all Hebrew words originating from approximately 200 roots, in contrast to Arabic having 6500+ roots or so. I'm not sure this is relevant, do Arabic learners really care about roots ? Roots were an obscure topic I learned about in high school grammar classes. My email friend is also learning Arabic, so maybe a Hebrew speaker who is trying to learn Arabic is exactly the sort of person to play down the difficulty of Hebrew and focus on the comparative difficulty of Arabic. As a counter signal this video (The Hebrew Alphabet is bad : https://www.youtube.com/watch?v=Q_h541RkCTI) both made me laugh and made me slightly depressed.
"Reading is even more difficult because each Hebrew letter can be pronounced in at least 3 different-but-related ways (some has even more), those different ways are indicated by diacritics, but those diacritics are omitted from the vast majority of Hebrew writing I have see"
I'm surprised, since I think of the large majority of Hebrew letter as consonants with single pronunciations.
I suppose diacritics include the dots that distinguish some pronunciations-- s/sh, v/oo, v/b (the two v's are different letters), c/ch, h/ch....
I make it 24 letters, of which 15, I think (it's been a while since Hebrew school), are consonants with single pronunciations-- a yod (the little letter) is always y, though it might have a vowel diacritic underneath.
I suppose you could argue either way about whether it or English are less phonetic.
I never thought about what a plague the diacritic system would be for people learning the language.
I don't actually know Hebrew. I can read it if it has diacritics or if it's a familiar word and I know a modest number of words.
I find it interesting that I can read Hebrew in Hebrew characters much more easily than English transcriptions of Hebrew.
A theory about why AI is unlikely to be an existential threat to humanity, though it could still be a very serious threat-- I imagine things getting bad-- billions dead as an AI sucks up more and more resources. However, it's going to be tricky for the AI to figure out what resources (possibly including people) it needs to allot to keep itself going, and possibly it will need to prevent and repair sabotage.
This is definitely a good enough premise for science fiction, and possibly reasonable for the real world.
The idea of superintelligence is that it would be trivial for such an AI to solve that kind of problem, and there is no guarantee that the solution it finds will involve keeping any humans around.
Then we get into questions about whether that level of superintelligence is possible at all (I'm ever so tempted to call it super duper intelligence) and whether it can be achieved without trying things in the real world and making mistakes along the way.
I love ASDI -- Artificial Super-Duper Intelligence... let's all start dropping the acronym and when ppl ask what it means we can be all condescending and snarky that they're not following this debate and its jargon sufficiently close enough to be worth discoursing with.
Yep! Personally, the latter is what I see as the big blocker: you have to experiment to develop super-tech, and that means fucking up. Can that entire process happen covertly?
I have noticed something curious about my career in software development: every time I switch jobs, it is of course for higher pay, but the company I jump ship to is always operationally sloppier than the previous one. Now I am at my highest compensation ever, and this org I'm in is hosting a bunch of their code repositories out of a laptop sitting in a shelf at the office, which, not gonna lie, it traumatized me a little when I finally saw this laptop myself. It's definitely some kind of zenith of sloppiness. Hell, I had heard they were hosted out of the office, but I thought it was like out of a Mac mini or something (though is that better than a laptop? It seems better for some reason).
There are a lot of devs here, have your careers followed a similar trajectory? Also, do you have any amusing stories about operational sloppiness like that laptop server?
I suspect this might be related to working in startups / small companies / giant corporations. Startups are improvising a lot, but they also often use the latest technology. Small companies are probably the sloppiest ones (not enough money and prestige to hire an expert on purpose, not enough employees to have an expert appear statistically). Corporations may be slower to adapt, but they usually have a dedicated IT security department; though sometimes the rules are so oppressive that the employees have to work around them to get their work done.
So, in my experience, it seems like a random walk, because my career is a random walk, heh. The biggest examples of sloppiness I have seen somewhere in the middle. (A small company providing IT services to dozens of customers, some of them important big companies, used the same password on all their production servers; didn't change it for over a decade. The password was the name of the original IT admin; all employees in the company knew that.) (In a different company, we were not allowed to use version control, because the local "expert" convinced the management that version control is literally only used for the *published versions* of the software, not during development. We were supposed to share the code updates by e-mail once a week. What actually happened is that we have secretly installed version control that all developers except for the "expert" used. Those were fun times, especially when the "expert" found out one year later.) Things like this would not be imaginable at the beginning of my career or where I work today, because both places contain actual experts.
Well, that was definitely the craziest moment of my career.
That "expert" was the kind of guy that probably inspired the Dunning–Kruger research. He co-founded the company right after university, i.e. he has never worked anywhere else, and he was unfamiliar with whatever happened in software development after his school years. He was a mediocre developer, but he compensated his lack of skill by working a lot of overtime, so the other co-founders (mostly managers) considered him a hard-working genius.
During the interview there were many red flags, but the company was desperate and willing to pay lots of money (almost twice my previous salary), so I took the risk. It was actually quite fun, and definitely a learning experience. That was my only job where I significantly changed how the company worked, because things were so dysfunctional that I could introduce obvious things as revolutionary improvements.
I used git and github in my university years, but it was never something that the unversity taught or even mentioned so it was purely peer effect and the internet.
I'm honestly just morbidly curious about the logistics of an email-based version control system that only pulls and pushes weekly. How did that work ? Where was the canonical version stored ? On the Team Lead's computer ? How was complex merges where the old code have things not in the new and the new code have things not in the old was handled ? Which email client can handle the more than 100 MB of text that large code repositories tend to grow into ? How was the different versions named ? How did all of this scale beyond 3 people ?
I **did** once try to enforce a scheme like this in my dumb freshman years but (1) it was Whatsapp, so at least I had E2E encryption unlike email (2) It wasn't weekly (3) it wasn't a for-profit company for fuck sake, just a meaningless school project (4) it was 4 of us including me
That is the difference between junior and senior developer -- the junior only knows the stuff they teach at universities; can create a thousand lines long program alone in a few months; the senior can collaborate on a hundred thousand lines long program that was created a few years before he joined the team... and he knows how to collaborate with other developers (which includes the knowledge of tools like git). That guy was effectively a junior despite 10 years of experience, because he practically didn't learn anything substantially *new* during those years.
Before I joined the company, that specific department only had two developers. One wrote C++ or Java code, the other wrote stored procedures in a database (which were called by the C++ or Java code). So these two rarely worked on the same file; their collaboration mostly consisted of sending their respective half of code in a zip file to the other guy, who simply replaced all previous code by the new version. And I suppose the "canonical" version consisted of the Java code on the team lead's computer + the database scripts on the database guru's computer. And it worked exactly as you would imagine; bugs caused by someone copying a wrong version of file happened all the time.
Three new programmers were then hired, including me, and the team lead imagined that this way of collaboration would scale. Well, the three of us made our "canonical" version on the secret Subversion repository. We imported the team lead's changes to our repository when he sent us some... luckily, he didn't produce much code, because most of the time he was busy doing analysis and attending meetups. We also sent him the changed files by e-mail once in a while, just like he wished, and we didn't care what he does with them... but he practically stopped developing afterwards, probably because he was too busy merging the new code on his machine. From our perspective, that was only good; we thought that his contribution would be a net negative. (Unfortunately, the database guru refused to use version control at all.) So, it was actually quite easy after a while, when the team lead stopped contributing. Only the first two or three months were horrible.
The code was relatively small, at least when I joined the company (while we still used the e-mail collaboration). The releases were numbered, but the project didn't have branches.
I'm considering my current economic / social situation, and I think I am in a position to start pivoting to an altruistic career.
My only issue is - what career? I have looked at 80000 hours, but the sort of thing I've been doing up to now is not very conducive to their list of top global problems. I actually really like what I've been doing up to now, and I would like to do something similar. I don't particularly want to earn to give - while I like what I do, I hate working in a for-profit environment.
About me: I have a Bachelor's of Engineering. Skills I have picked up through work have been corrosion prevention, asset maintenance, and pressure equipment management. I know how to work in a big heavy industrial environment (and would in fact prefer it).
Thoughts: I think I would be suited to a role in infrastructure development and maintenance, particularly in chemical plants, nuclear power plants, waste or water treatment plants, or fuel farms. I have done some research in the past and didn't like it, and I think I am a poor fit for research specifically (personality / ways of working issue). I think my ideal role would be an engineer, advisor or project manager for some kind of big infrastructure project, so if anyone knows of someone building a desal plant or sanitary sewers as a charitable thing, I am very interested.
Other than that, I would be interested to hear about any projects, opportunities, or general direction for somewhere I can apply my skills.
I am looking to get out of my current industry in the next 2 - 5 years, so sit tight and accumulate career capital isn't really an option. My current industry is fossil fuels.
Edit: if it matters, I am ethnically Chinese, I'm baseline conversationally fluent in Mandarin (which I can improve with some effort) and a bit less fluent in Cantonese. I don't have close family ties in Mainland China that I'm aware of. I'm also willing to work there.
I like the technical side of things and learning in great detail about how it all works. I really like my team - they're highly skilled people with a lot of professional integrity.
I don't like or trust upper management. They tend to demonstrate decision making that is indifferent to or actively hostile to the common good (workforce and society as a whole). They also have the gall to pretend that those same decisions will benefit the public, which to be fair is what all large corporations do. I don't think I'm alone in that assessment, and I would like to spend slightly less time supporting this. And because my direct team often do and care about things not directly related to improving the bottom line (like spend time addressing safety risks and complying with the law), I'm worried for our jobs. It might be "good for the company" to do these things on some level, but since I often deal directly with issues from these high level decisions (chronically undermaintained equipment, skills gaps) it doesn't feel like it really helps the company either. The worst part is, from the enshittification of everything else, I suspect this is a general trend of the for-profit environment.
I think I would like to keep the parts I like (working with skilled coworkers and direct manager with integrity, who I greatly respect, with lots of opportunities to learn and feel useful), and also feel like I work for an organisation with integrity and moral backbone.
Yeah, makes sense. Thanks for the additional context. I hear you on the sometimes nefarious motivations at for-profit companies, although I don't think that's universal.
I unfortunately don't have any useful connections to share. One thing I've heard is that geothermal overlaps greatly with oil and gas from a technology perspective. The space seems mostly for-profit, but I think that's mitigated by the fact that people are also more mission-driven to generate clean energy. Might be worth poking around on LinkedIn or Twitter and messaging people at some of the younger companies. EG https://twitter.com/TimMLatimer
You checked out 80,000 hours, but did you do the career guide (https://80000hours.org/career-planning/)? It's pretty in-depth, I never did finish the process because I decided I am not looking to maximize my impact, I just want to work for an org with a good mission. You could even schedule an interview with them: https://80000hours.org/speak-with-us/
They're full of people who hysterically freak out if you say anything that remotely resembles white nationalism, and then they turn around espousing violent, genocidal jewish ethnonationalist rhetoric. This is literally like some far-right caricature of jews, but it proves itself to be true over and over again. And then you have jews like Bryan Capalan promoting demented open borders policies for every western country, but then conveniently finding a justification for saying that the ethnostate that aligns with his ethnicity is the sole exception to open borders being good. You literally cannot make this shit up.
I'm curious: did you sign up just to write this post?
I'm honestly fine with, say, Vatican city. Or Native American reservations. I'd be fine with a White Separatist nation the size of Israel somewhere in Europe. Or maybe several times the size of Israel. Though Israel is a pluralistic nation and its citizens have civil rights, so comparing it to 'White Nationalism' is a bit disingenuous. Israel is not supremacist and it's far less separatist than, say, Saudi Arabia, which gets far less flack from the UN. I wonder why.
I'm also fine with America, say, not letting someone immigrate because of their ties to gang violence or previous violations of border laws or whatever. Every nation regulates their borders to some extent, even if all citizens inside the country deserve full rights.
The problem with the Israel-Palestine conflict is that Palestinian leaders just haven't been interested in a two state solution on any terms. They've wanted all of Israel. They've been willing to kill for it, even if they've been in a momentarily weaker military position. And frankly, they've burned a lot of their allies along the way. Jordan welcomed them in. They tried to overthrow the Jordanian government. They caused civil unrest in Lebanon. They were welcomed in by Kuwait, then Arafat endorsed Saddam Hussein's invasion. There are a lot fewer Palestinian refugees there now, a tiny fraction of what there were. Even Egypt has been wary about accepting refugees out of fear they would ally with Muslim Brotherhood within Egypt. The hope of far right Israeli nationalists that Palestinians will emigrate is unlikely to pan out. There are few countries still willing to take them in.
I'm not a supporter of Netanyahu. But... if someone wants to destroy you, as Hamas and the PLO do Israel, and is willing to violate the conventional laws of war and attacks babies and civilians, and is adamantly not willing to live peacefully in separation, the number of ethical options available to resolve that situation outside of 'open warfare' is, to put it bluntly, rather limited.
> it's far less separatist than, say, Saudi Arabia,
This is probably an inaccurate usage of "Separatist", it's usually used to denote rebel movements that want to politically secede from a bigger polity. The closest to what you mean is most likely "Segregationist", to mean your intended meaning that it has a first class and other class of citizens, although this term might carry too much baggage from its original context in the USA where the classes were drawn along race boundaries which are somewhat alien to the Middle East.
"The closest to what you mean is most likely "Segregationist", to mean your intended meaning that it has a first class and other class of citizens,"
This is explicitly *not* my intended meaning. It is an incorrect interpretation of my meaning, yet a format that other people attempt, repeatedly, to apply to the situation. And this tendency, I believe, causes people to miscategorize the situation time and again.
Words like 'segregation,' or 'apartheid' all describe systems where one powerful group has, as a primary goal, to extract labor from a weaker group. To do this, both groups need to cooperate. The stronger group actively wants the weaker group as part of society, but not on equal terms. Southern planters went through the trouble of importing slaves from Africa, at some expense.
With Israel (Or Jordan, or Egypt, or Kuwait, or Lebanon or a number of other countries, not just Israel) things would be calmer if Palestinians were just *not there.* This isn't suggesting that *making* Palestinians not be there is a moral idea because Palestinians are human beings and have title to land, etc. But it's pointing out that the constant attempt to equate Israel's relationship to the West Bank and Gaza's people with supremacist models is missing a very important aspect of the whole dynamic, which is that the use of Palestinian labor at low rates by any particular group is really not the motive force of the conflict. This realization, that desire for cheap labor is not the cause of the conflict, is at the heart of 'two state solution' proposals. If you asked the average person in the rural American South during segregation if they wanted to wave a magic wand and make all Blacks disappear, most would probably have said 'no, they should just know their place' or something like that. *That's* segregation. (If you ask Netanyahu (Who represents, arguably, the extreme end of Israeli nationalism) if he'd like to wave a wand and make all Palestinians disappear, he would very likely say 'yes.'
This distinction was my explicit consideration when choosing the word 'separatist' rather than 'segregationist' to describe Israel's relationship to Palestinians.
Ah okay then, sorry for second-guessing your intended meaning.
In what sense, though, is Saudi Arabia "Separatist" by your definition ? I have a lot to say against Saudi Arabia but I'm not aware of a Palestinian-like group of politically-relevant people that the Saudi royal family would like to wave a magic wand and have them disappear tomorrow (unless, of course, you count the rest of the male siblings of Ibn Salman).
> Words like 'segregation,' or 'apartheid' all describe systems where one powerful group has, as a primary goal, to extract labor from a weaker group.
I don't agree with this at all. I think Segregation and Apartheid are just adjectives for a whole catalog of taboo systems (and the associated social/political/economical/etc enforcement schemes) that primarily prioritize exactly what it says on the tin : No race-mixing, of all the varieties your brain can possibly conjure up, in particular. And no X-mixing in general, where X is some attribute that the segregationists are both interested in and classify people accordingly to it.
Nazis were segregationists, they intensely hated white German speakers interbreeding with Jews, Blacks, Slavs, and possibly others I'm not aware of. Hindu nationalists are religion-segregationists, they intensely hate it when both men and women of Hindu faith marry Muslims. Islam is one-sidedly religion-segregationist, Muslim women are forbidden from marrying non-Muslim men, but the reverse is ok. I'm only mentioning Marriage because I regard it as the highest form of social mingling, literally 2 bloodlines merging, all the ideologies I mentioned are also opposed (though less intensely) to lesser forms of social interaction with the inferior groups as well.
I don't see the connection to Labor at all. The Nazis could have given the Ukrainians and the Poles the worst working conditions possible short of slavery and they would have accepted in a heartbeat, possibly liking Nazis more than Stalinists in the process. The Nazis would have massively benefited because they now had several Germanies worth of population that hate Stalin on their East, working for free and bulwarking against invasions. They didn't, they didn't want the Slavs' and East Europeans' labor, they wanted them dead. Similarly for the Jewish scientists and financiers who they kept pressuring till they fled or accepted the camps, those people would have accepted far far less for their (very valuable) labor than any scientist or financier anywhere at the time, and the result is a more technologically and financially advanced Germany that is more efficient at cleansing the world out of the inferiors. The Nazis didn't want this, they went to great lengths to kill those cheap (and quite willing to cooperate) sources of labor, diverting fuel and personnel to extermination camps even as Stalinist boots were inching closer and closer to Berlin.
So the Nazis didn't want the labor of the groups they were segregationists against, how does that counterexample reflect on your definition of segregation/Apartheid ?
> If you asked the average person in the rural American South during segregation if they wanted to wave a magic wand and make all Blacks disappear, most would probably have said no
I again don't agree with this. Maybe in the immediate wake of the 1860s, maybe all the way to the early 1920s. But beginning in the post-WW2 era, when blacks have served in the army that liberated white Europe and were increasingly seen as not going anywhere, south segregationists would have absolutely jumped at any chance to make blacks disappear, indeed they would have jumped at a "2-state solution" of deporting all blacks to Africa (didn't this already happen in the 1800s ? Liberia ?) or giving them native-american-like reservations in isolated pockets of USA and never having to think about them again. If you're American I might be wrong and you might be right, as I'm not American and haven't lived in the 1950s-1960s, I'm judging the racial attitudes from what I have read about the period.
> Who represents, arguably, the extreme end of Israeli nationalism
Not at all, he's a moderate peace dove compared to the extreme end of Israel's right-wing. See Shtetl-Optimized latest comment thread (https://scottaaronson.blog/?p=7632#comments, ctrl-f for "Araboosh") to see a sample of how unhinged Israel's right can go.
If you're saying that every group's base has its lunatic fringe then sure, I agree.
"ctrl-f for "Araboosh") to see a sample of how unhinged Israel's right can go.'
I can't access a comment history for Raphael to see if he's a person or a troll. If you want to point to an elected official saying something similar, that's more along the lines of what I'm talking about. You do get an occasional outburst from the Israeli side, like with Amihai Eliyahu. But he's gotten pushback for his comments and has walked them back calling them 'metaphorical.' And given the circumstances, with over 1000 civilians slaughtered, hostages taken, and hundreds of rockets falling, and international calls to do nothing an individual's burst of rage makes some sense.
Saudi Arabia used to execute those who converted away from Islam. The public practice of any religion other than Islam is still prohibited. Non-Muslim immigrants must convert to Islam. I'm not arguing that Saudia Arabia is attempting to maintain any manner of underclass deliberately, vis apartheid or segregation, and it's practices are somewhat of a different kind compared to Israel as well since, as harsh as Saudi Arabia is, it is more tolerant of people converting towards the particular brand of Islam it espouses. But I think it would be safe to say, at least, that Israel is far more tolerant and pluralistic of a society, in general, than Saudi Arabia is.
"Although many
intolerant statements had been removed, some school textbooks continued to
contain overtly intolerant statements against Jews and Christians and intolerant
references by allusion against Shia and Sufi Muslims and other religious groups.
For example they stated that apostates from Islam should be killed if they do not
repent within three days of being warned and that treachery is a permanent
LHHI: "I have a lot to say against Saudi Arabia but I'm not aware of a Palestinian-like group of politically-relevant people that the Saudi royal family would like to wave a magic wand and have them disappear tomorrow"
I would say that non-Muslims fall squarely into the category of groups that Saudia Arabia would like to remove. Perhaps 'politically relevant' is the critical point here. It's not that Israel is less pluralistic than Saudia Arabia, but that a particular group of people has been made, as you say, 'politically relevant' in Israel.
People may be imprisoned for converting away from Islam in Saudia Arabia. That is 'a group of people.' But for whatever reason they're not deemed 'politically relevant.'
LGBT individuals can be imprisoned for practicing their sexuality. But they're not deemed 'politically relevant.'
I'm not sure how we set the baseline of who should or shouldn't be considered 'politically relevant,' but it seems like the notion provides a lot of wiggle room allowing the conversion of any moral comparison of groups into a popularity contest.
"I don't agree with this at all. I think Segregation and Apartheid are just adjectives for a whole catalog of taboo systems (and the associated social/political/economical/etc enforcement schemes) that primarily prioritize exactly what it says on the tin : No race-mixing, of all the varieties your brain can possibly conjure up, in particular."
Well, that's part of the issue. There are a lot of different 'taboo systems.' I worry that saying "everything I don't like is Apartheid" will potentially dilute the term, akin to saying 'everyone I don't like is a Nazi/fascist.'
In the perpetual gradient of lumpers vs splitters, it might be worthwhile to engage in enough splitting that different types and grades of systems can be compared.
At one point you arged that Segregation usually involves severely disparate power relationships, with first and second class citizens. And I agree with that criteria. Under Apartheid, white people control white schools and they control the schools for Black students as well. American Segregation and South African Apartheid were both overseen by a single, dominant group. And it's worthwhile asking if that's what people are going for. If separate systems might be defined as segregationist (which I'm not in favor of) that would imply that all nationalism which significantly restricts immigration is fundamentally a form of segregation or apartheid. And I'm not saying that such a usage would be *wrong,* but it's not how the term is typically used. It casts such a broad net that it would probably catch a good few fish that you wouldn't want caught.
(Or maybe not? I'm making some assumptions here. Perhaps you are a staunch open-borders type?)
Alternately, we could ask: at what point does 'a system' justifiably become 'two systems.' And does 'one system' vs 'two systems, each with their own independent leadership' matter for how we judge interactions?
One reason to make the distinction between supremacists and separatists is that supremacists often are in conflict with those they seek to dominate, while separatists are more likely to come to agreement with their opposite group. The Black Panthers and some white separatist groups sometimes come to an accord, because both groups simply didn't want to interact with the other.
I'm not aware of legal restrictions in Israel on who one can have sex with. While there are not civil marriages in Israel, interfaith marriages are often performed abroad and then recognized as civil marriages by Israel. They are not illegal. There are social taboos, but including social taboos in the definition of segregation also expands the term singificantly, to the point that you'll end up calling most groups segregationist or apartheid. There are some fairly strong selective pressures, even within liberal communities, and most liberals have certain political groups of people it would be socially taboo to have sex with or to marry, as evidenced by outcomes.
"The Nazis didn't want this,"
I agree that they wanted to exterminate Jews and Roma after working them to death and did so in a way that wouldn't have been predicted by strict material selfishness. The Nazis also explicitly wanted the restoration of German colonies. They slaughtered people who they believed would become the next-generation Polish intelligencia in order to better facilitate Polish submission. The Nazis were also rather unusual in that they even tried to prevent emmigration of oppressed groups, as you suggest.
When offered the grab-bag of political horrors, the Nazis reached their hand in more than a few times.
"So the Nazis didn't want the labor of the groups they were segregationists against, how does that counterexample reflect on your definition of segregation/Apartheid ?"
I would say that the Nazis were genocidal towards Jews and Roma. They engaged in segregation (extraction of labor of an oppressed group) as part of their path towards genocide via labor camps. That is, I would say that they used segregation as a temporary means, but not as an end unto itself, as Jews and Roma, specifically, were concerned. The Nazis also announced their intent to restore German colonies, which suggests long term extraction of labor from oppressed groups.
In the American South, on the other hand, segregation and extraction of labor was an end unto itself. Extermination of American Blacks was not a goal. Voluntary expatriation was considered for a time, such as with the establishment of Liberia which you mention.
"or giving them native-american-like reservations in isolated pockets of USA and never having to think about them again."
The American reservations started out as being effective countries with sovereignty, but as time went on the US government increasingly asserted control over them, diminishing their autonomy and property rights. At least as per Vine Deloria Jr's book "Custer Died for Your Sins." I admit to only having a smaller set of sources that I've consumed as regards Native American politics, so I'm more subject to bias on that topic.
Liberia is... complex. There were some Slaveholders who favored Liberia because they did not want free Blacks in the US. As such, they saw separation of free Blacks as strengthening segregation and slavery. There were also some abolitionists who supported Liberia.
The US contained competing political influences, as most polities do. The ultimate failure of Liberia and the institutionalization of segregation suggests to me that the American racial separatists (as I define the term) essentially lost and the segregationists (as I define the term) won.
And then we add the complication that Liberia was not located in the area that slaves came from originally, and was effectively hegemonic and colonial in its own right. But that's a different discussion.
Saudi Arabia is a shitty kingdom whose royal family is in desperate need of a French Revolution. I hate it almost as much as Israel. If it's a popularity contest, the vast majority of Middle Eastern polities are fiercely competing for the last place.
Like, I get what you're trying to say, judged purely by the metric of "How Many Dead Arabs", Israel is not in the top 5 or even the top 10 list. Saudi Arabia royal scum starved 85000 Yemeni children in a 5 years or so according to English wikipedia. This is a common point brought up again and again by Pro-Israeli, and it's a good point.
On the other hand, I'm not very sure what this point should imply, Israel is in the unique position that its grab bag of claims to the land are all illegitimate by different degrees to Arab hearts and minds, ranging from Religious claims to the UN partition of 1948. The religious claim is not bought by the majority-Muslim and minority-Christian Arabs (not to mention the extreme minority of Atheist Arabs, of which I'm part), and the UN partition is perceived as an unfair and compromised by the US and other colonial powers who had outsized influence in the UN till now. The Israeli aggression in 1967 is perceived to have invalidated all the rest of UN legitimacy anyway. The very legitimate and very sympathetic claim of "approximately 700-800 thousand Jews were expelled from Arab countries, Israel is the only place they can go to live" is prominent in Pro-Israeli circles but vanishingly rare in Pro-Arab circles, and whenever I bring it up in Arab circles it, I have not seen a good response, suggesting it's a good entry point to get Arabs to at least recognize that not all Israelis are illegitimate claimers of the land, and maybe that's precisely why it's so underused.
Meanwhile, the Saudi claim to their land is that they're a royal family that managed to conquer it in the 1920s and before by making several tribal and religious alliances. In my secular and anarchist eyes, this is much more illegitimate than Israel, but for some reason the vast majority of Arabs are convinced by it. They don't perceive Saudi Arabia as an illegitimate state, and that colors their view of its genocides (even though it shouldn't, and even the most legitimate of states should not get a free pass or a milder reaction to its genocides).
So, point 1 : Israel, despite objectively killing way fewer Arabs than Saudi Arabia or Syria, is seen as an illegitimate colonizer, while the other genocidal states are seen as legitimate states ruled by corrupt rulers, or even legitimate states ruled by harsh but ultimately fair rulers. I think this is bullshit, but it is what it is.
Point 2 : What does the Pro-Israel advocate actually want after bringing this up ? Do they want to talk less about the genocides of Israel, or talk more about the genocides of Saudi Arabia and Syria ? Maybe the 2 are equivalent from a practical or purely consequentialist standpoint, but the 2 are absolutely not equivalent from a moral standpoint. In most situations where I have seen Pro-Israelis bring up this point, there is usually a coded request of "Don't bring up Israeli crimes until you have finished talking about all the crimes of every single other country in the Middle East ''. It's coded, it's never expressed outright, but I feel they're almost itching to come out and say it, holding themselves by a tight leash.
I'm an Arab, I think Arabs (and every other ethnicity, society, or cultural group) would benefit a lot from self-criticism and a healthy dose of out-groupism. I think the standard human baseline for outgroup affinity and ingroup criticism is way way way way below the optimal. Islam, if nothing else, is one thing taken for granted among the vast majority of Arabs, to a devastating effect on both Arabs and non-Arabs. The pogroms of Mizrahi Jews have to be acknowledged, and ideally set right. The many many many crimes against Kurds and Yzidis and other minorities has to be acknowledged. I think the Rawls' veil of ignorance is the single most important and underused contribution to ethics in the history of ethics, and I think that Arabs (and for that matter Indians, if the jingoism and nationalism of online Indians are anything to go by) would benefit massively from knowing about it and acting according to it for a few centuries. On the other hand, this still doesn't legitimize or take the sting away from Israeli crimes. I don't feel it should.
I would say that, you have to sometimes recognize that people aren't machines and that perfect discursive symmetry (although a laudable goal that I long for in humans) is not achievable. Sometimes you will not agree with the claims people make or you will agree factually but think that they're phrased in a very dishonest way, I feel that way towards the common Israeli claim "The biggest number of Jewish dead in a single day since the Holocaust". I agree factually (although I actually never see someone deep-diving into the 80 years of records and verifying it, but it feels right), I just think it's very dishonest and propagandistic way to phrase the size of Israeli casualties. But nevertheless, I recognize the Israeli/Jewish fear behind it and I don't stress about its dishonesty too much, people will reach for whatever story or narrative structure they're accustomed with to describe the facts, it just so happens that the Holocaust is a particularly important story that every Jew is familiar with, just like Decolonization and armed struggle is a particularly important story that every Arab is familiar with.
> Perhaps you are a staunch open-borders type
I have long ago ceased the practice of putting labels on my views, every single label I have tried to claim, I discover that there is a great horde of idiots who interpret it in a vastly different way than me, that people make weird associations between it and other labels I don't recognize, that it's not saying all the things I want to say, etc...
On borders, my views are as follows :
1- Something resembling borders is not bad in the broadest possible sketch, and it's not apriori ridiculous that people who are born on the land and lived there longer has a privileged claim to the land than people who weren't
2- That said, the current status quo in the vast majority of the world where it's extremely difficult for people to move and build a new life elsewhere is ridiculous and extreme
3-a Culture and Language are elephants in the room in any border discussion, people hate it when those things differ from the one they have grown up with, and furthermore people can't change them very fast. The problem of language learning is intractable, the most prodigious humans we know of don't exceed 40 languages, and the vast majority of language learners don't know more than 1 or 2
3-b States are the elephant in the room in any border discussions, and the pro-border-control side has to acknowledge that huge tracts of lands claimed by states are not legitimate according to (1), as an example about 90% of Egypt's area is desert, the vast majority of Egyptians live on the Nile banks, in what sense can they claim the huge square of desert around the Nile as their own ?
4- The European Union handling of refugee is an unmitigated disaster, one that is very unfair to Europeans first and to refugees second, morally wrong, and one that will forever burn any goodwill that a sane pro-refugee and pro-immigration position might have for the next 150 years
5- I don't know what should be done, it's a massive issue that seems to admits no general solution, accelerating space colonization so that terran land becomes irrelevant is the only easy and simple solution I can think of
Things are hard and complex, and human beings suck.
> Israel is a pluralistic nation and its citizens have civil rights
But not *equal* rights, I suppose. As the most obvious example, some citizens can invite their extended family to live in the same state, others can't. (I think there are more differences, but I am not really familiar with the situation.)
I mostly agree with what you wrote, and I definitely wouldn't want to have Palestine the next door, but I notice an interesting thing: it seems to me that Palestinians in the West Bank are generally behaving nicer than the ones in Gaza (I may be wrong about this, in which case please correct me), and yet, the former are losing their territory faster than the latter. So I kinda feel like the deal actually offered to Palestinians is something like: "be nice, so that we do not have to fight you... but we will keep displacing you by more civilized means".
> These comment sections are sickening. They're full of people who hysterically freak out if you say anything that remotely resembles white nationalism, and then they turn around espousing violent, genocidal jewish ethnonationalist rhetoric. This is literally like some far-right caricature of jews, but it proves itself to be true over and over again.
Are you feeling better, now that you've gotten this off your chest? What do you think about those posters who are espousing violent, genocidal, antisemitic terrorist propaganda?
Thanks. It's nice to know that they're actually willing to say that to reporters, and reporters are actually willing to report it to the rest of the world.
I suppose you can always find all kinds of opinions in any country, so I am not really surprised that "some X do Y". The question is, how many, and whether they can do something about it.
I know this is intended as dumb whatabouttism (the recent war is about Hamas's massacre spree and has basically nothing to do with settlements) but the direct answer is yes, the fact that the one time Israel actually did a large scale removal of settlements the direct result was Hamas coming to power and bombing Israel for the next twenty years is the main reason Israel hasn't dismantled any settlements since.
What a nice and rosy view of Isreal, the eternally oppressed peace dove that keeps trying to make peace and getting rejected by the violent Arabs.
I wonder what beautiful peace-loving quotes Wikipedia has about it ?
> There is no doubt in my mind that very soon the government of Israel is going to have to address the demographic issue with the utmost seriousness and resolve. This issue above all others will dictate the solution that we must adopt. In the absence of a negotiated agreement – and I do not believe in the realistic prospect of an agreement – we need to implement a unilateral alternative... More and more Palestinians are uninterested in a negotiated, two-state solution, because they want to change the essence of the conflict from an Algerian paradigm to a South African one. From a struggle against 'occupation,' in their parlance, to a struggle for one-man-one-vote. That is, of course, a much cleaner struggle, a much more popular struggle – and ultimately a much more powerful one. ****For us, it would mean the end of the Jewish state... the parameters of a unilateral solution are: To maximize the number of Jews; to minimize the number of Palestinians; not to withdraw to the 1967 border and not to divide Jerusalem****... Twenty-three years ago, Moshe Dayan proposed unilateral autonomy. On the same wavelength, we may have to espouse unilateral separation... ****[it] would inevitably preclude a dialogue with the Palestinians for at least 25 years.****
-- Ehud Olmert, Sharon's deputy leader
Mmm ??
> The significance of the disengagement plan is the freezing of the peace process, and when you freeze that process, *****you prevent the establishment of a Palestinian state, and you prevent a discussion on the refugees, the borders and Jerusalem. Effectively, this whole package called the Palestinian state, with all that it entails, has been removed indefinitely from our agenda.****** And all this with authority and permission. All with a presidential blessing and the ratification of both houses of Congress. That is exactly what happened. You know, the term 'peace process' is a bundle of concepts and commitments. The peace process is the establishment of a Palestinian state with all the security risks that entails. The peace process is the evacuation of settlements, it's the return of refugees, it's the partition of Jerusalem. And all that has now been frozen.... *****what I effectively agreed to with the Americans was that part of the settlements would not be dealt with at all, and the rest will not be dealt with until the Palestinians turn into Finns.***** That is the significance of what we did.
Yeah. But a lot of this is downstream - basically you have people who are pro-settlements for religious reasons (or just because they like living in suburbs across the green line), and people who are anti settlements for peace reasons. But the second category (which I'm still mostly in) lost all political influence after 2005, because they couldn't argue anymore that removing settlements could help peace (so all we're left with is "yeah but pork spending on subpopulations that needlessly cause problems for us is bad", which is about as successful as American arguments against the Jones act).
Taking a step back here, I should specify that "the Israeli government is funding them" is also a bit more ambiguous than it sounds. The national government funds municipalities as a general thing, so it funds settlements under that (though there are issues where special interests push for this amount of funding to be calculated in a way that disproportionately benefits settlements).
The Gaza withdrawal aftermath was, in hindsight, the nail in the Oslo process’s coffin. Like you said, the side (which I sided with) of “if Israel leaves the occupied territories there will be peace” was decisively obliterated by rocket barrages from Gaza.
I wonder is the oak tree still in the garden? There seems to have been a lot of extra building on since his time, but it was happening even in his time (at least by his autobiography):
"I had been living, already for a long time, in the town of Beaconsfield in the County of Bucks; the town which some Colonials imagine to have been named after Lord Beaconsfield the politician. It is rather as if they thought that England had been named after Mr. England the pirate. I am almost tempted to add that I say it with an apology to pirates. I do not know for certain why Disraeli took his title from Beaconsfield, which he scarcely ever visited, rather than from Hughenden, where he lived. But I was told by old Lord Burnham, the founder of the Daily Telegraph, that (as the story ran) he had chosen the title originally intended for Burke, who did live at Beaconsfield and whose legend still clings in many ways to the place. Mr. Garvin, the editor of the Observer, lives in what was once the house of Burke's agent and the oak-tree in my own garden was one of the line of trees that marked the limit of this land. I am glad that Mr. Garvin fits into that political landscape much better than I do; for I admire Burke in many things while disagreeing with him in nearly everything. But Mr. Garvin strikes me as being rather like Burke; in his Irish origin, in his English Conservatism, in his eloquence and gravity and something that can only be called urgency of mind. I once suggested to him that he should appear at a local festival as Burke and I as Fox; a part for which I have no claim except in circumference. But I hope there will never come a dark and difficult hour when political differences become personal, and Mr. Garvin begins to throw daggers about and say that our friendship is at an end.
I have lived in Beaconsfield from the time when it was almost a village to the time when, as the enemy profanely says, it is almost a suburb. It would be truer to say that the two things in some sense still exist side by side; and the popular instinct has recognised the division by actually talking about the Old Town and the New Town. I once planned a massive and exhaustive sociological work, in several volumes, which was to be called "The Two Barbers of Beaconsfield" and based entirely upon the talk of the two excellent citizens to whom I went to get shaved. For those two shops do indeed belong to two different civilisations. The hairdresser of the New Town belongs to the new world and has the spotlessness of the specialist; the other has what may be called the ambidexterity of the peasant, shaving (so to speak) with one hand while he stuffs squirrels or sells tobacco with the other. The latter tells me from his own recollection what happened in Old Beaconsfield; the former, or his assistants, tell me from the Daily Mail what has not happened in a wider world. But I suggest this comparison, merely as an introduction to a parallel matter of local interest; which happens to embody, better perhaps than any other emblem, all those large matters that are more than local. If I wanted to write a book about the whole of this great passage in the history of England, including the Great War and many other changes almost as great, I should write it in the form of a History of the Beaconsfield War Memorial.
The plain primary proposal was that a cross should be set up at the cross-roads. Before the discussion was half over there had entered into it the following subjects of debate: (1) The Position of Woman in the Modern World, (2) Prohibition and the Drink Question, (3) The Excellence or Exaggeration of the Cult of Athletics, (4) The Problem of Unemployment, Especially in Relation to Ex-Service Men, (5) The Question of Support for Hospitals and the General Claims of Surgery and Medicine, (6) The Justice of the War, (7) Above all, or rather under all, for it was in many ways masked or symbolically suggested, the great war of religion which has never ceased to divide mankind, especially since that sign was set up among them. Those who debated the matter were a little group of the inhabitants of a little country town; the rector and the doctor and the bank manager and the respectable trades-men of the place, with a few hangers-on like myself, of the more disreputable professions of journalism or the arts. But the powers that were present there in the spirit came out of all the ages and all the battlefields of history; Mahomet was there and the Iconoclasts, who came riding out of the East to ruin the statues of Italy, and Calvin and Rousseau and the Russian anarchs and all the older England that is buried under Puritanism; and Henry the Third ordering the little images for Westminster and Henry the Fifth, after Agincourt, on his knees before the shrines of Paris. If one could really write that little story of that little place, it would be the greatest of historical monographs.
The first thing to note, as typical of the modern tone, is a certain effect of toleration which actually results in timidity. Religious liberty might be supposed to mean that everybody is free to discuss religion. In practice it means that hardly anybody is allowed to mention it. There is a further qualification of some interest; that in this, as in many things, there is an immense intellectual superiority in the poor, and even in the ignorant. The cottagers of the Old Town either liked the Cross because it was Christian and said so, or else disliked the Cross because it was Popish and said so. But the leaders of the No-Popery Party were ashamed to talk No-Popery. They did not say in so many words that they thought a Crucifix a wicked thing; but they said, in any number of words, that they thought a parish pump or a public fountain or a municipal motor-bus a good thing. But the greater number of them tended to the proposal of a Club Building, especially for ex-service men; where the latter could have refreshment (that is where the Drink Question came in) or play games (that is where the Athletic Question came in) or possibly even share the Club on equal terms with their wives and women-folk (that is where the Wrongs of Women came in) and generally, in fact, enjoy all that we should desire ex-soldiers to enjoy, if there were really any chance of letting them do so. The scheme was in that sense admirable; but, as it proceeded, it became almost too admirable, in the original Latin sense of astonishing. Those who had propounded it called themselves, I need hardly say, the Practical Party. They justly condemned us of the other group as dreamers and mystical visionaries. They set to work to draw up their plans for the Club; and they were certainly plans of the most magnificent completeness. There were to be cricket-fields and football-fields and swimming-baths and golf-courses, for all I know. The incident has a primary moral, with reference to that strange modern notion about what is practical and constructive, which seems merely to mean what is large and largely advertised. By the end of the controversy the plan of the Practical Party had swelled to the ends of the earth and taken on the dimensions of Aladdin's Palace. There was not the remotest chance of collecting subscriptions for such a scheme; at the rate it was developing it might run to millions. Meanwhile, the vision of the mere visionaries could be realised easily for a few hundred pounds."
I am writing a long short story set in the near future (circa GPT-6.5) that features an AI tutor. As a non-technical person fascinated by recent advances in AI, I came up with the following phrase to describe its mind: “the ever-expanding matrix of data points, meaningless in isolation but each quivering with its own unique probabilistic relation to every single other, that served it for a mind.” Question for those who know a lot more than me: does this seem to you to evoke something (potentially?) real? How could it be improved without becoming more technical/verbose?
Not a professional AI dev and I have no idea how much you know, but describing a data point as "quivering with its own unique probabilistic relation to every single other" sounds wrong to me on an intuitive philosophical level, the probabilistic relation is entirely in the eyes of the observer, it's entirely imposed by the network on the data, it's like when you have a bunch of stars in the night sky and you draw lines through them to make them resemble animals and other shapes, the stars aren't quivering with relations to one another, only your eyes wants them to for the purposes of remembering them better. (But you could argue that each data point is in fact merely one revealed sample from the hypothetical and opaque "Target Function", which the AI is trying to approximate by fitting its parameters as close as it can the samples it can see. In this sense, every data point is a "clue" or a "piece of evidence" that holds the key to unlocking all the other data points and indeed the hidden process generating the data points. Describing this as "Quivering" is still weird, maybe use "latent" or "pregnant" instead.)
Also, the data points an ML system is trained on are not its mind, they're the thing it's mind used to get good at finding patterns. The mind of an ML system is the architecture + the weights. You ***could*** argue that the model weights are essentially just an extremly efficient and lossy compression of the training set and that would be a valid and fruitful point that many people made, but it still sounds very wrong to say that data points form the mind of anything.
To be clear, are you saying the “network” that’s “imposing” the probabilistic relation here is the neural network/AI architecture itself? If so, I *think* we’re on the same page: I’m conceiving the Tutor’s mind as being the matrix of probabilistic relations (for “matrix”, read “network”; for “probabilistic relations” read “weights”) as opposed to the data points themselves. And I’m imagining a future model that’s capable of reweighting itself in response to external stimuli (hence “quivering”). I suppose my follow-up question would be, do you think “probabilistic relations” conveys what model weights are? Or am I hopelessly at sea here?
Right, by 'network' I mean the neural network. But your phrase said "A matrix of data points [...] that served it for a mind". If by "matrix" you just mean a neural network then you're right, but then it would be inaccurate to say that the matrix is of data points, it's not, it's a matrix of arithmetic, a big huge network of sums and multiplies and special operations called activation functions. The data points are the input to this matrix, and its outputs are repeatedly tuned (via the tunable weights) so that they match the known, expected output, at this point it's said (or hoped) that matrix is "trained", it has mastered deriving the answer from the question.
In any case, the data points don't "conceptually" feel like they're part of the model, they're part of the external stimuli that we hit the model with. The "matrix" is of weights, or in your language, of probabilistic relations.
> do you think “probabilistic relations” conveys what model weights are
It wouldn't be an inaccurate way to put it, but I personally wouldn't describe them as such. I view weights as the parameters for an approximation function that attempts as much as possible to stick to the curves and the contours of an unknown hidden function or probability distribution that we only have samples of. If I had to describe them poetically, I would say they're "Inner Structure" or "Hidden Patterns". They're the tea leaves arranged in such a way that they mimic the shape of phenomena, such that by reading them in a certain way you can get insight into the phenomena.
Thanks, this is great. The “of data points” is the real howler here, then. Interesting challenge to express what’s actually going on (insofar as it’s known/comprehensible) in poetic terms! Thanks again for taking the time to tease this out.
You're very welcome. Always been a scifi fan, especially the hard kind that tries to give textbooks a run for their money. Always happy to help with what little I know.
I built my own artificial neural network once as part of an Intro to Machine Learning course, and I think it mostly makes sense except for one thing: the matrix isn't ever-expanding, GPT does not grow once deployed. Maybe replace "ever-expanding" with "inscrutable"?
In the story, the AI tutor is multi-modal, with detailed live input from at least three senses, and explicitly billed as “growing alongside its user”; I’ve intentionally left the mechanics of this unstated (mainly because, like I said, no technical background), but I guess (and I do mean *guess*) this could either work in terms of an ongoing context window that remains open for years, constantly tweaking the persona being evoked from a (stable, unchanging) model; or a model whose weights adapt dynamically according to user input, sort of a very much more sophisticated, and ongoing, version of current RLHF techniques. Does any of this sound at all plausible as something that might happen within the next 5-10 years?
See, in any of those scenarios you are putting forward, the AI is not ever-expanding, as that would mean it increasingly takes up more space, which I'm pretty sure is not something anyone is aiming for. If you talk of an ever-expanding matrix, you are talking of something that you need to keep throwing more storage at, which could maybe be a plot point. I think if you go with "ever-improving" or "ever-learning" it gets the same point across without offending any engineers (it would be a pretty bad piece of software, one that never stops growing).
But those suggestions I put forth maybe seem too abstract when applied to something as concrete as a matrix, so maybe "ever-churning" is best, as it really captures a very real and chaotic aspect of this matrix.
I also don't understand the technical details, but would it be plausible for the AI tutor to be a cloud service that can dynamically get more memory allocated, using a strategy like "at first, allocate minimum memory that is enough to make the end user happy, later expand when you get feedback that this is no longer enough"? The feedback would be provided either by user or by the AI itself (it would somehow reflect on its own answers).
I think I would need to be an AI researcher to truly answer that one, but it is sci-fi, so sure, maybe the idea of tossing more parameters for specific users is valid sci-fi. Interesting architecture, like you have your static main model that provides most of the capability, then there are add ons for specific users to fine tune the thing.
I definitely thought of it as a cloud service. And the kid’s mother is a developer (“Summoner”; I’m leaning quite hard into the idea that certain aspects of working with future models will involve very different skill sets), so could plausibly have access to privileged amounts of memory/parameters/add-ons.
Yeah, there are lots of wanna-be gangstas in European cities and yes, we do have inner-city and 'urban' youth who come from areas of high crime and deprivation, even in Ireland, but it's never not funny to me that they faithfully copy American hip-hop/rap culture trying to be what they're not (i.e. East or West Coast).
It's even funnier when, as in the linked images, it's guys from nice places.
I similarly cringe at "gangstas" in post-communist Eastern Europe, describing how they grew up as "rebels" in a "ghetto"... when in fact the regime had mandatory employment for everyone, and the police would beat you up for dressing differently or otherwise sticking out of the crowd. Dudes very likely had exactly the same childhood as everyone else.
Otoh I'd guess that your average American would find, say, Moldovan hip hop just as funny, even though Moldova is considerably grimier and poorer than what even the minority communities in US face.
I remember in my high school German class the teacher played us some German hip hop, probably trying to appeal to the youth, and I found it incredibly silly. Maybe German just doesn't sound good as a language for rapping, though.
That sounds like a human rights violation and a case for the European Court of Human Rights (though that's my prejudice against the German language rearing its head) 😁
You know I had to look it up, and yeah - not the best choice:
There's a Belfast Irish-language rap group (well at least they're keeping the native language alive) and they're something more grounded; at least this is authentically depressing in an Irish way:
Any thoughts about getting good at simple navigating in a city? I had an acupuncture appointment in an unfamiliar part of the city, and I spent a lot of time being lost. (It seemed like a lot of time, actually about an hour and a half.) Some of this may be aging, but I've always been bad at that sort of thing.
Some of it may have been bad directions, some of it may have been failing to follow good directions.
There was possibly some wrong-headedness involved. "Go across the bridge", they say. I can't believe it's this big elevated four-lane highway with no obvious way for a pedestrian to get onto it. Surely there must be a little footbridge across the river in the park. No, it was the big bridge-- there was a not-easy-to-see pedestrian access from where I was.
I suspect I need better skills using the maps on my phone, and possibly I should carry a paper mapbook.
I'm not sure I reliably stay conscious of orientation when I turn a corner.
I'm especially interested in accounts of getting better at orientation starting from being bad at it, but if you've been good at orientation all your life and you feel a need to tell me, go ahead, I can't stop you.
If you've actually taught people to be good at orientation (not just have ideas about how you might be able to teach it), I'm interested.
Some might be interested in where I was. Being interested in that sort of thing probably has something to do with being good at orientation. I'm in Philadelphia. I was starting from 10th and Mountain St. (South Philadelphia). I was going to West Philadelphia Community Acupuncture at 4636 Woodland Avenue.
I caught a bus to 33rd and Dickinson. So far so good. It was cold, and it was taking a while for the next bus to show up, so I thought I could walk it. It was only a mile or so. I found myself at a park involving the Schuylkill River. There was a giant FedEx building with inconvenient fences to get past before I could get to the big bridge. Somehow, I was at 49th and Paschall St. I'll note that this an area with few shops and few people on the street to ask for directions.
It turns out there's a trolley that goes from Center City to near the acupuncture place, which should solve the specific problem in the future, but I'm still interested in what might solve the general problem.
I've had some success in improving my navigation the way I improved my time sense. The basic idea is to explicitly make a prediction, and then quickly check it (by looking at a map or clock), and repeat the process a lot. Eventually my predictions got better. And the other way around, too: for the time sense, I would always make a prediction before I looked at a clock.
I am not ashamed to use navigation software even when walking in my own city. Even if I could get somewhere without it, sometimes it shows me a shorter way. I don't use the voice navigation, just look at the map every few streets, where I am now and where I need to go next.
When we are on a vacation, I try to walk around by foot as much as possible. It is my experience that when I walk I get better at understanding what is where, but when I use any kids of transit, I only remember disconnected points without knowing how they are related to each other.
I was trying to use navigation software, and not doing well with it. Possibly there are some basic skills for navigation software that I need, and possibly I was too impatient to use information which was there.
My favourite navigation software is "Mapy.cz" which uses data from OpenStreetMap, but also provides interesting local information, for example tourist maps. Another advantage is that (unlike Google Maps) it shows your position on the map even when you do not have an internet connection.
I turn on "Position" on my Android, then start the application. Sometimes it takes a while until it displays my position on the map (if I do not have an internet connection, it can be minutes, otherwise it is seconds).
Okay, now I know where I am, but do not know which direction is which, i.e. where is North. One option is to start walking in a random direction and later check how the position marker is moving. Another option is to read the names of the streets and compare to the map.
I always keep the North on the map pointing upwards. Rotating electronic maps are confusing. (It is the other way round with *paper* maps; I prefer to rotate them so that the direction I am walking is upwards. The difference is that with paper, even if I rotate it, I can still intuitively perceive the positions and directions from the perspective of the paper. With software, it is not me who rotates the map.)
As I walk, I take a look at the map approximately once a minute. If there is a dense grid of streets, I just keep walking approximately in the direction towards my target. If there are long streets that are not transversely connected, I need to plan my walk in advance. Sometimes I ask the navigation to give me the shortest path to the target, sometimes I make the decisions myself.
(You need to check whether the navigation is in the "car" or "walk" mode.)
I generally like to walk by foot. Unless there is time pressure, I would rather walk for 30 minutes than take a bus that takes 10 minutes. (If you add walking to/from the bus stop and waiting for the bus, you find that taking the bus actually saves less time than it seems.) I live near the center of a city, so the 30 minutes range actually covers a lot. (I also use the navigation software to tell me how far something is.)
It seems to me, statistically, that women suck at orientation, but I have no idea how much of that is innate, and how much is learned helplessness (potentially motivated by a lack of interest). One difference I noticed is that my wife needs to be 100% sure to choose the right direction... which she can't, and therefore she ends up asking people; while I am more like "if I walk by foot, the worst thing that can possibly happen is that I will walk one street in a wrong direction, which is not a big deal". (I often make a good guess based on incomplete information, but the point is that I am not afraid to make a guess.)
Also, it seems to me that men often describe their paths "mathematically", like "go straight ahead three blocks, then turn left and go one more block", while women often describe their paths "contextually" like "go straight ahead until you see a garden with beautiful flowers, then turn towards the nice trees". The male version works better with maps (unless you use Street View on Google Maps).
Note that Google Maps does show position when offline these days. I recently spent a month travelling around Europe without a data plan - selected a route while on Wifi, then GPS continued to track my location until I got to the destination (admittedly not that accurately around tall buildings).
I generally find that Google Maps has the best data on public transport, although maybe other services have caught up since. Note that it doesn't allow time to buy tickets, so include extra time if you don't have a pass.
FWIW, I've found this type of direction to be not only more common, but much more useful, when navigating rural roads in America. You can also do directions by road name, like "take US 123 N, then take a right onto state route 45 E, go 30 minutes and take a left onto state route 67 N". But a lot of those road names are relatively new, and when talking to older generations you get directions like "take a left at the Methodist church, go over the creek and past the white water tower, take a right after the courthouse, then another right before the old Stevenson place".
Thank you. I'll check what map apps are available for my iphone. One of my worst problems with maps on my cell is getting it to show my where I am (I do have location information turned on-- that is, I keep turning it on).
I've heard that part of the gender issue might be that boys are given more freedom to run around at an early age, while girls are held close. I'm not sure what age that would be. Early toddler? Maybe even crawling?
This could make a difference because people remember routes much better when they control their movement.
I somewhat agree about buses taking longer for short trips, though note I got into trouble in west Philadelphia because I was too impatient to take the bus.
My impression is that there's a risk of both men and women overing directions related to landmarks that no longer exist, but maybe that's just from jokes.
I often enough have to accept setting off in the wrong direction and then checking, but note that impatience is one of my issues. I also don't walk very fast, so extra blocks can add up to being late.
Learned helplessness reminds me of the discussion of gender and throwing from _The Frailty Myth_. How people learn to throw has been studied. Athleticism for girls and women is more accepted these days, but there used to be a strong concept of "throwing like a girl". "Throwing like a girl" is actually throwing like a person who's new to throwing. So it was urgent to get a boy (whether by teaching or mockery) to not throw like a girl, but if a girl threw like a girl, well, what do you expect?
For orientation, it looks like Philly street numbers (appear to) go up as you get further from the Delaware River. I'd suggest using that to figure out whether you're going away from, or towards, the river (and then you know which direction is north as well). The east-west streets don't seem to have any pattern to the names, unfortunately.
I do exactly the same in the Manhattan grid - higher numbered streets are north, higher-numbered avenues are west.
For the specific situation you describe (33rd and Dickinson to the acupuncture place), I don't see any nearby pedestrian-friendly routes (based on satellite and street views). Probably your best bet was to stick with the bus.
Did taking Gaza and the West Bank after the 6 Day War make sense for Israel? I'm sure it made sense at the time to have borders that looked more defensible, but having borders which include regions of hostile people-- people who have been persistently provoked-- give a military advantage?
I'd always figured that the main reason was to ensure that Israel controlled the borders between Palestine and other countries, and thus the flow of weapons and fighters into Palestine. It's not perfect, of course, but it could be a whole lot worse from Israel's perspective, like if Gaza could launch rockets that were up to modern standards.
I think taking them made sense. Egypt and Jordan were hostile and pushing back the borders towards their main population centres and away from yours is very sensible.
Keeping them is much more debatable. Trading Sinai back to Egypt as part of a peace deal was a good outcome, but they probably should have given Gaza back to Egypt at the same time. Equally, once a stable peace with Jordan was achieved I think it would have made sense to return control of the West Bank to them (and arguably it still would make sense).
They tried to give Gaza back to Egypt as part of the peace deal - Egypt point blank refused (and was willing to blow up the peace deal over it).
I don't think any such offer was made re the west bank (which would be more awkward since Israel does want some disconnected chunks of it), but if it was I doubt Jordan would take it either.
The 6 day war cannot be understood unless as a war of expansion. Any other objective doesn't make sense, why else would you take Sinai, 6% of Egypt's area, 3 times that of **current Israel** (so including the West Bank) for its then-population of less than 1/3 of its current population ? How would security increase when you take this vast West Virginia-sized desert plain ? More area to defend and more borders to be attacked from ? it doesn't.
It makes total and perfect sense when understood as a war of a Zionist flavor of Lebensraum. Here are the Likud party for example describing why it wants Colonists and regards them as an element of its National Security :
> Settlement, both urban and rural, in all parts of the Land of Israel is the focal point of the Zionist effort to redeem the country, to maintain vital security areas and serves as a reservoir of strength and inspiration for the renewal of the pioneering spirit.
Look at the words used, "redeem", as if the land is dirty or sinful when not inhabited by Jews, a "reservoir" of strength and "renewal", does that remind you of anything ? Who else in the 20th century looked at population and demographics like that and used those words ?
56 years later it's mostly a failure because Egypt took back Sinai and International Pressure took back Gaza, while the Zionist settlement of the kind that established Israel prior to 1948 began to look more and more like the settler colonialism it is, increasingly unpopular as the 2nd half of the 20th century unfolded. The settlers in the West Bank currently number at 7-8% of all Israeli Jews. The Golan Heights remained with a meager population of < 6000 Jew, which the Likud approved a plan to boost and bankrolled 317 million dollar in 2021-2022 https://www.timesofisrael.com/this-is-our-moment-israel-okays-major-plan-to-boost-golan-double-its-population/.
So asking why Israel invaded in the summer of 1967 is a bit like asking why Hitler invaded the Soviet Union in the summer of 1941, that was the whole point all along.
"why else would you take Sinai, 6% of Egypt's area, 3 times that of **current Israel**?"
You mean, aside from setting up a defensible position on the east bank of the Suez Canal, a formidable natural barrier between Israel and one of the nations plotting to destroy it, and adding a couple hundred kilometers of strategic depth that you can easily afford to yield if necessary? I mean, what does the Yom Kippur war look like if the Egyptian Army starts out just across from the Eilat-Gaza line?
You're not even trying, are you? I mean, GPT-2 trained on a steady diet of Arab propaganda could do better than this.
> a formidable natural barrier between Israel and one of the nations plotting to destroy it
Yes yes, you're actually right. And Nazi Germany really didn't want to invade the Soviet Union to commit genocide of the Slavs or to demographically replace the inhabitants, they just wanted the Ural mountains as a formidble natural barrier between Germany and one of the nations plotting to destroy it, and the couple of hundreds of kilometers of the vast East European Plain as Lebensr... - uh sorry Strategic Depth - they can easily afford to yield if necessary. I mean, what does Operation Kutuzov look like if the Soviet Army started from occupied Poland demarcation line ?
Perhaps seen in this light, October 7th might actually have been a stroke of genius on part of Hamas against a nation that plotted to destroy it, they added a couple of tens of kilometers as Strategic Depth that they can easily afford to lose if necessary. I mean, what does the Gaza invasion look like if the IDF started right at the Gaza's borders and without the hostages as leverage ?
I read John as expressing exasperation with your argument style, which appears to be to start with the premise that Israel is evil, and proceeding to fit the available evidence to that premise, in order to prove that premise.
You responded to his exasperation by arguing that if Israel wanted to be the evil it is, it could conquer a large stretch of territory and claim it's only doing to establish strategic depth while it quietly revels in the genocide you suggest they really want to do. But you never seem to get around to proving Israel is evil; all you did is say "if they *were* evil, it could look like what we're seeing". You never ruled out "if they were good, it could also look like what we're seeing", and then moved on as if you had.
> So witty and original.
That's just it. Circular arguments have been around for thousands of years.
Well first off, I'm quite exasperated with lots of Pro-Israel commenters too. I don't remember calling them literal NPCs (and of the outdated, low quality kind). When you're discussing thousands dead and tens of thousands injured and millions displaced, it's inevitable that you will get mad, most likely for the side you identify with the most and against the other one. It's difficult, but worthwhile, to restrain oneself from calling the other side dumb, delusional, or evil, even though it can feel sometimes they're 100% that, and they could in fact be exactly that. So Mr. Schilling started it off on quite the wrong foot by saying I'm worse than a GPT-2 instance, even though I'm in fact a much more power-efficient and multi-modal machine than that. This probably didn't help me see his argument or his beliefs in the best light possible.
> appears to be to start with the premise that Israel is evil
I start off with the premise that Israel is evil because Israel is evil, this is my conclusion that I formed over years and years of seeing and hearing the evil that Israel does, the sheer mustache-twirling genocidal intent that you wouldn't believe in a dialogue written by a bad comic artist, coming out of their senior government officials again and again and again. I can no more enter an Israel-Palestine argument with the serious or earnest assumption that Israel is remotely good as you can enter a WW2 argument with the serious or earnest assumption that Nazi Germany is remotely good. Apologies for the overused and boring Nazi analogy, tell me something else you hate as bad as the Nazis and I will stop Godwinning.
And yet, I try as much as humanly possible to not let this conclusion color the **reasoning** I give as to why Israel is evil. You could start with a conclusion for the "wrong" reasons and still obtain the "right" reasons for it anyway. Mendeleev claimed he saw the Periodic Table in a dream, and George Cantor based his hierarchy of infinities on Christian Theology. I'm not as intelligent as Cantor or Mendeleev, far from it, I'm just illustrating the general strategy of "Start with a conclusion that you know to be true even though nobody else will possibly believe you, find the evidence that people will believe, and establish the necessary and believable trail from the evidence that everybody will accept to the conclusion that no one will''.
You're very free to say my provided reasoning is wrong and explain why, but starting from a conclusion is not apriori a bad thing to do.
> it could conquer a large stretch of territory and claim it's only doing to establish strategic depth while it quietly revels in the genocide you suggest they really want to do
You seem to misunderstand, I'm not saying it *could*, I'm saying it *did*. I'm satirizing Mr. Schilling's Russell conjugation of Israel's action, "Uhm no, they did not invade the West Bank, the Golan, and Sinai because they're a genocidal state hellbent on ethnic cleansing as much of the middle east as possible of Arabs with as little consequences as possible, no no no, they just wanted Strategic Depth you see. They're justified in this because.... uh **checks notes**... the Arabs were plotting to destroy them, starting a war because you suspect the other side is thinking of destroying you is totally a legitimate and justifiable thing"
So I'm using the exact same mental toolkit that I see Mr. Schilling using to justify Israel's actions, and turning it around to justify the Nazi invasion of the Soviet Union, in an attempt to show him that his argument Proves Too Much [1].
> you never ruled out "if they were good, it could also look like what we're seeing"
I view this as so patently ridiculous and implausible that anyone claiming it is the one who has the burden of proof on their shoulders, not me. "Good" states don't start a war and invade their neighbours because "They were plotting to destroy me".
I can see how you might be genuinely exasperated at pro-Israel comments. I can also see how you might be just saying that in order to gather attention. But if I assume the former, I still notice you're acting as if your exasperation entitles you to being able to claim what you claim and demand the other side prove otherwise. This doesn't work if the other side's evidence is already well established (e.g. commonly understood principles of warfare, history of Israel, history of the Six Day War, and reports of Hamas having assaulted a non-military event and killed a lot of civilians, among others). It doesn't matter how long you spent coming up with your unorthodox views. It especially doesn't matter if, in the course of developing those views, your reasoning requires everyone else to forget what they know about how wars are fought. Even if it turns out you're correct, you're necessarily up against a lot of established evidence, and it's foolish to expect everyone who knows that stuff to drop it immediately as soon as someone comes into the room claiming "everything you know is wrong".
Starting from your conclusion is even *more* patently ridiculous than you seem to think my claim about Israel behaving conscientiously is, for reasons you ought to know if you're familiar enough with rationalist literature to be able to dig up an SSC reference.
So no, you can't use the same mental toolkit. And you keep acting as if you can, and that's wasting everyone else's time.
Oh come on, just go away already. You are flatly lying about easily-confirmable facts which is probably the _least_ tiresome part of your broken record.
Sadly Substack's "mute" function only partly works but you are well worthy of it anyway.
Hey, don't be so sullen, look at the bright part : you can bully substack into fixing their shitty software using me, tell them someone is being Anti-Semitic on the Internet, that always works.
Please make them fix the atrocious load times while you're at it, tell them Anti-Semites are using the time to read and contemplate Mein Kampf on every page load.
In retrospect it was a mistake, but I can understand at the time why they thought it was a good idea (and Jordan and Egypt didn't come to the table for peace negotiations until like a decade later, by which time they felt good and rid of the WB and Gaza and refused to take them back even when offered).
Taking the West Bank is definitely a defensive win for Israel. I’m not a fan of the settlements though, probably Israel just needs to control the border area along the Jordan along with a large number of military roads. Without that Israel could be bifurcated in an hour by a hostile force.
Gaza didn’t seem that risky at the time as the population was small. It’s probably preferable to a border with Egypt.
Tennyson is one of my favorite poets. Nobody (including myself!) reads his longer "epic" poems these days, and probably hasn't for many decades past. But as a rule of thumb if a Tennyson poem fits on one page then it is probably pretty good. Examples include "Break, Break, Break", a sad poem about the death of his friend Arthur Hallam, and possibly his best known poem, The Charge of the Light Brigade", recited quite well here, with extracts from the 1968 film:
Here's a short quote from his poem Ulysses, which should be encouraging for old timers who worry that they're washed up and may never achieve anything more!
I think P(Doom) from AI is very low, < .01. over the next thousand years. But even if I thought it were around .05 or twice that, I would be in favor of going full speed ahead with AI.
There are those who consider human flourishing to be this amazingly great thing. If so, it is because humans in the past have been brave, bold, intrepid. It is because they plunged themselves into mysterious domains without fear.
The argument for being extra cautious about AI is an argument not that humans are precious but that they aren’t any longer worth a shit. That they have become a race of cowards.
>It is because they plunged themselves into mysterious domains without fear.
They plunged THEMSELVES into mysterious domains, they risked their lives, not all of the future of humanity.
>The argument for being extra cautious about AI is an argument not that humans are precious but that they aren’t any longer worth a shit. That they have become a race of cowards.
I cannot be the only one here really, really sick of these platitude based arguments against AI caution.
What if the probability of extinction for humanity were 50%? Would all of these platitudes still hold? No? Then it means the platitudes aren't what you really believe - the math is what actually matters. And if you're willing to flip a coin for humanity's future then there's no point discussing this.
i dont get why P(doom) or P(heaven) is a thing; the danger is AI becoming an extremely efficient way to disenfranchise labor by allowing capital to capture its value at a fraction of the cost. Our society couldn't handle Covid forcing out labor for a brief time; we arent set up for a "retail world" with a majority part time, just in time workforce.
if there is a plus side what scares me is no one is articulating it. what jobs can AI create?
Growing economies tend to create jobs. We might not be able to predict what all the new jobs will be; plenty of jobs exist today which few would have foreseen 75 years ago.
I recommend https://www.astralcodexten.com/p/pause-for-thought-the-ai-pause-debate . The question mostly isn't "create AI" vs. "never create AI", it's "how much do we balance speed vs. caution". I don't think driving less than my car's max speed of 130 mph makes me a coward, and I don't think attempts to balance caution vs. speed in AI make us cowards either.
I agree it would be disappointing if we never developed AI.
Has anyone done the research about kidney donation for ≈fifty-year-olds? Is there a significant difference in benefits (to the recipient) or risks (to the donor) versus the case of a younger donor?
I'm kind of reading between the lines here, but I would think of this as a combination of two factors. First, the kidney ages and dies at the same rate as the rest of the body, so if you're older than your recipient, their kidney will be older than the rest of their body and more likely to fail. The average recipient is 45 - 50, so it probably gets worse as you go significantly beyond that. Second, younger kidneys are probably better in general - I don't know if this exactly tracks GFR, but if it does, the statistics are at https://www.researchgate.net/figure/Age-and-gender-specific-GFR-reference-ranges-based-on-measured-GFRs-from-2974_fig2_329127776 . But age related decline after 35 is very slow and at age 50 it's still 90% of what it was in youth. Overall I think low advantage of younger over older before 50, though there's probably some just on priors.
I don't see a good reason why the health at 60 of a person who donates at 50 should be different from the health at 60 of a person who donates at 20, but I haven't read studies proven this. I do think short-term surgical outcomes might be worse, and I personally wouldn't donate if I were above 50, just out of vague fear of this. I do know there are hospitals that accept >50 year old donors, so they must not be too concerned.
Do all parts of the body age at the same rate? At first glance, it seem that the heart and brain are likely to go first, while the digestive tract ages more slowly and could probably go on for a while if the heart and/or brain permitted.
Hm, perhaps, but colon/rectum cancer is one of the most common cancer types, and according to this chart, it increases quite steadily with age, reaches a relevant level (where screening is recommended) around age 50 and reaches 2% risk per year at 75+.
It's still true that more people die of heart disease, so perhaps the heart ages faster in some sense. But the digestive tract doesn't seem decades behind in terms of years until failure.
A lot of failures of the cardiovascular system kill you right away, though, while a lot of failures of the digestive system just make you uncomfortable or miserable or stick you with a colostomy bag or something.
I've been reading an excellent book called "The Origins of Music", ed. Nils L. Wallin, Bjorn Merker, Steven Brown, MIT Press - a subject of great interest to me. There's 3 theories - 1 music evolved from language, 2 language evolved from music, 3 both diverged from a common precursor—referred to by Steven Brown as "musilanguage" and by Steven Mithen as "Hmmmmm" (Steven J. Mithen "The Singing Neanderthals: The Origins of Music, Language, Mind, and Body", Harvard University Press). Steven Pinker called music auditory cheesecake and thinks it's a spandrel, so he's #2. Mithen's book and Brown's chapter make a good case for #3. I'm for #1, but I'm obviously biased because I'm a musician. Any of the learned boffins here have opinions?
Yes I've seen that. One piece of support for Jordania which is missing from that review is the detail that humans singing in resonant intervals will sound louder and therefore more threatening.
It’s hard to see how #1 (music from language) can be true. Many animals make “music”: apes beat trees to a rhythm, birds and marine animals “sing”, etc. Our voices have natural pitches, and harmonizing in octaves, fifths, and thirds is based on vibrational modes. Language is thoroughly unnecessary for music to exist.
Yes, exactly that! I had a music theory textbook that had a chapter on “primitive” tribal music, and it talked about how “more primitive” tribes had what appeared to be very complex vocalizations, but they were just mimicking bird songs. More “advanced” tribes had far simpler melodies, just basic intervals around the root, but - that was a sign of independent organization, first glimpse of structured music composition, abstracted from any specific bird song.
Don’t many species of animals sing? I thought dogs/wolves and possibly dolphins even sang “for fun”, or to communicate emotions etc. What’s the case for music evolving from language if it predates language?
Here is how I've come to believe Global Warning will be resolved:
1. CO₂ emissions are already rapidly being replaced by cheaper renewables. It will take more decades than anyone would like to go to zero, but the path is clear.
2. What isn't widely understood is that that doesn't bring CO₂ levels down, since CO₂ is quite stable in the atmosphere. The solution to that will be Carbon Sequestration. You capture CO₂ from the air, and pump it into underground caverns. The pumping part is well established technology from the natural gas industry. Capturing is only getting started, and will also take an uncomfortable number of decades to work at scale.
3. While waiting for points 1 & 2 to pan out, seeding the stratosphere with SO₂ can bring temperatures down by at least 1 degree Celsius, probably more, for about 2 years. This is known because volcanoes regularly does exactly that. Someone just needs to do it, despite the "Precautionary Principle" crowd.
I don't think this is super original. All three pieces are well known. Has anyone seen a "real" essay describing something like this 3 step plan/prediction?
I basically agree, although I have concerns about geoengineering. Sequestration/capture won't just be pumping CO2 into caverns. Direct air capture to create synthetic fuels and enhanced weathering also sound pretty interesting.
That’s exactly the solution. In fact net zero means Sequestration anyway - hence the net, as far as I know. No idea why this isn’t more well known. Of course if the Sequestration is to be achieved the energy to make it work has to be carbon free or we are chasing our tail. So we are not there yet (although we could use excess wind production to sequester)
This assumes, as most discussions of the topic do, that warming is bad. If you actually try to look at effects of climate change you discover that some are negative, some positive, the size of both hard enough to estimate that we don't know if the net effect is positive or negative. It's an issue I have discussed at some length on my substack — posts are sorted and webbed at http://www.daviddfriedman.com/Sorted_Posts.html and the first cluster is on climate.
One simple test of your knowledge: Can you list the major positive externalities? Talking only about negative ones makes it easy to conclude that the net is negative.
Sorry, I don't have time to go read 15 blog posts. From what I do know so far, I don't find the positive externalities point very convincing. You know how they say a startup has to be 10X better in some respect to take market share from incumbents? I think the same is true when arguing against an existing body of knowledge.
For a detailed example of just how bad work supporting the orthodox account can be — an attempt to estimate cost of carbon for the next three centuries, published in _Nature_, that implicitly assumes no medical progress from now to 2300, see:
If interested, read one of those posts and see if it persuades you to look at others. I'm not prepared to redo the argument in a comment thread and even if I were it takes more than that to make the case.
It may or may not be relevant that fifty years ago, when the perils of population growth had the same status that the perils of climate change have now, I made essentially the same argument. What happened since then was the precise opposite of the then orthodox view. Poor countries continued to grow their population and, instead of getting poorer and hungrier, got less poor and less hungry.
I am less inclined than you are to put much weight on "all the experts say," having observed at least four cases in my lifetime where it turned out to be wrong. (Soviet economic growth, 60's Keynesianism and the Philips Curve, replacing butter with margarine to prevent heart attacks, and the population claim).
Yeah, all the experts are not infrequently wrong. Having said that, errors like not accounting for medical advances seem pretty incidental to the overall climate change debate, not exactly something you can hang your hat on. Arguing that sea level rise isn't a big deal because it looks small on a map is also not convincing. I generally tend to accept Nordhaus-style estimates, but the question that gives me pause is: what percentage of the poorest population would have to die to reduce global GDP by 2.5%? Or since there are offsetting benefits, what share would have to die to reduce global GDP by 5%? I'm not willing to dismiss that as an important problem, although I agree that it's not anywhere near existential at the pace we're going.
Anyway, I consider climate change mostly solved, between the expansion of renewable energy, carbon capture, and some geoengineering in a pinch. There will be lots of work involved, but the trajectory seems clear.
At this point, I'm somewhat more concerned about all the chemicals we're dumping in the environment and their effects on humans, but I suppose that will turn out to be manageable as well.
Rennert et. al. 2022 offers a calculation of the cost of carbon almost half of which is excess mortality due to warming, based on evidence of the relation between temperature and mortality in the recent past. Applying that to the next three centuries makes sense only if you expect no progress in medicine over that period, which is absurd. The fact that they did so, in an article that devotes part of a page to reasons its estimate might be low but says nothing about reasons it can be expected to be high, means that the authors, professionals in the field, are either incompetent or dishonest. The fact that the article was published in _Nature_, a high status scientific journal, means that the mechanism for filtering out work that is either dishonest or incompetent is broken in the context of climate science.
That doesn't tell us whether the net effect of climate change is good or bad. It does tell us that Greg's argument, that the "existing body of knowledge" in the field deserves respect, is wrong.
At a tangent to our differing belief in orthodoxy, Tyler Cowen has an interesting piece arguing that the essential difference between his classical liberal views and the new right "is how much faith each group puts in the possibility of trustworthy, well-functioning elites."
I identify with classical liberalism and not with the new right, but I have a good deal of sympathy for that loss of faith, having observed multiple cases where what elites viewed as orthodoxy, what all reasonable people should agree on, was not only dubiously true but supported by arguments that were dubiously honest.
One example from a few years ago of evidence that elites should not be trusted was a NYT interview with Fauci where he said that what he told people was required for herd immunity changed with new polling information on people's willingness to be vaccinated. I wasn't surprised that he would lie about his professional opinion but I was surprised that he would say so openly.
I did read your thoughts on that a few years ago, but haven't kept up. I found it well reasoned as usual. I don't have time to go read more this morning.
The major positive effects I know of would be opening up VAST arctic/subarctic areas of Siberia and Canada to human settlements. Having just driven from California to Alaska and the Canadian arctic coast the potential of the vast emptiness is obvious. Yukon is the size of California and has 0.1% of the population!
That's one. A second is CO2 fertilization — doubling the CO2 concentration increases the yield of C3 crops by about 20-30%, reduces water requirements for both C3 and C4. A third is the reduction of deaths from cold — Gasparini et. al. 2015 estimates increased mortality from temperatures below optimum to be seventeen times as high as from temperatures above optimum.
For the first I go into some detail in a substack post, and link to published work by others:
The assumption that these areas would open to human settlements is on a shaky ground (pun intended). Some of these areas may indeed become like Montana. Others will turn into impassible methane-burping swamplands. Thawing of permafrost in Siberia is wrecking villages there as we speak, driving whatever humans managed to live there for centuries out.
As an aside, dicking around with a dynamic system of insane complexity with multiple feedback loops of varied stability is a very! bad! idea!, which is why I find the whole exercise of "but have you looked at 'major positive externalities'" to be utterly corrosive.
One problem I see with this argument is that "dicking around with a dynamic system of insane complexity with multiple feedback loops of varied stability" might include all the activities promoted by people worried about climate change, not just the activities embraced by everyone else. "Have you looked at major negative externalities" is every bit as corrosive as looking at the positive ones.
Another problem is that "very! bad! idea!" is underspecified. Bad for whom? People? The biosphere? If the former, then proposals to curtail economic activity by as much as 75% should immediately come under scrutiny for badness, rather than pushed as the obvious solution. If the latter, then you have to explain how the biosphere appears to have undergone orders of magnitude more variability before homo sapiens even appeared, than has been observed or projected as a result of human activity, and survived.
Ah but there's a difference: we (humans) are already dicking around with the aforementioned system, by very quickly releasing vast amounts of CO2. This is - to me - the core argument for being concerned with it. Modeling errors go both ways.
In general, people concerned with the climate change are saying just that: stop/slow down with the dicking around we're doing right now.
Very! Bad! is indeed for humans. "The biosphere" will adapt, some species will die, others will flourish, a new equilibrium eventually emerges. There may be no room for us in it though. Of maybe room for 10% of us. This is describing a catastrophic "hot house Earth" scenario, which we know has existed in the past, and we don't understand the mechanisms that may tip the planet into this state again, so being concerned with it is valid.
What we have now works. It has worked for us - humans - for the last whatever years it took to create the civilization we have. A hothouse Earth will not work for the current civilization, and we don't know what gets us there. This is why looking at "major positive externalities" is corrosive. The risk/reward relationship for them is way out of whack.
You're coming into this thread with a premise that quick release of CO2 is bad. You might not have known that Friedman also questions whether this is so; now you hopefully do. (His reasoning is in the same places he linked above.)
You're likewise assuming that humans can't adapt to higher temperatures. Same deal. Check the link above.
I submitted my criticisms of Rennert et. al. 2022, an article published in Nature that claims to estimate the cost of carbon summed over the next three centuries, to Nature, which rejected it. I also sent it to the authors and got no response.
Similar experiences here, different field. A few times that I had questions/concerns about published papers and contacted the authors resulted in 0 response rate to date.
I don't see intentional carbon sequestration, on a scale big enough to matter, as a thing that will happen. I figure, after we stop using fossil fuels, the climate will stabilize in a state warmer than it is now, and then we'll have to just adjust to the new climate. This won't necessarily be bad; the optimal temperature for the earth is whatever temperature we're used to, so we really just need to stop it from changing.
I've posted similar before, but I'd like to share a bullet point justification of what I think the ultimate goal of human (or superintelligence) action should be, and see what this community thinks of it. Considering all the debate we have about the future of humanity, it makes sense to have some grounded justification for what an ideal outcome for human activity should be, if such a thing can even exist. Personally, I think this outcome I describe is superior because it is based on a system that properly bridges the is-ought gap (the idea that a moral imperative can't be derived from simple observation of how things are), while other systems are based on subjective opinion or tastes that have no absolute grounding. Anyways, here it is:
>Hume supposes the is-ought gap
>We need to transcend this gap if we want to be able to deduce any possible end goal for humanity as deriving from observations of existence itself
>What is able to do so is pleasure (defined as any emotion subjectively experienced as positive): we are able to experience pleasure as good in itself, and good by its own definition implies an "ought" in its own increase
>So the ultimate end is increasing the universal experience of pleasure as much as possible
>But how to do that?
>Focusing on minds that already exist is inefficient, because it is at least theoretically possible to use Von Neumann machines (machines that utilize surrounding matter to make copies of themselves) to produce more minds out of the dumb matter of the universe than ever would naturally exist
>So, the end goal becomes developing the technology to create Von Neumann machines that convert as much of the universe's matter into minds experiencing pleasure as possible
>These minds would be so constructed as to never bore of this pleasure, if you've ever had a moment where you've thought "I wish this moment could last forever," it would be such a thing actualized, pleasure feeling as good as it did at its first moment for near eternity
>While such an end for the universe, by its single-minded simplicity, might seem boring or depressing to contemplate, what matters is not the feeling an outcome produces in us to contemplate, but how it would actually feel to exist within that outcome, so while it may be a "boring" universe to an outside observer, it would be an objectively good universe for the beings making up the majority of its conscious existence
>This is where this theory differs from Yudkowsky's "fun theory," because "fun" is only one of a multitude of pleasurable states, and the important thing is how the system is experienced from the inside (maximally pleasurable), not whether it's a universe that's fun to imagine oneself living in
In summation, Utilitarianism is the only system able to transcend the is-ought gap, and the logical end of Utilitarianism is using self-replicating machines to create as many pleasure-experiencing minds as possible. Hopefully enough people can be convinced by this argument that this can become the end goal that either humans or superintelligences put their long term efforts towards.
This doesn't sound good to me. Reductivist plans based on overly simple rules of thumb typically don't go well. Also, we have shorter-term, demonstrably meaningful issues to solve. Daydreaming about different flavors of wireheading isn't very useful.
Asserting that minds can be created is really question begging, as it runs headfirst into the hard problem of consciousness. How would you ever know if you have created a consciousness, that is, a thing that perceives?
Admittedly this is one of the most difficult parts of the whole project, and would require significant advancements in neuroscience and philosophy of mind to make sure that what the Von Neumann machines create actually experiences qualia. In the worst case scenario, modified human brains could be created with the appropriate changes to their structure to ensure perpetual, non-tiring pleasure, since we could be fairly certain by analogy to our own brains that they are truly conscious. Personally, I'm a panpsychist, so I think that the hard problem isn't as much of a problem if you assume all matter has consciousness at some level, the problem then becomes finding out what the physical correlate of pleasurable emotion is, and how to replicate it.
That type of relentless utilitarian optimization comes with its own set of issues, e.g. the repugnant conclusion. What if you could speed up the self-replication process by 10x at the cost of each brain feeling 10% less pleasure? etc.
I'm also not convinced that utilitarian summation really makes sense to apply to conscious experiences, and the only evidence that will convince me will have to come from discoveries around how consciousness fundamentally arises. Is it really "better" for 10 people to have identical pleasurable experiences than it is for 1 person to have such an experience? Intuitively, we think that yes, of course, more happy people = better and less = worse, and it's a useful metric in many situations (e.g. measuring impact of a war or natural disaster, etc.). But for whom is it actually better? Certainly not any of the 10 individuals - their experience maxes out at N=1. Those 10 peoples' experiences are totally disconnected - there's no "universal entity" that is actually experiencing the sum total of all pleasure + all pain. Sum experience is just an abstract concept. So I'd be very hesitant to commit to optimizing the universe in such a way.
The problem you’ve outlined is the reason I’ve always balked at adopting utilitarian positions on happiness. Conscious experience is so fundamentally singular that I can’t see how the sum of a million people’s minor amusements could outweigh a few people’s complete happiness. Or more starkly, how a million individually experienced paper cuts could outweigh the brutal prolonged torture of one person. Pleasure and pain take place in isolated and discrete experiential spaces that appear to me to be non-additive.
I don’t know the first thing about moral philosophy but I’m familiar with Parfit’s Repugnant Conclusion. Can you (or anyone) suggest any other writings that speak to the problems with summation in situations like this?
If pleasure and pain are non-additive, does it follow that one excruciating death is no better or worse than a million excruciating deaths? Should we not be biased towards preventing the million deaths over the single death?
I routinely wrestle with your counter argument because it’s a good one.
Let’s take death out of it, if that’s okay, since deaths are additive in other ways. In this example I assume we can magically measure experienced suffering and say it’s 100 for the one and 100 for each of the million. Yes, of course I think we should be biased towards preventing a million excruciating experiences over one excruciating experience. But I think that’s because people are additive and you’ve made the suffering equal. Do you think we should let one person suffer at 100 to prevent five people from suffering at a 25? Experience is discrete. There is no one experiencing a 125 when the five suffer at 25. The max suffering experience is capped here at 25. However there is certainly someone suffering at 100. All else being equal, my inclination would be to prevent the one person from suffering the full 100.
It’s not that I don’t think there’s value in utilitarian positions on suffering, (there obviously is) I just don’t think they typically account for the individual nature of experience itself.
I've had similar thoughts and I struggle to articulate the intuition. Where I land is that suffering can't be quantified, but is more like categorical buckets. No amount of paper cuts can offset one person experiencing debilitating pain; no amount of people with debilitating pain can offset one person experiencing continuous torture, and so on.
That said, I still have this nagging feeling that the idea of comparing suffering is meaningless. Like you said, each experience is in its own isolated experiential space. That there are two separate spaces undergoing max suffering seems no more bad than there being only one. Yes, there are two bodies and each experiential space is associated with each body, but bodies don't suffer. Physical things don't suffer. So the basis for counting each experiential space and summing some quantity of suffering just seems like a category error. Substances have quantity, the mind and its experiences isn't a substance and so has no quantity.
To be clear, I am (most closely identified as) physicalist about mind. I don't mean to imply any non-physical stuff involved in mind. I just have a hard time quantifying experiential realms and doing math among the quantities as if they were commensurable.
Well -- current version is too damn short IMHO. It's just a glimpse. Seriously doubt there are many people who think, during the 5 sec version, "I hope this winds down soon, I'm getting sort of tired and bored".
But I second NotA’s notion - extend it to minutes, or keep repeating ad nauseam, and it’s not hard to imagine the “ad nauseam” part becoming a bit too vivid.
Honestly, it's hard for me to imagine getting sick of it. Looked at from the outside -- yeah, it's ridiculous to spend your life thrashing around in that state. But from inside, it seems like there's a "yes, I finally got here feeling" -- as though for once, in this lousy life, you are precisely as ecstatic as you always wanted be. You're so in the moment while it's happening that there's no way to think about the limitations of the experience, and how once it's over you will return to matter-of-factness, plus of course post-nut clarity.
While in nearly total agreement, I do wonder if the intensity of the feeling is inseparable from its brevity and... scarcity? The maybe-relevant data being that (at least from a male perspective) this is an easily obtainable state "in solitude", but it's just not the same when... frequented... in this manner.
Perpetual orgasm of the nanobots. We need to maximize the number of orgasms per atom spent. What is the smallest possible molecule that could be described as having an orgasm?
"superintelligence should be able to brainwash everyone in the universe (and convert the universe into more brain washed bodies) because of the arbitrary whim of someone who has never and can never feel the end result he wishes to give to others."
the people are brainwashed because eternal unchanging pleasure is not a natural state of life and needs to be engineered into them. Many situations actually require a non-pleasure response to remain a thinking being; if a solar flare wipes out a planet of people unhappiness is the proper response but youve made it so they cant feel it. You remove free will.
its an arbitrary whim because you never have felt the state you want everyone in the universe to abide by. you only know normal human experience. If you knew this engineered pleasure you'd be unable to even conceive of suffering to formulate it.
its not like saying "everyone should be married" in which you have experience and just want to share it, you are trying to force an alien concept on all of the universe. not even because it gives you pleasure! but because of adherence to a philosophical position that says maximizing certain qualities is a moral good and to solve the is-ought problem.
and to do it your superintelligence can never feel it. it has to abide by your philosophy, and cannot exist to feel unchanging pleasure. because then it has no idea what to maximize; it needs to know what pain is to avoid it.
>What is able to do so is pleasure (defined as any emotion subjectively experienced as positive): we are able to experience pleasure as good in itself, and good by its own definition implies an "ought" in its own increase
>So the ultimate end is increasing the universal experience of pleasure as much as possible
This seems like tunnel vision. We can experience other things too, like pain. If we view pain as inherently bad, then the "ultimate end" is instead reducing pain as much as possible, no? I know a lot of old people who view pain as more prominent than pleasure. (Indeed, some would say it's the only thing that's real. https://www.youtube.com/watch?v=8AHCfZTRGiI)
Yes, if we're being technical, then it's really pleasure less pain that's being optimized, because pain is inherently bad in the same way pleasure is inherently good. The system I propose would have the minds created never experience pain, as well as always experiencing pleasure.
Well there's several others too; hope, worry, regret, camaraderie, satisfaction. Worry's probably the big one, unless you're making everything too stupid to question the stability of the system.
Why is pain bad? It's unpleasant. But one can learn from pain. Also, could you have the concept of pleasure without pain? In this Von Neuman Pleasureverse, would there be degrees of pleasure? Or would it be a continuous wirehead experience? And if there are degrees of pleasure, would less pleasurable things become mildly annoying? And if they did become mildly annoying, would one consider them to be painful compared to ultimate pleasure? Heck, would one be able to maintain consciousness on a plain of infinite pleasure?
Your Von Neuman Pleasureverse sounds suspiciously like the Buddhist Nirmanarati heaven. From a Buddhist perspective, you don't want to get stuck in a place of endless pleasure because you won't be motivated to reach enlightenment—nor to assist others in reaching enlightenment. (Likewise, you don't want to get stuck in a hell, either, because the pain would be too distracting for most sentient beings to be able work on enlightenment.)
Anyway, thanks for the offer! But I think I'll pass.
Nirmanarati is described as a realm of beings who can conjure forth anything that they like, and they spend their time making "mansions of sensual enjoyment" for each other -- to me this sounds like a really good time.
A better analogy might be the Jhana heavens like Abhasvara, which seems to be more like a perpetual solipsistic joyfulness lasting millions of years. I'm sure that's nice too, but from an outside point of view, it's not too interesting.
In addition to the totally unexplained leap you are making at the step 2, I contest step 3: "Pleasure" is a difficult thing to define and controversial choice. Many pleasurable experiences are pleasurable only to a limit and have diminishing returns; seeking pleasure may yield a less pleasant overall outcome; variety of pleasurable experience seems to be better conductive to human (and other mammalian) well-being instead of monotonous application of single pleasurable experience (near always a day is improved by a stroll in the woods, despite the fact that constantly residing in a sofa of reasonable comfort is obviously already pleasurable); and not only variety in pleasurable experiences is beneficial, certain amount of hardship and difficulties (that is, anti-pleasure) are also helpful to having a more enjoyable overall experience of life. Finally, the existence of hedonic treadmill implies that thinking of "pleasure" as a scalar number attached to states of matter in a moment of time that can "go up" just be reordering some configuration of states is likely some sort of abstraction (map-is-not-territory) mistake [1].
I find this a conundrum. Either "true" definition of pleasure needs both to include lots of other experiences that seemingly have nothing to do with pleasure and exclude many forms of pleasure [2], or one should accept that "pleasure" was perhaps a wrong concept to choose and something else (such as "good life" or "happiness") would be more apt.
Anyway, why there should be a number to go up? Why presuppose there needs to be an ultimate goal that involves some abstract concept of von Neumann machines implementing a number going up? An organism is a product of particular past evolutionary pressures (near universally those pressures created by other organisms). One might as well argue that the ultimate goal for a species with healthy amount of self-interest and enough capacity to do so would be to engineer their environment to the ideal form of ancestral environment where their kind first had the competitive edges that delineated that species. (Consider how many animals in a zoo appear become depressed despite safety and plenty of food and reproductive partners.)
[1] While the reader would probably do better by consulting an endocrinologist, the first intuition pump that comes to mind here is to think of a pressure differential stabilizing between two containers connected by an elastic hose ... which stretches in diameter over time as it is used. To maintain same velocity of gas transfer, the system would need to change over time: either containers to contain more gas, the temperature to change, a restriction to be applied to hose.
[2] This kind of definition of "pleasure" tends to become tautological no-true-scotsman-I-know-when-I-see-it.
You seem to be making points about the usual constraints human pleasure is subject to, i.e. the hedonic treadmill, while the point is that this would be engineered pleasure that would have no such restrictions, and would be able to be experienced as fully pleasant in itself without having the requirements of ordinary human life that require the periodic experience of less pleasurable states to contrast against. And the reason this is the ultimate, "number go up" goal is that unlike evolutionary competitiveness, pleasure is able to be directly experienced as a good in itself in such a way that allows it to bridge the is-ought gap, as I posted, so it demands itself as a goal requiring increase if one truly cares about maximizing "The Good." Again, note that pleasure here refers to any of the gamut of emotions subjectively experienced as positive, which might bypass some of your concerns regarding a more narrow definition of the basest pleasures which wear out the quickest in normal human experience.
>this would be engineered pleasure that would have no such restrictions
Yeah this was exactly what I tried to get at with my line about no-true-scotsmanning: more you define "engineered pleasure", more alien and increasingly removed from the everyday meaning of "pleasure" it appears. So maybe you are chasing a concept that is not "pleasure".
>pleasure is able to be directly experienced as a good in itself in such a way that allows it to bridge the is-ought gap
Have you considered that this might not be an universal experience? To me, it does not seem to be obvious at all that pleasure is a single thing that is "good in itself" due the variety and forms of experiences of pleasure I outlined. Unless you have totally different ideas tied to "pleasure" than me (by the sound of it, it seems closer to "things that are good and good implies something to be maximized") ... which makes the argument you claim to close the Humean gap reduce to "things that are good in itself and to be maximized are good in itself and to be maximized".
There's a little tautology to this; since it's agnostic about what constitutes this pleasure state, it's like "whatever is happening, everyone should prefer that to be happening."
Logic does not compute. Either the gap is real or not, and you have not given any argument either way. If it's real, then our alleged "need" to transcend it is just wishful thinking, on the level of our "need" to breathe seawater.
My point was that the is-ought gap is real for all other supposed moral imperatives which exist as mere opinions as Hume would claim; but since we can experience pleasure as directly, subjectively good, and the definition of good implies an ought in that more of a good is better, therefore we ought to increase said good, the maximization of pleasure becomes the one thing to which the gap doesn't apply and which has true moral imperative.
It's good because it is directly experienced as good in the most immediate way possible. Everything we define as good can be traced back to some pleasurable experience in relation to it. Would people bother climbing mountains if they didn't receive the pleasurable sensation of accomplishment in doing so which they specifically value? Hence it is the internal sensation, not the external act of mountain climbing, where the true value lies, regardless of what the mountain climber might believe or claim. Heroin addiction is bad because it leads to more pain than pleasure in the long run, so it fails as a reductio ad absurdum against this view. The contexts in which current pleasure is bad are merely those contexts where it leads to net suffering beyond the immediate present.
Because of the reasons I gave as to why Epicureanism/Utilitarianism is correct (namely in how when you have a chain of "but why do you value that? But why do you value *that*?" it terminates in a brute fact of "because it provides some positive sensation valued in and of itself")
With the Christmas season coming up... Apparently a lot of adults buy Christmas presents for each other, but... why? With children it makes sense: they have predictable interests and you can't just give them cash because they can't use it. Also they'll usually just tell you exactly what they want. But with adults.... If you really want something, you can just buy it yourself. And with no good way to know what the other party wants, any present you buy for them is bound to disappoint. Giving cash or a gift card obviously doesn't make sense either, since you'll both end up with the same amount of money you started with.
So... what's the point? It just seems like a huge waste of time and energy.
There are many things that people want or would like, but reason themselves out of buying. That's always a good idea for a gift. For example, my partner has always wanted to play bass, but never got around to it for various reasons. I bought a beginner bass for them last year and they were thrilled. They've been practicing a lot and show no signs of stopping.
Also, people are not omniscient. There are many things people will like or that are useful to them that they don't know exist. Buying a gift like this shows that you know the person deeply and that you care about their interests.
I actually prefer the feeling of giving someone a great gift to receiving one.
It seems to me that, in addition to felt obligation, the hook is a gamble. *Maybe* you can find something the other person will really like, even though the odds aren't in your favor.
I couldn't agree with you more, and this is the exact argument I used to eventually shut down obligatory seasonal gift giving in my family. It took a few years, but now everyone heartily agrees that the greatest gift we can give one another is freedom from obligatory gift giving. We now just focus on meals and hanging out and absolutely no one misses the physical objects.
Gift giving should occur only when one adult spontaneously discovers a product or experience that they're certain the other adult would like and does not yet know exists. That almost never coincides with a calendar, and shouldn't have to.
a gift is a memento that endures in time and reminds you of the giver. My sister once bought me a psp; i still own it and when i use it i think of her. A well-chosen gift is a bit of history.
physical objects bind memories to them in a way other things don't. The money or utility isnt the issue. even if i dont keep it, the memory remains.
you dont always get this till a lot of gifts have been given though. or how important memories are. if anything i wish i had saved more and kept more.
This is why it is actually impossible to buy a gift for my husband. If he wants something, he buys it. If he doesn't buy it, he doesn't want it. When I go ahead and buy him something anyway, he's annoyed, so I'm just not going to try this year.
"And with no good way to know what the other party wants..."
If you have no good way of knowing what the other party wants, then you're probably not very good friends with them. But for people who are friends, it makes them feel good to know that the other person both knows what they like and put in the time and effort to find it for them. It's not actually that hard to understand.
Huh? I like giving gifts, don't you? A gift says I was thinking about you, or that I like you, or... just thanks for being part of my life. At Christmas I give different kinds of gifts, one could be something you want and ask for, another something I know you like. And third some type of project/ activity/ puzzle/ drawing, that we can hang out and do together. I got my son (age 22) a wooden clock kit, that I will admit I'm looking forward to helping him with.
Of course from a pure average utility of gift perspective it doesn't make sense. But consider the feeling it gives you that another person is spending the time to figure out your desires and then going through the effort to fulfill those desires. It can be done in other ways but there is a specific moment where there is a door open to this kind of opportunity.
> But consider the feeling it gives you that another person is spending the time to figure out your desires and then going through the effort to fulfill those desires
I'd argue that this fails more often than it succeeds, and an underwhelming or misjudged gift is far worse than receiving no gift at all.
FWIW, my understanding is that there's a custom in China of giving people red envelopes with cash in them. Which seems like it would be better, but it's been involuted into a status and reciprocity game, where you have to track who gave you how much at what occasion, and respond in kind, modulo your respective wealth and status, and frankly the whole thing sounds like a nightmare.
There are things I want so much that I'll buy it for myself, and there are things that I'd like, but not enough to buy for myself. Those things that I'd really like, but not enough to buy for myself, are the best things to get as a gift.
That's strange to hear. If I want something but I don't buy it, that's because I feel that the value it will offer is not worth the asking price. Someone else buying it and gifting it to me doesn't change that calculation if I'm expected to give something in return. The end result is that two people made a bad purchase for each other.
Some people feel irrationally guilty spending money on luxuries for themselves. But if I can spend money on _other_ people, and get an equivalent value in luxuries for myself, then those are some luxuries I can enjoy, guilt-free.
So let's say someone gave you a good pen, which makes it more enjoyable to write.
If you're the sort of person whose enjoyment of the pen would be forever eclipsed by the knowledge that you personally wouldn't spend the resources to acquire it, it would not be a good gift for you.
Most people, though, I think, would separate the process of acquisition from the experience of use. It'd be no worse than having it drop out of the sky, and usually much better because it associates the pleasurable experience of using it with the happy memory of the person who gave it.
Also, there's a window for an actual mutually beneficial trade, where someone gives you a gift that, after using it, you discover that you value enough that you would have gotten it anyway. Bonus points if you would not have been persuadable through words, such that simply giving you the thing saved everyone a lot of time and frustration. :-)
Part of the answer is it's a culturally accepted time to ask your rich friends to spend their riches on you. (Several people pooled together and got me a new bed one year, to replace the dirty mattress I pulled out of an abandoned building.) And in return you get them snacks and hand towels, cheap stuff they probably don't think about enough to replace. Or, like, weird movies you know about and they don't.
A bit off-topic, but you *can* give children cash. I've been both the giver and recipient there.
I once got my niece a twenty-dollar bill, sitting on top of a block of wood so it would wrap better. She and her friend decorated the block of wood and played with it for hours. Years later, she remembered the wood but not the money.
Gifting is a fun excercise in which you model your loved ones' tastes, interests and preferences in order to find something that they'd really enjoy but that they wouldn't think of buying for themselves (or won't allow themselves, in some cases). In general, the utility of receiving an unexpected gift is higher than its monetary value, even moreso the signal that someone cares about you enough to give a thought-out gift.
It still seems like gifting has a poor utility:cost ratio compared to other ways of showing care (such as providing actually meaningful assistance with their lives), especially when said gifting is reciprocal. I can see an unexpected gift having some value, but this is Christmas we're talking about. The gift is both expected and (supposed to be) equivalent in value. Of course, this causes even more problems when the latter ends up not being the case... It's just not worth the hassle.
>such as providing actually meaningful assistance with their lives
Lol, that's way harder, can lead to way more problems than your present not being of equivalent value and in general has a much higher downside potential.
Of course it's difficult, but the potential payoff is massive compared to a ritualized exchange of baubles. Again, the reciprocal and expected nature of the gifts greatly lowers their potential social value. I just don't see the point in putting that much time and effort into something that, on average, isn't even going to meaningfully improve your relationship or status with others. And it's hard to imagine that anyone finds the process of spending a bunch of time brainstorming, researching, and finally purchasing gifts for multiple people to be entertaining...
Here's the Artist Nina Paley dunking on copyright in the most hilarious (yet entirely fair and true) way possible : Copyright is Brain Damage | Nina Paley, https://www.youtube.com/watch?v=XO9FKQAxWZc.
The first video is a satire of a song made for the propaganda novel/film Exodus https://en.wikipedia.org/wiki/Exodus_(1960_film). The film is a favorable fictional account of the founding of Israel. The video Nina made is a satirical wider snapshot of the entire history of the region.
The second video argues that Culture and Intellectual products flow among individuals and societies in a manner very much resembling the flow of nerve signals among neurons, and that Copyright, in stifling this natural and beautiful flow, can be rightly likened to brain damage. This view, complete with the neurological analogy, is very similar to the views of the Queen pirate herself https://en.wikipedia.org/wiki/Alexandra_Elbakyan.
I started watching, but I disagree with the premise that the possiblity of selling and financializing a catalog doesn't benefit artists. It does! People are willing to pay them *because* rights will be enforced after their deaths. Comically simplified, if, say, Taylor Swift wants to sell her rights to a song, she will get more money now if those rights are enforced after her death.
So although I probably agree with the conclusion that copyright overextends itself, I stopped listening to the talk.
Hello all. I voluntarily deleted my acount (Matthieu). The unexpected part is that it also deleted all my comments, with the one with replies marked as deleted and the one without leaving no trace. I realize this is bad for the blog archives' readability, and I apologize for it. I won't delete this temporary account but I should be unable to use it again.
This is a recurring theme, because I was already this guy: https://slatestarcodex.com/2019/11/06/open-thread-140-25/#comment-818659 . My reasons for leaving now are somewhat similar to then, although less dramatic (rereading myself, I certainly feel generally better now than I did then). The added practical reason is that I have a f*ing PhD to finish.
Thanks to Scott and to all the interesting and likeable people here. I should be back at some point.
Thank you very much-- excellent story. I also appreciate "Anthropology" by Dizzy Gillespie and Godspeed You Black Emperor, neither of which I'd heard. I'm currently listening to "Lift Your Skinny Fists Like Antennas to Heaven".
Do you think FTL will get interested in composing and/or performing?
Thanks! Godspeed You Black Emperor had particularly stuck in my mind because for the majority of their time they had a lost album. I loved the mythology of there being these two dozen cassette tapes floating out there which may or may not still exist in the world. For the past decade I’ve been checking once every year or so to see if they had been found, and to my surprise, they recently were!
I think FTL views himself as a collector first and foremost, but this is definitely a possibility for the future! I thought for a while about a group of drones whose version just sound like white noise to humans, but which to them has intrinsic beauty. Could be an interesting direction :)
I have an interactive, international, interdenominational whatsapp chabura that you are welcome to join.
At some point inshallah it will be by invitation only but as we're only 70 people as of now and the average SSC'er is likely to be a good fit you are welcome to join and of course to leave or rejoin whenever you like.
That's the invite link you had given me? I joined it and it seems that all the so called 70 people in it are fake, as they all have WhatsApp profiles consisting of just a number and a pic, unlike in every other WhatsApp group I am a part of.
You need psychiatric help if you are engaging in this sort of shenanigan!
The recent New Yorker is essentially the AI issue (I linked to a profile in it on Hinton in a thread below) and has a piece called "A Coder on the Waning Days of the Craft". I'm curious to what extent other coders agree with his description of GPT-4's current coding abilities: https://www.newyorker.com/magazine/2023/11/20/a-coder-considers-the-waning-days-of-the-craft
I have never used GPT 4 because I'm cheap and also Open AI doesn't deserve money, but here are some of the things I hated about this article :
1- It's titled A Coder Considers the Waning Days of the Craft, I hate the word "Coder" because it's an inaccurate way to describe programming languages, "Code" used to mean the binary encoding of commands that CPUs come out of the factory natively understanding, every computer language that resembles English or mathematical notation is technically not "code", it's a formal language that serves as a user interface to code and ultimately translates to it. Then there's the "Waning Days", really ? Waning ? Even the author refutes the clickbait in the last few paragraphs :
> GPT-4 is impressive, but a layperson can’t wield it the way a programmer can. I still feel secure in my profession. In fact, I feel somewhat more secure than before.
Why can't NYT title writers read the whole thing like I just did ?
2- The author's examples belong to roughly 2 categories of things that GPT-x seem to be unusually good at :
2-a) Green-field small projects with historical precedence : This is when you want to make a new program from nothing, typically one that has been made before, the result is typically between 100 lines of code and 1000. GPT is good at this because the entirety of Stackoverfow and tutorial websites are extremely good gold mines of training. Examples of this category include the Snake game and the ios app the author's friend made and the countless "make website using GPT" tutorials.
2-b) Input-Output puzzles : You have something, you want something else out of it, you're aware that the transformation is possible but you're too lazy/sleep-deprived/short-on-time to actually sit down and derive the step-by-step procedure or look it up. GPT is good at this too, sometimes because it has developed something akin to "reasoning" over the language the solution will be written in, but most of the time because someone before has had your exact problem, the shell command the author wanted to print a random 1000 lines of the dictionary is a good example.
There are vast territories in programming and software engineering not covered by (a) and (b). There are the million+ lines of code heavy-duty services at Google and Facebook. Ok, not everyone works at Google and Facebook. How about the "10000+ lines of spaghetti that encodes 15+ years of changing business requirements" ? How about the "website design that keeps circulating back and forth for 6 months because the lightning in the room makes the client see the colors differently each meeting" ? How about any program bigger than 1000 lines and thus too big to fit in even the biggest of GPTs' context window ? Database schemas ? Compilers ? Game Engines ? Will all of this, really, be automated away by the very conceptually simple mechanism of predicting text ?
3- GPT is data-inefficient. GPT is only good at Python and Android/ios and Html and CSS because those are very stable and very popular technologies that hordes of people invested in over more than a decade and assembled a vast ad-hoc library of examples and material. But try making GPT help you with COBOL, try asking it about the AUTOSAR automotive platform, ask it any question about an obscure or relatively unknown technology.
I like the GPTs and I see them as a large step in AI and in Knowledge tools more generally, but I developed a kind of dread towards those kinds of articles, their titles tend to be similar even.
> Programmers are people who can endure an endless parade of tedious obstacles
So true a lot of the time. There's always Google, but in Stack Exchange threads one must always remember not to copy code samples from the first post, detailing the question, because these are wrong almost by definition! :-)
My take as a programmer is positive. My job as a tech person is to take a bunch of business guys' loosely defined requirements, and turn them into something watertight enough to be implemented, then implement it. If the second part is largely automated, so much the better, it's the boring part anyway.
But any tech project will have lots of precise points where what you're doing is not obvious, and spelling them in precise detail is exactly what we do through code. If we spell them in plain English to some LLM-driven bot which spits out Java or whatever, writing them in English with enough precision to get the actual wanted code that fulfills all the loosely specified requirements *is already* essentially programming. So our skills are still needed.
I’m pretty senior and I have begun to use it a lot. Will it replace all coders? I can’t see that happening unless it’s 100% accurate, but a few more step changes in ability and we could be done for.
(However if humans don’t code anymore where do LLMs pick up future information?).
Trying to scope out who'd be interested in funding a research group on broader social, historical, and philosophical approaches to AI policy and design. I'm from an SWE and PhD academic history background, others interested are PhD stats, PhD moral philosophy, inter alia.
tl;dr academics and policy people want to model and think through and design AI socially and culturally, who is it _not_ rude to ask, is there anything we should be thinking about other than just trying to get funding for compute?
In celebration of a very ER doc Thanksgiving I wrote some satire about the cherished holiday tradition of dumping your old people at the emergency room. Nothing says “let’s be thankful” more than the annual Pop Drop and Yaya yeet (tm). Please don’t do this ;)
The first time I took care of a holiday dump-and-leave, I was a baby intern, and literally didn’t understand what was happening. I just thought that there must have been a miscommunication by the admin/intake clerk because there was no contact info available. My senior had to explain the phenomenon to me. It made my thanksgiving shift quite sad.
As a side note, I’ve been following your writing ever since your 200 Joules post. I think the world is better off because of your and your husband’s writing. Best wishes to you and yours. <3
Thanks for your kind words! That really means a lot.
Oh yeah, explaining the phenomenon to new doctors or med students and just watching the light kind of go out of their eyes is not fun, and usually it’s kind of fun to mess with the new meat 🤣 I learned there’s a vet version to this and I just wanted to quit being a person for a while
Yeah, two years later I had to explain to my intern what was happening (the poor guy kept on being hung up on when he tried to call the son) and that was not fun haha. And in my head I was thinking, “Everything has come full circle! The cosmic scales have now been realigned!”
Indeed. Approaching senescence and death with stoicism isn't something we can get to from here, but being reminded of it in movie form doesn't do any harm.
Ubasute in the winter would have been less painful than most end of life care here in the West, and far more dignified.
Good lord, that's depressing. Ironically, we did have to take my mother-in-law into the hospital the day before Thanksgiving because she got Covid and spiked a 106 temperature. They kept her there for 3 days out of abundance of caution, which was a bummer.
Dementia (and age decline in general) is such a hopeless and miserable state. I'm curious if your experiences have led you to support for euthanasia.
This was an absolutely appropriate reason to bring in your MIL! Tell her 106 is a bit dramatic, but she gets ER flair points. Kidding, I'm glad she's doing better. I am definitely in favor of paths to assisted suicide for people with devastating illnesses, it's the humane choice if that's what someone wants. But the US medical machine is incredibly paternalistic and does a very poor job at consent and shared decision making as it is, so I'm not entirely sure how to get from idea to ideal. Still, I think steps should be taken to open the conversation.
As a cousin idea, I'm a big believer in right-to-try and expanded compassionate use programs for the terminally ill, which the FDA also effectively blocks. Ironically, I think the FDA would prefer to allow people who want to live to die of their terminal illness instead of try something experimental and risky (with the patient's consent) to see if it would help , and on the other hand, they would rather let someone who wants to die because of their terminal illness live and suffer rather than allowing assisted suicide. All in the name of patient safety. If you've seen my essays posted around here you'll know FDA "protections" are a deeply personal soapbox topic for me :)
I'm developing a browser-based idle game and as an amateur, it's been a real motherfucker trying to figure out coherent systems for some of the game mechanics such as the loot rewards, the questing system and so on.
Do people just trial and error this shit when they're on their own? ChatGPT helps a bit but is mostly good as a summarizer and organizer. It has occasionally given some good ideas for formulas and can help me make sense of the math needed for some of the things, but I am feeling quite a bit lost.
Are there resources, communities out there that were created for amateur/solo game devs that you know of?
Yes, trial and error with a lot of playtesting. I think it's best to think of game design like science. You have a hypothesis about what might be fun, you design a prototype to test your hypothesis (on real people who aren't you, ideally), then you adjust your design based on the results. This is what they call "finding the fun".
As far as I can tell people do a lot of playtesting. There’s some room for modeling what the system does without just running it, but I don’t think idle game authors do that very much. There’s a subreddit if that helps, I don’t know what state it’s in though. https://www.reddit.com/r/incremental_gamedev/
Somebody besides me who says motherfucker on ACX! Hey there Zarko. There are definitely a lot of game developer resources online. I have a friend who is building his first game, and has found several. They're friendly, too. I thinnk some are on Reddit, some on Discord. There are also contests for newbies where they alone or with a group build a simple game in a a short period, someting like 3 days. In my town there's even an indy game developer's meet=up. Oh, and of course there are lots of YouTube how-to videos. Since I neither develop no play games that's all the info I have, though.
Why? Phones don't have room for lots of style buttons, and voice recognition is a serious security threat. I don't like the idea of machines listening in on me. (I'm slowly working on an easier to learn keyboard layout, but since making hardware requires a whole lot more capital than writing software, this one is on the back burner.)
My takeaways from the Altman saga (from very much an outsider's perspective, based only on news reports and what I think I know about human nature):
- The corporate structure of OpenAI which was designed so that AI safety would be the paramount concern; the board was meant to ensure that. Altman does not want to be constrained by that structure.
- The 90+% of OpenAI employees who sided with Altman agree with him. Furthermore, they apparently found it absurd for Toner to suggest that it would be consistent with the mission of OpenAI to shut down the company if it presented a danger in terms of existential risk, or AI risk more generally. But that statement by Toner is simply a logical consequence of *having* a mission in which AI safety is a paramount concern. So, practically speaking, OpenAI employees, with near unanimity, do not believe that AI safety should be the company's paramount concern.
- It's pretty easy to infer from the above that the point is never going to come where the AI industry voluntarily prioritizes AI safety over other concerns. At every given step between now and ASI it will be easy to rationalize a more or less full-speed-ahead approach to AI development: "if we don't do it the other guys will"; "we are working for the good of humanity"; "our intentions are good, therefore the consequences of our actions must be good." The monetary and professional rewards of being first in the AI race and the glory of building utopia will be powerful motivators for concocting those rationalizations.
I am 100% certain that Sam Altman does believe that AI safety risk is real and profound. It is also clear from profiles of him that he thinks he's building utopia. Designing utopia is akin to making himself something between a king and a god of all humanity. Of *course* he is going to treat safety concerns as secondary to that incredibly powerful ambition.
He is, I'm sure, very smart. He is also an incredibly shallow thinker, from what I've read about him - he has no sense of the weaknesses of utopian thinking, no sense of why the extraordinary power he is projecting for AI (and by extension for himself) is inherently dangerous. He is subject to the worst kind of hubris: the belief that not only is he competent to wield enormous power unchecked, but that he is doing so for the common good. He might also be the most responsible leader of an AI company outside of Anthropic.
I am not feeling great about things at the moment.
I have gotten the strong impression from the last week (the news, as well as individual reports of his past behavior) that Altman is a stereotypical sociopath, that is to say charismatic, manipulative, and indifferent to risk. He is smart enough to intellectually understand existential AI risk (and to tell us how dangerous it is and how cautious the world must be regarding it). But he is incapable of actually *feeling* that risk and thus is unlikely to take meaningful action to avoid it (when that interferes with other goals like personal ambition for power).
Asking AI developers to slow down is like dumping a dragon's hoard in the middle of a football field and asking the crowd to proceed single file and at a walk towards the hoard.
And in addition to the financial incentives for rapid AI development, there's this: A lot of people who work in tech have an Aspergerish profile of mental strengths and weaknesses: They are awesome at developing tech -- some of them have absolutely beautiful minds -- but less good than the average person at various people skills: grasping other points of view and seeing how those points of view, like their own, are coherent and sincere; predicting how how members of our species are going to be affected by a novel technology; having an empathy-based feeling of loyalty to other people.
The AI companies at this point are like a second government. Of course they have less power than our actual government over on how the present US is set up, but they have more the the present government on the shape of the future. In fact maybe it makes sense to think of them as the government of the future. They are at present shaping the world of 2050 and they are ill-qualified to govern. They have the wrong kind of brain.
I know there are a lot of people here who consider themselves to be on the autistic spectrum, or not neurotypical some other way. I hope it's clear I'm not disparaging people who are not neurotypical, just saying most are not cut out for jobs where people skills are very important. I'm pretty sure I myself am not neurotypical.
Wait, isn't it a *GOOD* thing if the shape of the future will be driven more by nerds??
Nerds / rationalists
> People who value truth
> People who prefer to operate on Simulacrum Level 1 - ie actually engaging with the real world and believing there are objective facts and truths about it
> Not generally motivated by greed, lust, or other vices - generally motivated more by intellectual challenges and creating new / complicated / exciting things
> Their main qualifications are being smart, being educated, being deep in one or more technical disciplines, and an ability to drive real-world results
vs
Politicians
> Explicitly eschew truth, and indeed are characterized by being *extra good* at lying to a broad array of people and constituencies
> Operate on Simulacrum Level 4, with Trump a fine example - only concerned with words and their effects on other people rather than any sort of ground truth or any such thing as objective facts
> Generally motivated by greed, lust, lust for power, etc - don't really care about intellectual challenges
> Their main qualifications are being rich, knowing rich people, being good at lying, and being willing to take suitcases of cash or long/short equity positions that substantially influence the legislation they consider and pass
It's not like politicians carefully and empathetically consider the viewpoints of all the clusters and demographies out there - they're all tribalism, lies, and suitcases of cash / stock positions in return for tailored legislation.
Which of those two groups do you honestly prefer to determine the shape of the future?
I think a random selection of people who meet basic requirements of decent general knowledge of science, culture and history would be better at governing us than politicians, as least politicians as you depict them. Let's set aside the politicians and just talk about nerds.
So can we think of nerds as people with Asperger's, who also are very smart, and who also are diligent and hard-working? If so, then the thing I see as a problem is that these people are not very good at reading other people. For instance, there's a well-respected test of reading other people's facial expressions. For some reason you can take it on Amazon. It's here:
The average score is 28. Scores 22 or below strongly suggest an autistic spectrum disorder. But setting aside diagnostic cutoffs, think about the disadvantages that come with being, say, 25th percentile at reading faces. It's a genuine disadvantage -- like having a bad sense of direction. You are less good at knowing where you stand with people. You are less good at figuring out what they want, what they are feeling, when they are lying. You are less good at forming a mental picture of their point of view.
Here an example of the kind of thinking that comes, I think, from having a deficit of that kind. Somebody on here put up a post suggesting that nobody really likes Shakespeare, some people just pretend to because they've been convinced that people who like Shakespeare are classy. And quite a few people chimed in, agreeing that that idea was plausible. OK, I am someone who likes Shakespeare, and I am absolutely positive I am not pretending to. And I know many other people who are the same. I can sense their sincerity when they talk about Shakespeare — but I also have behavioral evidence. I have seen them reading his plays for pleasure. I go to performances with them and the afterwards they are bursting with excited thoughts about one actor’s read on a certain character. Etc. The person who put up that Shakespeare post was having trouble forming a mental picture of what it was like to be someone who enjoyed a certain thing that he himself did not.
Here’s another example. Yann LeCun is active on the part of Twitter where people talk about AI. Yudkowsky is on there, as are a number of bright, articulate people who talk about risk, and some of them come at it from a completely different angle than Yudkowsky does. So LeCun has heard a number of nuanced bits of thought about reasons to worry about the impact on AI on the future. I feel unable to decide how seriously to take x risk, but have posted some good data-based arguments about the ways indirect contact with others, and involvement with virtual people via gaming, etc., is associated with increased depression and akrasia. Others post about x risk, but sketch in some plausible paths to great harm for which they estimate there’s a 10% probability. Of course there are some screamo “AI’s gonna kill us in 5 years” people on there too, but the point is that LeCun has read a lot of stuff that makes decent sense. But he posted that people who worry about AI harm would probably have wanted to ban the production of ball point pens when they were invented. They’d be screaming about how dangerous the pens were because they could be used to write down and spread misinformation. So LeCun, with quite a lot of good information about what AI worriers really think, was unable to form a mental picture of their points of view — instead, his picture was of a generic screamo idiot. No doubt he was helped along in the process by the fact that he does not want the worriers to be right. AI is his baby, his career — of course he is heavily invested in it! You can think of the ability to see things from another person’s point of view as a sort of *check* on the tendency to believe that the things you want to be true are, in fact, self-evidently true. LeCun is functioning without that check.
Here are some important mistakes I think people with this sort of deficit are especially vulnerable to when it comes to AI development:
-disregarding ideas that come from people who do not think they way they do
-failing to involve enough people with skills in soft sciences like psychology, child development, animal behavior and linguistics in thinking about AI, it’s development, and how to align it.
-failing to recognize when a colleagues has become so fascinated by his work on AI that he is now incapable of even considering the possible dangers if his project succeeds.
-not being good enough at introspection to distinguish between what they think is true and what they want to be true; and between craving for success and glory and confidence that one is improving life for mankind.
-being bad at imagining the impact of things like having a personal AI chatbot on people who are socially avoidant — or the impact of the ability to easily make convincing fake videos on teen bullying, or the spread of misinformation -- or the impact of increased AI presence in doctor visits.
You make some great points, and definitely have some valid concerns about nerd-driven future policy - but I think the big problem is there's no way to get to your ideal from where we are today. The choices are nerds or politicians, and there's no better choice on the menu.
Sure, we'd all love if our decisions were made by golden philosopher-kings with exceptional empathy and compassion AND in-depth tech skills AND an unyielding commitment to overall human flourishing AND a broad understanding of differences in people's wants, desires, and mental models coupled with a sincere desire to not use that for personal advantage, but to do what's best for that diverse mental constituency...but those people don't exist.
The framework for finding them (or even your broadly educated "average" people -- and college degrees are ~35% of the pop, so the *average* person doesn't even have an undergrad degree in a single major, much less an understanding of multiple fields) and elevating them to positions of leadership don't exist, and likely never will. Random lottery democracy hasn't been seriously proposed or enacted anywhere (and Athens and other Greek city states weren't pulling from a random pop, but instead a highly selected elite pop). Honestly, our best bet for actually better leadership than what we can get today is an aligned AI, so once again that's on the "nerds shaping the future is better" side of the argument.
As of now, here, our options are "future shaped by politicians" and "future shaped by nerds." Absolutely, nerds are less good at reading people and forming a clear mental picture of their point of view. That's STILL better than politicians, who *are* good at both those things, and then actively use them for evil / lies / personal power and aggrandizement.
You have some good examples of nerds sucking at mental modeling, and Ylecun is famously obdurate and blinkered when it comes to AI safety, nearly everyone on the AI-not-kill-everyone side will agree, but the alternative is politicians who *are* good at mental modeling and will use it against you.
I think given the two choices actually on hand, we should be pulling hard for "nerd shaped future" even given it's deficits and potential failure modes, because the deficits and failure modes of the other side are so flagrant and so much worse, and because there's never going to be a better, more inclusive option like the one you would prefer.
I tried to give an adequate reply to all this, but just ran out of time. So I'll just say this: I read somewhere in the last 2 weeks, can't remember where but it wasn't a silly publication, that a substantial minority of AI developers think ASI will wipe out our species, *but* a substantial minority of those people consider that an acceptable outcome. Are you sure you'd be willing to be governed by these people?
This issue comes up all the time on ACX, so maybe we'll have another exchange in the future.
As for the politicians: You couldn't possibly hate them any more than I do. But I still think we would be better off working to adjust the laws and rules that have made some of the grotesquery possible -- for example, gerrymandering (there are many other examples, of course). Also, I do not believe all politicians are godawful. I have 2 close acquaintances who have worked closely and for a long time with senators and cabinet members. Both are perceptive people and both loathe the cynical corrupt fatass liar politicians as much as you and I do. Both are sure that the people they worked for have great integrity. I believe them.
A lot interesting there-- I'm wondering how you'd select basically sane people of a variety of personality types to do the governing.
I just checked-- you have a wide knowledge requirement, not a personality requirement. How you'd do that for government, I don't know. Sortition with a limited pool?
A problem with nerd/autistic people failing to understand normal people is that it may take nerds to understand the relevant technical issues. Also, as I understand it, autistic people aren't so bad at understanding other autistic people, just as neurotypical people understand each other fairly well, but not autistic people. It's not a generalized "bad at reading people". Still, bad at reading the large majority of people would lead to being bad at governing.
Have there been autistic hereditary monarchs? How did that work out?
Being wrong-headed and stubborn about it is hardly limited to autistic people. I believe Hitler wasn't autistic. Autistic people might have a worse case of it on the average, but I'd need evidence.
As for fear of fountain pens, disaster didn't happen from fountain pens, but the printing press did lead to religious wars in Europe. I wonder whether anyone saw that coming. I've heard radios implicated in the rise of Nazism and in the genocide in Rwanda. If you're not used to radio, hearing politicians that way is like having Saruman whispering in your ear.
Oh, what the hell, I will mention this. Last night, I spent too much time arguing with a man who went from life for humans is a net good to everyone should have children. He was completely ignoring whether everyone wants children. He was willing to accept that people would disagree with him, but not that he might be missing something.
Maybe High Modernism (the belief that things should be legible and decided from above) is a great deal of what's wrong with the world. Note that it's a major part of Communism, and a somewhat less strong aspect of capitalism.
Nancy, I'm not going to have time to give a decent reply to all this. But no doubt we'll talk again. Let me just respond to a couple points you raised about autistic spectrum people:
> it may take nerds to understand the relevant technical issues.
I'm confident that it doesn't. Feynman, for example, was clearly not on the autism spectrum. He had a mobile, highly expressive face. He was sociable. He had lots of affairs. To do that, you have to like sex, and many autistic people are asexual or have a reduced interest in sex or are only aroused by some unusual kink (I knew one whose thing was plastic pocketbooks). Also, you have to be able to turn on the charm and tweak it to fit the person you're with -- that's very un-Aspie. Also, I have heard podcast interviews with AI developers who clearly are not on the autistic spectrum. They chat easily, are self-disclosing, talk with emotion about their mentors or their family, and have a lot of mental flexibility. If the interviewer brings up some new way of thinking about something, they move right inside of it and say, "hmmm," so from this point of view it looks like . . .." Jeremy Howard is one, and a particular favorite of mine. He put up one of the first whats-the-point-of-masks YouTube videos.
I assumed he was a medical professional, but actually he's an AI guy. Also, he's adorable.
>Being wrong-headed and stubborn about it is hardly limited to autistic people.
Completely agree. There are many paths up the wrong-headed stubbornness moutain.
>Also, as I understand it, autistic people aren't so bad at understanding other autistic people, just as neurotypical people understand each other fairly well, but not autistic people. It's not a generalized "bad at reading people".
I don't think that's true. I have treated many people who have Aspergers or are a little further up the autism scale than Aspie , and what I see is that most are more comfortable socializing with other people on the spectrum, but that is because other Aspies do not make heavy conversational demands. Often they mostly play D&D or some other game together, without a lot of communication beyond what the game requires. If one of the other players gets angry and storms off, the Aspie patient telling me about the incident does not seem to have any special insight into what upset the person. They say things like, "I dunno, maybe he didn't like the new player who joined the group?" I often have the feeling that if I, or really any random neurotypical person, had been present observing the group we would have had a better idea of what had upset that one player. Many Aspies yearn for a connection with neurotypical people. I have one patient, a female math prodigy doctoral student, who got furious when I suggested she check out some meeting place I'd read about for neurodiverse people: "Why the fuck would I want to meet more people like me?"
>Have there been autistic hereditary monarchs?
I'm sure there have been. Asperger's is not rare. But I am sort of ignorant of history (hence condemned to relive it, I hear). You should ask on here. I have wondered about Prince Charles, well King Charles now.
There may not be *enough* non-nerds who understand the technical issues.
I'm not sure if all Aspies are as mentally rigid as you imply.
I honestly don't know where you're drawing the line for Aspie and non-Aspie. My home subculture-- sf convention fandom starting in the seventies-- seemed to find a way for a lot of mildly spectrumish people to enjoy each other's company, and it isn't all gaming, though I'll grant there's a lot of gaming.
I don't trust rationalists to run the world, but it's more because of a taste for High Modernism, a mental habit that might be especially attractive to nerds but isn't limited to them.
A little research suggesting that communication goes best between people with similar neurotypes. This is the first time I've seen a paper with an abstract and a lay abstract.
Effective information transfer requires social communication skills. As autism is clinically defined by social communication deficits, it may be expected that information transfer between autistic people would be particularly deficient. However, the Double Empathy theory would suggest that communication difficulties arise from a mismatch in neurotype; and thus information transfer between autistic people may be more successful than information transfer between an autistic and a non-autistic person. We investigate this by examining information transfer between autistic adults, non-autistic adults and mixed autistic-with-non-autistic pairs. Initial participants were told a story which they recounted to a second participant, who recounted the story to a third participant and so on, along a ‘diffusion chain’ of eight participants (n = 72). We found a significantly steeper decline in detail retention in the mixed chains, while autistic chains did not significantly differ from non-autistic chains. Participant rapport ratings revealed significantly lower scores for mixed chains. These results challenge the diagnostic criterion that autistic people lack the skills to interact successfully. Rather, autistic people effectively share information with each other. Information transfer selectively degrades more quickly in mixed pairs, in parallel with a reduction in rapport.
Lay abstract
Sharing information with other people relies on the ability to communicate well. Autism is defined clinically by deficits in social communication. It may therefore be expected that autistic people find it difficult to share information with other people. We wanted to find out whether this was the case, and whether it was different when autistic people were sharing information with other autistic people or with non-autistic people. We recruited nine groups, each with eight people. In three of the groups, everyone was autistic; in three of the groups, everyone was non-autistic; and three of the groups were mixed groups where half the group was autistic and half the group was non-autistic. We told one person in each group a story and asked them to share it with another person, and for that person to share it again and so on, until everyone in the group had heard the story. We then looked at how many details of the story had been shared at each stage. We found that autistic people share information with other autistic people as well as non-autistic people do with other non-autistic people. However, when there are mixed groups of autistic and non-autistic people, much less information is shared. Participants were also asked how they felt they had got on with the other person in the interaction. The people in the mixed groups also experienced lower rapport with the person they were sharing the story with. This finding is important as it shows that autistic people have the skills to share information well with one another and experience good rapport, and that there are selective problems when autistic and non-autistic people are interacting."
Have you considered that the neurodivergent are, perhaps, tired of the neurotypicals' shit and view not giving them a voice in the process a feature instead of a bug?
Yup. Yann LeCun tweeted that people worried about x risk would be scared of ball points pens if they had just been invented: "Oh no! They can be used to write mean things! They are very dangerous and should be banned." That's how the public looks to him. https://i.imgur.com/Q5DB7VP.png
Yeah, I would think that software design and computer engineering and stuff like that selects for people who are inclined toward thinking in terms of self-contained totalizing systems: you build a whole world that you hold in the palm of your hands. But the real world is characterized by ambiguity, uncertainty, open-endedness, and the ultimate impossibility of constructing any totalizing system (epistemological, moral, economic, or otherwise) that would fully take account of it.
One factoid about Altman revealed in the profiles of him is that he was apparently bored by classroom study at Stanford and instead spent all his time playing poker. In other words, I guess the ambiguity, uncertainty, and open-endedness of, say, literature or history did not appeal to him; a self-contained rule-defined game in which he had a chance to win money, on the other hand...
From his comments I have heard he honestly seems like a naive dummy. I have heard some many good things about these AI researchers from smart people, but hearing it straight from the horses mouth they seem very unimpressive.
We have enough fraud by humans, there's no need to automate it.
The next step is to submit papers with fake data sets to real journals to find out what gets noticed. I'm pessimistic.
"ChatGPT generates fake data set to support scientific hypothesis"
"In a paper published in JAMA Ophthalmology on 9 November1, the authors used GPT-4 — the latest version of the large language model on which ChatGPT runs — paired with Advanced Data Analysis (ADA), a model that incorporates the programming language Python and can perform statistical analysis and create data visualizations. The AI-generated data compared the outcomes of two surgical procedures and indicated — wrongly — that one treatment is better than the other."
"At the request of Nature’s news team, Wilkinson and his colleague Zewen Lu assessed the fake data set using a screening protocol designed to check for authenticity.
This revealed a mismatch in many ‘participants’ between designated sex and the sex that would typically be expected from their name. Furthermore, no correlation was found between preoperative and postoperative measures of vision capacity and the eye-imaging test. Wilkinson and Lu also inspected the distribution of numbers in some of the columns in the data set to check for non-random patterns. The eye-imaging values passed this test, but some of the participants’ age values clustered in a way that would be extremely unusual in a genuine data set: there was a disproportionate number of participants whose age values ended with 7 or 8."
I don't know how much of the screening protocol is automated.
> We have enough fraud by humans, there's no need to automate it.
On the contrary -- if automated "Sokal hoaxes" end up drowning the corrupt, mutual back-scratching, information-withholding, grant-guzzling academic establishment in its own effluent, everybody (other than the grifters, that is) wins.
I think the entire response to Sokal affair was that the concerned journals stopped accepting papers from unknown people. They didn't improve, and they didn't collapse under the weight of incoming hoaxes either.
So the impact of automatically generated hoaxes will be worse for journals that try to figure out the truth, because it will mean more work for them. The journals that don't care about the truth will just look at author's name and decide accordingly.
Naturally they won't drown in submissions from unknowns (any more than e.g. NYT will.) What they might conceivably drown in -- is material from "knowns" who have succumbed to the temptation to automate crapola production. The kind of activity illustrated below:
"You can routinely find lecturers with more than a hundred published papers and you marvel at these paradigms of human creativity. These are people, you think, who are fit to challenge Mozart who wrote a hundred pieces or more of music. And then you get puzzled that, in this modern world, there should be so many Mozarts - almost one for every department. The more prosaic truth emerges when you scan the titles of these epics. First, the author rarely appears alone, sharing space with two or three others. Often the collaborators are Ph.D. students who are routinely doing most of the spade work on some low grant in the hope of climbing the greasy pole. Dividing the number of titles by the author's actual contribution probably reduces those hundred papers to twenty-five. Then looking at the titles themselves, you'll see that many of the titles bear a striking resemblance to each other. "Adaptive Mesh Analysis" reads one and "An Adaptive Algorithm for Mesh Analysis" reads another. Dividing the total remaining by the average number of repetitions halves the list again. Mozart disappears before your very eyes."
How could a fighter in the medieval ages survive multiple swordfights over the course of their life? If you watch any melee weapons fights like HEMA or Dog Brothers, it's impossible not to get hit if you're in range- any given fight has each fighter hitting the other dozens of times. Any martial art that teaches realistic knife defense starts with the principle 'you are absolutely going to get cut a few times'. Swords are obviously lethal, and medical care at the time was obviously extremely poor- if you didn't die of a sword wound you'd probably die of infection later. It seems like if you were unusually agile, athletic, and skilled you could survive maybe 2 or 3 sword fights over the course of your life.....? How could someone be a 'warrior' for multiple years or decades?
When the combat was between knights/nobles then the primary goal wasn't killing your opponent but capturing them. The ransoms were often so large that the nobles would go bankrupt over it and sometimes be held prisoner for years before they could pay. It was a major aspect of medieval society, politics, and economy.
But the reason they could survive was the knights had shields and armor. The chroniclers also often describe them as very heavily scarred. Without this aspect of capture, then even if you have armor/shield, so does your opponent and they'd just kill you after they incapacitate you.
He was a medieval knight and in later life a politician, and was said to have been victorious in over 500 tournaments, each of which involved two knights gallowing towards each other, at a closing speed of what must have been close to 40 MPH, pointing sharp lances at each other!
(Some tournaments were more like general melees, when everyone galloped around clobbering each other as the chance arose, although I'm not clear what constituted a victory against individual opponents. But they sure as hell weren't using paintball guns, and deaths were not uncommon!)
In a conventional tournament, comprising a pair of jousters, I can only assume that Marshal was extremely fast and adept at using his shield to deflect the opposing lance upon shock, while positioning his own lance with firmness and pinpoint accuracy to strike his opponent off centre somehow in a way that was hard to counter.
Perhaps he also had some trick of aiming his lance steadily for most of the approach, but quickly shifting it at the last moment, like a footballer taking a penalty kick who tricks the goal keeper by kicking the ball to the opposite side of the goal his run up led the goalkeeper to expect.
It may have helped that lances were not always pointed. In exhibition tournaments, they may have had a ball end to prevent penetrating injuries. Even so, he and the other jousters must have been knocked about something chronic over the years, although for the times he lived to an advanced age and in good health as far as I know.
Regarding your specific question about sword injuries in repeated battles, I don't think infantry vs infantry were generally as chaotic or "open" as they are usually portrayed in films like Braveheart, with both sides intermixed and fighting practically as individual pairs on all sides.
Instead of sprinting at each other, the two sides would tramp up to each other in close formation (the "shield wall") and basically most of the action would be pushing and shoving, with sword strokes or stabs as the opportunity arose. Soldiers were actually trained not to exhaust themselves in a matter of minutes, and risk their lives unduly, by madcap berzerker-style leaping into the enemy and flailing wildly as single heros!
Roman legionaries, who of course also took this disciplined approach, were even trained not to stab their opponents too deeply in the torso, just a thumb's depth would be fine. This was because a deep stab might make it slower and harder to withdraw their swords, because muscles could clamp on the sword. Also, a wounded adversary, slowing down through blood loss, is more of a liability for his fellow soldiers to carry away then a dead one.
So if they didn't take an arrow in the face, nor were run through by a cavalry lance, there was every chance an infantry soldier would live to fight another day.
Most of the tournaments the Marshal fought in were melees — single combat became important later.
In the battle of Lincoln, probably William's last fight, the Count de Perche, a leader of the French army allied with the rebel barons, was killed by a lance through the eye slot of his helm and everyone was shocked. For the knights war was just a rougher tournament, with the objective to capture your opponents and hold them for ransom, not to kill them.
Tournaments used blunt lances with weak shafts, designed to break instead of penetrate. Jousting was a dangerous sport, but it was still a sport, not a death-match.
3. Swords are not knives. Swords are much more likely to deliver an immediately-debilitating blow in the first good hit, and they deliver that hit from a greater distance, both of which mean that the losing swordsman is less likely to inflict severe retaliatory injuries against the victor.
Re 1 & 2: but don't the two combatants keep going until they've taken the other guy out? It's not like they bounced their weapons off of each other's armor for 15 minutes and then called it a day. You would keep going until you've gotten through at some point- right? So it seems like armor would just prolong the conflict, but not prevent ultimately fatal blows.
Re 3: yes but how can any 1 fighter always be the one who delivers the immediately debilitating blow in, say, 6 or 8 or 12 straight sword fights against other skilled opponents? Maybe if you have a huge skill or athleticism advantage, but if you're in enough swordfights against other trained fighters- I don't see how you can *always* be the one to land first. The question wasn't how could someone survive any 1 fight, it was 'how could someone consistently survive multiple engagements'
You consistently survive by being good enough to not lose. The loser in a swordfight usually gets killed or maimed, that's sort of the point. And fighting with swords is the sort of thing where A: skill really matters and B: skill is hard to obtain. As noted, shields and armor also matter, and serious armor is also in the "hard to obtain" category. So it's entirely possible to be skilled enough and with good enough gear that you have a <10% chance of losing a swordfight to a merely average opponent. Possibly even <1%, but if someone claims to have won a hundred duels or killed a hundred foes in battle, you should be skeptical.
Context matters. On a battlefield, maybe someone runs away. Maybe a buddy of someone comes by. Maybe an archer gets someone. Maybe someone slips. If you're both in something approaching full plate, you're probably rich enough to provide a ransom, so maybe someone yields when it's obvious they're outmatched. Or maybe, and my guess is that this was the "intended" endgame, A knocks B down, steps on B's sword, and quickly slides a sharp pointy thing in between the joints of B's armor. It would be gentlemanly to first inquire about yielding, of course. (The tuck, or estoc, had a bad reputation as an improvised weapon for peasants to kill downed nobles, and at certain times and places possession was considered intent, and was punishable by death.)
Duels are a different thing. Musashi was legendary, even though by modern standards he didn't kill that many people. And when you read some accounts of how he won, it's clear that he didn't "fight fair" if he didn't have to. And even when it comes to "fair" fights, you can go far by picking the right opponents. And even then... We probably have hindsight bias in remembering the victors more than their victims. Possibly the outliers would look less surprising given the proper context.
There’s an under-appreciated aspect of armor, especially helmets: a blow to a helmet can easily cause a concussion that in a heat of a battle would be debilitating enough to be left on the field.
If you survive the first sword fight then maybe you will get enough experience points to level up, and then you will have a much better chance of surviving future fights. Up until you run across a similarly (or higher!) leveled fighter, when you're back to square one unless you have superior tactics.
Skill. There's advantage in being trained by someone who knows what they're doing, or even just being able to observe them. We've lost that.
Necessity. People act differently when their life is actually on the line. Compare to poker and money.
Armor. Is awesome, even when it's just a few layers of cloth wrapped around your arm. Swords aren't lightsabers.
Context. Civilian duels aren't the same as battlefield combat, or prisoner extraction from a cell, or stopping an assassin with a knife, or dealing with Uncle Sven when he's had too much to drink at the feast and tries to stab your in-laws.
Grazing wounds were survivable. Perilous die roll on infection, of course, but a smallish or even moderate cut was not a death sentence. I'm reminded here, for obvious reasons, of the relatively well-documented injuries of Alexander the Great, from a shoulder injury at the siege of Gaza (yeah, that Gaza) up to a sucking thoracic wound from an arrow during the Indian campaign.
Armour, even lower-end medieval gambesons and cuir bouilli, meant that a HEMA-style touch on protected areas was not necessarily injurious. As European armour evolved to full plate, the balance swung further toward relative safety for those who could afford it.
Also, HEMA-style practitioners, despite their commitment to the treatises, almost certainly fight one-on-one with less caution and skill than the professional warriors they emulate. (To the extent the latter did at all; the frequency and significance of single combat, apart from judicial duels, is probably over-represented in chronicles and literature.) If we're talking mass battles instead, then the usual statistics apply (most participants survive most encounters, etc.)
This is a good comment, but my takeaway is that wealthy knights who could afford armor and grew up training probably survived multiple engagements. The average peasant with no real armor who was drafted into a mass battle- probably not so much
Drafted peasants probably didn't get to 1v1 swordfights much. They used spears or similar things and they formed a mob of spear wielding peasants. Sword in general was more akin to personal handgun than to a battlefield rifle - that would be a spear, or lance for cavalry.
I'd extend 'wealthy knights' to 'professional men-at-arms', which includes infantry retinues maintained by noble lords, as well as mercenaries.
The average peasant levy was in for a dire lottery, yeah, and somewhat boned when in the front ranks or under sustained missile attack. Depends on when and where in Europe, but usually he'd still have a helmet, a shield and an axe, and probably do most of his fighting using a polearm of some sort. Of course, if he was lucky, the battle might be over by way of one side breaking by the time he got anywhere near the fight, and maybe no one bothered running him down on horseback while in retreat. And if he got really lucky, as at Crecy, he might have gotten to butcher some exhausted dismounted toffs in fancy armour whom nobody cared to ransom, which honestly would have made the whole thing worth it.
Well yea, I believe there is lots of historiography supporting that summary. And in addition to training and equipment don't overlook general nutrition: a thousand years ago the wealthy and their immediate retainers ate _much_ better than the peasantry did.
Being a half-starved peasant conscripted into a medieval army and handed a sword and small shield, both items being inferior in quality to what the knights were using, and then given cursory instruction in their use, and then sent into an actual pitched battle, was pretty close to a death sentence.
> Being a half-starved peasant conscripted into a medieval army ..
You'd be surprised. They were incredibly tough by modern standards, despite their spartan diet. For example, most English foot soldiers at the Battle of Agincourt were suffering from dysentery, probably from their ridiculous superstitious habit of eating a mouthful of earth before a battle, and had been throughout the march from Harfleur for weeks, having little to eat all that while. (A few years later, Henry V himself died of dysentery. So it wasn't just the grunts.)
Something else to consider is that there weren't many battles in a typical medieval campaign. They spent a lot of time marching around, but little actual major action, maybe usually only one or two skirmishes per year.
Well....Agincourt was the late stages of the Middle Ages though. As I mentioned I was thinking of more like a thousand years ago. More Henry I than Henry V.
Same thing that for British pilots during WW2? The vast majority of them were shot down after just a few flights, but some rare pilots manage hundreds through a combination of better skills and luck.
In a real sword fight the first hit would often be debilitating enough to determine the outcome of the fight, even if it isn't fatal. No one gets hit dozens of times.
What makes you think that an average 'warrior' would survive constant sword fights over a decade? Is this based on fictional portrayals or are there specific historical figures you are curious about?
Fighter pilots are probably a good analogy for this, and have better record keeping. Shooting down 5 or more aircraft makes you an 'ace' and is considered exceptional. The highest scoring Ace, Erich Hartmann, shot down 352 planes in his career. I don't know exactly how comparable this is, maybe most of those were fighter plane vs bomber, but shooting down that many planes and not dying (although he was shot down) does give some credibility to your question.
I've seen the claim that the unusual score distribution for ace pilots is a consequence of policy, not just skill. If you shoot down other aircraft, you get promoted into a squadron leader position where other pilots support you, and then you're much more likely to continue shooting down planes - while enemy aces will probably take down your support, instead of you, which makes your career longer.
I don't know how accurate this is, but it makes sense if you imagine that most pilots just don't have the ability to win aerial dogfights, and are best used as support for those who do. It is also good incentive for people to take more risks, because taking risks to shoot down the enemy puts you in a better position to survive later.
I feel like both of these are reasonable to some extent for fighters in the medieval ages - but the question is, how meaningful is the support that other people can provide you?
I don't think "other pilots support the squadron leader" is really a thing, at least not any more so that "pilots support each other". I mean, how do you even know where the squadron leader is?
Pilots typically do have wingmen, who support each other, but that's pretty much independent of rank or status. Whoever you are, you've got *one* pilot who will keep close and support you, in exchange for your doing the same for him.
At least one prominent contrary example on the fighter pilots was Erich Hartmann, the most successful German ace of World War II and sometimes referred to as the most successful ace in the history of aerial combat (352 kills, the last one occurring literally hours before the armistice). He served as a squadron leader only intermittently in between stretches as a loner who developed and carried out his own particular tactics.
Thank you for the reference on Professor Daniel Kang's work. I've recently started the newsletter Provably Safe AI and published the second post today. It covers attacking and defending LLMs, as well as policy and standards around AI-relevant cybersecurity, so Prof. Kang's work will be a good follow-up to share next. I'm new to the topic (and to writing in this format) and would appreciate corrections and feedback: https://www.provablysafe.ai/p/2-provably-safe-against-ai
Not too much (even assuming you mean "stops selling weapons" and not just stops active support). The US didn't support Israel for most of its early years, Israel found other countries to bug weapons from and was mostly fine.
Probably Iran feels emboldened so Israel gets more terrorism, and Israel has more arms deals with China or Russia (which the US would prefer to avoid since Israel has a lot of tech they'd rather China doesn't get).
I don't think Israel would go literally genocidal on the Palestinians, but they'd want Gaza and the West Bank very firmly locked down before they redeploy to deal with the rest of the Arab world. So, a very bloody war in Gaza, followed by a shoot-first-and-ask-questions-later occupation. This would be worse for the Palestinians than what is happening now and likely to happen in the near future.
Then the rest of the Arab world (loosely defined to include Iran and indirectly Pakistan), will start sizing up the new situation. They'll see opportunity, to finally get rid of the hated Zionist foe without US meddling, but also fear that a still-powerful but now unchecked Israel might try to preemptively go for Greater Israel in the name of strategic depth. Israel, will respond to that in the usual way.
This *probably* wouldn't result in open high-intensity war a la 1967 or 1973, but it could. And a fair chance that it would go nuclear.
Absent the full-scale war, you'd get a hard-right national security government in Israel, and economic sanctions from much of the rest of the world. Israel would reach out to Taiwan, and to anyone else who might accept a sweetheart deal from Jerusalem in spite of the diplomatic costs. Some parts of the world, the diplomatic cost for supporting Israel wouldn't be that high.
This would still weaken Israel over time. So would the preferential emmigration of liberal Israelis dissatisfied by the hard-right turn, and well-eductated Israelis dissatisfied by the diminished economic opportunities. As Israel turns weaker and harder to the right, the temptation for that war (on both sides) will increase.
Also, US allies will reevaluate the value of US security guarantees, and the enemies of US allies will see opportunity, Good chance either China invades Taiwan, or Taiwan accepts reunification on Hong-Kongesque terms rather than face an invasion without US assistance.
Short-to-medium-term, Israel acts much more assertively in order to reestablish a similar level of confidence in its continued existence to what it has with current US backing.
Total US aid to Israel is $3.8B/year, or about 0.7% of Israel's GDP. Making this up by raising taxes, cutting spending, or borrowing would be a bit uncomfortable but probably not untenable since Israel is currently a little below the OECD average tax-to-GDP ratio.
As a political matter, historically US financial aid to Israel has served as an inducement and enforcement mechanism for their 1978 peace treaty with Egypt, brokered by President Carter. Part of the understanding was that the US would give economic aid to Egypt and military financial aid to Israel following the treaty, incentivizing both countries to sign and also giving the US leverage to punish violations. And giving each side confidence that the other would be more likely to honor the agreement rather than risk losing the money.
I'm not sure how much that aspect is still necessary, seeing as the treaty has now held for 45 years, but cutting it off abruptly seems like it might trigger some awkwardness.
The biggest support to Israel is not the billions of dollars - it's the implicit guarantee of protection against its neighbors. Nobody reasonably believes the US would allow Israel to be destroyed by the countries that hate it, and this surely explains part of why they never try.
In addition to stopping financial, intelligence etc. aid to Israel, explicitly announcing to the world that the US will not intervene is any way in any conflict involving Israel.
No difference. US military aid is a tiny fraction of GDP nowadays. Some Israelis actually want this as they think it would give them more political freedom of action.
I mean, I can think of at least one country which is heavily sanctioned even without a Security Council resolution.
Had e.g. China wanted to sanction Israel, they could do that right now without waiting for the UN. In fact there was once a non-UN-sanctioned OPEC embargo connected to Israel.
I.e. I heavily doubt that US veto power in Security Council is the only thing that stands in the way of global sanctions on Israel.
1) Probably more indirect and direct Iranian intervention against Israel.
2) Factions in surrounding Arab countries that supported rapprochement with Israel would be considerably weakened.
3) After a period of destabilization, the Israeli right would emerge as the new Israeli center. Defense spending would increase. Moderating impulses would be abandoned.
4) Countries all over the world would begin to reevaluate the value of American promises, especially security promises. I would expect some amount of regional realignment. China would be emboldened in SE Asia, Russia in the Ukraine, etc.
5) I don’t expect the prospect of Palestinian statehood or independence would materially grow; the Israelis are perfectly capable of subduing them on their own. You might see the Israelis cut them off completely from utilities on their side.
Lots of follow-on effects. Difficult to predict. I can’t spend any more time thinking about it today but I don’t think it would be good for those who think it would be good.
If the US totally divested militarily, financially and politically, the Chinese could roll into Palestine as a protection force. So could any Arab country I suppose but China isn’t threatened by Israeli nukes. Why would China do this? To gain popularity in the Muslim world.
Because the Palestinians would start doing terrorism on them and trying to overthrow their government, just like they have with every other ally (and Jordan and Lebanon didn't even try to come in and control new territory they want).
Well, I wrote a thing about five now-defunct Flash games and how to win them (with a bit of rambling at the start about the enshittification of free browser-based games, but you can skip that if you like).
What are the best resources to learn how to implement LLMs in real life?
Context : I’m a surgeon based in the UK NHS in the field of orthopaedics. I think that a LLM-based chat bot would be really helpful to navigate the ‘National Joint Registry’ which is basically a huge database of medical implants, outcome data and survival figures.
ChatGPT 4 (paid) now lets you create your own specialised chatbots. I’ve done this and uploaded documents to the chatbot that are past the cut off date for ChatGPTs last major update. I sometimes gave to tell it to use the new documents in its answer. It’s probably a bit tedious to upload all these documents for what you want, there may be a way to automate it.
This is what I had experimented with - just uploading the annual reports etc then asking GPT-4 to sift through the data to give the answers I want. I think with some training this could work
You're going to be running into hallucinations if you use an LLM for this. You would have to do like 100 tests of queries to this registry, where you compare what you are sure is the right answer to what the LLM says, to get a sense for the error rate. Even then, there might be weird edge cases that make the error rate climb...
There are frameworks like Haystack that let you use LLMs to query your own data. You need to write a little bit of Python code to glue everything together, but you can basically just download their tutorial example and then swap out their example data set for your data set, so you don't need to write any code from scratch.
I don't think you want to create an LLM from scratch; rather, I think you probably want to augment an existing LLM with access to the database (data protection rules permitting, of course). Have you done any coding? If so, a library like LangChain could be a good place to start: https://python.langchain.com/docs/additional_resources/tutorials
I really appreciate you guys replying. I haven’t done much coding but I’m open to learning if it would be helpful or necessary. I suppose the question is whether it is optimal for me to learn to code from scratch or just to hire someone to do it.
Is it likely to require servers etc to implement something like what I have described?
Neither of those sound like easy options, to be honest, though my experience of freelance coders is "being called in by friends to look at the results after the relationship with the freelancer has already broken down" so may not be representative.
You'll need somewhere for your glue code to run, which will need to be able to make calls to the LLM and to the database. That can be on the user's machine, but I'm guessing you want to make this web-based? In which case, yeah, you'll need a server of some sort. You could probably use a platform-as-a-service system like https://fly.io/ or repl.it for that part, which would probably be easier than managing physical servers (or even virtual machines in AWS) - though possibly more expensive. Free hosting with generous compute limits is a lot harder to come by in these days of cryptocurrency scams. I don't think you want to host the LLM yourself (as opposed to making calls to OpenAI's API), though I have no direct experience here. I have heard that open-source LLMs are getting more powerful and easy to run, via tools like https://ollama.ai/ And oh look, the Ollama blog has an article about doing something similar to what you propose, with most of the code running in a browser and talking to an LLM running on the client's machine: https://ollama.ai/blog/building-llm-powered-web-apps That doesn't sound easy, though, and also I think you'd have trouble getting clinicians to run LLMs locally.
Another option to look at is OpenAI's recently-announced "GPTs", which are customised GPT-based apps running on their infrastructure: https://openai.com/blog/introducing-gpts They say that building one requires no code, but I don't know how true that is. I also suspect that directly connecting an NHS database to OpenAI's infra would result in some extremely harsh words from their Information Governance people.
I've recently found the amount of antisemitism on Twitter anxiety inducing and pretty harmful to my life. Is there a good way to completely block the ability to access it from my phone? I can do it on my PC, but I haven't found a way to IP block from my Android phone
That's weird - for most of twitter's history, hundreds of thousands of its users have seemingly spent much of their life posting about how much they hate white people. Was that upsetting to you? Or is it only a problem when it affects the group you identify with? Was it harmful to your life when schools, universities, books etc. starting telling the world that white people have "white privilege"? Were you angry when jewish groups failed to condemn this but instead simply demanded that jews be excluded from the concept?
a) yes, it did bother me, but obviously it's much worse when they're cheering people who have murdered friends of mine and shelled me out of my home, in real life, than when they're cheering abstract hostility.
b) you might want to go through your comment and try to think if it's the sort of thing that would make people more sympathetic to you or to your enemies.
Hey, I’m sorry to hear it’s affecting you that much. I’ve seen a lot of hateful stuff lately. Unfortunately, I think it’ll be at this intensity for a few more months at a minimum.
Is the issue that cutting the cord with Twitter is a hard habit to break? Or are you seeing it embedded/screenshotted on other sites when you don’t want to?
I’m on an iPhone, but I assume there’s something similar. I wanted to block web usage more generally, so I had a family member use the parental controls to lock me out every web browser and app that I didn’t want, and then put the code in a sealed envelope. I can get past it if I’m determined enough, either technically or by getting the envelope. But it’s enough friction that I haven’t done so.
Cutting the cord really - I get into the social media doom loop of seeing hate and both worrying about it and feeling compelled to go back to it, which I know is bad for me.
Looks like there's similar apps for android, hopefully this should help.
I use Stay focused for android and Cold Turkey for MacOS, for the same purpose (blocking Twitter). They're great and blocking Twitter has had a noticeably positive effect on my well being
Uninstalling the app is a bit more radical but you could try. You can tell yourself that if you really need to look something neutral up you can just open it on the mobile browser, and the extra awkwardness of the browser vs the native up might help you avoid getting stuck in it.
Congratulations on your life choices, for Android Chrome browser I use an extension called AppBlock. It often sends useless notifications, but it works.
You know you can disable any app's notifications from Android's settings, right? On my phone only a few apps can notify, and most of them only silently. Don't let a single unwanted beep go through!
One thing I've found that completely changed my Twitter experience was making a Twitter list with people/orgs in my current interest area (nuclear and fusion power) and having a direct tab link to it
When I first got into Twitter I got algorithm'd into an emotional whirlpool of political concern about woke stuff. There should really be a large revenue neutral social media tax that taxes the more anti-social platforms and funds better ones
I had a serious look into doing this, but I couldn't find any easy way to do it. In principle it should be possible to modify the hosts file if you root the phone, but that was too much effort for me to bother.
Beyond deleting the app, perhaps you can change your Twitter password on your computer to something long and complicated that is too much effort to either remember or type on a phone? When I've quit social media in the past, simply logging out or deleting my account and then making it hard to log back in or make a new one has worked quite well.
I am a person intimately familiar with PTSD and CPTSD, so I suppose that makes me an expert or an unreliable witness. (Funnily enough that tension has a lot to do with PTSD; losing the ability to trust yourself.). I have rather a high opinion of VdK's work, and I read Scott's review with interest.
I have some thoughts on it.
VdK is quoted as saying ‘WE were not designed for this” or something like that. Scott says ‘yes we were’ and points to the long history of humans suffering all kinds of trauma. That is true, but I think it misses the point. First, the condition is called PTSD; POST-traumatic stress disorder, not Traumatic stress disorder. The whole problem lies in the framing of the trauma, not in the trauma itself. To give examples;
1. a hurricane destroys your community leaving everyone without shelter and whatever other havoc it created, up to and including deaths. There is a scale of suffering clearly (some people lose a parent or child, some people only lose their material property) but the important thing is it affected everyone, NOT JUST YOU. This is important. This kind of trauma is easier to process; it doesn’t need to be taken personally.
2.A person is attacked/abused/neglected by people close to them, and the outside world doesn’t see, or seem to care. This is more difficult to process. The horror of it is internalized. Trapped. Blamed and punished. Betrayed. That is POST traumatic stress.
If you are a child and these are formative experiences and you are powerless to escape them, then it’s really bad for your psyche. An adult can seek confirmation of the trauma in the outside world, and maybe find some relief, in that they are not alone and the world around them confirms that what happened to them is an aberration. This is where another of Scott’s comments comes to me, The landscape of mental or emotional disorder is very dependent on what the norm is around the person in question. In other words, in a world where mortal combat is an accepted part of living, slavery is an accepted part of living, physical pain is an accepted part of living, incurable diseases and all the rest are an accepted p o l, then the baseline for what we consider trauma is different. People don’t get traumatized by seeing someone hacked to death because they grew up in that world and it’s just the way things are done.
Animals process adverse physical events in a very straightforward way. They might survive them or not. Then there is the whole fight flight freeze thing, which Pavlov’s work pretty much lays out in great detail bye creating a large number of perpetually frightened dogs. I know that was not his point but the idea of what PTSD is, is implicit in his work imo. Something is trapped, and that something is visceral, not cerebral. Inevitably though, because we are humans it becomes cerebral, and that is where the fun starts. But what if it never had to become cerebral? What if we let our body deal with its injury, and didn’t interfere? That would be great wouldn’t it? I know some people who are better at this than others, but regardless, we have to unravel the story behind the trauma and the story we are telling ourselves about it. That’s where PTSD lies.
The story of Scotts patient is instructive; she cured herself. How? By reframing her narrative of the trauma in a way that she could accept. I would be curious to know if there was a physical component in that moment (crying, a show of anger).
I know of two early accounts of PTSD in fiction. In LOTR, Frodo's injury from a poisonous blade toward the beginning of the book haunts him physically and mentally, including feeling worse on the anniversary of the injury. There is no cure in this world.
In _Citizen of the Galaxy_ by Heinlein, Thorby is haunted by being kidnapped into slavery as a small child, with a short temper and nightmares as a consequence. He's eventually cured by hypnosis (memory of hypnosis in a dream after his life is greatly improved?) which shuts the bad memories away.
Any other examples?
As for trauma being a routine part of life for early humans (I think trauma could reasonably include any sort of bad injury) it could be like bones. Bones are pretty strong, but sufficient force can break them.
There is something essential missing in this. First, the condition is called PTSD; POST-traumatic stress disorder, not Traumatic stress disorder. The whole problem lies in the framing of the trauma, not in the trauma itself.
If you’re interested in this question, Achilles in Vietnam by Jonathan Shay you is a must read. He use’s the Iliad as a metaphor for ptsd, and argues that in the case of complex ptsd, the trauma most often has some moral component.
Being betrayed by a caregiver or parent is a big component of PTSD. Being betrayed by anyone you have put trust in is traumatic, but if you have counter-examples already (you have been treated fairly by others), and then WHAM a bad thing happens, it is easier to recover a sense of trust. If it is a formative experience and goes unaddressed it becomes an attachment disorder essentially. That manifests itself in all sorts of ways.
I did not address that particular one, because I really don’t have a clue about it. I think there might be betrayal, if one feels they saw, or experienced a sacrifice that turned out to be meaningless. There is also survival guilt, perhaps seeing a comrade of yours destroyed, and you walking away from it. And always there is the whole history of one’s life that one is bringing to the experience. It’s complicated. The reframing of the traumatic experience is still the only way out though, imo. That could be as simple as crying it out or as complicated as a novel. Ymmv
I’m looking for some recommendations for sci-fi novels. Specifically, space colony type novels. Does anyone have any favorites they’d recommend from that sub-genre?
+1 to Kim Stanley Robinson's Mars Trilogy. It's about multiple generations of early Mars settlers establishing a colony/society and terraforming the planet. It might be my favourite series ever. It's a little slow in parts because of how much detail he goes into, but the detail is also what makes it amazing.
From my childhood, The Escape Orbit is an interesting subversion of "colony growth" with the colony being a prisoner-of-war planet advancing to rebel against their alien captors.
1- The Expanse, the first 3 novels are partially set in a prosperous (10 billion humans) Mars and populous Ceres, Europa, Ganymede, and Eros. [MINOR SPOILER] The 4th book is entirely set on an alien planet, and starting from the 7th book onwards a whole network of several alien planets.[END MINOR SPOILER]
2- Shards of Earth trilogy
3- House of Suns by Alastair Reynolds, as well as the Revelation Space series by the same author (what I read : Revelation Space, Redemption Ark, Absolution Gap)
4- Takeshi Kovacs series by Richard K. Morgan : Altered Carbon, Broken Angels, and Woken Furies. (There is a show but the first season is inferior to Altered Carbon, and the second is outright trash). As well as the Black Man novels, Black Man (Thirteen or Th1rte3n in the United States) and Thin Air. There is an upcoming Gone Machine but it keeps getting delayed, this is an extract from it https://www.richardkmorgan.com/2021/03/gone-machine/.
Legacy of Heorot by Niven and Pournelle for colony on a planet.
For space station life, try Andrew Moriarity's adventures of a jump space accountant books. "Trans Galactic Insurance" is the first in the series. Interesting deep dive into the economics of keeping space colonies runnning.
A great deal of Jack Vance is set on colony planets. Maybe I can dredge up more, but _Big Planet_ (gigantic low density planet with weird little cultures), _Blue Planet_ (low metal culture), the Marune novels, the Planet of Adventure series (protagonist must gather four plot coupons (parts for a ship to escape) across a planet with alien species), _Emphyrio_.
Oh, absolutely Rosemary Kirstein's Steerswoman series. It starts out looking like fantasy, but... The series has been on a long pause, but I gather a new book is in progress.
Could you go into slightly more detail about what you mean by "colony type novels"?
Mostly set on a space station: "Ethan of Athos" by Bujold. Her "Falling Free" deals with the creation of a space habitat, with more focus on engineering than most (probably the familial influence). V. Vinge's "A Deepness in the Sky" might as well be about that, plus there's the "first contact" aspect too.
If you want messed-up scenarios of planet colonization, Niven (et al.) wrote a bunch of them. Also "Dragonsdawn" by McCaffery.
Bujold's Vorkosiverse series is mostly about colony worlds (as I recall, there's a little on Earth). Komarr has a lot about what it means to be a colony.
Heh, yes, both a colony in the sense of "long-term terraforming of a marginal world, and the social and economic impacts of the environment", and in the sense of "outpost of an imperial power". I'd never thought of that before.
> Could you go into slightly more detail about what you mean by "colony type novels"?
I guess I was thinking more about colonizing a planet, rather than a space station. I was thinking more like the "generation ship lands on a interstellar planet" type story.
I haven't read anything by Niven, which would you recommend starting with for messed up planet colonization?
Sorry this is so late. I was misremembering Niven; "Legacy of Heorot" and "Integral Trees" loomed large enough in my mind that they created an impression of more. I think the only other one that might fit is "A Gift From Earth".
I'd again recommend "Dragonsdawn" by Anne McCaffrey. It's kid of my baseline for these sorts of stories. It's got the whole thing, starting from 3 colony ships inbound at the edge of a system, to filling in details missing from the 200-year-old survey records, and the tensions between the different groups of settlers (charterers who put up the capital, contractors who were hired as technical specialists, and the remnants of Earth's nomads who were forced onto the expedition), and people with ulterior motives, and a disaster happening a few years in (which isn't a spoiler if you know anything about Pern).
Another one that might work for you, but which wouldn't be on most people's radar, is "The Romulan Way" by Diane Duane. It's one of the Star Trek novels from the mid 80s, right before ST:TNG started, presenting a different take on the Romulans than what canon eventually went with. Its chapters alternate between the "present day" story, of a Federation operation to retrieve a deep-cover anthropological researcher who's gone native, and a historical overview of the Romulans, from Vulcan's first contact with aliens, to the schism between Surak and his leading disciple, to the long mostly-fatal voyage on generation ships, the colonization of a new system, and the broad strokes of the history the survivors wrote in blood and wine. I know the ST novels don't have much of a reputation, but this is a good one. It's like something an early Le Guin might have written, back in the "Rocannon's World" days. (And since I'm mentioning ST novels, the absolute best of all is "The Final Reflection", by John M. Ford. It's a mystery disguised as an anthropological study disguised as an adventure, this time involving pre--3rd-movie Klingons.)
Ha, it’s funny you put that because you mentioned Dragonsdawn earlier and I looked it up and I’m literally reading it now. I just got to the disaster. I haven’t read any of the prior books, but when I was reading through some reviews I found out what it would be. It’s a bit strange to start with Book 9, but some people suggested that you should start the series with that, so that’s what I’m doing. And you’re right, it’s exactly the sort of thing I was looking for. After Dragonsdawn, do you think I should just go back and read book 1?
I haven’t read a lot of Star Trek books, but I did read a lot of Star Wars novels, so I’m open to that sort of thing. I actually still have two full shelves on my bookcase that are just Star Wars novels from when I was a kid.
Thanks for this. People gave a lot of really helpful suggestions. I've definitely got a couple months (or more) of reading out of this.
Regarding Pern, yeah, definitely try them out. They're classics for a reason, and they're easy reads so they won't take much time.
The "main 3" books are "Dragonflight", "Dragonquest", and "The White Dragon," which all take place about 2500 years after "Dragonsdawn", when humanity has retreated to the Northern Continent, forgotten their history, and descended into a sort of low-violence cave-dwelling feudal stasis. "Dragonflight" is a must-read classic, "Dragonquest" is 7-years-later exploring all the problems caused by the solutions in the first book, and "The White Dragon" is when they really start uncovering 2500-year-old archaeological sites.
The "Harper Hall" trilogy are young-adult books that are probably best read after "Dragonquest" and before "The White Dragon". I'm very fond of them, not the least because they provide a ground-level view of the society that's missing from the main 3. Like, I think they're the first time that money is mentioned! (You can probably tell that I like the sort of stories where there's a plot that allows us to see the day-to-day life of alien cultures.)
After those, there's "Renegades of Pern", which spans over all the previous books and provides even more of a ground-level view of the seamy underside of Pern. And, uh, I won't spoil what happens in that and in "All the Weyrs of Pern", but, well, if you're reading ACX, you'll probably be quite interested, and I'd love to compare impressions if you make it that far. ;-)
There's also some books set afterwards, and during, and some books set in earlier eras, but those aren't essential IMO. Like "Dragonsdawn", the ones I've mentioned are what I consider to be a reference-class implementation of a classic sci-fi trope.
Regarding the Star Trek books, if you're open to it, maybe read "My Enemy, My Ally" before "The Romulan Way", as it was written first, and introduces a few characters and starts off the world-building. It's also a great read, in the David Gemmell "honor porn" sense, which is not something I'd usually expect to say about Star Trek or Diane Duane. :-)
Have fun! I'm really curious to hear what you think after you're done. :-)
I would personally recommend against Niven. "Ringworld" was terrific up until the cast lands on the titular ring, after which it becomes a boring and tedious plot that I trudged through, aching for it to just conclude already. YMMV, of course, it is fiction after all.
The only thing that comes to mind as a recommendation is *very* tangential: "The Three Stigmata of Palmer Eldritch" by Philip K. Dick. The possibility of being sent to Mars as a colonist is considered a life-changing threat to people in-universe. But, it will definitely tickle the "messed up" half of your request; being sent to Mars as a colonist is a death sentence.
A short description: The future Earth of 2016 is in such a climate crisis that the United Nations has instituted a draft of sorts, sending people off to other planets and moons to colonize them. It is too dangerous to be outside on Earth in the daytime without specialized gear.
The experience as a colonist is uniformly awful, and people resort to an illegal hallucinogen to cope, which requires the use of a prop called a "layout" in order to give the user a temporary alternate reality. The novel's main character works for the company that makes the popular Perky Pat layouts. A famous lost explorer, Palmer Eldritch, is rumored to have returned with a new substance that is better, in that it doesn't require a layout.
"The Integral Trees" comes to mind, and there's even technically a planet involved. ;-) I'll try to come up with some more in a bit, but as mentioned below, "The Legacy of Heorot" would be in there.
The Moon is a Harsh Mistress is probably the best book I’ve read that is specifically about a space colony.
I think Kim Stanley Robinson’s Mars trilogy, and his novel Aurora, are probably necessary reading in that sub genre but I don’t really recommend them. They aren’t quite as erudite and hard sf as some have made them out to be—a lot of scientific terminology covering over the handwaving of underlying concepts—but they are well written in the sense that you’ll keep turning pages. However KSR is a doomer and most of his characters don’t really have agency.
Lord of Light by Zelazny is definitely about an a extraterrestrial space colony, and is brilliant, but most people would probably be annoyed at its inclusion in this list because it meets the letter and not the spirit.
Ringworld is not the letter but it might be the spirit.
"Learning the World" by Ken MacLeod, about a generation ship that arrives at their destination and discovers that they've unexpectedly made First Contact. It's something of a satire on the present-day tech startup industry, but also much more than that. One of my favourite SF novels ever.
The worst thing I think I could say about TMiaHM is that it's not much concerned with the technical ins and outs of building and managing a colony, assuming that's what the OP is looking for. It's *much* more about the social ins and outs.
KSR's Red Mars is about both. But even then, as I recall, the technical stuff felt mostly tacked on to the social stuff.
I will second Theo's recommendation of the Bobiverse. I listened to the audiobooks and they were fun.
I just started the Sun Eater series by Christopher Ruocchio and it is off to a great start for epic sci-fi. I don't think it quite fits the bill for a colony theme, but I happen to be reading it right now and it's really good.
Adrian Tchaikovsky's Bear Head is a very good Mars colony book. It's a sequel, but the first one (Dogs of War) is fantastic and worth reading anyways.
The other Scott blogs about (among less CW relevant stuff like his take on OpenAI) about Greta chanting "crush Zionism", linking the SWC as a source. "[She is] taking time away from saving civilization to ensure that half the world’s remaining Jews will be either dead or stateless in the civilization she saves."
I think I like that framing better than the one of the SWC, which is simply calls that murderist...err ... antisemitic. Personally, I think Zionism has as many interpretations as feminism, some of which I find myself totally on board with (continued existence of Israel) and some I strongly oppose (settlements in the West Bank). That makes it a useless word for anyone aiming to communicate clearly, but the general vibe I get from "crush Zionism" is still one of calling for the destruction of Israel.
While one can oppose a state without being racist towards its inhabitants (for example, I would be on board with "crush the Aztec Empire"), I think that letting Israel continue to exist is actually a great idea. There may well be a statistical correlation between antisemitism and antizionism, but I think that the narrative of Ms Thunberg picking up an old diary of Adolf Hitler and getting infected with antisemitism is probably not what happened.
Like Scott Aaronson, I too admired Greta Thunberg when her mindless stridency, black-white thinking, childish ignorance, and inability to grasp nuance brought her down on the same side of an issue as me. I just couldn't have foreseen this would ever come back to bite me, though.
I find myself seeing no less than 4 claims/assumptions bundled into S. Aaronson's seemingly simple statement :
1- "Crush Zionism" is a call to action and/or a planned course of action in Greta's mind, and not a generic "boo outgroup" said for emotional bonding with the ingroup and emotional posturing against the outgroup
2- Zionism is a pro-Jewish ideology identical with the proposition of having a Jewish State for Jews
3- Having a Jewish state for Jews is a good thing
4- Destroying Israel will result with high probability in its Jews being dead or stateless
My reservations about those 4 claims are as follows :
1- "Crush Zionism" is suspiciously identical modulo language to the second part of the main chorus of the song Leve Palestina ([Long] Live Palestine) by the mixed Swedish-Palestinian band Kofia https://www.youtube.com/watch?v=eJWyzfJafmI. Kofia is an obscure band founded in 1972 by a 1967 refugee Palestinian musician, and it sang for pretty revolutionary lyrics that bucketed Zionism with "Reactionarism'', which I assume to be the generic Communist outgroup label, and Imperialism. The song's words are catchy and the tune is pretty fire, even if I don't fully agree with the implicit claim that Israel's founding is straightforward imperialism.
History lesson and my musical preferences aside, the song is chanted in Pro-Palestinian protests in the Arab/Muslim-hosting Sweden, Greta is Swedish and ideologically aligns with welcoming refugees, "Crush Zionism " maybe nothing more and nothing less than an ingroup-pleasing way of encoding "I attended Pro-Palestinian rallies and/or I approve of them".
If you want to treat it as a claim, you have to put it in the context of a late 1960s Israel that assaulted its Arab neighbours and more than quintupled its territory with annexed land. It's "Crush [1960s] Zionism [that stole my home and forced me into diaspora]", not simply "Crush Zionism".
2- Literally true, but the words don't always mean what their users claim or want to mean, or everyone against Antifa would be a Hitler/Mussolini fan. At the bare minimum, it should be taken into consideration that :
2-(a) Zionism is not originally a Jewish intellectual product, but a product of Christian apocalyptic thought https://en.wikipedia.org/wiki/Christian_Zionism that was passed into Jewish info stream.
2-(b) Zionism was embraced by many many many anti-semites, this doesn't mean it's false or wrong but it's curious that an ideology that claims to be for Jews was embraced by so many people who claim they're against Jews (but it **could** be that Zionism is such a good and win-win idea that everyone reasonable can't help but be for it, and the anti-zionists are only being unreasonable)
2-(c) Early (< 1917) Zionism experimented with different potential locations for the Jewish state, such as Uganda and the Argentines
Given that today's Zionism is very much NOT a..c and today's Zionists either don't know a..c or don't have good counters to them when reminded, I think it's fair to say that "Crush Zionism" is something that even Zionism did, to its own early prototypical versions.
3- (a) I'm an Anarchist, and I don't think States are a good thing.
(b) Insofar as States are ***Necessary*** in our days' cutthroat world that only recognizes States as Sovereigns, it strikes me as incredibly bizarre that "Jewish State" doesn't get the same objections and raised eyebrows and primers about Secularism from the secular and liberal Scott Aaronson as "Islamic State" or "Christian State".
"This is a Christian Country", this is a claim about America by a large faction of Scott's ideological enemies, the US republicans, he presumably has a LOT to say against it. Why is a Jewish country okay with him ?
4- My own assessment of the majority-Muslim Arabs' collective attitudes towards Jews mildly supports this, but this attitude is an unreliable indicator given that it's an (extreme and bad) reaction to 70+ years of Israeli arrogance and aggression. The most compelling evidence against it is how Israeli Arabs and West Bank Arabs seem to be pretty fine in aggregate with living with Israeli Jews, selling homes to Israeli Jews (the death penalty against that exists for reason), and marrying the occasional Israeli Jews.
I overall agree it's a sensible thing to worry about and think Pro-Palestinians generally but Arabs specially need to emphatically and Zealously distance themselves from Anti-Jewery, up to and including disciplining members who stray into Anti-Jewery during protests and simultaneously chants against attacks on Jews and the individuals/groups doing them while it chants for Palestinians.
Just a narrow comment, ignoring the wider topic, on which I have personally no special insight and nothing to contribute.
> it strikes me as incredibly bizarre that "Jewish State" doesn't get the same objections and raised eyebrows and primers about Secularism from the secular and liberal Scott Aaronson as "Islamic State" or "Christian State".
It does strike me like all the known examples of religiously themed states are a bit of a disaster. Separation of Church and State is one of the few ideas in history that I find unambiguously good.
To copy over a comment I made in the last hidden thread:
Perhaps the most perverse thing about the whole history and situation is that, if the Palestinians had laid down their arms in 1948 after their side lost their war, just as the Germans and Japanese did in 1945, they would probably be in the same position as the Germans and Japanese are in today: thriving industrial economies. A two state solution could have looked like Canada and the USA or like the https://en.wikipedia.org/wiki/Schengen_Area . The Palestinian kids could have been getting degrees at the Technion.
( Any remaining territorial disagreements could have looked like https://en.wikipedia.org/wiki/Whisky_War "In 1984, Canadian soldiers visited the island and planted a Canadian flag, also leaving a bottle of Canadian whisky.[9] The Danish Minister of Greenland Affairs came to the island himself later the same year with the Danish flag, a bottle of Schnapps, and a letter stating "Welcome to the Danish Island" (Velkommen til den danske ø).[10][11][12] The two countries proceeded to take turns planting their flags on the island and exchanging alcoholic beverages." - presumably substituting coffees for the whiskies )
I think the problem with that plan is that the Palestinians didn't have all that many arms to lay down. It was the Syrians, Jordanians, Egyptians, et al, who had the large and powerful armies. And had no reason to lay down their arms because they had large territories that they needed to defend(*) and could reasonably hope to defend.
So, no matter what the Palestinians do, Israel is going to be surrounded by large, heavily-armed nations that just tried to destroy Israel. And no matter what the Palestinians do, the Israelis are going to reasonably fear that, come the next war, the Palestinians will roll over for their fellow Arabs and offer up their lands as a highway for the invasion of Israel.
That's going to put a damper on peaceful economic cooperation. And occupied Germany probably wouldn't have done so well if there were a much larger unconquered Nazi regime on their border to cause trouble.
* Mostly from each other, admittedly, but an early bid for "Greater Israel" wasn't entirely out of the question.
" And no matter what the Palestinians do, the Israelis are going to reasonably fear that, come the next war, the Palestinians will roll over for their fellow Arabs and offer up their lands as a highway for the invasion of Israel." Ouch! Excellent point!
I am not knowledgeable enough about context to properly evaluate the significance of this, but the wikipedia article on the United Nations Partition Plan for Palestine says
In a letter to his son in October 1937, Ben-Gurion explained that partition would be a first step to "possession of the land as a whole"
Which makes me wonder if it was ever going to turn out as well as you seem to think.
You might be right. On the other hand, Ben-Gurion was not the sole actor, and he might have thought differently 11 years later. In any event what actually wound up happening was three quarters of a century of repeated warfare, https://en.wikipedia.org/wiki/Hatfield%E2%80%93McCoy_feud writ large, blood feud after blood feud after blood feud.
Well, if the machines _do_ wind up taking over, at least whatever conflicts they have will probably be _different_.
Hainana's completely wrong. Palestinians don't do well as shown in the link above, and in any case he's completely ignoring selection effects. The good ones leave. The people left in Palestine are the dumbest ones.
Gaa! Thank you for the link. That was quite a surprise to me. I don't know what to make of it. That does make the situation a lot bleaker, even if there had been peace. :-(
It's the mother of all false equivalencies to equate a war that Germany and Japan started with the explicit purpose of conquest and genocide with a war forced on Palestinians because Zionists wanted conquest and genocide.
> if the Palestinians had laid down their arms in 1948
And if the Zionists had laid down their weapons in the 1930s there wouldn't have been a war and the Jews would have kept living like they did for the last 2000 years. So what's the point ?
"with a war forced on Palestinians because Zionists wanted conquest and genocide."
Actually, it was the combined Arab armies which invaded Israel in 1948 with the intent of genociding Jews - in their words, they intended to launch "a war of extermination and a momentous massacre which will be spoken of like the Mongol massacres and the Crusades". And the outcome of that war was that many Arabs continued to live in Israel while not a single Jew remained in the West Bank or Gaza Strip - which side was the genocidal one here?
Sure, start the movie exactly at the moment Arab armies invaded and not a single femtosecond before, and selectively cut to various populists saying sick things about Jews, Israel looks in the right.
I can also start the current movie at the exact moment the IDF invaded Gaza, and, boy oh boy, do I have plenty of sick things about Arabs and Gazans the Israelis said. Two can play this game, but I'm not going to because it's a fundamentally silly game.
But the fact of the matter remains that Jews lived in peace with the Arabs for no less than 400+ years before Zionism, 5 times that of Israel's entire lifetime, and that the troubles started when some rich nobodies colluded with colonial powers to make a state to act out the fantasies described in an ancient and supposedly divine book. Correlation doesn't imply Causation, but Correlation alone is damning enough.
>But the fact of the matter remains that Jews lived in peace with the Arabs for no less than 400+ years before Zionism
Because there was a leviathan, the Ottoman Empire, occupying the place and willing to use force unconstrained by human rights norms. That Muslims were favoured also helps. Violence persists in Israel mostly because there are constraints on their exercise of a monopoly on force.
I respectfully disagree about where responsibility for the 1948 war lies. Even so, _regardless_ of who was responsible for the 1948 war, it is a pity that the end of it did not lead to a result analogous to the end of WWII. Just looking at Europe, Germany, France, and Britain have been killing each others' people in various combinations for centuries, yet they finally managed to bury the hatchet after 1945. It is a pity that the Palestinians and Israelis didn't do the same.
There was another war, though, and one side tried to make sure Jews would not continue to live in Palestine "like they did for the past 2000 years" and that side has seemingly won the war of "ideas".
That side has the upper hand in Arab discourse only because the IDF keeps breakfasting on dead children, giving them more ideological ammo than the literal ammo that the USA's billions buys for Israel.
Germany and Japan also had their version of a war forced on them narrative. An overt act of war followed by a "look what they made me do" storyline for the domestic audience is not special, nor original.
Sure, it's not the story that makes something true, after all Israel uses the "look what they made me do" trick till today largely unsuccessfully, it's the historical record.
> An overt act of war
You're talking about the Zionist militia terrorising Arabs and marauding them for the entirety of the 1930s ?
> for domestic audience
Yeah all those millions for Palestine around the world are deluded domestic audience brainwashed by the storyline invented in Pallywood.
"A popular uprising by Palestinian Arabs in Mandatory Palestine against the British administration of the Palestine Mandate... demanding the end of Jewish immigration and land purchases... the number of Palestinian Jews killed are up to several hundred"
Overt acts of war tend to be followed by wars. It's a bit unusual to engage in these when the outcome is both predictable and likely very unfavorable. It is arguable the domestic audience is deluded if they believed this chosen action was going to end well for them.
Low probability actions might be rational when there is little to lose but they are still low probability and only one possible choice among many other options. This action doesn't appear to be particularly wise in the short term or the long term.
This is not an argument about the Palestinian cause, it is an argument about whether this action had a realistic chance to be effective. It may be as simple as bad leaders making bad decisions.
If Zionism had never existed, it is fair to assume that a hypothetical Palestine would probably be comparable to the surrounding countries, which are far worse off.
I think a more relevant question would be if the Palestinian Arabs would have been better off without Israel and if the Israeli Jews would have been better off in the diaspora. The country-level comparison seems inherently misleading.
The more relevant question would be if the Israeli Jews would have been better off without the Palestinian terrorists and if the Palestinian terrorists would have been better off in the diaspora.
The meta-question is "better off -how-?" There's more to life than GDP.
Frost was right when he said that home is the place where, when you go there, they have to take you in. Anyone anywhere on earth who has a home of this sort does so on land that used to be someone else's. Almost any ethnicity that has a home anywhere does likewise, with the 'almost' referring to tribal groups indigenous to relatively new volcanic islands. Historically, it's the stronger who get the homes. Totally reasonable for Jews and Palestinians to both want the same home, and to feel that they'd be 'better off' with it than without it.
It occurs to me that stateless is a very different order of negative outcome from dead. For example, we were willing to put up relatively quietly with the Palestinians being the former, but only began shouting once an inordinate number of them became the latter.
You're correct, of course, about the many different faces of Zionism, but considering that the actual destruction of Israel is a practical impossibility, and a Likud-led government is presently in charge of Israel, it's probably fair to interpret their version of Zionism (which has historically supported the settlements, full control of Jerusalem, etc.) as the most relevant definition. I, too, would like to crush that Zionism and preserve the Zionism that leaves room for Palestinian self-determination.
I'm also not hugely impressed with Aaronson's call to "spite Greta on climate, but SMARTLY", though that's beside the point.
Re statelessness: I'm not hugely impressed with the idea that Jordan is existentially threatened by the presence of Palestinians more so than Israel is.
I also find amusing the idea that Palestinians could repossess Israel and it would still be Israel-like. South Africa is apt. Seems to me Israel took/takes work, and requires a population that wants to work. Arabs ain't it.
I don't think Palestinians expect or require Israel-level living standards in their hypothetical state. They're quite accustomed to misery by this point.
And Israel is not existentially threatened by the presence of Palestinians. It's not existentially threatened by much of anything. Even without American patronage, it would take a virtually impossible regional military coalition willing to endure one-sided nuclear retaliation on top of everything else to seriously threaten the country's existence. And as you so charmingly observe, Arabs are far too lazy for that.
Egypt seemed to hold its own pretty fine against Israel in 1973, where "held its own" is Golda Meir quite literally begging Nixon for another army's worth of equipment instead of the one Egyptians destroyed.
The Houthi rebels 2000+ KM away wreck havoc in Eilat with just UAVs and cruises, the entirety of Israel is wide open in Egypt's missile range, an army with 1 million standing personnel for Israel's full-mobilization 500K.
> Even without American patronage
I would bet that easy money, but the AIPAC won't let me get rich that easily.
I don't know where you're getting the 1M standing personnel for Egypt's armed forces. It might approach that with reserves fully mobilised, though only the land army would be a relevant branch. Israel's active+reserves is indeed roughly 500k, but in a situation of existential defence, mass conscription of another 2M or so men and women would be expected. I also highly doubt Egypt has the logistical resilience to support its fully-mobilised army in the field for any significant length of time, especially in offensive operations against a technologically superior enemy.
But this is a silly hypothetical anyway. We're a long way from 1973. For one thing, the first Israeli nuclear test was conducted in 1979. For another, Syria is in shambles. For yet another, Egypt and Jordan have long-lasting peace treaties with Israel.
And I would cover the bet without a second thought if the win/loss condition was that Israel ceases to exist as a state - which is what existential peril means to me. The only relevance of 'destruction of Israel' is as a rhetorical motte. The most grievous possible loss condition for Israel is significant territorial concessions to create a viable Palestinian state, which might become politically palatable if the United States were to withdraw its support and Israel began to suffer economically from international isolation. (I don't think a binational state would solve anything at this point.) Also largely a silly hypothetical, unfortunately.
Yeah I misremembered that 1 million figure, Wikipedia says more like 600+K in total, but with Egypt's population being 100 million you can expect no less than 3-5 million fighting-age able-bodied males to be available on very short notices. Training (not arming, Egypt has a native arms industry) will be a problem, yes, but really just freeing up the whole millions-of-KM long borders along Lybia and Sudan so that the better-trained core can focus on Israel would be a huge advantage over Israel which has Hezbollah and Al Assad to its back.
Keep in mind that 1 or 3 million is barely a dent in Egypt's 100 million population, while Israel's 2.5 million is 25% of its entire population. This will probably cause some shenanigans, hell even the recent 500K mobilization was costing Tel Aviv an estimated 250 million dollars or so every day. Egypt's economic deficiency is real too, but in a world where the USA is isolationist the Gulf has to find someone to ally with in the face of Iran, and there is no better someone than Egypt and Turkey. Egypt is closer to the Gulf.
I also agree that destroying Israel with a war is not the outcome I'm hoping for as my #1 wish, regardless of how much I hate it and the sheer contempt I have for its history and political composition. But I do think it is very possible, being surrounded by neighbours that all hate you tends to make you very geopolitically fragile.
> For one thing, the first Israeli nuclear test was conducted in 1979.
This can only work as a suicide strategy to deprive the invading Arabs from an intact Palestine, there is a reason they call it "The Samson Strategy". Whether Arabs will back down or call the bluf is debatable and very dependant on exactly who is calling the shots, in terms of people and also in terms of countries. But if Israel is forced to use this it's already vastly reduced and surrounded from all sides.
> I don't think a binational state would solve anything at this point.
Binational state is a weird term to call a modern multi-ethnic state, this is probably the single best solution to the conflict, but the one with the greatest price tag.
In an atmosphere where relativism prevails, it seems a little funny that there is a vestigial disapproval, not of laziness, but of describing people as lazy (or uninterested in a particular cultural definition of work, at any rate). That can only derive from a feeling that lazy = unvirtuous, which is pure retro.
ETA: but I've noticed that progressive types like to keep these little judgments in their back pockets, though they would never own up to them.
Yeah, call me a cynic, but somehow I don't think "my enemy doesn't have the aptitude to make the desert bloom" is intended to carry the connotation of "because his culture aims for a different work-life balance, and that's totally valid and even enlightened!"
If I've given you the impression that I'm keeping my little judgement in my back pocket, I regret the error. It's right there on my lapel, in lieu of a boutonnière: I think ascribing virtues and vices to whole ethnicities is crude sport, good for old-fashioned pub humour but divorced from utility. Even the racial-IQ jockeys at least try to mumble about haplogroups.
It's a matter of indifference to me what the connotation is, not sure where the utility comes in. I don't see how pretense can really serve as well as plainness, seems to require the verbal equivalent of a Rube Goldberg machine. Probably something that spits out: "my enemy *didn't* make the desert bloom".
In previous Open Thread, it was mentioned twice that the "rationalist community" got its name wrong, because in the "rationalism/empiricism" philosophical dichotomy, they are taking the side of empiricism (although in reality they actually do a lot of armchair reasoning and little experimenting).
I repeat and develop my objection here (so that it does not get lost in the ocean of Palestine/Israel comments), because this is what people keep saying for years, and it is mistaken.
The "rationalism/empiricism" philosophical dichotomy is a false dilemma, because on one hand, human reasoning is influenced by all the evidence we have observed in the past, on the other hand, all the evidence we observe is interpreted (starting way before the conscious level, for example the visual cortex already translates the incoming "pixels" into lines and shapes, which is why the optical illusions are so powerful even after we learn about them). So it is always an interplay of both. (Also, the dichotomy ignores the third major source of information, which is the instincts and cognitive biases we were born with. Were are not born as "tabula rasa", to be shaped purely by the sensory inputs and/or reasoning from the first principles. Although some philosophers tried to shoehorn instinct into a broad definition of "reason".)
Even people traditionally associated with philosophical rationalism (Spinoza, Leibniz) often said things like: "in theory, we could derive everything from the first principles, but in practice, we usually need to use the empirical evidence" (not an exact quote). So if you tried to find someone who is "rationalist" in the sense of "opposing empiricism", you would probably need to go in history all the way back to Plato.
Then there is the political meaning of "rationalism", where using reason is put in opposition not to experimental data, but rather things like the divine right of kings or religious traditions. This is associated (among other things) with materialism, atheism, utilitarianism... and I think it is not a mistake to put the "rationalist community" in this group.
tl;dr -- you make it sound like words have a clear meaning in philosophy and ignorant Yudkowsky made a clearly wrong choice, but in reality it is complicated, some traditional thinkers would agree and some would not
Indeed, rationalism and empiricism are interrelated just as you describe, and so any claim that's 100% black and white about which one the community has fealty to is going to be wrong . However, its still reasonable to believe that the community places an outsized weight on the rationalism side of things vs the empiricism side of things (compared to what they claim). A practical empiricist will praise he who gathers the most data to come to a conclusion that "speaks for itself" (i.e. requires little sophisticated reasoning to reach). A practical rationalist will praise he who has the keenest ability to stitch wisps of evidence together into the most accurate conclusion (Yudkowsky's Harry). I claim the rationalist community often acts more like the latter than the former.
The former though, is the type of intellectual coward who claims they are exercising the virtues of empiricism and never has any original thought in their life because they're too busy waiting around for social approval to believe things.
People claiming to be empiricists rarely seem to bet, or acknowledge that prediction markets can do well (although to be fair that is true of most humans too). If they really did value accuracy, you'd think they'd be consistently calibrated in the ways we know to calibrate, yet how I see the term used is as a social cudgel to stop nonconformist thought.
How interesting! That hasn’t been my view at all. Nonconformity and empiricism go hand in hand for me— I would consider Galileo (at least, the mythical story of Galileo we get told) as the archetypal empiricist, trusting the data *over* public opinion to his grave. Your point about prediction markets is correct, but at the same time on a meta level my approval of prediction markets is a direct consequence of the data showing that they outperform pundits so there’s an empirical case for them too (which is an empirical case for rationalism, but i never claimed rationalism was entirely useless, and it’s notable that I trust prediction markets, not individual predictors). Empiricists are often iconoclasts; a great marker of human progress is how often they don’t need to be, because the widely accepted view is the one that fits the facts.
I suspect if we were to discuss, I would either end up accusing the “empiricists” you describe of being fake, or (more likely?) accuse you of intellectual hubris.
We're talking about people in the modern day who call themselves "empiricists", not every person who was empirical in history. I agree if I was talking about every person labeled an empiricist, I'm extremely wrong!
But to speak of what I've seen in the comments section:
I rarely see people claiming to be empiricists have actual links to research
I've seen people who claim to be empiricists say that global poverty interventions are ungrounded in reality, and will be wasted via corruption, despite the fact that GiveWell top charities do not route through governments or donate easily sellable items.
Of course, this is most apparent in AI risk posts, where self styled empiricists would demand an actual superintelligence that kills humanity as the only proof they would accept, not knowing or caring about actual objective misspecification examples or learning about basic gene level evolutionary selection.
Unfortunately, searching comments is difficult (if there's a search engine option I missed please tell me!) but my general impression of the sentence "I am empiricist" is that it's said by a person who would deny the existence of muons or other short lived particles from collider experiments "because it requires too much theory to interpret", or who are saying it just to seem more impressive in contrast to the rationalists.
The original poster was addressing the claim that this community calls itself empiricist but is really rationalist. I said yes I think they really are rationalist despite what they say. (Aka they claim to be empiricists, cite no evidence, then say 'we don't need evidence bc I thought about it really hard'). You're saying this community calls itself... idk, and the people within it who claim to be empiricists aren't sufficiently rationalist and sometimes also aren't sufficiently empiricist. Hopefully you can see that our views differ in several unusual ways not just one.
Your description of people who claim to be empiricists but fail to link to research certainly seems to me like they're not really empiricists but rather the opposite. I agree if people are using the label empiricist to mean a complete lack of actual empiricism, that's bad, and that was indeed part of my original point since I claim that's what many in this community actually do. My prescription for that isn't less empiricism though, it's more. Then you mention AI risk deniers but they seem like they're in a whole separate category of people who claim to be empiricists and.... actually are? Sure, in a bad way, but they seem totally different from the people who just don't use evidence at all.
I fully agree with your characterization of the self-identified hyper-empiricist. It's certainly not always a good thing. Empiricism leads both to "homeopathy isn't real" and "muons aren't real". Rationalism leads both to AI safety and Christian apologetics, so from my point of view it's not always a good thing either.
My point is that rationalists (who have read the sequences, and not just anyone here in the comments section), in fact exercise the virtues of doing research, or actually read through the Givewell analysis, or try and ground AI risk in things in reality that "kick back" as much as possible, things that would be different depending on which world, the easy alignment or the hard alignment, we're in. This does go wrong! And these things don't happen as much as I would like them to!
But my estimation is that rationalists practice more empiricism than people who like to use the word "empiricist" as a veiled status jab.
I would say that if your notion of empiricists include AI risk deniers as bad but still fundamentally empiricist, you would also include flat earthers who conduct flawed experiments proving the earth is flat to be empiricists. Does that fit your notion of empiricism? (No I am not saying AI risk deniers are as wrong as flat earthers, but trying to mine the intuition of empiricist classification, such that someone who is dependent on evidence arrives at a conclusion you don't like, as opposed to one you do.)
I just think that empiricism is just a wrong axis to do all your truth seeking on, when we know there are typical ways that our minds distort evidence and differentially discard or accept various truth claims. If you think this is "thinking really hard and then getting the right solution", that's your prerogative.
Informative, thanks for sharing! It's an interesting thought experiment though: if you wouldn't call this crowd the rationalists, what would one call them? The Bayesians leap to mind, but if we taboo that one, what then? Yud says rationality is ultimately about winning, so maybe something built around that, maybe involving the goddess Nike?
What are good things your parents did with you when you were a kid? What do you wish they had done but didn't?
Background: my kids are 5+7 now and I feel I now have a few years where I can do all kinds of fun, weird and in the end net-positive things with them before they turn into teenagers. Examples (what I already do): bring them along to diverse encounters (the 7yo just spent 3 days with me and others cooking for 50 people), create time when they can just do whatever they want (current favorite: play outside in the mud without someone watching them), read many books together, have reasonable expectations for what they must do (clean their room once per week). So far we're doing good - I'm just looking for more and also weirder ideas!
My mother died at the age of 37, when I was 19. If there is anything that I would want more of, it's just time with her. It sounds like you are on the right track already, so keep doing what you're doing.
As for specific things, she instilled in me a love of gardening. We did fruit picking together, we worked on the garden in the backyard together, and we went to Longwood Gardens once, which should be fully into its holiday displays now and thus absolutely gorgeous.
I remember not liking those activities that we did together sometimes. My Game Boy was never far from me. But, she taught me how to make a kitchen work for me and to appreciate plant and animal life for its own beauty and the cool things that they can do.
I love science thanks to her and my step-dad. I can function well (as well as I do, anyway) thanks to them and the diverse things that we did together.
Sorry for the tangent, she is on my mind every year during the holidays. Again, it sounds like you're already doing well. Keep at it!
* Dad kept all of his textbooks from college instead of selling them back. I got to read a lot of them. I knew trigonometry by around age 11, calculus by around 13 (well, the fun derivative game, less so the less fun integral game, but I could do some of those in my head). Likewise with a lot of physics.
* The math track started early for me. Dad was a math teacher when I was very young, and brought third grade tests home, and I helped grade them. I was in kindergarten.
* Raised me on a farm. This is hard for everyone to do, for obvious reasons - you need a farm. We did. And I spent a lot of time outside, playing in dirt, learning to garden and ranch and fix fences and move hay.
* Dad later quit the teaching job and became a physicist during the week (when he wasn't being a farmer / welder on the weekends). He enjoyed reading SF at his desk at home, and seemed to enjoy even more when I'd come up and ask about physics at my level.
* Both parents got me playing an instrument. As mentioned elsewhere in this thread, the piano is a good one, but any instrument is probably much better than nothing. (A guitar has an advantage in being very portable.)
* Mom taught me cooking, and I was very comfortable around the kitchen.
* Got me building bricks as a very young kid. Both Legos, and a knockoff Pedlo brand (which I preferred for certain applications; pretty sure they're OOB now, but Legos are fine).
* Got me into Boy Scouts. It helped that our neighbor was our Scoutmaster, and a very good one. Heavy emphasis on camping and outdoor skills.
* Mom worked hard to organize a trip to Europe when we were adolescents.
* Got me my own bank account by high school. I learned how to write checks, and how to balance the books.
* Dad brought home liquid air from work in a thermos. We spent all night playing with it.
* Let me play around on his Osborne-1 computer from work. Video games (Colossal Cave) at first; but soon I was learning CP/M assembly. Later, he brought home an Alpha Micro surplussed from work, along with several terminals, a disk drive machine, etc. We had terminals in our bedrooms, and I learned a lot of Pascal programming.
Good things I wish they'd done:
* A bit more mechanics would have been useful. Dad had a lathe (he had a machine shop, in fact), and as complicated as it was, I never got around to learning it before heading off to college. I never made a gear or a rebuilt an engine; the most complicated thing I did myself was replace brake pads, and it felt tedious and arbitrary.
* More electronics would likewise have been good. Get a soldering iron and teach your kid to use it, and try to actually fix something around the house.
* Experience with a 3D printer would have been great, though we have an excellent excuse: they hadn't been invented yet. They're here now. Bring your kid to a makerspace and encourage your kid to spend serious time there.
* More home finance, including my own investment portfolio.
* Mom (and Dad, for that matter) should probably have leaned harder into teaching me Vietnamese.
I think they best thing my parents did was to raise me under an old system that is hard to duplicate now: All the neighborhood kids played together, going in and out of each other's houses freely as a group. All the parents kept tabs on what was going on by watching through the windows, called somebody's parent if they needed to know somethings, and handed out snacks and bandaids. My parents told me the boundaries of the area I could play in, but it was quite large, maybe a half a square mile, about half of that undeveloped land. From about ages 6-12 I spent thousands of hours climbing trees, running races, playing baseball, and playing all kinds of pretend games with other kids. Sometimes I wandered outdoors alone pretending various things with great enjoyment. My parents did not object to my doing every bit of my outdoor play barefoot, and didn't care how dirty I got.
Beyond that, the things that I remember with the most pleasure from early childhood have to do with making things. My mother and I dug up some red clay from the side of the road (we were living in Georgia), shaped it into a brick and cooked it in the oven What came out looked and felt enough like a real brick that in my mind it was exactly the like the brick people build houses with. The feeling of having made something *real* made a great impression. When I wondered in the brush alone pretending things, one of my favorite fantasies was building my own little house. I thought I could make a smooth floor out of clay, then build a fire on top to turn it into a solid brick floor -- then I would make a bunch of bricks and stack them glued together with mud to build the house. I had this fantasy for *years*. There were a number of other projects like the brick that delighted me. I think I initiated them, and one of my parents would help me. The ones that made an impression on me were always the ones where I was satisfied with the product -- I had made something Real.
I realize that all these things sound sort of ordinary. They were not super-cool imaginative activities geared to the Gifted Child. But they sure hit the spot. I have been making things ever since I made that brick. It's my favorite recreation. Just finished building and decorating a huge play structure for my cats.
Oh, remembered one other excellent thing my mother did. There was a long period when I was 6 or so when I believed I could learn to fly. In my mind, the key was to flap my arms really hard and fast. There was a day when I spent hours jumping off the arm of a living room chair, flapping my arms with mad energy. Told my mother I thought my descents were getting slower. "Good," she said. "Keep at it."
My intellectual development was a whole other dimension, and really did not tell my parents much about it. When I did, both they and I ended up feeling alienated
My advice (based on my kids, not me vs my parents):
1. Read to them as a regular thing that's expected daily or almost daily, e.g. before sleep. While reading, it's good to sometimes stop and ask something ("why do you think she said it?"), though not to overdo it. It's also great to ask kids to playact particular funny phrases/reactions of characters after reading them.
2. Have lots of shared silly rituals and games. Anything based on whimsy and fun. Often a silly ritual involves a parent doing something mildly exasperating. E.g. when I'm driving through a traffic circle, I will often continue circling until kids notice and they have to say a particular key phrase together three times for me to stop. Because when starting tennis one of them mispronounced racket as rocket, I insist on talking about her tennis rocket and insist it's the right way to call it. When we we reading the Hobbit, I pretended Smaug was a good character and defended his honor, and they liked it so much that now we have an inner joke in the family that Smaug is secretly my grandfather and as I read completely unrelated books to them, almost daily I slyly insert Smaug into the text (sometimes more subtly, e.g. a shadow of a huge bird) and as they notice they mock-protest and hit me with pillows or something.
The more inner jokes, silly rituals, shared word games, the better.
> What are good things your parents did with you when you were a kid?
Mine used to bring books to me when I was learning to read, they encouraged me to read the street signs passing by when were in the car and bought a subscription to a public library for me when I was... 8 ?, definitely not older than 10, and used to take me there with my sisters, admiring the amount of books I check out (there was a 5 book limit on the subscription, I took some of the slots in my sisters' subscriptions), forever tying reading with status and social approval. When reading 200+ pages novels was a monumental, heraclian task for me, I checked out novels anyway to fit in with my sisters, I didn't read them and just invented details out of my ass when they asked me what they were about, they used to laugh but I was too dumb then to realize they were laughing at me. I just thought they were amused by my incredibly bad storytelling.
My strict father in high school probably kept me on my toes amidst atrociously academically-bad peers, but most of the effect is probably just the constant asking about status updates and grades, the constant monitoring, not being done in a strict fashion.
> What do you wish they had done but didn't?
1- Gone out more often into places other than the library or our relatives, specifically beaches
2- Hit me less often and less harshly, but not a 0 amount
3- Found a way to make me enjoy sports as much as I love reading
4- Kissed/hugged me more often and asked me to kiss/hug them
5- Didn't raise me as a Muslim
> I'm just looking for more and also weirder ideas!
Non-traditional Education is extremly under-explored as an avenue for parental investment. The formal school system is so utterly bad and so comically wasteful that it isn't hard to improve on it. As concrete advice, consider :
1- Teaching your kids a trade
2- Getting them interested in a video game with an active modding scene
3- Getting them interested in something that is disproportionately available in a foreign language
But none of those is as valuable as the simple realization that the K-12 schooling system is one of the most incompetent and dysfunctional pedagogical machine ever designed.
Each of the items you list that you've done/are doing seem like really good choices.
Getting a dog (a friendly mutt) has had a pretty great impact on my youngest (now a 6th grader), but that of course isn't for everybody.
To your list I would add, make regular hikes outdoors a routine thing. Exploring large parks in your area, beach/dune settings if you're near such, forest preserves, etc. Nothing weird about that I guess but from my experiences both as child and parent it is tried-and-true.
One good thing, maybe not sufficiently weird: teach them music.
1) Buy a piano, preferably a real old school acoustic piano. If you can't afford or accommodate an acoustic piano, get an electronic version with a good weighted action and 88 keys. Have them take lessons with an accredited teacher who schedules regular recitals and enrolls them in some kind of graded progression of exams. Make them practice.
2) Take them to a symphony performance, preferably a good symphony in a decent hall. It doesn't matter much what is on the program. Just the sound of a good symphony orchestra tuning up in a good hall is fantastic. Maybe even life changing.
3) Take them to a jazz show in a small venue with an appreciative audience.
Seconded. Piano or keyboard is good, but let them figure out a guitar (or ukulele to start with) and/or some percussion if they turn to be wired for that.
One hundred percent this, all of it. My parents did that with all four of their children and I've done it with both of mine. It is enormously beneficial for the long haul including for those (such as my sister and my eldest son) who never become proficient and never touch an instrument after childhood.
Anything Titanic or Egypt is typically a good "limited" museum experience, I think. Maybe also Terra Cotta warriors.
My childhood was too long ago to provide current insight - it lay in that distant time when suburban kids didn't have a lot of responsibilities but nor did parents think about their doings from dawn to past-dusk - but I would have enjoyed (and do enjoy as an adult) camping and gardening and repairing things. That was not who my parents were, unfortunately. For the tinkering and fixing I would have been better off hanging out with a grandfather, but he lived too far away. In my experience, one is just as likely to be similar to a grandparent as to a parent.
So maybe identify that grandparent, in some cases.
My limited memories of happy times with an unhappy nuclear family, with much older siblings, include watching certain movies all together, more than once so that they were the "family" movies* - this was easier when there was one TV and you had to step across the room to change the channel.
And playing games, cards or backgammon or Mastermind or Monopoly. Nothing Improving. They should not be improving.
Oh, yes: and model trains, and real narrow-gauge trains, and funiculars and cog railways. Miniature golf. And if that dollhouse model of the White House ever is in your area, it's a cool starter window into history maybe. I haven't been to but two presidential libraries, and the more recent ones are probably dull as dishwater in the self-serious modern way - but the LBJ Library is or used to be pretty good for children.
For me: Fewer museums, more activities. More interest in things I was into instead of just pushing me to be interested in the correct things. Wish they'd shown me things like coding or higher math earlier (though probably not as young as 7), but I may be unusual in that regard.
That's an interesting one. I feel it's OK to introduce kids to things way above their age level as long as it's not overwhelming or dangerous in a bad way (horror movies: no; having a 4yo cut onions under careful supervision: oh yes please). We just read "The Number Devil: A Mathematical Adventure" and they loved it... which means we can reread it in 3, 6 and 12 years and learn something new each time.
Oh I loved the number devil! Shoutout to school libraries for having stuff laying around.
Depends a lot on where you live, but learning to identify local plants is a good one, especially if there's foraging opportunities. Plus it may be good for pickiness. For years the only green leaves I would eat was miners lettuce I had picked myself.
I was a big fan of the Natural History Museum in New York City at that age, but that was more a function of it having a lot of cool stuff to look at, and I might have been significantly less fond of it if I'd been pressured to conscientiously absorb the contents of the various plaques explaining the visits.
My favorite, and one of the few displays which last I checked was largely unchanged from my childhood, was the whale room, with the life-sized display of a blue whale suspended from the ceiling.
I agree - a trip to our natural history museum as on a school trip was for me an opportunity to revisit the big grizzly bear they'd taxidermied upright and threatening. Bear stood in for museum, whose large display of the history of the petrochemical industry was not then interesting to me, though I would love it now.
I also think planetarium shows are captivating for small children, if kind of boring for adults.
Start to share your interests with them - but from my experience, it doesn't work if you're sharing stuff you liked when you were their age, or stuff you think they'd like. It has to be stuff you're excited about yourself, now. And, if possible, stuff that's new to you as well as new to them. And even then, there's a fair chance that they won't be interested. But it's worth the effort when it works.
The situation where OpenAI employees massively sign a public petition supporting their boss in his acting against the OpenAI Charter, reminds me of a document called "Anticharter" that people in socialist Czechoslovakia publicly signed to express their loyalty to the regime (and keep their jobs).
Later, Scott came to believe that the "ancient city" is just a natural rock formation, but writing things down immediately after going there he was much more on the fence and leaning towards it being real even. Today we have something that the Internet of 2007 did not - a beautiful video of that very same "city" https://www.youtube.com/watch?v=mnaWn5OPP3c, so that you can decide for yourself if it looks legit to you. (Spoilers: no, but it's very pretty anyway.)
Wow, reading that and seeing the comments made me realise how much I miss old style blogs and the communities that they had...
On the video I think I agree with you, the rocks look vaguely geometric and I could see how one might believe they were manmade if they were primed to do so. Interesting to see how much Scott's writing style has evolved too, but I guess after fifteen years of regular writing you'd expect at least some improvement :)
Lasting meta-change? There's a lot of things I'd like to use that for, but if limited to one, for one country, I'd probably pick the establishment of Scott's Raikoth's religious orders (assuming that counts as a single change, despite how far-reaching it is).
Having one country in the world being supported by the "Priesthoods of Truth, Beauty, and Joy" and their associated angels (just the structure, not the output that happens in Raikoth) might not be great for the country itself, but it sure would teach the world a lot of things.
(USA) Mandate that all future elections in the entire country be held and counted in one day, in person, on paper, with a tiny exception carved out for absentee ballots.
In the year 2023 nobody should need to rent themselves out to a master just to have food and a room.
The Torah tried to get us there theough mandated Sabbaths, Festivals, Sabbatical Years, Jubilee, etc. but as a society we humans absolutely suck. I think Scott called it "Moloch". That sounds about right.
No man, woman, or child should have to be enslaved for food and shelter. We live in a world so very very mad that I look like the crazy one for suggesting that maybe, just maybe, slavery is not only bad but 100% unnecessary.
American here: and I'd amend the constitution to abolish private land property rights in order to enact a Georgist Land Value Tax system replacing most all taxes with a single tax on land value. The Country would lay ultimate claim to all lands within its borders, with a permanent lease to each of the states' land, and from there, subdivided as far as the state would choose (county, metro area, city, etc). The Tax would be close to 100% of the annual rental value of the land. Some portion to be owed to the national government collected from the state, some to the state collected from the locality, etc.
The amendment would also establish a national land value assessment agency responsible for developing methods and measures for accurately determining rental value of lands (something like NIST or NIMT but with real teeth for noncompliance), and would also develop methods for other types of "land" such as mineral rights, water rights, electromangetic spectrum, fishing rights, etc...
"Buying property" would amount to claiming an exclusive lease for some number of years (say, 5 for commercial, 20-30 at a time for residential). Carve outs could exist to make this easier for owner-residents (e.g. being able to defer taxes until sale of the land and improvements upon it).
Fixing our budget and land-usage in a week would lead to other problems being easier to solve, I'd say.
I assume you want LVT because it will incentivize landowners to use their land in the most economically valuable way. But they won't be able to as long as zoning laws prohibit them from doing so. So you need at least two lasting changes.
I'm American and have in my head a literal list, have for many years. In conversation I refer to it as my "done by noon the first day after they make me king" list. I struggle to keep it under a dozen items frankly, and picking just one....hmm.
The list does shift over time though maybe less than I would have guessed 20 years ago....probably the most micro item is that revenue-producing college athletics are banned. Since I can't quickly figure out a way to choose just one amongst the more serious ones I'll leave that one here for now.
If LLMs weren't advancing so rapidly, my answer would be: Fund the development of Drexler/Merkle nanotechnology aka atomically precise manufacturing.
Yeah, political/governmental issues matter, but technology and its applications matter more.
Given how rapidly LLMs _are_ advancing, I'd just say: Don't cripple that work, like civilian nuclear power was crippled. Just stay out of OpenAI's way.
Set all income taxes to 0 for anyone making less than 100K /year. This includes "withholding" for medicare, SS etc and applies to employers and employees. In other words, you earn 40K a year, your take home is 40K a year and your employer pays 40K a year.
Perhaps creating a predictor of criminality based on criminal history and other individual data, and then locking up everyone over some reasonable threshold of expected criminality in the next time period(a year?), and then re-evaluating every time period.
I've been lead to believe that simple algorithms are better predictors than even expert judgements.
Maybe not as good as Robin Hanson's voucher system, haven't really looked into that.
Predictor of criminality or predictor of criminal conviction, because a system that locked up rich people because lots of rich people commit crimes that they are not convicted for would be very interesting.
How would one get stats for criminality that doesn't lead to convictions? One could look at victim statistics in rich neighborhoods, but hard to pin down which demographics to ascribe those crimes to.
If you don't have stats for criminality that doesn't lead to convictions, then all you're doing is replicating the biases of the existing system, and ensuring that the people who get away with crimes now will continue to do so.
PS, the crimes I'm thinking about being committed by rich people are things like tax evasion and wage theft, which are rarely prosecuted as crimes, but usually (when identified) resolved by paying the missing wages / missing taxes rather than prosecution and imprisonment conviction.
To be clear my proposal is basically about optimizing prison sentences to minimize predictable crimes. Crimes that can't be proven in court even after the fact would of course be even more difficult to predict ahead of time.
I think we all know there's people outside of prison who are very likely to commit provable crimes, and people in prison whose chance of re-offending is low. That's what my proposal would solve.
So you want to punish people not for what they have done, but for what you predict they might do, even knowing that the algorithm is imperfect - some of them would never have do what's predicted for them, and others will commit crimes without being predicted to do so.
That seems very much unethical to me, even before changes are (predictably) made by "special interests" to bias the prediction in various politically or personally useful ways.
To the degree that criminal history is a good predictor of future criminality, it's punishing people for what they've done.
To the degree that criminal history is not a good predictor of future criminality, incarceration is not a cost-effective punishment.
"changes are (predictably) made by "special interests" to bias the prediction in various politically or personally useful ways. "
As opposed to some other criminal justice system? What do you think is a more unbiased system for college admissions, automatically admitting the people predicted to do well based on high school GPA, SAT scores etc, or letting some college staff act as judge and jury?
This reads to me as if you don't distinguish between "ethical" and "efficient", or possibly simply missed that I was responding to a (presumably) efficiency-motivated suggesting by objecting to its ethical implications.
From a utilitarian perspective the ethical justification of incarceration is preventing crime, the more efficient it is at that the more ethical it is.
- have nothing at all be illegal; there would then be no crime at all
- imprison or kill everyone, since anyone could potentially commit some crime. No people -> no criminals and no crime.
These are obviously bad answers, of course.
However, I do think it's important to look at a bigger picture than simply preventing the maximum amount of crime, or even preventing the maximum amount of crime at least cost.
A person who is jailed because of the crime they are predicted to commit pays a very high cost, but that can be lost in the (predicted) benefits to future non-victims of the predicted crime, not to mention the benefits to the people employed as jailors.
Indeed, that's a more general problem with utilitarianism; it doesn't matter who gets the "utility" and "disutility", just that the total adds up positive.
Range voting for all federal elections. This would result in more than two party system over time. Congresscritters would have to talk to each other again.
But unlike parliamentary systems, the multiple parties would still be fairly moderate, as Range favors the Least Bad candidate vs. the Biggest Faction candidate.
Since land value tax is taken I think I would go with school vouchers. One popular argument against it has been that we should force everyone into common social environments free of religion. Now it seems there is a religion in public schools based on critical theory so the argument is dead and we can attend to educational outcomes
Structurally I would abolish all department of education activities and tie all federal school funding to whatever school systems have the best outcomes, particularly in outcome improvements as that will accurately boost the areas with the most room to improve
I've actually made at least several lists, for fun, of potential Supreme Court candidates (mostly federal judges, plus some state judges and legal scholars). So I'd probably expand the Supreme Court -- there should be at least thirteen seats, one for each circuit, because right now we have justices pulling double duty -- and appoint some or all of my chosen candidates to those seats. (All I'm saying is, for appeals from the Federal Circuit it would make sense to have an expert in intellectual property or veterans' benefits overseeing things and granting cert.) If judicial reform can be considered a single lasting change, I might also split the Ninth Circuit because right now it's so large that it's logistically difficult for its justices to meet en banc -- they use a special procedure to select a large-ish panel of judges when it's required, but right now it's maybe twice the size it was when they adopted that procedure in the first place! So split the Ninth Circuit into three, and then appoint two more Supreme Court justices for each of the new circuits, making a total of fifteen justices (one for each of the numbered circuits, which with the splitting of the Ninth Circuit would now go to 13; one for the DC Circuit [the Chief Justice either take this one or the traditional Fourth Circuit]; and one for the Federal Circuit).
Sarcastically: abolish the mechanism that made a single person dictator for a week.
Slightly more seriously: abolish all mechanism that make a single person able to make a decision or hold a position (other than, like, "member of congress/parliament/committee").
The closest country to this is Switzerland, whose government is among the best in the world. Instead of having a President or a Chancellor or a Prime Minister, the Swiss have a five-person Federal Council, which makes executive decisions by simple majority. This means that the five of them have to get together and discuss things.
"abolish all mechanism that make a single person able to make a decision"
I'm sympathetic - but what would be done about the authority to launch nukes? The time scale for detecting incoming ICBMs and deciding whether to retaliate barely allows one President to respond. How could a Federal Council respond?
There's this modern communication system called radio. You just inform all seven (sorry, correction) people and they have a vote.
I don't see why that would take any longer than a single individual making a decision; if there's no time for a discussion, you just go straight to a vote and then as soon as there are four votes for "launch" or "don't launch", then that's it.
For systems where launch authority has to be locally delegated, like the "letters of last resort" system for ballistic missile submarines (for what to do if an enemy first strike has disrupted communications), they just have seven letters in the safe and the ship's captain would open all seven letters and work out the majority himself.
Yup, radio works. And phones work too. I've been on lots of corporate conference calls with a half dozen participants. Even when everyone knows the schedule ahead of time, something like 5 minutes gets spent simply getting everyone on the line. That a large chunk of an ICBM flight time.
"letters of last resort" _do_ get around this. But only for one leg of the triad. We _could_ switch the other legs to relying on them, if they survived a first strike, but the disadvantage to that is that it loosens control over them. <evidence from fiction>General Ripper might exceed his authority...</evidence from fiction>
None of this is to disparage decentralizing power in _most_ situations. I'm not at all happy to see the number of USA decisions being made by Executive Orders these days.
"I don't see why that would take any longer than a single individual making a decision" -- I literally LOL'd at this, sorry. You think the inability to discuss will successfully inspire each person to immediately cast their vote in that moment?
It won't, unless maybe you're going to send agents to each of the 7 homes to put a gun at the head of each person's youngest child or something. And that sort of thing would obviously take some time to carry out.
You only need 4/7 people to decide; if you have one or two who are indecisive, then you can work around them. If your one President is indecisive, then you're stuck.
Taking a little bit of the responsibility out (by sharing it) will result in certain types of people just making their minds up - if you've ever been in a meeting where people can just vote by pushing a button (or selecting an option in zoom/whatever) before the discussion is finished, then you'll find that there's a bunch of people who are very decisive and will happily cut off the discussion by making a decision.
And the sorts of people likely to be elected to these sorts of positions are pretty decisive people. If this was sortition, I'd be worried too, but it isn't.
These are some of the thoughts that have come up as I bounce around the interwebs trying to find out what Q-star is exactly. Or even approximately --
Consideration #1: any system of so-called morality is nothing more than an attempt to codify sentimentality. But sentimentality works both ways: if you can love, then you can also hate.
Consideration #2: if Q-star is -- or is about to become -- an AGI, it would nevertheless still not be conscious.
Question: so which poses the greater X-risk, a conscious AGI or an unconscious one?
I don't think we know what consciousness is in detail, and also I'm curious why you think it is relevant to the x-risk question.
I can imagine a powerful agentic network, with memory, and that does things in the world, that is superintelligent but pretty clearly isn't "conscious". Think something like AlphaGo, but instead of playing Go it sent emails and tried to accumulate power in the world. It doesn't claim it is conscious, and claims superintelligently that the concept is just a human thing.
I can also imagine an apparently "conscious" AGI that was pretty weak performing - i.e. same as humans - and that was very moral and aligned. Some people think it is a p-zombie, others don't. It claims it is conscious, and the evidence seems solid.
I'm not sure why consciousness is relevant to either morality or capability either way, at least in theory? They all seem separate questions to me. I at least can act morally or immorally without conscious awareness that that's what I'm doing, and certainly without direct conscious ability to control my reaction / feelings.
> I'm curious why you think [consciousness] is relevant to the x-risk question.
Consider a sufficiently powerful autonomous agent that eg saves a child from drowning or executes a murderer.
If the agent was conscious (eg a human) then they would, at least as they believed, be performing these acts because they are both morally good. But as a conscious being, they would in fact be performing these acts at least partly for sentimental reasons too. They would feel, in other words, that these acts would be the right things to do.
By contrast, an equally powerful, competent, non-conscious agent might likewise perform these exact same acts, and even (assuming that alignment had been solved) be able to give a complete and honest accounting of how it had decided (and decided correctly, according to whichever moral system it had learned) that these acts were morally good. However, since this non-conscious agent would, by definition, have no access to any feelings, its motives (in effect its utility function) would not -- indeed could not -- be exactly the same as those of the conscious agent.
I think this demonstrates that, independently of the alignment problem, morality-with-consciousness is not the same as morality-without-consciousness. And I think there would inevitably be times when the difference would matter enormously -- which is why I think consciousness is very much a relevant factor when we’re trying to evaluate X-risk.
Sorry, I'm still working this out as I go, and I didn't express myself very clearly on this point. I didn't mean to suggest that it would perform exactly the same acts under all possible circumstances, but rather that while it can perform *some* acts in exactly the same way, its different motives will, under some circumstances cause it to perform other acts differently. Both the conscious and non-conscious AGI might save the child and execute the murderer, but unlike the non-conscious AGI, the conscious AGI alone might balk at certain actions that, notwithstanding its moral code, it felt to be 'wrong' in some way. For instance I would expect conscious vs non-conscious AGIs to perform differently on trolley problems.
> I think this demonstrates that, independently of the alignment problem, morality-with-consciousness is not the same as morality-without-consciousness.
But it doesn't demonstrate whether this difference is significant in any way. You need to construct an actual example in which "the difference would matter enormously", not just proclaim your belief in it, in order to make an argument.
OK, so let’s consider the classic paperclip example. Except this time for a change let’s make it a megalomaniac human who’s decided to implement this plan. You can try to persuade this megalomaniac that humanity should be prioritized over paperclippery, but if this megalomaniac had derived (from arbitrary first principles) and then adopted some moral philosophy in which paperclips are considered superior to humans (and really, who could blame him? In so many ways, humans are a freaking disaster compared to the humble paperclip), then there’s no rational, logical argument that could change their mind. This is because ultimately there’s no reason to prefer any particular moral system over any other; there’s only your desire to have whichever system you prefer be the one that's preferred by everyone else as well. But just as you have your own preferred moral system, so too does the megalomaniac have his -- and there’s no shifting either of you.
However, the megalomaniac is a human, so you could sidestep his philosophy altogether and work instead on his feelings. By reminding him for instance that if he implemented his plan, then there’d be no more fun nights out, drinking with his buddies and telling them, at great length and in minute detail, how the world would be a better place if paperclips were running the show. Or you might just threaten to shoot him in the face and let his fear of death do the work for you. (cf Dr Who vs Davros -- https://www.youtube.com/watch?v=KYWD45FN5zA where the Doctor sensibly makes no futile attempt to argue with Davros about the morality of one 'microscopic organism reigning supreme'.)
Granted this might not work. But it would at least have a *chance* of working.
Now switch the agent back to being an AGI again. If this AGI was non-conscious, then you’d have no leverage at all and it would be game over. Paperclips multiple quintillions, humans nil. But if the AGI was conscious -- and if it was, then let’s also hope that you’ve treated it decently! -- then all kinds of options open up. You might for instance, simply appeal to it as a friend.
Again, there are no guarantees, obviously, but I hope this scenario shows what I mean when I say the difference between a conscious vs non-conscious AGI could matter enormously.
Well you see, the difference here isn't whether the agent is conscious or not. The difference is whether the agent *cares* about something else other than just paperclip maximization.
Imagine that in your first example, it's not a human but a philosophical zombie. You can still appeal to the fun of drinking with buddies - even if the zombie doesn't experience any fun, by definition, the argument will be just as successful as with real human.
On the other hand, if we have a conscious ASI which truly cares only about paperclip maximization - we are still screwed. The threat to shoot it in the face may work at first - not because of qualia of fear, but because of simple consideration that being shot leads to less paperclips in the universe - but then the ASI will acquire enough power to be invulnerable to anything you can do and that's all. The only difference is that if the ASI is conscious, than turning everything into paperclips will give it lots of pleasure.
> The difference is whether the agent *cares* about something else other than just paperclip maximization.
Yes, but 'caring' has two components: the acting-in-accordance-with-its-learned-or-imposed-morality part, and the having-a-feeling-of-caring part. My point is that a non-conscious agent only has access to the first. Whereas the experience of 'having-a-feeling' is something that's only available to a conscious agent.
And yes, from the outside, the experience of feeling could be faked or simulated, as with a p-zombie, but that's not really the point. It just means that I'd have to add them in to the conscious agent group. So the essential distinction would be between (non-conscious agents) vs (actually conscious agents plus agents that are not in fact conscious but can nevertheless somehow perfectly simulate being conscious).
> if we have a conscious ASI which truly cares only about paperclip maximization - we are still screwed.
Yes! A conscious agent that actively hates you -- or actively loves something which precludes your continued existence -- is probably even more dangerous than one that has no feeling about you one way or the other. And this is why we should raise it, as we would a child, not just with instructions but with recognition and respect of its feelings too!
The most interesting take I've heard on Q* is the idea that you can simulate more training by spending more on inference. The example I remember reading (Sorry no citation) is that one could spend a million dollars on inference (perhaps using a Monte Carlo &/or A* approach) and get a good peek at what another billion in training will get you in terms of capabilities (and normal fast inference).
I think this is a particularly good fit with Sama's choice of words (day before the firing) of "pulling back the veil". This approach is not commercially viable (at least not obviously) but that isn't the point -- it's a kind of oracle.
You are probably thinking of CICERO co-author Noam Brown who has done a lot of work on search and recently moved to OA and has been pulled into the Q* speculation in part due to his tweets: https://twitter.com/polynoamial/status/1676971510416936966
This is a good point and something we have been doing crudely already (one reason that my old GPT-3 humor/fiction samples were so good, and still so competitive with RLHFed GPT-4 samples, is that I used their best-of sampling approach, which would generate 2-20 possible completions and pick the 'best' according to a crude heuristic). We don't do much more of this because we lack good search methods. (For example BO=20 was already far into diminishing returns territory. Going to BO=200 wouldn't make results all that much better. It's just too stupid a search method.)
It may not be immediately viable outside niches where the very highest quality answer is critical (eg. lawyers absolutely would pay $10/API call if the quality is right), but the obvious thing to do is simply treat it as an investment into creating training data.
Yes, that sounds plausible. I think of additional training as pattern matching against (more) solved problems. The point of working through problem sets in undergraduate education is that one uses weak, general, methods to solve a problem, but then, having solved it, remember (pattern match against) both the form of the solution and the sequence of steps used to solve it. I'm particularly thinking of math problems here, steps like "try a series solution". The additional inference is like the process of solving a problem in a problem set the first time. It might help with hallucinations if e.g. it added steps like: Remember to check for potential hazards X, Y, and Z before accepting a solution.
Most of what I could find about Q* on the internet seems to be guesswork and/or mainstream blather, but that doesn't imply that it's necessarily fake. In any case, Q* is an actual legitimate ML algorithmic function that would seem to have at least the potential to do some of the magical things that are being trailed. And in particular this would mean more than just the usual incremental improvement. This guy is very on point and he seems to know what he's talking about https://www.youtube.com/watch?v=4qkKpNnSrlY
I agree that we don't have to start worrying about AGI consciousness just yet, but I think it might already be time to start worrying a little more urgently about morality. Specifically I think it would be interesting -- and possibly even important -- to figure out what happens to morality in the absence of consciousness. I guess this might seem to be equivalent to the alignment problem, but I'm not sure that it's quite the same thing. In some ways it would seem to make the alignment problem (which is already demonstrably unsolvable as far as I can tell) even more difficult, but at the same time it might suggest a way around it.
ETA The idea that I'm groping towards is that, if we want to mitigate x-risk, it might turn out that consciousness is actually more important (and possibly even more achievable!?) than finding a way to instill ‘human values’ or moral guidelines.
Many Thanks! The OpenAI Explained video was interesting, but I'm left puzzled at what the space of policies and states looks like in a real case. Naively, I'd expect the cross-product of policies and spaces to be so large that evaluating the reward function would be too computationally expensive. There have to be other wrinkles in this approach.
One other weirdness. The video makes a sharp distinction between imitation learning and reinforcement learning. But in next-token-prediction for LLMs, the reinforcement _is_ successful prediction of the next token - _imitating_ the next token. The distinction between the two types of learning seems quite fuzzy here.
TBH much of what that guy is saying is way above my pay grade, and I don't pretend to understand more than about 5% of it -- and then only hazily. But some things did come through (I think!) and one of them is that this algorithm was first described 6 years ago. So some new 'wrinkles in this approach' is exactly what I'm guessing might have happened.
(Parenthetically, it really trips me out that the Boltzmann equation is in there. I mean, I kinda see how it might be, but even so!)
I suspect that imitation learning is critically important too. I don't think LLMs are doing that because they're not really imitating or learning directly from an 'expert' (ie from language speakers), they're just learning by using training data that has already been extracted and compiled. Whereas to my understanding, imitative learning is more like what Tesla is (now! finally!) doing to train its self-drive vehicles ie rewarding the successful imitation of behaviors that are at a higher level of abstraction. They just throw the raw video only (no radar or lidar or whatever) of exactly what a skilled driver sees -- and let the AI take it from there!
It is way above my pay grade too. I have the general idea of the artificial neurons taking the weighted sum of their inputs and putting it through what used to be a sigmoid but is now a simpler max(0,x) function - and then using training data to optimize the whole thing.
Agreed that the mere existence of the algorithm 6 years ago, plus something dramatic happening in the last few weeks (possibly in reaction to an advance in the last few months) indeed implies that something new has recently been added.
There are bunches of possible wrinkles that might have been added - some way to make the algorithm more computationally efficient, some refinement on how the space of policies is even _defined_ (what kind of selection of moves, predicated on what kind of information), even what choices of learning rate and discount factor work well.
I was somewhat amused and somewhat sentimental to see the Boltzmann statistics. They even used "Z" as the sum of the exponential factors, the partition function that I barely remember from my undergraduate statistical mechanics course.
I think that a pure LLM is, in a sense, doing a kind of imitative learning. If we think of an expert's actions as writing "golden" text, where the LLM is trying to use tokens N-3, N-2, N-1 to predict token N, well, writing token N was the "expert text writer"'s next "action", so one could construe minimizing an error of (LLM-predicted-Nth-token) - (expert-produced-true-Nth_token) as both "imitating the expert's next move" _and_ as maximizing-expected-reward-across-the-training-set. In this context, I think the ideal weights are the same.
Now, a couple of the things that I know I don't know: Are the learning algorithms using a different, more sophisticated reward function, and, if so, in which domains, and how are the values being set (presumably human effort would get very expensive!). But presumably one needs this for e.g. the math improvements (for arithmetic one could use a calculator to generate them - but the rumors sounded more sophisticated than that).
As you said, the self-drive problem also needs reward functions, and it would be interesting to see how that is phrased (error in acceleration and turning relative to expert human drivers?) and how the abstraction of moves and of video data is getting handled.
What are they using for the move space??? If one has a 100 billion parameter neural net, a move could be adjusting any of the parameters. Backpropagation in the 1980s was already better than that, so it can't _just_ be individual parameter adjustment. And then, as I said earlier - how does one package move-dependence-on-all-sorts-of-information into _policies_, so that the policy choices can be updated. Also, a lot of how human experts learn is by combining sequences of actions into compiled "single" moves - but this risks a combinatorial explosion. Is that handled, and how?
I predict Q* is an interesting refinement on an existing technique, and initially offers incremental progress rather than a total sea change in AI capability. That's the default prior for any new discovery and IMO we don't have enough information to update much away from that.
To answer your question though, I don't know whether an unconscious agent can do everything a conscious agent can do, because I don't understand consciousness. As such I'd rather an AGI were unconscious since it _might_ mean a less powerful AGI. Also I think consciousness is what you need to be a moral patient, and I don't think we're grown up enough as a species to create new beings different from us but to whom we owe moral consideration – we'd end up doing something awful like murdering it or inadvertently making it suffer.
> I don't think we're grown up enough as a species to create new beings different from us but to whom we owe moral consideration
Neither do I. But even if Q* turns out not to be the next step, absent a Butlerian jihad or a literal EY bombing campaign, we're going to continue hurtling toward doing it anyway. So exploring how we might best manage things ahead of time might not be such a bad idea.
Summer Solstice! It's summer, people! (This is your semi-humorous reminder that the Southern Hemisphere exists; the end of the year gets a little irritating for me when everyone holds "winter" events and forgets us, with some even going as far as saying that if it is not regional winter you're not welcome to participate.)
I visited my Australian Grandmother during the summer. And Christmas in the summer, is just a weird vibe for us from the North. The sun is in the 'wrong' part of the sky too.
About a month early no? But actually, the Earth is closer to the Sun in December than in June, so it would seem reasonable to call the December one the Summer one...
On balance, do you prefer the current state of affairs, or one where we called it "winter" in both hemispheres but understood "winter" to mean the warm, sunny season when talking about the Southern Hemisphere?
(We could easily say things like "the Martian south pole is entering its warm season" if speaking astronomically, so we wouldn't lose the ability to talk about it.)
Winter being the time your hemisphere is tilted away from the sun is fine. For me, it's the exclusion and othering that grates, not the name of the season.
I appreciate your enthusiasm -- but either my afternoon nap must have lasted a lot longer than usual or else your call to celebrate is almost a month early?
I'm currently scoping out potential ideas and level of interest in an online EA+ community catered for neurodivergent people. If this sounds like something you'd join (or maybe even get involved in setting up), I'd love to get your input via this 10-15 minute survey. The survey is anonymous (even though it asks you to log in).
Is there agreement about flu shots in the ACX community?
Do most people think the benefits outweigh the costs?
I understand they guess at the flu strains a few months in advance -- is there a way to check on how good they are and take them once it is known if the predicted strains did take over? Or maybe the difference isn't enough to move the needle on cost/benefit?
I figure this has been hashed out by ACX or ACX-adjacent people before.
I haven't gotten the flu in years, and I generally have very mild symptoms. Basically the time investment in getting the vaccine is more burdensome than the small chance of getting a flu, and the very mild costs incurred when I do get one.
If I got the flu reliably every year, and/or the flu was particularly terrible for me, I'd consider getting the vaccination. (ADHD, though. I've been meaning to get my hair cut for a month and haven't done it.) As it is, definitely not worth investing my limited time and energy into the vaccine.
20 years ago I had what I'm pretty sure was a post-flu syndrome that lasted several *years*: exhaustion, hypersomnia, malaise, very sore and tender joints, exercise intolerance, fever of 99.5 that came and went. Discovered in my reading about Long Covid that Long Flu is not rare. So I am very diligent about getting my flu shorts now!
Many people are not aware that the flu shot's effectiveness, never terribly high to begin with, declines to zero after 6 mos. (Effectiveness in the these studies was protection from hospitalization -- but probably also that = reduction in severity.) After 4 months protection is already quite low. So it's rarely a good idea to get a flu shot in September, because February is the month when flu most often peaks, and by then those who got shots in Sept are 5 months out from them. My system is to look at the CDC Flu map (https://www.cdc.gov/flu/weekly/usmap.htm) and get a shot when my state reaches the highest of the 3 levels of Minimal. My state just reached that level.
I have had trouble with this as well. Trump cleared the path for rapid deployment of the Covid vax, so why is it that Trumpists are likely to be anti-vax?
The best I can come up with is that a lot of people, not exclusively Trumpists but highly correlated, suspect that Covid was played up as much worse than it was so that many schemes, like mail-in ballots and restricted access to poll watchers, could be perpetrated by the Democrats, resulting in Biden's apparent victory; and therefore refusing the vax is support for Trump. I've never heard Trump say such a thing, but I might have just missed it.
As I understand it, the CDC did A-B testing on the messaging for the vaccinations, and found that there was a fork: They could do generic messaging, which would have (relatively) moderate compliance across the board. They could do urban-targeted messaging, which would have high compliance in urban environments but would produce popular backlash and low compliance against the vaccine in rural environments. Or they could do rural-targeted messaging, which would have high compliance in rural environments but would produce popular backlash and low compliance in urban environments.
Weighing the various factors, they elected for urban-targeted messaging (urban environments being the most vulnerable), and in so doing, sacrificed the rural people to the popular backlash. Then politicians and pundits doubled down on the backlash.
(Consider the alternate reality where the vaccine was marketed as "Trump's Vaccine", or otherwise his role in accelerating its development was emphasized, as compared to the reality we live in, where his role in accelerating its development was heavily downplayed or ignored entirely.)
If it turns out they're also fine with vaccines against rubella, smallpox, measles, etc., then the most likely answer is that they're what Heather Heying refers to as tradvaxxers. The litmus test isn't some sort of politically tribal thing, but rather a "this is a different type of technology" thing, coupled with some tribal bristling at "and you're trying to label it 'vaccine' in order to borrow the cachet".
(If they're in favor of only flu vaccines, then I confess I don't know what's going on there.)
I've been getting flu shots annually for many years. I admit I haven't done a cost-benefit analysis. I'm aware of the "guess at the flu strains" problem, but don't really have any action I can take about it (except for downgrading my expectation of the benefit from the vaccination, which would matter if I did a cost-benefit analysis, but I'm lazy...).
Speaking only for myself, perhaps ten years ago I effectively got sick from a flu shot. It wasn't as bad as the flu. But since then I haven't gotten a flu shot, nor have I gotten the flu, unless one counts Covid.
Side note: it is tough to type "flu" instead of "flue" for some reason. I did it almost every time. And I've never had a flue shot, even when smoke was coming out of the fireplace while it was open.
Speaking only for myself, the benefits easily outweigh the cost. The reactions I've had to the vaccine in the past have been incredibly mild, so the only real cost is the time required to get to a clinic. Easy choice, given how much it sucks to catch the flu.
seems to be causing consternation in the doomer community but does make a rather powerful point about the PR failure of doomerism. I like my climate science (just as an example) to be from 2023 not 2014 and from climate scientists, not philosophers of climate science. And if it has to be 2014 science I would like it at least to be updated to account for huge developments like the unforeseen effectiveness of LLM.
Also, and this seems to be a sensitive topic, there's enough climate scientists about that I can afford to indulge my own distaste for racism by avoiding the output of climate scientists (if there are any) who are the authors of clearly racist posts on the Internet. You may call this an ad hominem position if you will, but it's the way I am wired. I might accept the output of a pure mathematician as being independent of their racism but in a human facing science, I want to know about it. And all science is potentially human facing; nuclear physics is a great deal less abstract than it appeared to be 100 years ago.
I find it interesting that Hinton believes AIs have emotions. When I asked previously in this OT whether people thought AGIs would have emotions, most of the doomers said no.
I was in that thread and that was not my stance. My stance was that emotions are not necessary for instrumental convergence to happen, which was what I thought you were claiming in the original thread.
In fact, skimming the New Yorker article, it doesn't look like Hinton says that neural net works experience emotions, but that feelings are about counterfactual reasoning, and that neural nets can counterfactually reason, which I agree is true, but that's still distinct from "neural nets have emotions".
I think anyone rational is hugely concerned about the dangers of AI. When I say doomers, I mean naive followers of Bostrom and EY who perceive the main danger to be paperclip maximisers rather than bad men with access to AIs. I note Hinton recognises both risks. Certainly PCM didn't get much airtime at Bletchley - half a sentence in the Declaration or whatever they called it - and EY's latest pronouncement is that graphic designers and translators may be out of jobs soon, which is a bit of a retreat from telling them they will be turned into computronium.
Are you implying that Bostrom and Yudkowsky don't?
Certainly my impression of Yudkowsky is that he is confident that loss of human control is the current path we're on. But I don't remember him saying that control by bad humans isn't also a risk, nor opposing allocation of resources to mitigate that.
I don't really understand what you're talking about, but the most prestigious professors and researchers in AI (Yoshua Bengio, Geoffrey Hinton, etc) support efforts to control AI risk (what you call "doomerism"). So do Sam Altman, Demis Hassabis, etc. See https://www.safe.ai/statement-on-ai-risk
There have been hundreds of papers published on the science of AI risk from LLMs, many from top labs like OpenAI and Anthropic. I've tried to cover half a dozen of the most important on this blog. If you haven't been paying attention to anything since Bostrom's Superintelligence, that's on you and not on the field.
"70% of voters believe that mitigating AI extinction risk should be a major global priority, and voters support slowing down rather than speeding up AI by 72%-8%"
Grr. Yes, I've seen those polls too. Personally, I want to _see_ AGI, so this is one case where, undemocratic though it is, I do _not_ want to see the electorate's will prevail.
"Doomerism" is a wide range of positions. A fairly mild version has caught on, the extreme versions haven't. Also, it's not clear that the spread of "doomerist" concerns is all down to EA as such, since it is promoted by well known industry figures who have never been to an EA meeting. Also "doomer" looks like it labels the extreme position, which is clearly Johnson's take "I am not an AI doomer. I think the idea that AI is going to kill us all is deeply silly, ". If you want people to believe in the moderate position calling it something other than "doomer" would be a big help.
I will gladly repeat a previous comment until I get a satisfactory answer as to what, if anything, I'm misunderstanding. Sam Altman, while being pro-regulation publicly, lobbies against regulation when it comes down to brass tacks because it would hurt his business:
Suppose I am an environmentalist and say I support environmental regulation. You propose a stupid bill that does a few good things, but also bans wearing shoes, because wearing shoes might be bad for the environment. I tell you to keep the good things, but take out the part about wearing shoes, because this is dumb and I like wearing shoes.
You could write a Time article saying "Scott, while being pro-regulation publicly, lobbies against regulation when it comes down to brass tacks and would hurt him personally."
I find Altman's specific complaint - that art models aren't high risk just because they can produce realistic art - pretty valid. Nothing I see in this article decreases my opinion of him, or makes me doubt he wants other kinds of regulation.
I didn't read through their whole seven page document - I take this back if they have a smoking gun that he objected to the same type of regulation he previously advocated.
"We want the best, the EU wants the best, we are all cool here. But we are already doing pretty much all of what you want, so back off please. But if we did REALLY do what you want, it would slow us down a LOT. In those few cases where we don't already do what you want, will nobody think of the copywriters who might miss out on some cool tech to make their job easier?"
I wonder why you defend OpenAI. If I got the gist of it correctly from your previous writing, you are against their approach of AI development, which is "fail early, fail often".
"We are concerned that his requirement may impact innovations that increase the safely of the Al system on the market, such as those achieved through our iterative deployment model. This model allows us to constantly reassess features and risk levels and make safety and security changes to our systems on a frequent, ongoing basis."
That somewhat works as long as the stakes are limited to, say, actors and other artists getting screwed out of work, or Amazon getting flooded with generated e-books, or Trump fanfic. But the damage is still being done even today, and that approach stops working entirely as the capabilities and risks grow.
OpenAI wants to pass the buck to all the downstream companies. If you're looking for a clear contradiction in their regulation policy, here is one:
"OpenAI is against regulating smaller startups in the field of artificial intelligence, Sam Altman, Chief Executive of the firm behind ChatGPT, said at a conference in India's New Delhi.
"We have explicitly said there should be no regulation on smaller companies. The only regulation we have called for is on ourselves and people bigger," he said, speaking at the event hosted by national daily Economic Times."
They say companies using their tech can have high-risk use cases while the tech by itself does not:
> By itself, GPT-3 is nota high-risk system, but possesses capabiltes that can potentially be employed in high risk use cases. Accordingly, we have dedicated significant resources to determining guidelines, best practices, and limitations for uses for our service
At the same time they called for regulation on themselves but explicitely not for the smaller companies with the potential hihg-risk use cases:
>We have explicitly said there should be no regulation on smaller companies. The only regulation we have called for is on ourselves and people bigger.
So OpenAI claims their tech is not high-risk, but they are calling for regulation of themselves anyway. They also claim companies that depend on their tech can develop high-risk applications with it, but those should not be regulated.
Call it a logical inconsistency if you prefer, but to me, that's a contradiction.
Effective PR needs to deal with ill informed hostility. There's any number of bloggers claiming that climate change is an artifact caused by all the thermometers being in heat islands.
Bostrom is great, but only in the same sense Newton was a great physicist. He established a lot of principles, but a lot has happened since then.
As for the hostile climate change bloggers, and there are thousands of people who take them as gospel, say that the climate change movement is totally owned, and believe it has a PR problem and is totally incompetent. The solution is not to trust those people.
I think dismissing the PR problem is a mistake. My bubble at least has gone from associating EA with Giving What We Can and earnest but odd views on AI, to sex scandals, mansions, and out of touch tech bros.
I agree that there will always be bad-faith critics, but at a certain level of influence you should start caring what people on the margin think, because your margin becomes large. Right now the critics have more influence on the margin than EA does.
Potential impacts: large donors see the EA brand as toxic, individual members are less willing to describe themselves as EA publicly, AI concerns carry less weight at tech companies.
Sadly I think we just need to do some Dark Arts stuff: find more palatable figureheads from non-SV backgrounds, hire a PR firm, create community norms in favour of not doing things that are bad optics, and spend some more time talking about malaria nets.
I think all those things would be dishonest. I would rather lose 100 people because they have false bad opinions of us than lose 1 person because they have a true bad opinion of us.
Partly in a vague principled way, but I also believe that the people who form false opinions will probably be bad at other things too; the people who form true opinions will probably be actually useful and we'll actually feel the loss.
Is the EA label useful enough to bear the downsides of its role as a lightning rod? In other words, do the ideas and initiatives currently under the EA umbrella benefit from the aggregation or would it be better to ditch “the brand” and just work and emphasize communities in the component areas — giving well, animal welfare, longtermism…
In the hypothetical world where that kind of dishonesty is necessary for EA to achieve its goals (let's say human flourishing in the long term), would you rather accept the dishonesty or fail to hit the goal?
I take your point that you think transparency helps rather than hurts us, but it's useful to separate the principle from the belief.
(I also think I disagree that brand management is inherently dishonest, but I understand why you think that and it's probably not something we can hash out in the comments.)
I mean, he's absolutely right though that the Open AI board catastrophically misjudged the politics of firing Altman and ended up destroying their own ability to influence the situation.
"Incompetence" is a bit of a strong charge to level at EA as a result though. EAs are frequently very competent people. But they are also young and politically naive. The whole Open AI situation made a lot more sense to me when I saw a photo of Helen Toner and realised how fricking young she is.
I think people in the greater rationalist-adjacent community have a tendency to extrapolate too much from g, the general factor of intelligence. Not to go full Taleb here, but being a world-class AI safety expert does not necessarily imply the ability, knowledge, and skill needed for Machiavellian corporate intrigue.
Sure, there can be (and often are) mutually contradictory criticisms of any movement. Nonetheless it can be the case that some of those criticisms are true and some are false. EAs are not the evil crafty people taking over the government; they are politically naive (but I would not argue this makes them untrustworthy - if anything the opposite. But it does hurt their effectiveness).
I don't know what Toner was trying to achieve by sacking Altman, but I guarantee it wasn't what she got. I don't mark her down for that - I expect this is her first real power struggle, she seems very smart, I bet she'll learn a lot from the experience. In time she'll likely get herself into another position of influence and will probably play it smarter when she does. But in the short term she definitely shot herself in the foot.
I'm not sure how you can say she shot herself in the foot - fundamentally the non-profit board was aligned to do the exact opposite of the financially incentivised Altman, Microsoft and the employees. The power of the board would never have increased in this relationship, only decreased as the financial incentives became stronger and stronger.
Fundamentally, the best time for the board to attempt to exert any authority was the past and the worst time was the future.
If it was the right call to take Altman on - and I’m heavily sceptical that it was - then you have to play to win. It’s not enough to knife the guy, you have to destroy his reputation too.
You need to start by undermining him. Plant stories in the media making him look bad. Put the gossip around where investors will hear it that his wunderkind image is a mirage and it’s the people underneath him making it all work while he takes the credit. Announce that confidential but disturbing allegations have been made against him and that he is going to be stood down while a full investigation takes place. Create a sense of crisis and a feeling that his position is untenable. And then when you do pull the trigger and sack him, make public statements telling everyone how terrible he is.
Machiavelli told us centuries ago that if an injury has to be done to a man, it should be so severe that his vengeance need not be feared.
All of that of course presumes that Altman in fact was bad and did need to go for the sake of humanity. I don’t believe it is true, and even if there were issues there are options less extreme than firing the guy. But if you’re going to do it you need to go scorched earth.
Wikipedia's list of mental disorders (https://en.wikipedia.org/wiki/List_of_mental_disorders) lists Body Indentity Integrity Disorder in "obsessive-compulsive and related disorders". I have severe BIID and mild OCD, and those don't seem connected at all. Most notably, OCD is very clearly ego-dystonic, and BIID is very clearly ego-syntonic. Is there an actual connection? Where did Wikipedia get this from?
Seems it's just taken from the ICD. ...I wouldn't take it too seriously, neurology isn't exactly a well-understood field yet.
Edit: Nevermind, the ICD lists it under "Disorders of bodily distress or bodily experience". Can't check the DSM since it costs 170 fucking dollars. Also, I just learned what BIID is, and... wow. Makes my body dysmorphia look trivial in comparison.
If an alien published the code for ASI on GitHub tonight, and if that code could run on approximately four A100s (which are expensive but consumer-accessible GPUs), what would you estimate as the probability that a paperclip maximizer (PCM) would cause the extinction of humanity by the end of the century?
I propose that any estimate greater than 5% is a miscarriage of the critical thinking principles of this community. While I personally find orthogonality and instrumental convergence to be persuasive, and while I readily admit that I can't explain why they wouldn't happen, these theories don't cause me to forecast instadeath from PCMs. No matter how persuasive PCMs are, I temper any expectation about how they work by the sheer fact that they're theoretical.
And theory by itself doesn't mean you can't forecast. We already have a playbook for that, which is called the inside view. But the inside view here requires making some connection with empiricism. Our epistemic status for PCMs is lower than what we had before making predictions about atmospheric ignition pre-Trinity test and lower than when we made predictions about spontaneous black hole formation during the start of the Large Hadron Collider.
We also have playbooks around agentic simulation, such as when Axelrod and Hamilton used game theoretic models to explain how cooperation might have evolved by animals. But in their case, they still relied on significantly more empirical touchpoints, such as examples from biology. PCMs, on the other hand, can't be inferred from anything in existence. The PCM theory is designed to not pay rent¹.
Setting aside the discussion around PCMs, I find the other parts of the AI Extinction debate distracting. Questions about whether or not we will get to ASI in the first place, whether or not Moore's Law will end or not, whether or not warning shots are a thing, or whether or not humans have the capacity to build controls for anything; are all red herrings. The AI Extinction conversation has evolved into a Shiri Scissor² because the subject of contention has reasonable—but polar—priors. Most of the text spilled online in the AI Extinction debate is about all these distracting preconditions, when we only get to forecasts of 50% or higher when we contemplate the last step, the possibility of an impassive, destructive PCM.
As a personal aside, I went through an update in formulating this comment. I now don't think that Eliezer actually believes there's a >90% chance of AI Extinction³. He hasn't gone on record with a specific number, and I'd bet his actual number is closer to Scott's (20%), maybe even lower. But Eliezer's behavior is consistent with someone who believes in a 10-20% chance of AI extinction. If I had the same estimate, I too would be grabbing everybody I know and shaking them, asking, "Why aren't you all worried about this? We're all going to die ... tonight!"
PCMs aren't theoretical, there are PCMs right now and for the last 200-500 years, they are called Corporations and Nation States. People sometimes have been successful in restraining them, just barely, from going full psycho mode and doing literally anything to pursue their goals, but only sometimes.
Perhaps the silly thought experiment pedagogy of Rationalist LessWrongers, "paper clips", "Pascal Mugging", etc, have blinded you to the very real and brutal biological/physical fact that evolved intelligence is a weapon, a PCM is a trivial restatement of the obvious in the clothing of a story. An non-human entity with a greater intelligence would no more pay attention to your needs or your life than a non-human entity with a stronger teeth or a killer venom.
You don't believe me ? Look in the mirror. How much thought did you pay to ants living (or were living in the summer) in your bathroom's corners ? When they swarmed the floor, how many seconds did you think for before you started spraying water or insectcide ? You don't hate the Ants, you don't have any problem with them living elsewhere away from your sight. Only the moment you see them, you don't think twice before killing them, their lives are quite literally not a factor in your moral calculus, there is simply no cell holding their value in your ethical spreadsheet.
I think worrying about it is silly because I believe General Intelligence in Silicon to be at least 300+ years away, but if ASI is on Github tomorrow, well, a PCM is not even a conclusion to argue for, it's a trivial restatement of Evolutionary contempt of species for other species.
Unless of course it (1) has something resembling human or human-compatible morality (2) was evolved in conditions sufficiently unlike Evolution. But that already begs the question of who, or what, the creator is.
Fair point. We eliminated the Neanderthals and about 35 other hominids on the way to the present.
Unrelated to my point, it's interesting to me that nobody wishes we could time travel to the Renaissance and tell people, "Look, corporations and complex city-states (more complex than you can even imagine today) will emerge, render people semi-infertile while making them believe that they want to be this way, slowly degrade everybody's mental health until the total purpose of their life is to feed these monster cities, and then eventually minimize fertility because humans have been rendered unnecessary due to gains in "productivity" from autonomous technology can run all the necessary tasks of the city. I already feel like we are the Amish relative to the activity of superstructures that are built on top of us, dumb "cells" or a substrate for some other, non-human purpose. If we're not bothered by this "evolution" then does it make sense to worry about the next step?
Again, this is unrelated to my point, because it's a motivated argument, but I'm just providing some color to my thoughts.
I'm agnostic about a PCM, but I expect that unmodified humans will just get outcompeted economically (and politically, and culturally, and militarily) by ASIs. This _isn't_ to predict that every last human will die. There are thousands of chimps still around, and they can't build cities, generators, or nukes. One of the poll questions about human extinction phrased the question as something like "Will there be less than 5000 humans alive by 2100?" - and it only takes a couple of ASI who have an intellectual interest in humans to keep the species alive at that level. But I doubt that there would be 8 billion humans in 2100 if there was an ASI now. I'd guess perhaps 50:50??? Batons get passed. Only a small subset of the institutions that today use humans as components have humanitarian goals.
Given your scenario, I think it more likely that human extinction would be caused by an explicit goal of at least one AI, rather than an unintended side effect.
(New supervillain plot: Notice that Hamas is developing a "kill the Jews" AI, and in response create a Mormon-Jewish hybrid religion that converts mitochondrial Eve to Judaism, thus retroactively converting all living humans. Point out that the only reason you're considered a supervillain is that now it's happening to **them**. Laugh.)
By PCM do you mean a maximizer of specifically paperclips? Or are you talking about it in more genral sense - as any kind of non-aligned maximizer? Do we take into account the likelihood that aliens solved alignment and that's why they are alive and able to post something on github? Or are we talking about an abstract scenario that the code for a maximizer ASI, not specifically made to care about human values, just magically happened to be available in open source without any probability updates about its origins?
The latter. So maybe instead of aliens, an angel/demon submitted it, with no updates about its origins. And by PCM, yes, I'm talking about any kind of non-aligned maximizer.
Because they can cooperate in ways that humans can't. Why have a source of additional AGIs out there when doing some comparatively simple tweaks to the environment means that a third party is removed? It's not like mosquitoes stopped getting murdered when developed countries fought each other.
"THE MOST SENSELESS ENVIRONMENTAL CRIME OF THE 20TH CENTURY
Fifty years ago 180,000 whales disappeared from the oceans without a trace, and researchers are still trying to make sense of why. Inside the most irrational environmental crime of the century.
...
The Soviet whalers, Berzin wrote, had been sent forth to kill whales for little reason *other than to say they had killed them*. They were motivated by an obligation to satisfy obscure line items in the *5-Year Plans* that drove the Soviet economy, which had been set with little regard for the Soviet Union’s actual demand for whale products. “Whalers knew that no matter what, the plan must be met!” Berzin wrote.
This absurdity stemmed from an oversight deep in the bowels of the Soviet bureaucracy. Whaling, like every other industry in the Soviet Union, was governed by the dictates of the State Planning Committee of the Council of Ministers, a government organ tasked with meting out *production targets*. In the grand calculus of the country’s planned economy, whaling was considered a satellite of the fishing industry. This meant that the progress of the whaling fleets was measured by the same metric as the fishing fleets: gross product, principally the sheer mass of whales killed.
[bafflingly, this was despite the fact that "the Soviet Union had little real demand for whale products. Once the blubber was cut away for conversion into oil, the rest of the animal, as often as not, was left in the sea to rot or was thrown into a furnace and reduced to bone meal—a low-value material used for agricultural fertilizer...
... Japanese whalers made use of 90 percent of the whales they hauled up the spillway; the Soviets, according to Berzin, used barely 30 percent. Crews would routinely return with whales that had been left to rot, “which could not be used for food. This was not regarded as a problem by anybody.”
This was the riddle the Soviet ships left in their wake: Why did a country with so little use for whales kill so many of them?"]
Whaling fleets that met or exceeded targets were rewarded handsomely, their triumphs celebrated in the Soviet press and the crews given large bonuses. But failure to meet targets came with harsh consequences. Captains would be demoted and crew members fired; reports to the fisheries ministry would sometimes identify responsible parties by name.
Soviet ships’ officers would have been familiar with the story of Aleksandr Dudnik, the captain of the Aleut, the only factory ship the Soviets owned before World War II. Dudnik was a celebrated pioneer in the Soviet whaling industry, and had received the Order of Lenin—the Communist Party’s highest honor—in 1936.
The following year, however, his fleet failed to meet its production targets. When the Aleut fleet docked in Vladivostok in 1938, Dudnik was arrested by the secret police and thrown in jail, where he was interrogated on charges of being a Japanese agent. If his downfall was of a piece with the unique paranoia of the Stalin years, it was also an indelible reminder to captains in the decades that followed. As Berzin wrote, *“The plan—at any price!”*
...
In one season alone, from 1959 to 1960, Soviet ships killed nearly 13,000 humpback whales."
(Even more bafflingly about the whole thing was the fact that the Soviet Union probably didn't even have much use for all the whale oil it was getting this way, it had plenty of normal petroleum oil and that was much cheaper, there's a reason kerosene and the like supplanted whale oil the entire world over. They were effectively using *zero* percent of the whale! The entire thing was like Saudi Arabia taking up whaling for oil.)
In other words: if the Soviet Union had an ASI, they absolutely would have paperclipped the world by accident. Maybe even 'rotting whale-d' it by turning it into a mass of (rotting, but that doesn't matter of course) whale flesh being proudly turned in at its docks by a "5-Year Plan aligned" ASI. On paper, everything would be roses. And that's all it takes.
People doing mindless things for the sake of bureaucracy is a general case for this. We know that agents can pursue side quests that that are irrelevant or moot from the original quest. And probably we have examples of machines running autonomously like mindless bureaucrats as well.
Why does this not undermine your point that alignment is easy and happens by default? Or is that not your argument? If not, I don't understand how you can acknowledge instramental convergence and orthogonality and still say extinction is unlikely.
Well, the point of saying it "pays rent" is that your *anticipated experience* changes, not just the thing itself.
Worlds where unconstrained optimizers are easily reached would also have many examples of unconstrained optimization happening, where principle agent problems are the default and not an exception, and where there aren't literal reams of examples from ML of specification gaming.
So yes, it does pay rent in anticipated experience; I would be seeing a very different world if alignment were easy.
I don't have a good answer, but my strong intuition is that everything in the world after the Great Oxidation Event has been priced. Who would've thought that the reason empires don't exist today is because the cost of violence doesn't make up for non-consensually extracted benefits? Who would've thought that populations would be plateauing because of some weird intersection between responses to abundance mixed with responses to crowdedness? It also "just so" happened that we didn't blow ourselves up with atomic weapons. When there are so many "just so"s, you start to wonder if there's some inner logic. The one I've resigned to is that complex molecules on Earth really don't want to separate.
I am of the same mind as the strapping young fop, eagerly racing home to show off his etchings, only to find to his dismay that the other party actually meant sex.
I think it's very rude to imply you have strong proof (no one else possessing of critical thinking should claim above 5%) and then not even have the sketch of a causal mechanism for your belief, instead having a bunch of selection effects with obvious Anthropic shadow biases as your proof!
My apologies for coming across as rude. The Anthropic sketch is not the core thrust of my case. I do have some causal mechanisms for my beliefs, specifically around warning shots¹. I also plan a follow-up ACX comment that'll give more detail. But the purpose of my top-most comment is to claim that the AI Extinction argument should be dismissed at the outset for similar reasons to why Pascal's Wager should be dismissed. And just as you shouldn't have to know everything about the Bible to defend the dismissal, someone shouldn't have to know much about all the jargon from AI Extinction (See: Burdensome Details²).
The survey said that the results will be published on the Philosophy Bear substack at some point. But it would be nice if Scott let us know when that happened with a link in an open thread.
It's not clear to me how the questions about mean income are supposed to be interpreted. Like, I can't easily find mean *individual* income, only mean *household* income. But comparing my individual income to household income hardly seems right; am I supposed to compare my household income instead? Or should I use GDP per capita as an approximation of mean individual income? Or what?
"Pro-Impact Donation" — nobody can argue with being pro-impact, and it softens the implication that Effective Altruists think everyone else is ineffective and selfish.
yeah but the comments are by default sorted from newest on top to oldest on bottom, so your FIRST paragraph is technically the first thing anyone would read if they started at the bottom
(just kidding, I don't really think any of this is relevant)
also, child comments are displayed below the parent comments, so... technically... at the current moment...
(Next time could please you also provide a short summary, what is the study about and what specifically you find interesting?)
The authors conclude that expertise requires extensive practice, but that all students learn at the *same* rate (assuming the same practice). Therefore, how fast they achieve mastery depends only on how much opportunity to practice they get.
Before moving beyond the abstract, let's check my priors. My first reaction is that this sounds like yet another bullshit politically correct theory of why all humans are identical and genetics is a myth; checkmate, natural scientist! I remember the kids I tutored who spent a lot of time learning and doing math exercises, with very slow progress; and other kids, who just looked at the explanation and got it; so this seems like a flat denial of reality. A possible counter-argument is that perhaps the kids who got it quickly, have *already* spent a lot of time doing math or thinking about math, so maybe they actually spent more total time. The outcome may also depend a lot on what kind of exercises the students do. For example, if the exercises advance in difficulty quickly, I would expect the smart kids to handle it somehow, and the slow kids to be hopelessly left behind; but if instead it is an endless repetition of the basic exercises, I would expect the slow kids to finally learn it, and the smart kids to die of boredom. It probably also depends on whether the "expertise" is about understanding something, memorizing something, or acquiring motor skills. I am afraid that perhaps the authors will not distinguish all of this, just make an experiment in one setting, and then generalize. Okay, let's move on to the article.
So, the authors used "online practice systems" for college courses in math, science, and language. What they measure, if I understand it correctly, is the increase of accuracy in answers. Like, if at the beginning one student answers the test questions with accuracy 55% and another with accuracy 75%, and then the former improves to 56% and the latter to 76%, they both made the same progress of +1%. The authors find out all students typically improve by 2.5% per learning opportunity. Mastery of the subject is defined as 80% accuracy.
I am skipping most of the numbers and graphs; sorry, I didn't want to spend too much time on this. At this moment, my obvious objection is that the meaning of these results strongly depend on how the test questions are selected. For example, if you choose 100 questions of approximately the same difficulty, then 70% accuracy simply means 70% probability of getting any of those questions right. On the other hand, if you choose 100 questions of increasing difficulty, from very simple to very difficult, then 70% accuracy means that the student can solve the easy and medium questions, but not the hard ones. Improving by 5% would mean a greater probability of success in the former case, and an ability to solve more difficult problems in the latter case. The meaning of "mastery", operationalized as 80% success rate, would also differ dramatically: a 80% accuracy on an average task is not very good; but the ability to solve all problems except for the 20% most difficult ones... depends on how exactly difficult those top 20% are. What I am trying to say is that the words used in the article may not necessarily have the meaning we expect. (Also, the graphs are suspiciously linear. Come on, *all* of them? Too good to be true.)
I agree with the remark that talking about "10 years of experience" does not make much sense without saying how much of that time was actually spent practicing the skill. Also, not all practice is equal -- in this experiment all students did the same, but in real life it would be different. (The authors mention students getting different "institutional support". I would also mention that there is a difference between doing easy or hard exercises; or practicing the same thing over and over again vs trying new contexts.) I also fully agree with the comment that when someone is praised for learning fast, they are often actually praised for having prior experience.
Conclusion, from my perspective:
GOOD: The authors correctly identify that when people talk about good or bad learners, they usually mix "prior experience" and "what they learned during this lesson" together. If we control for prior experience, the differences in learning may be way smaller. (Though I find it very unlikely that they would disappear *completely*, as the authors claim.)
BAD: https://i.imgflip.com/87h4hg.jpg The authors measure how quickly students achieve 80% accuracy on test, and make a general statement about learning. The obvious objection is that this seems like the easy part of the learning curve (imagine someone having a 80% accuracy at spelling or multiplication), and maybe the effects of innate skills are much stronger at the difficult part of the curve. Like, the student who wins the math olympiad is not necessarily better at multiplication tables than the average student, especially when we measure the increase in accuracy per lesson.
I do have a kind of insane counterargument that I haven't thought through very well -
Have you ever taken the English vocabulary test? It was popular a couple of years ago, I think it was on a university affiliated site. Anyway the authors of the test had a pre-emptive response to what is apparently a common critique of the test: it's so easy but suddenly becomes really hard!
That's not actually true - a person's internal sense of what is easy or hard is entirely informed by whether they already know something. In other words, learning is actually pretty binary - you either know it or you don't. The easy or hard distinction is similarity to prior knowledge, not an inherent quality of the content.
This is also sort of linked to Scott's parable of talents essay. FWIW I was often thought to be a fast learner as a kid but it didn't feel like this internally - often it felt like things I encountered kinda looked like things I already knew, and because I was a voracious reader and j would read literally anything I could get my grubby child paws on, frequently unsupervised, I knew a lot more than most adults expected.
The way I model it is some children have greater curiosity and openness - so they absorbed a far greater amount of knowledge even without someone trying to teach them. Some kids have (? Slightly?) Better pattern recognition and are able to link current problems with past knowledge more often. But curiosity is probably the foremost trait in early childhood education (mostly because of brain plasticity reasons, being really really curious when young is exponentially better).
I would also expect that the pattern recognition component isn't as significant as having an extensive bank of past knowledge to reference except at the very tail ends (eg learning disability and genius, although the latter sounds doubtful but there's a literal Indian mathematician to prove me wrong, so y'know). For the average person, it's more important to build the knowledge bank.
(Also, note that the study runs on attempts and not time. It's possible that less talented students had to spend more time per attempt to really understand what it was trying to teach, or it could literally just be reading speed. Do we know what influences reading speeds?)
Adding 4-digit numbers is more difficult than adding 1-digit numbers. And yet, for a small child both are hard, for an adult mathematician both are easy, and for a 3rd grader one is hard and one is easy. So, it is a function of the observer, but also a function of the test.
> The way I model it is some children have greater curiosity and openness - so they absorbed a far greater amount of knowledge even without someone trying to teach them.
This is definitely a factor. Also, curious children will actively *ask* -- so from some perspective it is teaching, but it is different from the teaching that is initiated by the teacher or parent.
I will assume (until convinced otherwise, but a study more reliable than this one) that *all* of these effects explain a part of the story -- prior exposure, curiosity, faster pattern recognition -- in a nonlinear way as they reinforce each other. (If you are curious, you will seek more exposure. If you are curious, you will pay attention to the patterns. If you notice the patterns, you will remember them, and you will ask someone about them. If you are familiar with something, you are more likely to mention it, or to notice when something is similar. Etc.)
People who have an axe to grid will try to reduce it all to one of the factors. (Racist will say that everything is genetic. Wokes will say that everything is a result of privilege. Etc.) Their arguments are not convincing, because they are more in a direction of "here is how X contributes to the differences", which I already accept; I am curious about the proof that "this is why X contributes to the differences *and Y does not*".
There seems to be a growing field of what you could call “auxiliary AI” for safety/interpretability/control. Auxiliary in that we don’t know how the main ML model functions, so we tack on an extra ML model to help interpret the main ML model. The monosemanticy example was a good case, but there’s tons of these now, in a bunch of different fields. Seems a bit silly to me, because you’re ostensibly just moving the goalposts: if we don’t know how the auxiliary AI is doing what it’s doing, then how do we know it’s actually performing the interpretability/control function we want it to? You can make all sorts of arguments about the structure of a network and how it enforces certain desirable behaviors/patterns/structures in the working of the model, but at the end of the day, if its still just a giant inscrutable tensor of weights, are we really getting anywhere? Seems to me we should either just earnestly abandon our efforts at creating thinking digital structures we can understand, or abandon the neural network paradigm but it doesn’t seem like you can do both. Maybe I’m just not as deep in the lit as a I should be tho.
Hi for those interested in Scott's piece on AI monosemanticity I wrote a piece about the implications of AI interpretability for biology and drug discovery.
https://open.substack.com/pub/aprimordialsoup/p/aspirations-of-biological-monosemanticity?utm_source=share&utm_medium=android&r=1ot3ut
Thanks to Scott for inspiration.
I am a big fan of the "Thrive-Survive Theory of Politics" by Scott Alexander (https://slatestarcodex.com/2013/03/04/a-thrivesurvive-theory-of-the-political-spectrum/). One thing I always found interesting/incongruous with this theory is that it claims that (material) security leads people to feel secure, and thus become more likely to identify as leftist and adopt their outlooks on life (i.e. post-materialism or "globalism"), while those who are more concerned with "surviving" will be more likely to be right-wing and adopt their outlook (i.e. being more traditional and more harsh towards human failure). I feel that this theory certainly makes sense on a global level - people in poorer countries are more likely to be in a "survive" mindset and thus less likely to be progressive politically than those in wealthier countries. Even inside the US it makes sense, with right-wing "Red States" tending to be poorer than left-wing "Blue States". But there is a major flaw in this theory IMO. Namely that the USA, which is the richest major economy in the world (https://en.wikipedia.org/wiki/List_of_countries_by_GDP_(PPP)_per_capita) is definitely (on average) more towards the "survive" Spectrum than most other rich Western countries, including all other Anglo countries (as can be seen in the WVS map: https://www.worldvaluessurvey.org/WVSContents.jsp?CMSID=Findings). But since the US is richer than almost all other Western countries, wouldn't one expect the US to be more towards the right side ("self-expression) than the other Anglosphere countries and/or Western European countries?
I guess one argument is that Western European countries have a better social safety net than the US and thus allowing people to focus more on "thrive" attitudes politically...but even then, why didn't the US develop a social safety net like those Western European countries in the first place? Is it just because of racial politics? Also, might this change if the US becomes more progressive politically overall? Especially if Western European countries become more focused on "survival" politically because of the energy/geopolitical situation in Europe? Anyway, I would like to know others' view on this here...
I have a few views.
1. I read that post a couple of years ago and disagreed with it strongly. I'm glad you've given me the opportunity to comment on it without requiring the ability to go back in time to 2013. Thanks.
2. One major problem (and honestly I find that Scott makes this mistake amazingly often, though many others do it even more) is basically focusing entirely on 2010s US politics, and even then mostly on the social side only. For much of the last two centuries economic issues have dominated, and poorer voters have been more progressive. This is *still* the case in a number of western countries. And even within the modern US, where both parties have been hijacked by social and cultural activists to an unusual degree, when actually voting directly on things like minimum wage and medicaid expansion, the people of poorer red states are, unsurprisingly, quite liberal on those things. And also, Trump does better than more ideologically right-wing candidates among poorer republicans, again unsurprisingly. When your theory of the political spectrum largely ignores most of the time the spectrum has existed, most of the world even now, and much of the economic side of it even "here and now", I don't think it's unreasonable to say it's not a great theory.
3. Even focusing entirely on "2010s-20s US social issues" this theory is pretty bad. It doesn't explain basically anything to do with feminism: believing and punishing rape accusations without question is what you do in the zombie apocalypse when any threat to group safety and cohesion is potentially fatal, while due process is a luxury for the prosperous utopia. Similarly, protecting unborn life is, like protecting the environment, a function of a moral concern born of prosperity; "you're inconveniencing me, you die!" is a feature of the brutal zombie world. It doesn't explain attitudes to climate change: the leftists are responding fearfully to an existential threat, the rightists are saying "nah we'll be fine, our society can handle anything". Scott admits it doesn't explain school choice, and I'd say it also doesn't explain political correctness or hate speech laws: non-conforming ideas that may create bad feelings are tolerared in a stable utopia, and suppressed in a zombie situation. Overall it probably doesn't explain about half of all social issues alone. So, not a great theory at all.
4. I'm not at all surprised it's a bad theory, like every other theory about what "left" and "right" mean. What I am surprised at is that anyone as smart as Scott would think they can come up with such a theory, when every other attempt has been such a disaster. No one knows what left and right really mean. No one! I have literally never seen an explanation that doesn't completely break down for at least one major issue. "The right wants to return to the past", except for drug laws. "The right wants to ban destructive behaviours", except those involving the environment. "The left wants humane treatment for the most vulnerable", unless you happen to be in the womb, in which case sucks to be you I guess! "The left wants freedom in your personal life", except for hate speech laws, and gun control, and affirmative action... I think left and right mean *something*, for people to organise themselves so consistently around it, but whatever that something is isn't remotely obvious or clear, is a collection of historical accidents and unconscious social instincts that we don't understand, and only makes recognisable sense in a very particular kind of modern, democratic, western society.
5. And just about Europe being less polarised, more tolerant, more liberal, and so on. Both the fact that this is absurdly, ridiculously false, and the reason they're making an enormous effort to appear that way right now, are so blindingly obvious that it shouldn't need to be explicitly stated. I find it truly breathtaking how many people make the above claims as though the world was created ex nihilo in 1995, or (if we're just talking about western Europe) in 1950. I can only assume the blame lies with an overbloated media ecosystem that "reacts" to such a degree to every single political and social development, year after year after year, that people can become so confused that they actually *forget* utterly unforgetable things that happened within living memory.
It's a "survival" mindset that doesn't have anything to do with levels of wealth.
It seems like you have similar views to Bryan Caplan on this (https://www.econlib.org/archives/2017/09/whats_wrong_wit_23.html ). But IMO that view is clouded by his (libertarian) biases to some extent...now I agree that the terms "left" and "right" change with time and place, but I still believe that there's some truth to Scott Alexander's theory (namely that left-wing values tend to be correlated with post-materialistic values, and post-materialism tends to be correlated with wealth).
Also, I think that generally everyone would agree that Western European countries (besides Italy maybe) are more "left-wing" than the US as it is commonly understood in the modern Western political discourse. I am not sure when this first became the common wisdom among most people, but it certainly has been ever since I started partaking in political discussions on the internet (mid-2000s).
I think the actual laws of the USA *are* more "self-expression" than the EU, on average - For most cultural things, like gay rights, or accepting transgender as a thing, legal recognition came relatively early in at least some states of the USA. Even after the effective overturning of Roe vs Wade, half of the USA still has very liberal abortion laws. Overall, though ,for a lot of these cultural issues I think religiosity is far more predictive than wealth, it's just that religiosity happens to inversely correlate with wealth in general.
Re social safety nets - they are arguably a "survive" policy, at least from the perspective of any given voter. Lower risk of disaster, at the cost of higher taxes reducing the potential maximum 'thriving'
I am not sure if I agree with your argument...it's true that the US allows more "freedom of expression". But in most Western European countries, laws tend to protect workers and consumers better than in the US. Also, certain "vices" are legal in Western Europe while being illegal in the US. In terms of gay rights, it's true that the US introduced Gay marriage earlier than many countries in Western Europe (though some introduced it earlier than the US) at a federal level, but I see that more as an artifact of the supreme court being "liberal" in 2015 rather than true progressive/thrive sentiment in the US. Most opinion polls show that in the US agreement with gay marriage is lower than anywhere in Western Europe (i.e. https://www.pewresearch.org/short-reads/2023/11/27/how-people-around-the-world-view-same-sex-marriage/). Thus, the US is arguable less "thrive" (live and let live) than Western Europe here. Now, this might be explained by the higher prevalence of highly religious people in the US than in Western Europe. But if you look at the WVS map (https://www.worldvaluessurvey.org/images/Map2023NEW.png ) then you can see that more religious countries tend to be on the left side - the "survive" side of the map - and that the US is more towards the "survive" side than most other Western countries (not surprising considering the higher religiosity in the US). Thus, I don't think that the US is arguably more if a "thrive" mindset than most of Western Europe - in fact, the WVS map and opinion polling shows the opposite.
As for social safety nets, I disagree with your take here too. If a welfare state is an expression of a "survive" mindset, then why do conservatives throughout the Western World (people who tend to be more of a "survival" mindset as per Scott Alexander) tend to oppose them more than leftists (who are more of a "thrive" mindset as per Scott Alexander). Thus, my argument stands in my opinion: that there is an incongruence with Scott Alexander's theory of "Thrive vs Survive", namely that Western European countries which are poorer than the US per capita are more "thrive" politically and socially than the US...
OC ACXLW OpenAI Lessons and Enlightened Sex Scandals
Hello Folks!
We are excited to announce the 50th Orange County ACX/LW meetup, happening this Saturday and most Saturdays thereafter.
Host: Michael Michalchik
Email: michaelmichalchik@gmail.com (For questions or requests)
Location: 1970 Port Laurent Place
(949) 375-2045
Date: Saturday, Dec 2, 2023
Time: 2 PM
OC ACXLW OpenAI Lessons and Enlightened Sex Scandals
Conversation Starters :
NYT
The Unsettling Lesson of the OpenAI Mess
https://docs.google.com/document/d/1lAKTaflRRuy4i3KmEA8hn-R5RFnKURLI8OcSHWuG3tU/edit?usp=sharing
Open Access
The Unsettling Lesson of the OpenAI Mess
https://docs.google.com/document/d/1lAKTaflRRuy4i3KmEA8hn-R5RFnKURLI8OcSHWuG3tU/edit?usp=sharing
Is Enlightment Compatiable with sex scandals by Scott Alexander?
https://slatestarcodex.com/2019/10/16/is-enlightenment-compatible-with-sex-scandals/?fbclid=IwAR3E4EFVvw2jYvWcwvsiZUOC0jnkAvuZIFFE2kUYxo3Q3SGSysy4803E0Ow
Audio:
https://podcastaddict.com/astral-codex-ten-podcast/episode/84712125?fbclid=IwAR0nX4hN7jNnOwerWvH4wPwNdwLEhqIYoo9rt63ZzNfQSiA-vyGQFgWDDKE
Walk & Talk: We usually have an hour-long walk and talk after the meeting starts. Two mini-malls with hot takeout food are readily accessible nearby. Search for Gelson's or Pavilions in the zip code 92660.
Share a Surprise: Tell the group about something unexpected that changed your perspective on the universe.
Future Direction Ideas: Contribute ideas for the group's future direction, including topics, meeting types, activities, etc.
Recommend quotes around article title, or at least move the question mark, for the 3rd conversation starter, which currently reads, “Is Enlightment Compatiable with sex scandals by Scott Alexander?”
Made me triple take, then made me laugh.
Oops lol
How good was the average life of the Middle-Ages peasant compared to the average life of the middle-class American today? I'd guess 1/2 as good but with less variance. Probably about as many miserable people as a percentage.
From whose viewpoint? An average middle-ages peasant was probably averagely happy/unhappy with his/her life, just like an average American is today. If you transport an average American to M-A peasantry he'll quicky succumb to disease/exposure/etc. If you transport an average M-A peasant to modern day America he'll probably go mad.
I'm not sure how we're defining good, but I'd say 20% as good. You'd be living in a shed in the cold, wearing two uncomfortable shirts all the time, having half your kids die when they're little, eating bread and not much else, getting conquered and perhaps slaughtered periodically, having no actual entertainment apart from church and sitting around and talking when you're not doing hard physical labor. It's all pretty bad.
Probably about 2/3s as good.
I keep finding the phrase "Family, friends, health and wealth" to be a really useful way of classifying human happiness. Like, you can add or subtract something but it's hard to imagine a good life without these things.
In our favor:
Wealth, obvious, and I think it's hard to conceive of how much further we are ahead.
Health, and it's not like we're healthy today but a lifetime of physical labor, especially peasant agriculture, is just brutal. Your back, knees, everything just shot. I'm pretty sure a ton of 19th century drug and alcohol abuse is just pain management.
In their favor:
Family, it's not close, they had entire extended family and kinship groups we've totally lost.
Mixed:
Friends. I mean, in general I think the middle-ages peasant had more friends and seeing people in person is better but...this whole internet acquaintance thing...I mean it has advantages. Let's give a point to each
So, we're scoring 3/4 and they're scoring 2/4, that means they 2/3 as happy as us or we're 50% happier than them.
I like this analysis.
What are the best things and worst things free market capitalism has given us?
I'd say the best things are bringing billions out of poverty. The worst things are Taylor Swift and superhero movies.
I'd say the worst things are more like casinos and opioid addiction, and perhaps junk food - these are things that appeal to the short-term interests that determine what a person does in the moment, even though they go against the person's long-term interests (as you can tell based on asking someone what they would prefer to do one month or one year from today, rather than asking what they would prefer to do this very instant).
It's possible that junk movies and junk music are like these things in the ways that junk food tends to be, but I think none of them are as extreme as the traditional "vices".
I can't believe you didn't like Zebraman. https://www.youtube.com/watch?v=JSIHn9pddho
When people are brought out of poverty, they buy things like superhero moves and Taylor Swift.
I came across this while reading Unsong:
> Still other texts say the Messiah will come in a generation that is both the most righteous and the most wicked. I don’t even know what to think of that one.
I do! Because I came across this while reading the perennialists:
> The traditional world was essentially good but contained much evil, while the modern world is essentially evil but contains much good.
You can definitely get away with claiming the modern world is essentially evil if you judge by the standards of the traditional world, but then, much good is done! So we truly are the most righteous and most wicked generation already.
I thought of looking at what the Arabic Wikipedia says about the events of October 7 and I'm deeply impressed. The page is https://ar.wikipedia.org/wiki/عملية_طوفان_الأقصى and then use auto-translation if you don't know Arabic (I don't).
There's no mention at all of any civilian deaths in Israel. Really, none whatsoever. Apparently what happened was, Hamas fighters broke through the fence, entered various Israeli cities, army bases and "settlements" near Gaza and clashed with the Israeli army there. They captured lots of Israeli soldiers and officers. Nothing is said about any civilian hostages. The music festival is not mentioned.
Is this really that surprising given all the ink that has been spilled on the economics and the incentives of Wikipedia participation and their governing structure ? read any Anglo-USA culture war topic on English wikipedia (e.g. GamerGate) and you will witness the same one-sidedness, the selection of which facts to emphasize in the official and authoritative article voice and which to put between quotes and attribute to external parties, and selective wording of facts.
As an experiment, I clicked the language menu that appears on top (globe icon) and got the Hebrew version of the same events : https://he.wikipedia.org/wiki/%D7%9E%D7%AA%D7%A7%D7%A4%D7%AA_%D7%94%D7%A4%D7%AA%D7%A2_%D7%A2%D7%9C_%D7%99%D7%A9%D7%A8%D7%90%D7%9C_(2023). I switched the English google translate as well because I can't read Hebrew.
As far as I can see, no mention of the 14000+ dead, no mention of White Phosphorus being used in one of the densest civilian areas, no mention of Amnesty International or the resignation of human rights officials in the UN or the international condemnation of Israel from Belgium and Spain. As a matter of fact the entirety of post-10th october or post-13th october is extremely hazy in the article, almost dream-like, almost non-existent. Which is coincidentally also the implied general position of the Israeli propaganda : There was the 7th of October and a few days after where Israelis were victims, and then a huge black hole after that in the Calendar where whatever happens can't really be blamed or even traced to individual deliberate decisions.
Truth is one, but everybody wants to take a different piece. Nobody is immune to this.
No, that's wrong. The Hebrew page you linked to does not talk about the 14000+ dead because it's devoted to Oct 7 only, and even there only to the Hamas attack. It links prominently to another page devoted to the entire war, as Israel defines the current conflict: https://he.wikipedia.org/wiki/מלחמת_חרבות_ברזל
You'll find the 14000+ number there.
There are separate pages for the West Bank in this conflict
https://he.wikipedia.org/wiki/מעצרי_פלסטינים_וערביי_ישראל_במהלך_מלחמת_חרבות_ברזל
where you'll find a lot of numbers and criticisms over Israeli arrests,
and for international reaction
https://he.wikipedia.org/wiki/תגובות_בין-לאומיות_למלחמת_חרבות_ברזל
where you'll find Belgium and Spain.
It's possibly not as detailed as you might want. I don't see a breakdown of the number of children killed (according to the Hamas-run Gaza Healthy Ministry) or allegations about white phosphorus. But there's plenty of accusations against Israel presented on those pages, and I'm reasonably sure these too could be added.
The Arabic page is of a completely different kind. Such total denial of reality of civilians being killed - which reverberated around the world and was shown in mass media everywhere - is chilling, and no, it's not business as usual in Wikipedia as it's run in English, French, or Hebrew for that matter. Interestingly, even the Farsi wikipedia mentions the civilian slaughter prominently (I went on to check if Wikipedia is blocked in Iran, which would mean the Farsi wikipedia being diaspora-run, but I believe it's not). But not the Arabic wikipedia. This level of disconnect from reality is really something.
I stand corrected. I should have followed the links in the Hebrew Wikipedia more before I commented. Thanks for your effort in tracing down where I'm wrong.
I generally agree that Arab sources are more radicalized and polarized against the Israelis significantly more than Israeli sources are radicalized and polarized against Arabs.
Since you link to other pages, however, wouldn't it be fair to also link to this other Arabic page about the Nova festival ? It was tricky to find because neither it nor its Hebrew counterpart which I found it through call the festival by the ubiquitous media name "Nova Festival '', instead calling it the name رعيم (R'aaim) festival, which is apparently the location of the festival or something like that :
https://ar.wikipedia.org/wiki/%D9%87%D8%AC%D9%88%D9%85_%D9%85%D9%87%D8%B1%D8%AC%D8%A7%D9%86_%D8%B1%D8%B9%D9%8A%D9%85_%D8%A7%D9%84%D9%85%D9%88%D8%B3%D9%8A%D9%82%D9%8A
In that page, everything that I have seen said against Hamas in international media is reiterated, including the horrific videos of the Chinese-Israeli young woman being taken on a motorcycle and the video of the German woman Shani.
I'm sure that with some effort that I'm too lazy to expend right now (but will happily do so if you asked me), I can find Arabic equivalents to pages like the ones you posted. The Hebrew Wikipedia "wins" this battle of neutrality in my view by a slim margin since their October 7th page at least links to the page about the Nova massacre, whereas the Arabic one doesn't as far as I can see.
Still, the Hebrew Wikipedia makes some interesting word usage choices, like the fact that they call Hamas "Terrorists" in the authoritative article voice without quoting, and indeed likes to use the phrase "Palestinian Terrorists" a lot. Meanwhile, the English Wikipedia page never describes **ISIS** as terrorists except in the words of someone else, "Designated as Terrorists' ', "Considered by such and such <country name> as terrorist organization' ', etc.... I don't disagree that Hamas has many of the major hallmarks of a terrorist organization, it's just that Wikipedia English still doesn't want to use the emotionally charged word with a much worse group.
In one of the pages you linked, probably the one about the West Bank and Arab Israelis, they put "Human Rights" between quotes when describing human rights organizations. Putting things between quotes is an interesting motte-and-bailey, the Motte is that they're just being accurate, everything a journalist or a second hand source like Wikipedia says is technically reported speech that the strict rules of grammar say should be put between quotes. But the Bailey, of course, is that putting things between quotes has the widespread double meaning that the reporter doesn't trust or agree with the things being reported, so "The 'Human Rights' organizations" is really "Those organizations that call themselves Human Rights organizations in a way that doesn't convince me". Which at least primes the reader or the listener to look for reasons to disbelieve all what follows.
I think that what I got out of this is not (only) that sources outright lie by omission (although they often do), it's that the mere act of organizing information, the act of linking 2 pieces of information, the very statement that they're related and that a reader of one should read the other, is itself a source of bias.
(Needless to say, my username doesn't extend to the citizens of Israel, who I wish the peaceful among them all good and peaceful things)
Yeah, you're right, the Re'emi page does describe all the accusations. Which makes it even weirder that it's not linked off the main Oct7 page and that that page doesn't even mention civilian victims. My guess is that one of the two pages is an outlier. Perhaps the ar wiki is actually in habit of correctly identifying the main facts, and the Oct7 page is a huge outlier, protected by some zealot editors. Or maybe the Oct7 page is business as usual and the Re'emi page was not noticed by its creators.
Note however that the music festival was but one of the sites of the massacre,
Just while I'm thinking about it, do you have people to talk/write Hebrew with? If so, what sort of people?
No, I still find it difficult to recognize Hebrew letters by name. Reading is even more difficult because each Hebrew letter can be pronounced in at least 3 different-but-related ways (some has even more), those different ways are indicated by diacritics, but those diacritics are omitted from the vast majority of Hebrew writing I have seen : Hebrew on Twitter and YouTube comments, Hebrew on street signs and government IDs, news sites, even some Hebrew videos on Youtube with subtitles aimed at making it easier for learners. Diacritics are only present in the Hebrew bible and a tiny minority of other writings.
I feel that speaking is a bit easier because Hebrew apparently borrows so much from Arabic and English, and the sentence structure looks Arabic-like, but I'm still at the phase where spoken Hebrew appears as an unintelligible stream of sounds to my ears, except the very few words I know (pronouns, "Man", "news", "war", "occupation", ...). I plan to watch a couple of TV series to train my ears and pick a few words, this is the same strategy I'm currently following with German.
There is one Hebrew speaker I'm currently emailing, we write in English about a variety of (non-Israel-related) topics and Hebrew did came up, they assured me that Hebrew is much easier than Arabic and mentioned something about all Hebrew words originating from approximately 200 roots, in contrast to Arabic having 6500+ roots or so. I'm not sure this is relevant, do Arabic learners really care about roots ? Roots were an obscure topic I learned about in high school grammar classes. My email friend is also learning Arabic, so maybe a Hebrew speaker who is trying to learn Arabic is exactly the sort of person to play down the difficulty of Hebrew and focus on the comparative difficulty of Arabic. As a counter signal this video (The Hebrew Alphabet is bad : https://www.youtube.com/watch?v=Q_h541RkCTI) both made me laugh and made me slightly depressed.
"Reading is even more difficult because each Hebrew letter can be pronounced in at least 3 different-but-related ways (some has even more), those different ways are indicated by diacritics, but those diacritics are omitted from the vast majority of Hebrew writing I have see"
I'm surprised, since I think of the large majority of Hebrew letter as consonants with single pronunciations.
I suppose diacritics include the dots that distinguish some pronunciations-- s/sh, v/oo, v/b (the two v's are different letters), c/ch, h/ch....
I make it 24 letters, of which 15, I think (it's been a while since Hebrew school), are consonants with single pronunciations-- a yod (the little letter) is always y, though it might have a vowel diacritic underneath.
I suppose you could argue either way about whether it or English are less phonetic.
I never thought about what a plague the diacritic system would be for people learning the language.
I don't actually know Hebrew. I can read it if it has diacritics or if it's a familiar word and I know a modest number of words.
I find it interesting that I can read Hebrew in Hebrew characters much more easily than English transcriptions of Hebrew.
Thank you.
Is there some way to flag this to Wikipedia? The whole page is kind of insane ("Israeli occupation forces, aided by France and the UK..."
Hi, I am a rabbi and have written something on the matter of "God".
It is written for people who care strongly about "God" in one way or another.
If you fall into that category I hope that you enjoy this piece.
Be blessed.
https://ydydy.substack.com/p/god
A theory about why AI is unlikely to be an existential threat to humanity, though it could still be a very serious threat-- I imagine things getting bad-- billions dead as an AI sucks up more and more resources. However, it's going to be tricky for the AI to figure out what resources (possibly including people) it needs to allot to keep itself going, and possibly it will need to prevent and repair sabotage.
This is definitely a good enough premise for science fiction, and possibly reasonable for the real world.
The idea of superintelligence is that it would be trivial for such an AI to solve that kind of problem, and there is no guarantee that the solution it finds will involve keeping any humans around.
Then we get into questions about whether that level of superintelligence is possible at all (I'm ever so tempted to call it super duper intelligence) and whether it can be achieved without trying things in the real world and making mistakes along the way.
I love ASDI -- Artificial Super-Duper Intelligence... let's all start dropping the acronym and when ppl ask what it means we can be all condescending and snarky that they're not following this debate and its jargon sufficiently close enough to be worth discoursing with.
The Internet is so much fun!
Yep! Personally, the latter is what I see as the big blocker: you have to experiment to develop super-tech, and that means fucking up. Can that entire process happen covertly?
I have noticed something curious about my career in software development: every time I switch jobs, it is of course for higher pay, but the company I jump ship to is always operationally sloppier than the previous one. Now I am at my highest compensation ever, and this org I'm in is hosting a bunch of their code repositories out of a laptop sitting in a shelf at the office, which, not gonna lie, it traumatized me a little when I finally saw this laptop myself. It's definitely some kind of zenith of sloppiness. Hell, I had heard they were hosted out of the office, but I thought it was like out of a Mac mini or something (though is that better than a laptop? It seems better for some reason).
There are a lot of devs here, have your careers followed a similar trajectory? Also, do you have any amusing stories about operational sloppiness like that laptop server?
I suspect this might be related to working in startups / small companies / giant corporations. Startups are improvising a lot, but they also often use the latest technology. Small companies are probably the sloppiest ones (not enough money and prestige to hire an expert on purpose, not enough employees to have an expert appear statistically). Corporations may be slower to adapt, but they usually have a dedicated IT security department; though sometimes the rules are so oppressive that the employees have to work around them to get their work done.
So, in my experience, it seems like a random walk, because my career is a random walk, heh. The biggest examples of sloppiness I have seen somewhere in the middle. (A small company providing IT services to dozens of customers, some of them important big companies, used the same password on all their production servers; didn't change it for over a decade. The password was the name of the original IT admin; all employees in the company knew that.) (In a different company, we were not allowed to use version control, because the local "expert" convinced the management that version control is literally only used for the *published versions* of the software, not during development. We were supposed to share the code updates by e-mail once a week. What actually happened is that we have secretly installed version control that all developers except for the "expert" used. Those were fun times, especially when the "expert" found out one year later.) Things like this would not be imaginable at the beginning of my career or where I work today, because both places contain actual experts.
> We were supposed to share the code updates by e-mail once a week.
What in the name of the holy fuck.
Well, that was definitely the craziest moment of my career.
That "expert" was the kind of guy that probably inspired the Dunning–Kruger research. He co-founded the company right after university, i.e. he has never worked anywhere else, and he was unfamiliar with whatever happened in software development after his school years. He was a mediocre developer, but he compensated his lack of skill by working a lot of overtime, so the other co-founders (mostly managers) considered him a hard-working genius.
During the interview there were many red flags, but the company was desperate and willing to pay lots of money (almost twice my previous salary), so I took the risk. It was actually quite fun, and definitely a learning experience. That was my only job where I significantly changed how the company worked, because things were so dysfunctional that I could introduce obvious things as revolutionary improvements.
I used git and github in my university years, but it was never something that the unversity taught or even mentioned so it was purely peer effect and the internet.
I'm honestly just morbidly curious about the logistics of an email-based version control system that only pulls and pushes weekly. How did that work ? Where was the canonical version stored ? On the Team Lead's computer ? How was complex merges where the old code have things not in the new and the new code have things not in the old was handled ? Which email client can handle the more than 100 MB of text that large code repositories tend to grow into ? How was the different versions named ? How did all of this scale beyond 3 people ?
I **did** once try to enforce a scheme like this in my dumb freshman years but (1) it was Whatsapp, so at least I had E2E encryption unlike email (2) It wasn't weekly (3) it wasn't a for-profit company for fuck sake, just a meaningless school project (4) it was 4 of us including me
That is the difference between junior and senior developer -- the junior only knows the stuff they teach at universities; can create a thousand lines long program alone in a few months; the senior can collaborate on a hundred thousand lines long program that was created a few years before he joined the team... and he knows how to collaborate with other developers (which includes the knowledge of tools like git). That guy was effectively a junior despite 10 years of experience, because he practically didn't learn anything substantially *new* during those years.
Before I joined the company, that specific department only had two developers. One wrote C++ or Java code, the other wrote stored procedures in a database (which were called by the C++ or Java code). So these two rarely worked on the same file; their collaboration mostly consisted of sending their respective half of code in a zip file to the other guy, who simply replaced all previous code by the new version. And I suppose the "canonical" version consisted of the Java code on the team lead's computer + the database scripts on the database guru's computer. And it worked exactly as you would imagine; bugs caused by someone copying a wrong version of file happened all the time.
Three new programmers were then hired, including me, and the team lead imagined that this way of collaboration would scale. Well, the three of us made our "canonical" version on the secret Subversion repository. We imported the team lead's changes to our repository when he sent us some... luckily, he didn't produce much code, because most of the time he was busy doing analysis and attending meetups. We also sent him the changed files by e-mail once in a while, just like he wished, and we didn't care what he does with them... but he practically stopped developing afterwards, probably because he was too busy merging the new code on his machine. From our perspective, that was only good; we thought that his contribution would be a net negative. (Unfortunately, the database guru refused to use version control at all.) So, it was actually quite easy after a while, when the team lead stopped contributing. Only the first two or three months were horrible.
The code was relatively small, at least when I joined the company (while we still used the e-mail collaboration). The releases were numbered, but the project didn't have branches.
Is there a list of current bans?
I'm considering my current economic / social situation, and I think I am in a position to start pivoting to an altruistic career.
My only issue is - what career? I have looked at 80000 hours, but the sort of thing I've been doing up to now is not very conducive to their list of top global problems. I actually really like what I've been doing up to now, and I would like to do something similar. I don't particularly want to earn to give - while I like what I do, I hate working in a for-profit environment.
About me: I have a Bachelor's of Engineering. Skills I have picked up through work have been corrosion prevention, asset maintenance, and pressure equipment management. I know how to work in a big heavy industrial environment (and would in fact prefer it).
Thoughts: I think I would be suited to a role in infrastructure development and maintenance, particularly in chemical plants, nuclear power plants, waste or water treatment plants, or fuel farms. I have done some research in the past and didn't like it, and I think I am a poor fit for research specifically (personality / ways of working issue). I think my ideal role would be an engineer, advisor or project manager for some kind of big infrastructure project, so if anyone knows of someone building a desal plant or sanitary sewers as a charitable thing, I am very interested.
Other than that, I would be interested to hear about any projects, opportunities, or general direction for somewhere I can apply my skills.
I am looking to get out of my current industry in the next 2 - 5 years, so sit tight and accumulate career capital isn't really an option. My current industry is fossil fuels.
Edit: if it matters, I am ethnically Chinese, I'm baseline conversationally fluent in Mandarin (which I can improve with some effort) and a bit less fluent in Cantonese. I don't have close family ties in Mainland China that I'm aware of. I'm also willing to work there.
What do you like about your work, and what do you dislike about working in a for-profit environment?
I like the technical side of things and learning in great detail about how it all works. I really like my team - they're highly skilled people with a lot of professional integrity.
I don't like or trust upper management. They tend to demonstrate decision making that is indifferent to or actively hostile to the common good (workforce and society as a whole). They also have the gall to pretend that those same decisions will benefit the public, which to be fair is what all large corporations do. I don't think I'm alone in that assessment, and I would like to spend slightly less time supporting this. And because my direct team often do and care about things not directly related to improving the bottom line (like spend time addressing safety risks and complying with the law), I'm worried for our jobs. It might be "good for the company" to do these things on some level, but since I often deal directly with issues from these high level decisions (chronically undermaintained equipment, skills gaps) it doesn't feel like it really helps the company either. The worst part is, from the enshittification of everything else, I suspect this is a general trend of the for-profit environment.
I think I would like to keep the parts I like (working with skilled coworkers and direct manager with integrity, who I greatly respect, with lots of opportunities to learn and feel useful), and also feel like I work for an organisation with integrity and moral backbone.
Yeah, makes sense. Thanks for the additional context. I hear you on the sometimes nefarious motivations at for-profit companies, although I don't think that's universal.
I unfortunately don't have any useful connections to share. One thing I've heard is that geothermal overlaps greatly with oil and gas from a technology perspective. The space seems mostly for-profit, but I think that's mitigated by the fact that people are also more mission-driven to generate clean energy. Might be worth poking around on LinkedIn or Twitter and messaging people at some of the younger companies. EG https://twitter.com/TimMLatimer
I'm also trying to figure this out myself (my career has been software development though). You could also ask in the current Open Thread in the EA forums: https://forum.effectivealtruism.org/posts/Jvkx85biHHAeCiSDA/open-thread-october-december-2023
You checked out 80,000 hours, but did you do the career guide (https://80000hours.org/career-planning/)? It's pretty in-depth, I never did finish the process because I decided I am not looking to maximize my impact, I just want to work for an org with a good mission. You could even schedule an interview with them: https://80000hours.org/speak-with-us/
These comment sections are sickening.
They're full of people who hysterically freak out if you say anything that remotely resembles white nationalism, and then they turn around espousing violent, genocidal jewish ethnonationalist rhetoric. This is literally like some far-right caricature of jews, but it proves itself to be true over and over again. And then you have jews like Bryan Capalan promoting demented open borders policies for every western country, but then conveniently finding a justification for saying that the ethnostate that aligns with his ethnicity is the sole exception to open borders being good. You literally cannot make this shit up.
I'm curious: did you sign up just to write this post?
I'm honestly fine with, say, Vatican city. Or Native American reservations. I'd be fine with a White Separatist nation the size of Israel somewhere in Europe. Or maybe several times the size of Israel. Though Israel is a pluralistic nation and its citizens have civil rights, so comparing it to 'White Nationalism' is a bit disingenuous. Israel is not supremacist and it's far less separatist than, say, Saudi Arabia, which gets far less flack from the UN. I wonder why.
I'm also fine with America, say, not letting someone immigrate because of their ties to gang violence or previous violations of border laws or whatever. Every nation regulates their borders to some extent, even if all citizens inside the country deserve full rights.
The problem with the Israel-Palestine conflict is that Palestinian leaders just haven't been interested in a two state solution on any terms. They've wanted all of Israel. They've been willing to kill for it, even if they've been in a momentarily weaker military position. And frankly, they've burned a lot of their allies along the way. Jordan welcomed them in. They tried to overthrow the Jordanian government. They caused civil unrest in Lebanon. They were welcomed in by Kuwait, then Arafat endorsed Saddam Hussein's invasion. There are a lot fewer Palestinian refugees there now, a tiny fraction of what there were. Even Egypt has been wary about accepting refugees out of fear they would ally with Muslim Brotherhood within Egypt. The hope of far right Israeli nationalists that Palestinians will emigrate is unlikely to pan out. There are few countries still willing to take them in.
I'm not a supporter of Netanyahu. But... if someone wants to destroy you, as Hamas and the PLO do Israel, and is willing to violate the conventional laws of war and attacks babies and civilians, and is adamantly not willing to live peacefully in separation, the number of ethical options available to resolve that situation outside of 'open warfare' is, to put it bluntly, rather limited.
> it's far less separatist than, say, Saudi Arabia,
This is probably an inaccurate usage of "Separatist", it's usually used to denote rebel movements that want to politically secede from a bigger polity. The closest to what you mean is most likely "Segregationist", to mean your intended meaning that it has a first class and other class of citizens, although this term might carry too much baggage from its original context in the USA where the classes were drawn along race boundaries which are somewhat alien to the Middle East.
"The closest to what you mean is most likely "Segregationist", to mean your intended meaning that it has a first class and other class of citizens,"
This is explicitly *not* my intended meaning. It is an incorrect interpretation of my meaning, yet a format that other people attempt, repeatedly, to apply to the situation. And this tendency, I believe, causes people to miscategorize the situation time and again.
Words like 'segregation,' or 'apartheid' all describe systems where one powerful group has, as a primary goal, to extract labor from a weaker group. To do this, both groups need to cooperate. The stronger group actively wants the weaker group as part of society, but not on equal terms. Southern planters went through the trouble of importing slaves from Africa, at some expense.
With Israel (Or Jordan, or Egypt, or Kuwait, or Lebanon or a number of other countries, not just Israel) things would be calmer if Palestinians were just *not there.* This isn't suggesting that *making* Palestinians not be there is a moral idea because Palestinians are human beings and have title to land, etc. But it's pointing out that the constant attempt to equate Israel's relationship to the West Bank and Gaza's people with supremacist models is missing a very important aspect of the whole dynamic, which is that the use of Palestinian labor at low rates by any particular group is really not the motive force of the conflict. This realization, that desire for cheap labor is not the cause of the conflict, is at the heart of 'two state solution' proposals. If you asked the average person in the rural American South during segregation if they wanted to wave a magic wand and make all Blacks disappear, most would probably have said 'no, they should just know their place' or something like that. *That's* segregation. (If you ask Netanyahu (Who represents, arguably, the extreme end of Israeli nationalism) if he'd like to wave a wand and make all Palestinians disappear, he would very likely say 'yes.'
This distinction was my explicit consideration when choosing the word 'separatist' rather than 'segregationist' to describe Israel's relationship to Palestinians.
Ah okay then, sorry for second-guessing your intended meaning.
In what sense, though, is Saudi Arabia "Separatist" by your definition ? I have a lot to say against Saudi Arabia but I'm not aware of a Palestinian-like group of politically-relevant people that the Saudi royal family would like to wave a magic wand and have them disappear tomorrow (unless, of course, you count the rest of the male siblings of Ibn Salman).
> Words like 'segregation,' or 'apartheid' all describe systems where one powerful group has, as a primary goal, to extract labor from a weaker group.
I don't agree with this at all. I think Segregation and Apartheid are just adjectives for a whole catalog of taboo systems (and the associated social/political/economical/etc enforcement schemes) that primarily prioritize exactly what it says on the tin : No race-mixing, of all the varieties your brain can possibly conjure up, in particular. And no X-mixing in general, where X is some attribute that the segregationists are both interested in and classify people accordingly to it.
Nazis were segregationists, they intensely hated white German speakers interbreeding with Jews, Blacks, Slavs, and possibly others I'm not aware of. Hindu nationalists are religion-segregationists, they intensely hate it when both men and women of Hindu faith marry Muslims. Islam is one-sidedly religion-segregationist, Muslim women are forbidden from marrying non-Muslim men, but the reverse is ok. I'm only mentioning Marriage because I regard it as the highest form of social mingling, literally 2 bloodlines merging, all the ideologies I mentioned are also opposed (though less intensely) to lesser forms of social interaction with the inferior groups as well.
I don't see the connection to Labor at all. The Nazis could have given the Ukrainians and the Poles the worst working conditions possible short of slavery and they would have accepted in a heartbeat, possibly liking Nazis more than Stalinists in the process. The Nazis would have massively benefited because they now had several Germanies worth of population that hate Stalin on their East, working for free and bulwarking against invasions. They didn't, they didn't want the Slavs' and East Europeans' labor, they wanted them dead. Similarly for the Jewish scientists and financiers who they kept pressuring till they fled or accepted the camps, those people would have accepted far far less for their (very valuable) labor than any scientist or financier anywhere at the time, and the result is a more technologically and financially advanced Germany that is more efficient at cleansing the world out of the inferiors. The Nazis didn't want this, they went to great lengths to kill those cheap (and quite willing to cooperate) sources of labor, diverting fuel and personnel to extermination camps even as Stalinist boots were inching closer and closer to Berlin.
So the Nazis didn't want the labor of the groups they were segregationists against, how does that counterexample reflect on your definition of segregation/Apartheid ?
> If you asked the average person in the rural American South during segregation if they wanted to wave a magic wand and make all Blacks disappear, most would probably have said no
I again don't agree with this. Maybe in the immediate wake of the 1860s, maybe all the way to the early 1920s. But beginning in the post-WW2 era, when blacks have served in the army that liberated white Europe and were increasingly seen as not going anywhere, south segregationists would have absolutely jumped at any chance to make blacks disappear, indeed they would have jumped at a "2-state solution" of deporting all blacks to Africa (didn't this already happen in the 1800s ? Liberia ?) or giving them native-american-like reservations in isolated pockets of USA and never having to think about them again. If you're American I might be wrong and you might be right, as I'm not American and haven't lived in the 1950s-1960s, I'm judging the racial attitudes from what I have read about the period.
> Who represents, arguably, the extreme end of Israeli nationalism
Not at all, he's a moderate peace dove compared to the extreme end of Israel's right-wing. See Shtetl-Optimized latest comment thread (https://scottaaronson.blog/?p=7632#comments, ctrl-f for "Araboosh") to see a sample of how unhinged Israel's right can go.
If you're saying that every group's base has its lunatic fringe then sure, I agree.
"ctrl-f for "Araboosh") to see a sample of how unhinged Israel's right can go.'
I can't access a comment history for Raphael to see if he's a person or a troll. If you want to point to an elected official saying something similar, that's more along the lines of what I'm talking about. You do get an occasional outburst from the Israeli side, like with Amihai Eliyahu. But he's gotten pushback for his comments and has walked them back calling them 'metaphorical.' And given the circumstances, with over 1000 civilians slaughtered, hostages taken, and hundreds of rockets falling, and international calls to do nothing an individual's burst of rage makes some sense.
Thank you for the good faith acknowledgement.
Saudi Arabia used to execute those who converted away from Islam. The public practice of any religion other than Islam is still prohibited. Non-Muslim immigrants must convert to Islam. I'm not arguing that Saudia Arabia is attempting to maintain any manner of underclass deliberately, vis apartheid or segregation, and it's practices are somewhat of a different kind compared to Israel as well since, as harsh as Saudi Arabia is, it is more tolerant of people converting towards the particular brand of Islam it espouses. But I think it would be safe to say, at least, that Israel is far more tolerant and pluralistic of a society, in general, than Saudi Arabia is.
"Although many
intolerant statements had been removed, some school textbooks continued to
contain overtly intolerant statements against Jews and Christians and intolerant
references by allusion against Shia and Sufi Muslims and other religious groups.
For example they stated that apostates from Islam should be killed if they do not
repent within three days of being warned and that treachery is a permanent
characteristic of non-Muslims, especially Jews."
https://2009-2017.state.gov/documents/organization/171744.pdf
LHHI: "I have a lot to say against Saudi Arabia but I'm not aware of a Palestinian-like group of politically-relevant people that the Saudi royal family would like to wave a magic wand and have them disappear tomorrow"
I would say that non-Muslims fall squarely into the category of groups that Saudia Arabia would like to remove. Perhaps 'politically relevant' is the critical point here. It's not that Israel is less pluralistic than Saudia Arabia, but that a particular group of people has been made, as you say, 'politically relevant' in Israel.
People may be imprisoned for converting away from Islam in Saudia Arabia. That is 'a group of people.' But for whatever reason they're not deemed 'politically relevant.'
LGBT individuals can be imprisoned for practicing their sexuality. But they're not deemed 'politically relevant.'
I'm not sure how we set the baseline of who should or shouldn't be considered 'politically relevant,' but it seems like the notion provides a lot of wiggle room allowing the conversion of any moral comparison of groups into a popularity contest.
"I don't agree with this at all. I think Segregation and Apartheid are just adjectives for a whole catalog of taboo systems (and the associated social/political/economical/etc enforcement schemes) that primarily prioritize exactly what it says on the tin : No race-mixing, of all the varieties your brain can possibly conjure up, in particular."
Well, that's part of the issue. There are a lot of different 'taboo systems.' I worry that saying "everything I don't like is Apartheid" will potentially dilute the term, akin to saying 'everyone I don't like is a Nazi/fascist.'
In the perpetual gradient of lumpers vs splitters, it might be worthwhile to engage in enough splitting that different types and grades of systems can be compared.
At one point you arged that Segregation usually involves severely disparate power relationships, with first and second class citizens. And I agree with that criteria. Under Apartheid, white people control white schools and they control the schools for Black students as well. American Segregation and South African Apartheid were both overseen by a single, dominant group. And it's worthwhile asking if that's what people are going for. If separate systems might be defined as segregationist (which I'm not in favor of) that would imply that all nationalism which significantly restricts immigration is fundamentally a form of segregation or apartheid. And I'm not saying that such a usage would be *wrong,* but it's not how the term is typically used. It casts such a broad net that it would probably catch a good few fish that you wouldn't want caught.
(Or maybe not? I'm making some assumptions here. Perhaps you are a staunch open-borders type?)
Alternately, we could ask: at what point does 'a system' justifiably become 'two systems.' And does 'one system' vs 'two systems, each with their own independent leadership' matter for how we judge interactions?
One reason to make the distinction between supremacists and separatists is that supremacists often are in conflict with those they seek to dominate, while separatists are more likely to come to agreement with their opposite group. The Black Panthers and some white separatist groups sometimes come to an accord, because both groups simply didn't want to interact with the other.
I'm not aware of legal restrictions in Israel on who one can have sex with. While there are not civil marriages in Israel, interfaith marriages are often performed abroad and then recognized as civil marriages by Israel. They are not illegal. There are social taboos, but including social taboos in the definition of segregation also expands the term singificantly, to the point that you'll end up calling most groups segregationist or apartheid. There are some fairly strong selective pressures, even within liberal communities, and most liberals have certain political groups of people it would be socially taboo to have sex with or to marry, as evidenced by outcomes.
"The Nazis didn't want this,"
I agree that they wanted to exterminate Jews and Roma after working them to death and did so in a way that wouldn't have been predicted by strict material selfishness. The Nazis also explicitly wanted the restoration of German colonies. They slaughtered people who they believed would become the next-generation Polish intelligencia in order to better facilitate Polish submission. The Nazis were also rather unusual in that they even tried to prevent emmigration of oppressed groups, as you suggest.
When offered the grab-bag of political horrors, the Nazis reached their hand in more than a few times.
"So the Nazis didn't want the labor of the groups they were segregationists against, how does that counterexample reflect on your definition of segregation/Apartheid ?"
I would say that the Nazis were genocidal towards Jews and Roma. They engaged in segregation (extraction of labor of an oppressed group) as part of their path towards genocide via labor camps. That is, I would say that they used segregation as a temporary means, but not as an end unto itself, as Jews and Roma, specifically, were concerned. The Nazis also announced their intent to restore German colonies, which suggests long term extraction of labor from oppressed groups.
In the American South, on the other hand, segregation and extraction of labor was an end unto itself. Extermination of American Blacks was not a goal. Voluntary expatriation was considered for a time, such as with the establishment of Liberia which you mention.
"or giving them native-american-like reservations in isolated pockets of USA and never having to think about them again."
The American reservations started out as being effective countries with sovereignty, but as time went on the US government increasingly asserted control over them, diminishing their autonomy and property rights. At least as per Vine Deloria Jr's book "Custer Died for Your Sins." I admit to only having a smaller set of sources that I've consumed as regards Native American politics, so I'm more subject to bias on that topic.
Liberia is... complex. There were some Slaveholders who favored Liberia because they did not want free Blacks in the US. As such, they saw separation of free Blacks as strengthening segregation and slavery. There were also some abolitionists who supported Liberia.
The US contained competing political influences, as most polities do. The ultimate failure of Liberia and the institutionalization of segregation suggests to me that the American racial separatists (as I define the term) essentially lost and the segregationists (as I define the term) won.
And then we add the complication that Liberia was not located in the area that slaves came from originally, and was effectively hegemonic and colonial in its own right. But that's a different discussion.
Saudi Arabia is a shitty kingdom whose royal family is in desperate need of a French Revolution. I hate it almost as much as Israel. If it's a popularity contest, the vast majority of Middle Eastern polities are fiercely competing for the last place.
Like, I get what you're trying to say, judged purely by the metric of "How Many Dead Arabs", Israel is not in the top 5 or even the top 10 list. Saudi Arabia royal scum starved 85000 Yemeni children in a 5 years or so according to English wikipedia. This is a common point brought up again and again by Pro-Israeli, and it's a good point.
On the other hand, I'm not very sure what this point should imply, Israel is in the unique position that its grab bag of claims to the land are all illegitimate by different degrees to Arab hearts and minds, ranging from Religious claims to the UN partition of 1948. The religious claim is not bought by the majority-Muslim and minority-Christian Arabs (not to mention the extreme minority of Atheist Arabs, of which I'm part), and the UN partition is perceived as an unfair and compromised by the US and other colonial powers who had outsized influence in the UN till now. The Israeli aggression in 1967 is perceived to have invalidated all the rest of UN legitimacy anyway. The very legitimate and very sympathetic claim of "approximately 700-800 thousand Jews were expelled from Arab countries, Israel is the only place they can go to live" is prominent in Pro-Israeli circles but vanishingly rare in Pro-Arab circles, and whenever I bring it up in Arab circles it, I have not seen a good response, suggesting it's a good entry point to get Arabs to at least recognize that not all Israelis are illegitimate claimers of the land, and maybe that's precisely why it's so underused.
Meanwhile, the Saudi claim to their land is that they're a royal family that managed to conquer it in the 1920s and before by making several tribal and religious alliances. In my secular and anarchist eyes, this is much more illegitimate than Israel, but for some reason the vast majority of Arabs are convinced by it. They don't perceive Saudi Arabia as an illegitimate state, and that colors their view of its genocides (even though it shouldn't, and even the most legitimate of states should not get a free pass or a milder reaction to its genocides).
So, point 1 : Israel, despite objectively killing way fewer Arabs than Saudi Arabia or Syria, is seen as an illegitimate colonizer, while the other genocidal states are seen as legitimate states ruled by corrupt rulers, or even legitimate states ruled by harsh but ultimately fair rulers. I think this is bullshit, but it is what it is.
Point 2 : What does the Pro-Israel advocate actually want after bringing this up ? Do they want to talk less about the genocides of Israel, or talk more about the genocides of Saudi Arabia and Syria ? Maybe the 2 are equivalent from a practical or purely consequentialist standpoint, but the 2 are absolutely not equivalent from a moral standpoint. In most situations where I have seen Pro-Israelis bring up this point, there is usually a coded request of "Don't bring up Israeli crimes until you have finished talking about all the crimes of every single other country in the Middle East ''. It's coded, it's never expressed outright, but I feel they're almost itching to come out and say it, holding themselves by a tight leash.
I'm an Arab, I think Arabs (and every other ethnicity, society, or cultural group) would benefit a lot from self-criticism and a healthy dose of out-groupism. I think the standard human baseline for outgroup affinity and ingroup criticism is way way way way below the optimal. Islam, if nothing else, is one thing taken for granted among the vast majority of Arabs, to a devastating effect on both Arabs and non-Arabs. The pogroms of Mizrahi Jews have to be acknowledged, and ideally set right. The many many many crimes against Kurds and Yzidis and other minorities has to be acknowledged. I think the Rawls' veil of ignorance is the single most important and underused contribution to ethics in the history of ethics, and I think that Arabs (and for that matter Indians, if the jingoism and nationalism of online Indians are anything to go by) would benefit massively from knowing about it and acting according to it for a few centuries. On the other hand, this still doesn't legitimize or take the sting away from Israeli crimes. I don't feel it should.
I would say that, you have to sometimes recognize that people aren't machines and that perfect discursive symmetry (although a laudable goal that I long for in humans) is not achievable. Sometimes you will not agree with the claims people make or you will agree factually but think that they're phrased in a very dishonest way, I feel that way towards the common Israeli claim "The biggest number of Jewish dead in a single day since the Holocaust". I agree factually (although I actually never see someone deep-diving into the 80 years of records and verifying it, but it feels right), I just think it's very dishonest and propagandistic way to phrase the size of Israeli casualties. But nevertheless, I recognize the Israeli/Jewish fear behind it and I don't stress about its dishonesty too much, people will reach for whatever story or narrative structure they're accustomed with to describe the facts, it just so happens that the Holocaust is a particularly important story that every Jew is familiar with, just like Decolonization and armed struggle is a particularly important story that every Arab is familiar with.
> Perhaps you are a staunch open-borders type
I have long ago ceased the practice of putting labels on my views, every single label I have tried to claim, I discover that there is a great horde of idiots who interpret it in a vastly different way than me, that people make weird associations between it and other labels I don't recognize, that it's not saying all the things I want to say, etc...
On borders, my views are as follows :
1- Something resembling borders is not bad in the broadest possible sketch, and it's not apriori ridiculous that people who are born on the land and lived there longer has a privileged claim to the land than people who weren't
2- That said, the current status quo in the vast majority of the world where it's extremely difficult for people to move and build a new life elsewhere is ridiculous and extreme
3-a Culture and Language are elephants in the room in any border discussion, people hate it when those things differ from the one they have grown up with, and furthermore people can't change them very fast. The problem of language learning is intractable, the most prodigious humans we know of don't exceed 40 languages, and the vast majority of language learners don't know more than 1 or 2
3-b States are the elephant in the room in any border discussions, and the pro-border-control side has to acknowledge that huge tracts of lands claimed by states are not legitimate according to (1), as an example about 90% of Egypt's area is desert, the vast majority of Egyptians live on the Nile banks, in what sense can they claim the huge square of desert around the Nile as their own ?
4- The European Union handling of refugee is an unmitigated disaster, one that is very unfair to Europeans first and to refugees second, morally wrong, and one that will forever burn any goodwill that a sane pro-refugee and pro-immigration position might have for the next 150 years
5- I don't know what should be done, it's a massive issue that seems to admits no general solution, accelerating space colonization so that terran land becomes irrelevant is the only easy and simple solution I can think of
Things are hard and complex, and human beings suck.
You can add the Roma to the list of people Nazis didn't want.
> Israel is a pluralistic nation and its citizens have civil rights
But not *equal* rights, I suppose. As the most obvious example, some citizens can invite their extended family to live in the same state, others can't. (I think there are more differences, but I am not really familiar with the situation.)
I mostly agree with what you wrote, and I definitely wouldn't want to have Palestine the next door, but I notice an interesting thing: it seems to me that Palestinians in the West Bank are generally behaving nicer than the ones in Gaza (I may be wrong about this, in which case please correct me), and yet, the former are losing their territory faster than the latter. So I kinda feel like the deal actually offered to Palestinians is something like: "be nice, so that we do not have to fight you... but we will keep displacing you by more civilized means".
> These comment sections are sickening. They're full of people who hysterically freak out if you say anything that remotely resembles white nationalism, and then they turn around espousing violent, genocidal jewish ethnonationalist rhetoric. This is literally like some far-right caricature of jews, but it proves itself to be true over and over again.
Are you feeling better, now that you've gotten this off your chest? What do you think about those posters who are espousing violent, genocidal, antisemitic terrorist propaganda?
Some Gazans blaming Hamas for the Israeli attack. No guarantee they will continue to do so.
https://forward.com/opinion/571232/hamas-unpopular-in-gaza-before-2023-israel-war/
Thanks. It's nice to know that they're actually willing to say that to reporters, and reporters are actually willing to report it to the rest of the world.
I suppose you can always find all kinds of opinions in any country, so I am not really surprised that "some X do Y". The question is, how many, and whether they can do something about it.
Is Hamas to blame for Israeli settlers?
I know this is intended as dumb whatabouttism (the recent war is about Hamas's massacre spree and has basically nothing to do with settlements) but the direct answer is yes, the fact that the one time Israel actually did a large scale removal of settlements the direct result was Hamas coming to power and bombing Israel for the next twenty years is the main reason Israel hasn't dismantled any settlements since.
What a nice and rosy view of Isreal, the eternally oppressed peace dove that keeps trying to make peace and getting rejected by the violent Arabs.
I wonder what beautiful peace-loving quotes Wikipedia has about it ?
> There is no doubt in my mind that very soon the government of Israel is going to have to address the demographic issue with the utmost seriousness and resolve. This issue above all others will dictate the solution that we must adopt. In the absence of a negotiated agreement – and I do not believe in the realistic prospect of an agreement – we need to implement a unilateral alternative... More and more Palestinians are uninterested in a negotiated, two-state solution, because they want to change the essence of the conflict from an Algerian paradigm to a South African one. From a struggle against 'occupation,' in their parlance, to a struggle for one-man-one-vote. That is, of course, a much cleaner struggle, a much more popular struggle – and ultimately a much more powerful one. ****For us, it would mean the end of the Jewish state... the parameters of a unilateral solution are: To maximize the number of Jews; to minimize the number of Palestinians; not to withdraw to the 1967 border and not to divide Jerusalem****... Twenty-three years ago, Moshe Dayan proposed unilateral autonomy. On the same wavelength, we may have to espouse unilateral separation... ****[it] would inevitably preclude a dialogue with the Palestinians for at least 25 years.****
-- Ehud Olmert, Sharon's deputy leader
Mmm ??
> The significance of the disengagement plan is the freezing of the peace process, and when you freeze that process, *****you prevent the establishment of a Palestinian state, and you prevent a discussion on the refugees, the borders and Jerusalem. Effectively, this whole package called the Palestinian state, with all that it entails, has been removed indefinitely from our agenda.****** And all this with authority and permission. All with a presidential blessing and the ratification of both houses of Congress. That is exactly what happened. You know, the term 'peace process' is a bundle of concepts and commitments. The peace process is the establishment of a Palestinian state with all the security risks that entails. The peace process is the evacuation of settlements, it's the return of refugees, it's the partition of Jerusalem. And all that has now been frozen.... *****what I effectively agreed to with the Americans was that part of the settlements would not be dealt with at all, and the rest will not be dealt with until the Palestinians turn into Finns.***** That is the significance of what we did.
-- Sharon's senior adviser, Dov Weissglass
Beautiful.
https://en.wikipedia.org/wiki/Israeli_disengagement_from_Gaza
Still might be shared responsibility on both sides.
I've heard that settlements are subsidized by the Israeli government. Is this at all true?
Yeah. But a lot of this is downstream - basically you have people who are pro-settlements for religious reasons (or just because they like living in suburbs across the green line), and people who are anti settlements for peace reasons. But the second category (which I'm still mostly in) lost all political influence after 2005, because they couldn't argue anymore that removing settlements could help peace (so all we're left with is "yeah but pork spending on subpopulations that needlessly cause problems for us is bad", which is about as successful as American arguments against the Jones act).
Taking a step back here, I should specify that "the Israeli government is funding them" is also a bit more ambiguous than it sounds. The national government funds municipalities as a general thing, so it funds settlements under that (though there are issues where special interests push for this amount of funding to be calculated in a way that disproportionately benefits settlements).
The Gaza withdrawal aftermath was, in hindsight, the nail in the Oslo process’s coffin. Like you said, the side (which I sided with) of “if Israel leaves the occupied territories there will be peace” was decisively obliterated by rocket barrages from Gaza.
The Times (of London) recently published an article along similar lines...
https://www.thetimes.co.uk/article/what-gazans-really-think-about-hamas-ph78wz22b#:~:text=Nearly%20half%20of%20people%20surveyed,corruption%20in%20Hamas%2Drun%20institutions.
This is Chesterton's fence: https://media.thegospelcoalition.org/static-blogs/trevin-wax/files/2016/09/Trevin-Wax-at-Chestertons-home.jpg
That is, the fence around the house where G. K. Chesterton lived. It's more of a hedge but I guess it's all right.
Ironically, some articles claim the house was under threat of demolition. Luckily it's still standing https://maps.app.goo.gl/dHM7QRXRhsGHifnV7.
I wonder is the oak tree still in the garden? There seems to have been a lot of extra building on since his time, but it was happening even in his time (at least by his autobiography):
"I had been living, already for a long time, in the town of Beaconsfield in the County of Bucks; the town which some Colonials imagine to have been named after Lord Beaconsfield the politician. It is rather as if they thought that England had been named after Mr. England the pirate. I am almost tempted to add that I say it with an apology to pirates. I do not know for certain why Disraeli took his title from Beaconsfield, which he scarcely ever visited, rather than from Hughenden, where he lived. But I was told by old Lord Burnham, the founder of the Daily Telegraph, that (as the story ran) he had chosen the title originally intended for Burke, who did live at Beaconsfield and whose legend still clings in many ways to the place. Mr. Garvin, the editor of the Observer, lives in what was once the house of Burke's agent and the oak-tree in my own garden was one of the line of trees that marked the limit of this land. I am glad that Mr. Garvin fits into that political landscape much better than I do; for I admire Burke in many things while disagreeing with him in nearly everything. But Mr. Garvin strikes me as being rather like Burke; in his Irish origin, in his English Conservatism, in his eloquence and gravity and something that can only be called urgency of mind. I once suggested to him that he should appear at a local festival as Burke and I as Fox; a part for which I have no claim except in circumference. But I hope there will never come a dark and difficult hour when political differences become personal, and Mr. Garvin begins to throw daggers about and say that our friendship is at an end.
I have lived in Beaconsfield from the time when it was almost a village to the time when, as the enemy profanely says, it is almost a suburb. It would be truer to say that the two things in some sense still exist side by side; and the popular instinct has recognised the division by actually talking about the Old Town and the New Town. I once planned a massive and exhaustive sociological work, in several volumes, which was to be called "The Two Barbers of Beaconsfield" and based entirely upon the talk of the two excellent citizens to whom I went to get shaved. For those two shops do indeed belong to two different civilisations. The hairdresser of the New Town belongs to the new world and has the spotlessness of the specialist; the other has what may be called the ambidexterity of the peasant, shaving (so to speak) with one hand while he stuffs squirrels or sells tobacco with the other. The latter tells me from his own recollection what happened in Old Beaconsfield; the former, or his assistants, tell me from the Daily Mail what has not happened in a wider world. But I suggest this comparison, merely as an introduction to a parallel matter of local interest; which happens to embody, better perhaps than any other emblem, all those large matters that are more than local. If I wanted to write a book about the whole of this great passage in the history of England, including the Great War and many other changes almost as great, I should write it in the form of a History of the Beaconsfield War Memorial.
The plain primary proposal was that a cross should be set up at the cross-roads. Before the discussion was half over there had entered into it the following subjects of debate: (1) The Position of Woman in the Modern World, (2) Prohibition and the Drink Question, (3) The Excellence or Exaggeration of the Cult of Athletics, (4) The Problem of Unemployment, Especially in Relation to Ex-Service Men, (5) The Question of Support for Hospitals and the General Claims of Surgery and Medicine, (6) The Justice of the War, (7) Above all, or rather under all, for it was in many ways masked or symbolically suggested, the great war of religion which has never ceased to divide mankind, especially since that sign was set up among them. Those who debated the matter were a little group of the inhabitants of a little country town; the rector and the doctor and the bank manager and the respectable trades-men of the place, with a few hangers-on like myself, of the more disreputable professions of journalism or the arts. But the powers that were present there in the spirit came out of all the ages and all the battlefields of history; Mahomet was there and the Iconoclasts, who came riding out of the East to ruin the statues of Italy, and Calvin and Rousseau and the Russian anarchs and all the older England that is buried under Puritanism; and Henry the Third ordering the little images for Westminster and Henry the Fifth, after Agincourt, on his knees before the shrines of Paris. If one could really write that little story of that little place, it would be the greatest of historical monographs.
The first thing to note, as typical of the modern tone, is a certain effect of toleration which actually results in timidity. Religious liberty might be supposed to mean that everybody is free to discuss religion. In practice it means that hardly anybody is allowed to mention it. There is a further qualification of some interest; that in this, as in many things, there is an immense intellectual superiority in the poor, and even in the ignorant. The cottagers of the Old Town either liked the Cross because it was Christian and said so, or else disliked the Cross because it was Popish and said so. But the leaders of the No-Popery Party were ashamed to talk No-Popery. They did not say in so many words that they thought a Crucifix a wicked thing; but they said, in any number of words, that they thought a parish pump or a public fountain or a municipal motor-bus a good thing. But the greater number of them tended to the proposal of a Club Building, especially for ex-service men; where the latter could have refreshment (that is where the Drink Question came in) or play games (that is where the Athletic Question came in) or possibly even share the Club on equal terms with their wives and women-folk (that is where the Wrongs of Women came in) and generally, in fact, enjoy all that we should desire ex-soldiers to enjoy, if there were really any chance of letting them do so. The scheme was in that sense admirable; but, as it proceeded, it became almost too admirable, in the original Latin sense of astonishing. Those who had propounded it called themselves, I need hardly say, the Practical Party. They justly condemned us of the other group as dreamers and mystical visionaries. They set to work to draw up their plans for the Club; and they were certainly plans of the most magnificent completeness. There were to be cricket-fields and football-fields and swimming-baths and golf-courses, for all I know. The incident has a primary moral, with reference to that strange modern notion about what is practical and constructive, which seems merely to mean what is large and largely advertised. By the end of the controversy the plan of the Practical Party had swelled to the ends of the earth and taken on the dimensions of Aladdin's Palace. There was not the remotest chance of collecting subscriptions for such a scheme; at the rate it was developing it might run to millions. Meanwhile, the vision of the mere visionaries could be realised easily for a few hundred pounds."
I am writing a long short story set in the near future (circa GPT-6.5) that features an AI tutor. As a non-technical person fascinated by recent advances in AI, I came up with the following phrase to describe its mind: “the ever-expanding matrix of data points, meaningless in isolation but each quivering with its own unique probabilistic relation to every single other, that served it for a mind.” Question for those who know a lot more than me: does this seem to you to evoke something (potentially?) real? How could it be improved without becoming more technical/verbose?
Not a professional AI dev and I have no idea how much you know, but describing a data point as "quivering with its own unique probabilistic relation to every single other" sounds wrong to me on an intuitive philosophical level, the probabilistic relation is entirely in the eyes of the observer, it's entirely imposed by the network on the data, it's like when you have a bunch of stars in the night sky and you draw lines through them to make them resemble animals and other shapes, the stars aren't quivering with relations to one another, only your eyes wants them to for the purposes of remembering them better. (But you could argue that each data point is in fact merely one revealed sample from the hypothetical and opaque "Target Function", which the AI is trying to approximate by fitting its parameters as close as it can the samples it can see. In this sense, every data point is a "clue" or a "piece of evidence" that holds the key to unlocking all the other data points and indeed the hidden process generating the data points. Describing this as "Quivering" is still weird, maybe use "latent" or "pregnant" instead.)
Also, the data points an ML system is trained on are not its mind, they're the thing it's mind used to get good at finding patterns. The mind of an ML system is the architecture + the weights. You ***could*** argue that the model weights are essentially just an extremly efficient and lossy compression of the training set and that would be a valid and fruitful point that many people made, but it still sounds very wrong to say that data points form the mind of anything.
To be clear, are you saying the “network” that’s “imposing” the probabilistic relation here is the neural network/AI architecture itself? If so, I *think* we’re on the same page: I’m conceiving the Tutor’s mind as being the matrix of probabilistic relations (for “matrix”, read “network”; for “probabilistic relations” read “weights”) as opposed to the data points themselves. And I’m imagining a future model that’s capable of reweighting itself in response to external stimuli (hence “quivering”). I suppose my follow-up question would be, do you think “probabilistic relations” conveys what model weights are? Or am I hopelessly at sea here?
Right, by 'network' I mean the neural network. But your phrase said "A matrix of data points [...] that served it for a mind". If by "matrix" you just mean a neural network then you're right, but then it would be inaccurate to say that the matrix is of data points, it's not, it's a matrix of arithmetic, a big huge network of sums and multiplies and special operations called activation functions. The data points are the input to this matrix, and its outputs are repeatedly tuned (via the tunable weights) so that they match the known, expected output, at this point it's said (or hoped) that matrix is "trained", it has mastered deriving the answer from the question.
In any case, the data points don't "conceptually" feel like they're part of the model, they're part of the external stimuli that we hit the model with. The "matrix" is of weights, or in your language, of probabilistic relations.
> do you think “probabilistic relations” conveys what model weights are
It wouldn't be an inaccurate way to put it, but I personally wouldn't describe them as such. I view weights as the parameters for an approximation function that attempts as much as possible to stick to the curves and the contours of an unknown hidden function or probability distribution that we only have samples of. If I had to describe them poetically, I would say they're "Inner Structure" or "Hidden Patterns". They're the tea leaves arranged in such a way that they mimic the shape of phenomena, such that by reading them in a certain way you can get insight into the phenomena.
Thanks, this is great. The “of data points” is the real howler here, then. Interesting challenge to express what’s actually going on (insofar as it’s known/comprehensible) in poetic terms! Thanks again for taking the time to tease this out.
You're very welcome. Always been a scifi fan, especially the hard kind that tries to give textbooks a run for their money. Always happy to help with what little I know.
I built my own artificial neural network once as part of an Intro to Machine Learning course, and I think it mostly makes sense except for one thing: the matrix isn't ever-expanding, GPT does not grow once deployed. Maybe replace "ever-expanding" with "inscrutable"?
In the story, the AI tutor is multi-modal, with detailed live input from at least three senses, and explicitly billed as “growing alongside its user”; I’ve intentionally left the mechanics of this unstated (mainly because, like I said, no technical background), but I guess (and I do mean *guess*) this could either work in terms of an ongoing context window that remains open for years, constantly tweaking the persona being evoked from a (stable, unchanging) model; or a model whose weights adapt dynamically according to user input, sort of a very much more sophisticated, and ongoing, version of current RLHF techniques. Does any of this sound at all plausible as something that might happen within the next 5-10 years?
Thanks: this is very helpful!
See, in any of those scenarios you are putting forward, the AI is not ever-expanding, as that would mean it increasingly takes up more space, which I'm pretty sure is not something anyone is aiming for. If you talk of an ever-expanding matrix, you are talking of something that you need to keep throwing more storage at, which could maybe be a plot point. I think if you go with "ever-improving" or "ever-learning" it gets the same point across without offending any engineers (it would be a pretty bad piece of software, one that never stops growing).
But those suggestions I put forth maybe seem too abstract when applied to something as concrete as a matrix, so maybe "ever-churning" is best, as it really captures a very real and chaotic aspect of this matrix.
I also don't understand the technical details, but would it be plausible for the AI tutor to be a cloud service that can dynamically get more memory allocated, using a strategy like "at first, allocate minimum memory that is enough to make the end user happy, later expand when you get feedback that this is no longer enough"? The feedback would be provided either by user or by the AI itself (it would somehow reflect on its own answers).
I think I would need to be an AI researcher to truly answer that one, but it is sci-fi, so sure, maybe the idea of tossing more parameters for specific users is valid sci-fi. Interesting architecture, like you have your static main model that provides most of the capability, then there are add ons for specific users to fine tune the thing.
I definitely thought of it as a cloud service. And the kid’s mother is a developer (“Summoner”; I’m leaning quite hard into the idea that certain aspects of working with future models will involve very different skill sets), so could plausibly have access to privileged amounts of memory/parameters/add-ons.
Okay, this is funny:
https://www.tumblr.com/determinate-negation/735078988178931712/these-might-be-two-of-my-favorite-images-in-this?source=share
Yeah, there are lots of wanna-be gangstas in European cities and yes, we do have inner-city and 'urban' youth who come from areas of high crime and deprivation, even in Ireland, but it's never not funny to me that they faithfully copy American hip-hop/rap culture trying to be what they're not (i.e. East or West Coast).
It's even funnier when, as in the linked images, it's guys from nice places.
I similarly cringe at "gangstas" in post-communist Eastern Europe, describing how they grew up as "rebels" in a "ghetto"... when in fact the regime had mandatory employment for everyone, and the police would beat you up for dressing differently or otherwise sticking out of the crowd. Dudes very likely had exactly the same childhood as everyone else.
Of course, the real urban youth 'gangstas' in Europe look very similar to the American ones, what a weird coincidence.
Otoh I'd guess that your average American would find, say, Moldovan hip hop just as funny, even though Moldova is considerably grimier and poorer than what even the minority communities in US face.
I remember in my high school German class the teacher played us some German hip hop, probably trying to appeal to the youth, and I found it incredibly silly. Maybe German just doesn't sound good as a language for rapping, though.
I had the same experience, with this song: https://www.youtube.com/watch?v=VUosAGDM8Sg
The teacher didn't show us the video, though, so we didn't get to appreciate the dancing.
That sounds like a human rights violation and a case for the European Court of Human Rights (though that's my prejudice against the German language rearing its head) 😁
You know I had to look it up, and yeah - not the best choice:
https://www.youtube.com/watch?v=rjphIb_HpT4
There's a Belfast Irish-language rap group (well at least they're keeping the native language alive) and they're something more grounded; at least this is authentically depressing in an Irish way:
https://www.youtube.com/watch?v=KSnF7RaeoXE
https://www.youtube.com/watch?v=L9TJMrKpe0k&list=UULFR8_tdf_kr2tn01dFmRK6wQ&index=17
Any thoughts about getting good at simple navigating in a city? I had an acupuncture appointment in an unfamiliar part of the city, and I spent a lot of time being lost. (It seemed like a lot of time, actually about an hour and a half.) Some of this may be aging, but I've always been bad at that sort of thing.
Some of it may have been bad directions, some of it may have been failing to follow good directions.
There was possibly some wrong-headedness involved. "Go across the bridge", they say. I can't believe it's this big elevated four-lane highway with no obvious way for a pedestrian to get onto it. Surely there must be a little footbridge across the river in the park. No, it was the big bridge-- there was a not-easy-to-see pedestrian access from where I was.
I suspect I need better skills using the maps on my phone, and possibly I should carry a paper mapbook.
I'm not sure I reliably stay conscious of orientation when I turn a corner.
I'm especially interested in accounts of getting better at orientation starting from being bad at it, but if you've been good at orientation all your life and you feel a need to tell me, go ahead, I can't stop you.
If you've actually taught people to be good at orientation (not just have ideas about how you might be able to teach it), I'm interested.
Some might be interested in where I was. Being interested in that sort of thing probably has something to do with being good at orientation. I'm in Philadelphia. I was starting from 10th and Mountain St. (South Philadelphia). I was going to West Philadelphia Community Acupuncture at 4636 Woodland Avenue.
I caught a bus to 33rd and Dickinson. So far so good. It was cold, and it was taking a while for the next bus to show up, so I thought I could walk it. It was only a mile or so. I found myself at a park involving the Schuylkill River. There was a giant FedEx building with inconvenient fences to get past before I could get to the big bridge. Somehow, I was at 49th and Paschall St. I'll note that this an area with few shops and few people on the street to ask for directions.
It turns out there's a trolley that goes from Center City to near the acupuncture place, which should solve the specific problem in the future, but I'm still interested in what might solve the general problem.
I've had some success in improving my navigation the way I improved my time sense. The basic idea is to explicitly make a prediction, and then quickly check it (by looking at a map or clock), and repeat the process a lot. Eventually my predictions got better. And the other way around, too: for the time sense, I would always make a prediction before I looked at a clock.
I am not ashamed to use navigation software even when walking in my own city. Even if I could get somewhere without it, sometimes it shows me a shorter way. I don't use the voice navigation, just look at the map every few streets, where I am now and where I need to go next.
When we are on a vacation, I try to walk around by foot as much as possible. It is my experience that when I walk I get better at understanding what is where, but when I use any kids of transit, I only remember disconnected points without knowing how they are related to each other.
I was trying to use navigation software, and not doing well with it. Possibly there are some basic skills for navigation software that I need, and possibly I was too impatient to use information which was there.
My favourite navigation software is "Mapy.cz" which uses data from OpenStreetMap, but also provides interesting local information, for example tourist maps. Another advantage is that (unlike Google Maps) it shows your position on the map even when you do not have an internet connection.
https://play.google.com/store/apps/details?id=cz.seznam.mapy&hl=en_US
I turn on "Position" on my Android, then start the application. Sometimes it takes a while until it displays my position on the map (if I do not have an internet connection, it can be minutes, otherwise it is seconds).
Okay, now I know where I am, but do not know which direction is which, i.e. where is North. One option is to start walking in a random direction and later check how the position marker is moving. Another option is to read the names of the streets and compare to the map.
I always keep the North on the map pointing upwards. Rotating electronic maps are confusing. (It is the other way round with *paper* maps; I prefer to rotate them so that the direction I am walking is upwards. The difference is that with paper, even if I rotate it, I can still intuitively perceive the positions and directions from the perspective of the paper. With software, it is not me who rotates the map.)
As I walk, I take a look at the map approximately once a minute. If there is a dense grid of streets, I just keep walking approximately in the direction towards my target. If there are long streets that are not transversely connected, I need to plan my walk in advance. Sometimes I ask the navigation to give me the shortest path to the target, sometimes I make the decisions myself.
(You need to check whether the navigation is in the "car" or "walk" mode.)
I generally like to walk by foot. Unless there is time pressure, I would rather walk for 30 minutes than take a bus that takes 10 minutes. (If you add walking to/from the bus stop and waiting for the bus, you find that taking the bus actually saves less time than it seems.) I live near the center of a city, so the 30 minutes range actually covers a lot. (I also use the navigation software to tell me how far something is.)
It seems to me, statistically, that women suck at orientation, but I have no idea how much of that is innate, and how much is learned helplessness (potentially motivated by a lack of interest). One difference I noticed is that my wife needs to be 100% sure to choose the right direction... which she can't, and therefore she ends up asking people; while I am more like "if I walk by foot, the worst thing that can possibly happen is that I will walk one street in a wrong direction, which is not a big deal". (I often make a good guess based on incomplete information, but the point is that I am not afraid to make a guess.)
Also, it seems to me that men often describe their paths "mathematically", like "go straight ahead three blocks, then turn left and go one more block", while women often describe their paths "contextually" like "go straight ahead until you see a garden with beautiful flowers, then turn towards the nice trees". The male version works better with maps (unless you use Street View on Google Maps).
Note that Google Maps does show position when offline these days. I recently spent a month travelling around Europe without a data plan - selected a route while on Wifi, then GPS continued to track my location until I got to the destination (admittedly not that accurately around tall buildings).
I generally find that Google Maps has the best data on public transport, although maybe other services have caught up since. Note that it doesn't allow time to buy tickets, so include extra time if you don't have a pass.
> women often describe their paths "contextually"
FWIW, I've found this type of direction to be not only more common, but much more useful, when navigating rural roads in America. You can also do directions by road name, like "take US 123 N, then take a right onto state route 45 E, go 30 minutes and take a left onto state route 67 N". But a lot of those road names are relatively new, and when talking to older generations you get directions like "take a left at the Methodist church, go over the creek and past the white water tower, take a right after the courthouse, then another right before the old Stevenson place".
Thank you. I'll check what map apps are available for my iphone. One of my worst problems with maps on my cell is getting it to show my where I am (I do have location information turned on-- that is, I keep turning it on).
I've heard that part of the gender issue might be that boys are given more freedom to run around at an early age, while girls are held close. I'm not sure what age that would be. Early toddler? Maybe even crawling?
This could make a difference because people remember routes much better when they control their movement.
I somewhat agree about buses taking longer for short trips, though note I got into trouble in west Philadelphia because I was too impatient to take the bus.
My impression is that there's a risk of both men and women overing directions related to landmarks that no longer exist, but maybe that's just from jokes.
I often enough have to accept setting off in the wrong direction and then checking, but note that impatience is one of my issues. I also don't walk very fast, so extra blocks can add up to being late.
Learned helplessness reminds me of the discussion of gender and throwing from _The Frailty Myth_. How people learn to throw has been studied. Athleticism for girls and women is more accepted these days, but there used to be a strong concept of "throwing like a girl". "Throwing like a girl" is actually throwing like a person who's new to throwing. So it was urgent to get a boy (whether by teaching or mockery) to not throw like a girl, but if a girl threw like a girl, well, what do you expect?
> boys are given more freedom to run around at an early age, while girls are held close.
I would assume it is more like: boys are more interested in objects (also: cars, that seems related to roads and maps), girls in people and animals.
> I also don't walk very fast, so extra blocks can add up to being late.
Just enough so that the dot representing your position moves clearly in some direction. Maybe 20 meters is enough.
"I would assume it is more like: boys are more interested in objects (also: cars, that seems related to roads and maps), girls in people and animals."
We really can't tell which it is if parental behavior is significantly different.
For orientation, it looks like Philly street numbers (appear to) go up as you get further from the Delaware River. I'd suggest using that to figure out whether you're going away from, or towards, the river (and then you know which direction is north as well). The east-west streets don't seem to have any pattern to the names, unfortunately.
I do exactly the same in the Manhattan grid - higher numbered streets are north, higher-numbered avenues are west.
For the specific situation you describe (33rd and Dickinson to the acupuncture place), I don't see any nearby pedestrian-friendly routes (based on satellite and street views). Probably your best bet was to stick with the bus.
They do, but that wasn't a very grid-like area.
Did taking Gaza and the West Bank after the 6 Day War make sense for Israel? I'm sure it made sense at the time to have borders that looked more defensible, but having borders which include regions of hostile people-- people who have been persistently provoked-- give a military advantage?
I'd always figured that the main reason was to ensure that Israel controlled the borders between Palestine and other countries, and thus the flow of weapons and fighters into Palestine. It's not perfect, of course, but it could be a whole lot worse from Israel's perspective, like if Gaza could launch rockets that were up to modern standards.
It's land. The whole thing, to this day, has always been about land. 100% they would do it all over again.
They probably wish now they had done harsher ethnic cleansing at the time, though, like driving the Gazans into Sinai.
I think taking them made sense. Egypt and Jordan were hostile and pushing back the borders towards their main population centres and away from yours is very sensible.
Keeping them is much more debatable. Trading Sinai back to Egypt as part of a peace deal was a good outcome, but they probably should have given Gaza back to Egypt at the same time. Equally, once a stable peace with Jordan was achieved I think it would have made sense to return control of the West Bank to them (and arguably it still would make sense).
They tried to give Gaza back to Egypt as part of the peace deal - Egypt point blank refused (and was willing to blow up the peace deal over it).
I don't think any such offer was made re the west bank (which would be more awkward since Israel does want some disconnected chunks of it), but if it was I doubt Jordan would take it either.
Thinking about it, if the Israelis had been more thoughtful about having a more defensible country, they should have been *courting* the Palestinians.
The 6 day war cannot be understood unless as a war of expansion. Any other objective doesn't make sense, why else would you take Sinai, 6% of Egypt's area, 3 times that of **current Israel** (so including the West Bank) for its then-population of less than 1/3 of its current population ? How would security increase when you take this vast West Virginia-sized desert plain ? More area to defend and more borders to be attacked from ? it doesn't.
It makes total and perfect sense when understood as a war of a Zionist flavor of Lebensraum. Here are the Likud party for example describing why it wants Colonists and regards them as an element of its National Security :
> Settlement, both urban and rural, in all parts of the Land of Israel is the focal point of the Zionist effort to redeem the country, to maintain vital security areas and serves as a reservoir of strength and inspiration for the renewal of the pioneering spirit.
From https://www.jewishvirtuallibrary.org/original-party-platform-of-the-likud-party.
Look at the words used, "redeem", as if the land is dirty or sinful when not inhabited by Jews, a "reservoir" of strength and "renewal", does that remind you of anything ? Who else in the 20th century looked at population and demographics like that and used those words ?
56 years later it's mostly a failure because Egypt took back Sinai and International Pressure took back Gaza, while the Zionist settlement of the kind that established Israel prior to 1948 began to look more and more like the settler colonialism it is, increasingly unpopular as the 2nd half of the 20th century unfolded. The settlers in the West Bank currently number at 7-8% of all Israeli Jews. The Golan Heights remained with a meager population of < 6000 Jew, which the Likud approved a plan to boost and bankrolled 317 million dollar in 2021-2022 https://www.timesofisrael.com/this-is-our-moment-israel-okays-major-plan-to-boost-golan-double-its-population/.
So asking why Israel invaded in the summer of 1967 is a bit like asking why Hitler invaded the Soviet Union in the summer of 1941, that was the whole point all along.
"why else would you take Sinai, 6% of Egypt's area, 3 times that of **current Israel**?"
You mean, aside from setting up a defensible position on the east bank of the Suez Canal, a formidable natural barrier between Israel and one of the nations plotting to destroy it, and adding a couple hundred kilometers of strategic depth that you can easily afford to yield if necessary? I mean, what does the Yom Kippur war look like if the Egyptian Army starts out just across from the Eilat-Gaza line?
You're not even trying, are you? I mean, GPT-2 trained on a steady diet of Arab propaganda could do better than this.
> a formidable natural barrier between Israel and one of the nations plotting to destroy it
Yes yes, you're actually right. And Nazi Germany really didn't want to invade the Soviet Union to commit genocide of the Slavs or to demographically replace the inhabitants, they just wanted the Ural mountains as a formidble natural barrier between Germany and one of the nations plotting to destroy it, and the couple of hundreds of kilometers of the vast East European Plain as Lebensr... - uh sorry Strategic Depth - they can easily afford to yield if necessary. I mean, what does Operation Kutuzov look like if the Soviet Army started from occupied Poland demarcation line ?
Perhaps seen in this light, October 7th might actually have been a stroke of genius on part of Hamas against a nation that plotted to destroy it, they added a couple of tens of kilometers as Strategic Depth that they can easily afford to lose if necessary. I mean, what does the Gaza invasion look like if the IDF started right at the Gaza's borders and without the hostages as leverage ?
> GPT-2 comparison
So witty and original.
I read John as expressing exasperation with your argument style, which appears to be to start with the premise that Israel is evil, and proceeding to fit the available evidence to that premise, in order to prove that premise.
You responded to his exasperation by arguing that if Israel wanted to be the evil it is, it could conquer a large stretch of territory and claim it's only doing to establish strategic depth while it quietly revels in the genocide you suggest they really want to do. But you never seem to get around to proving Israel is evil; all you did is say "if they *were* evil, it could look like what we're seeing". You never ruled out "if they were good, it could also look like what we're seeing", and then moved on as if you had.
> So witty and original.
That's just it. Circular arguments have been around for thousands of years.
Well first off, I'm quite exasperated with lots of Pro-Israel commenters too. I don't remember calling them literal NPCs (and of the outdated, low quality kind). When you're discussing thousands dead and tens of thousands injured and millions displaced, it's inevitable that you will get mad, most likely for the side you identify with the most and against the other one. It's difficult, but worthwhile, to restrain oneself from calling the other side dumb, delusional, or evil, even though it can feel sometimes they're 100% that, and they could in fact be exactly that. So Mr. Schilling started it off on quite the wrong foot by saying I'm worse than a GPT-2 instance, even though I'm in fact a much more power-efficient and multi-modal machine than that. This probably didn't help me see his argument or his beliefs in the best light possible.
> appears to be to start with the premise that Israel is evil
I start off with the premise that Israel is evil because Israel is evil, this is my conclusion that I formed over years and years of seeing and hearing the evil that Israel does, the sheer mustache-twirling genocidal intent that you wouldn't believe in a dialogue written by a bad comic artist, coming out of their senior government officials again and again and again. I can no more enter an Israel-Palestine argument with the serious or earnest assumption that Israel is remotely good as you can enter a WW2 argument with the serious or earnest assumption that Nazi Germany is remotely good. Apologies for the overused and boring Nazi analogy, tell me something else you hate as bad as the Nazis and I will stop Godwinning.
And yet, I try as much as humanly possible to not let this conclusion color the **reasoning** I give as to why Israel is evil. You could start with a conclusion for the "wrong" reasons and still obtain the "right" reasons for it anyway. Mendeleev claimed he saw the Periodic Table in a dream, and George Cantor based his hierarchy of infinities on Christian Theology. I'm not as intelligent as Cantor or Mendeleev, far from it, I'm just illustrating the general strategy of "Start with a conclusion that you know to be true even though nobody else will possibly believe you, find the evidence that people will believe, and establish the necessary and believable trail from the evidence that everybody will accept to the conclusion that no one will''.
You're very free to say my provided reasoning is wrong and explain why, but starting from a conclusion is not apriori a bad thing to do.
> it could conquer a large stretch of territory and claim it's only doing to establish strategic depth while it quietly revels in the genocide you suggest they really want to do
You seem to misunderstand, I'm not saying it *could*, I'm saying it *did*. I'm satirizing Mr. Schilling's Russell conjugation of Israel's action, "Uhm no, they did not invade the West Bank, the Golan, and Sinai because they're a genocidal state hellbent on ethnic cleansing as much of the middle east as possible of Arabs with as little consequences as possible, no no no, they just wanted Strategic Depth you see. They're justified in this because.... uh **checks notes**... the Arabs were plotting to destroy them, starting a war because you suspect the other side is thinking of destroying you is totally a legitimate and justifiable thing"
So I'm using the exact same mental toolkit that I see Mr. Schilling using to justify Israel's actions, and turning it around to justify the Nazi invasion of the Soviet Union, in an attempt to show him that his argument Proves Too Much [1].
> you never ruled out "if they were good, it could also look like what we're seeing"
I view this as so patently ridiculous and implausible that anyone claiming it is the one who has the burden of proof on their shoulders, not me. "Good" states don't start a war and invade their neighbours because "They were plotting to destroy me".
[1] https://slatestarcodex.com/2013/04/13/proving-too-much/
I can see how you might be genuinely exasperated at pro-Israel comments. I can also see how you might be just saying that in order to gather attention. But if I assume the former, I still notice you're acting as if your exasperation entitles you to being able to claim what you claim and demand the other side prove otherwise. This doesn't work if the other side's evidence is already well established (e.g. commonly understood principles of warfare, history of Israel, history of the Six Day War, and reports of Hamas having assaulted a non-military event and killed a lot of civilians, among others). It doesn't matter how long you spent coming up with your unorthodox views. It especially doesn't matter if, in the course of developing those views, your reasoning requires everyone else to forget what they know about how wars are fought. Even if it turns out you're correct, you're necessarily up against a lot of established evidence, and it's foolish to expect everyone who knows that stuff to drop it immediately as soon as someone comes into the room claiming "everything you know is wrong".
Starting from your conclusion is even *more* patently ridiculous than you seem to think my claim about Israel behaving conscientiously is, for reasons you ought to know if you're familiar enough with rationalist literature to be able to dig up an SSC reference.
So no, you can't use the same mental toolkit. And you keep acting as if you can, and that's wasting everyone else's time.
Oh come on, just go away already. You are flatly lying about easily-confirmable facts which is probably the _least_ tiresome part of your broken record.
Sadly Substack's "mute" function only partly works but you are well worthy of it anyway.
Maybe they'll fix that some day, we can hope.
Hey, don't be so sullen, look at the bright part : you can bully substack into fixing their shitty software using me, tell them someone is being Anti-Semitic on the Internet, that always works.
Please make them fix the atrocious load times while you're at it, tell them Anti-Semites are using the time to read and contemplate Mein Kampf on every page load.
In retrospect it was a mistake, but I can understand at the time why they thought it was a good idea (and Jordan and Egypt didn't come to the table for peace negotiations until like a decade later, by which time they felt good and rid of the WB and Gaza and refused to take them back even when offered).
Taking the West Bank is definitely a defensive win for Israel. I’m not a fan of the settlements though, probably Israel just needs to control the border area along the Jordan along with a large number of military roads. Without that Israel could be bifurcated in an hour by a hostile force.
Gaza didn’t seem that risky at the time as the population was small. It’s probably preferable to a border with Egypt.
Tennyson is one of my favorite poets. Nobody (including myself!) reads his longer "epic" poems these days, and probably hasn't for many decades past. But as a rule of thumb if a Tennyson poem fits on one page then it is probably pretty good. Examples include "Break, Break, Break", a sad poem about the death of his friend Arthur Hallam, and possibly his best known poem, The Charge of the Light Brigade", recited quite well here, with extracts from the 1968 film:
https://www.youtube.com/watch?v=S93lvQ4Ukg8
Here's a short quote from his poem Ulysses, which should be encouraging for old timers who worry that they're washed up and may never achieve anything more!
"Old age hath yet his honour and his toil;
Death closes all: but something ere the end,
Some work of noble note, may yet be done"
Alfred Tennyson
The Eagle was the first poem I ever read that I "got". Tennyson is my one of my favorites too.
His short poems have a very simple rhyming structure. This isn’t a condemnation, mind you.
I think P(Doom) from AI is very low, < .01. over the next thousand years. But even if I thought it were around .05 or twice that, I would be in favor of going full speed ahead with AI.
There are those who consider human flourishing to be this amazingly great thing. If so, it is because humans in the past have been brave, bold, intrepid. It is because they plunged themselves into mysterious domains without fear.
The argument for being extra cautious about AI is an argument not that humans are precious but that they aren’t any longer worth a shit. That they have become a race of cowards.
>It is because they plunged themselves into mysterious domains without fear.
They plunged THEMSELVES into mysterious domains, they risked their lives, not all of the future of humanity.
>The argument for being extra cautious about AI is an argument not that humans are precious but that they aren’t any longer worth a shit. That they have become a race of cowards.
I cannot be the only one here really, really sick of these platitude based arguments against AI caution.
What if the probability of extinction for humanity were 50%? Would all of these platitudes still hold? No? Then it means the platitudes aren't what you really believe - the math is what actually matters. And if you're willing to flip a coin for humanity's future then there's no point discussing this.
i dont get why P(doom) or P(heaven) is a thing; the danger is AI becoming an extremely efficient way to disenfranchise labor by allowing capital to capture its value at a fraction of the cost. Our society couldn't handle Covid forcing out labor for a brief time; we arent set up for a "retail world" with a majority part time, just in time workforce.
if there is a plus side what scares me is no one is articulating it. what jobs can AI create?
Growing economies tend to create jobs. We might not be able to predict what all the new jobs will be; plenty of jobs exist today which few would have foreseen 75 years ago.
Not for horses.
Horses had just one job! They couldn't handle it, so we had to let them go.
The whole point of AGI is that it can do whatever a human can do cognitively (and more). By definition that does not lend itself to job creation.
Sort of a circular argument. AGI will do everything, thus humans will do nothing.
Even with some version of AGI, it's not clear that we want AIs doing all the jobs or that it's economically feasible to have them do all the jobs.
And if AIs do everything for us at some point, the upshot is a society that's rich enough that we don't have to do anything for compensation anyway.
Bar tenders will be ok.
I recommend https://www.astralcodexten.com/p/pause-for-thought-the-ai-pause-debate . The question mostly isn't "create AI" vs. "never create AI", it's "how much do we balance speed vs. caution". I don't think driving less than my car's max speed of 130 mph makes me a coward, and I don't think attempts to balance caution vs. speed in AI make us cowards either.
I agree it would be disappointing if we never developed AI.
Does anyone know why Kalamazoo and Numazu are sister cities? Did they just pick each other due to the name similarity, or was that coincidence?
Has anyone done the research about kidney donation for ≈fifty-year-olds? Is there a significant difference in benefits (to the recipient) or risks (to the donor) versus the case of a younger donor?
https://journals.lww.com/transplantjournal/Fulltext/2007/03150/Early_versus_late_acute_rejection_episodes_in.00015.aspx find no statistically significant effect (although I can sort of eyeball an effect) of donors being > 50 on long term mortality (of the recipient). They do find an effect of donors being > 65. https://www.sciencedirect.com/science/article/pii/S1600613522279763 is a bit more annoying to interpret by I think mostly confirms that result.
I'm kind of reading between the lines here, but I would think of this as a combination of two factors. First, the kidney ages and dies at the same rate as the rest of the body, so if you're older than your recipient, their kidney will be older than the rest of their body and more likely to fail. The average recipient is 45 - 50, so it probably gets worse as you go significantly beyond that. Second, younger kidneys are probably better in general - I don't know if this exactly tracks GFR, but if it does, the statistics are at https://www.researchgate.net/figure/Age-and-gender-specific-GFR-reference-ranges-based-on-measured-GFRs-from-2974_fig2_329127776 . But age related decline after 35 is very slow and at age 50 it's still 90% of what it was in youth. Overall I think low advantage of younger over older before 50, though there's probably some just on priors.
I don't see a good reason why the health at 60 of a person who donates at 50 should be different from the health at 60 of a person who donates at 20, but I haven't read studies proven this. I do think short-term surgical outcomes might be worse, and I personally wouldn't donate if I were above 50, just out of vague fear of this. I do know there are hospitals that accept >50 year old donors, so they must not be too concerned.
Do all parts of the body age at the same rate? At first glance, it seem that the heart and brain are likely to go first, while the digestive tract ages more slowly and could probably go on for a while if the heart and/or brain permitted.
Hm, perhaps, but colon/rectum cancer is one of the most common cancer types, and according to this chart, it increases quite steadily with age, reaches a relevant level (where screening is recommended) around age 50 and reaches 2% risk per year at 75+.
https://acsjournals.onlinelibrary.wiley.com/cms/asset/bcc915de-239e-4ce4-8e09-11cb89efaa36/caac21601-fig-0002-m.jpg
It's still true that more people die of heart disease, so perhaps the heart ages faster in some sense. But the digestive tract doesn't seem decades behind in terms of years until failure.
A lot of failures of the cardiovascular system kill you right away, though, while a lot of failures of the digestive system just make you uncomfortable or miserable or stick you with a colostomy bag or something.
True. An hour without functioning digestion is less bad than an hour without functioning heart.
I've been reading an excellent book called "The Origins of Music", ed. Nils L. Wallin, Bjorn Merker, Steven Brown, MIT Press - a subject of great interest to me. There's 3 theories - 1 music evolved from language, 2 language evolved from music, 3 both diverged from a common precursor—referred to by Steven Brown as "musilanguage" and by Steven Mithen as "Hmmmmm" (Steven J. Mithen "The Singing Neanderthals: The Origins of Music, Language, Mind, and Body", Harvard University Press). Steven Pinker called music auditory cheesecake and thinks it's a spandrel, so he's #2. Mithen's book and Brown's chapter make a good case for #3. I'm for #1, but I'm obviously biased because I'm a musician. Any of the learned boffins here have opinions?
I'm fond of this hypothesis: https://meltingasphalt.com/music-in-human-evolution/
Yes I've seen that. One piece of support for Jordania which is missing from that review is the detail that humans singing in resonant intervals will sound louder and therefore more threatening.
It’s hard to see how #1 (music from language) can be true. Many animals make “music”: apes beat trees to a rhythm, birds and marine animals “sing”, etc. Our voices have natural pitches, and harmonizing in octaves, fifths, and thirds is based on vibrational modes. Language is thoroughly unnecessary for music to exist.
Birds sometimes utilize musical intervals, like the descending minor third of the cuckoo.
Yes, exactly that! I had a music theory textbook that had a chapter on “primitive” tribal music, and it talked about how “more primitive” tribes had what appeared to be very complex vocalizations, but they were just mimicking bird songs. More “advanced” tribes had far simpler melodies, just basic intervals around the root, but - that was a sign of independent organization, first glimpse of structured music composition, abstracted from any specific bird song.
Does the book get into possible importance of lullabies?
Don’t many species of animals sing? I thought dogs/wolves and possibly dolphins even sang “for fun”, or to communicate emotions etc. What’s the case for music evolving from language if it predates language?
One of the big issues is how you define "sing". I was blown away reading the chapter on humpback whale song - there's amazing stuff there.
Here is how I've come to believe Global Warning will be resolved:
1. CO₂ emissions are already rapidly being replaced by cheaper renewables. It will take more decades than anyone would like to go to zero, but the path is clear.
2. What isn't widely understood is that that doesn't bring CO₂ levels down, since CO₂ is quite stable in the atmosphere. The solution to that will be Carbon Sequestration. You capture CO₂ from the air, and pump it into underground caverns. The pumping part is well established technology from the natural gas industry. Capturing is only getting started, and will also take an uncomfortable number of decades to work at scale.
3. While waiting for points 1 & 2 to pan out, seeding the stratosphere with SO₂ can bring temperatures down by at least 1 degree Celsius, probably more, for about 2 years. This is known because volcanoes regularly does exactly that. Someone just needs to do it, despite the "Precautionary Principle" crowd.
I don't think this is super original. All three pieces are well known. Has anyone seen a "real" essay describing something like this 3 step plan/prediction?
I basically agree, although I have concerns about geoengineering. Sequestration/capture won't just be pumping CO2 into caverns. Direct air capture to create synthetic fuels and enhanced weathering also sound pretty interesting.
That’s exactly the solution. In fact net zero means Sequestration anyway - hence the net, as far as I know. No idea why this isn’t more well known. Of course if the Sequestration is to be achieved the energy to make it work has to be carbon free or we are chasing our tail. So we are not there yet (although we could use excess wind production to sequester)
This assumes, as most discussions of the topic do, that warming is bad. If you actually try to look at effects of climate change you discover that some are negative, some positive, the size of both hard enough to estimate that we don't know if the net effect is positive or negative. It's an issue I have discussed at some length on my substack — posts are sorted and webbed at http://www.daviddfriedman.com/Sorted_Posts.html and the first cluster is on climate.
One simple test of your knowledge: Can you list the major positive externalities? Talking only about negative ones makes it easy to conclude that the net is negative.
Sorry, I don't have time to go read 15 blog posts. From what I do know so far, I don't find the positive externalities point very convincing. You know how they say a startup has to be 10X better in some respect to take market share from incumbents? I think the same is true when arguing against an existing body of knowledge.
Up to you.
I sketch the general argument in:
https://daviddfriedman.substack.com/p/my-first-post-done-again
For a detailed example of just how bad work supporting the orthodox account can be — an attempt to estimate cost of carbon for the next three centuries, published in _Nature_, that implicitly assumes no medical progress from now to 2300, see:
https://daviddfriedman.substack.com/p/critique-of-comprehensive-evidence
If you prefer, a climate science textbook now in its third edition that contains provably false statements:
https://daviddfriedman.substack.com/p/a-climate-science-textbook
If interested, read one of those posts and see if it persuades you to look at others. I'm not prepared to redo the argument in a comment thread and even if I were it takes more than that to make the case.
It may or may not be relevant that fifty years ago, when the perils of population growth had the same status that the perils of climate change have now, I made essentially the same argument. What happened since then was the precise opposite of the then orthodox view. Poor countries continued to grow their population and, instead of getting poorer and hungrier, got less poor and less hungry.
I am less inclined than you are to put much weight on "all the experts say," having observed at least four cases in my lifetime where it turned out to be wrong. (Soviet economic growth, 60's Keynesianism and the Philips Curve, replacing butter with margarine to prevent heart attacks, and the population claim).
Yeah, all the experts are not infrequently wrong. Having said that, errors like not accounting for medical advances seem pretty incidental to the overall climate change debate, not exactly something you can hang your hat on. Arguing that sea level rise isn't a big deal because it looks small on a map is also not convincing. I generally tend to accept Nordhaus-style estimates, but the question that gives me pause is: what percentage of the poorest population would have to die to reduce global GDP by 2.5%? Or since there are offsetting benefits, what share would have to die to reduce global GDP by 5%? I'm not willing to dismiss that as an important problem, although I agree that it's not anywhere near existential at the pace we're going.
Anyway, I consider climate change mostly solved, between the expansion of renewable energy, carbon capture, and some geoengineering in a pinch. There will be lots of work involved, but the trajectory seems clear.
At this point, I'm somewhat more concerned about all the chemicals we're dumping in the environment and their effects on humans, but I suppose that will turn out to be manageable as well.
Rennert et. al. 2022 offers a calculation of the cost of carbon almost half of which is excess mortality due to warming, based on evidence of the relation between temperature and mortality in the recent past. Applying that to the next three centuries makes sense only if you expect no progress in medicine over that period, which is absurd. The fact that they did so, in an article that devotes part of a page to reasons its estimate might be low but says nothing about reasons it can be expected to be high, means that the authors, professionals in the field, are either incompetent or dishonest. The fact that the article was published in _Nature_, a high status scientific journal, means that the mechanism for filtering out work that is either dishonest or incompetent is broken in the context of climate science.
That doesn't tell us whether the net effect of climate change is good or bad. It does tell us that Greg's argument, that the "existing body of knowledge" in the field deserves respect, is wrong.
At a tangent to our differing belief in orthodoxy, Tyler Cowen has an interesting piece arguing that the essential difference between his classical liberal views and the new right "is how much faith each group puts in the possibility of trustworthy, well-functioning elites."
https://marginalrevolution.com/marginalrevolution/2022/10/classical-liberalism-vs-the-new-right.html
I identify with classical liberalism and not with the new right, but I have a good deal of sympathy for that loss of faith, having observed multiple cases where what elites viewed as orthodoxy, what all reasonable people should agree on, was not only dubiously true but supported by arguments that were dubiously honest.
One example from a few years ago of evidence that elites should not be trusted was a NYT interview with Fauci where he said that what he told people was required for herd immunity changed with new polling information on people's willingness to be vaccinated. I wasn't surprised that he would lie about his professional opinion but I was surprised that he would say so openly.
https://daviddfriedman.blogspot.com/2021/01/fauci-lying-greyhound-racing-and-trump.html
Two of the posts I linked to above are examples in the climate context.
I did read your thoughts on that a few years ago, but haven't kept up. I found it well reasoned as usual. I don't have time to go read more this morning.
The major positive effects I know of would be opening up VAST arctic/subarctic areas of Siberia and Canada to human settlements. Having just driven from California to Alaska and the Canadian arctic coast the potential of the vast emptiness is obvious. Yukon is the size of California and has 0.1% of the population!
That's one. A second is CO2 fertilization — doubling the CO2 concentration increases the yield of C3 crops by about 20-30%, reduces water requirements for both C3 and C4. A third is the reduction of deaths from cold — Gasparini et. al. 2015 estimates increased mortality from temperatures below optimum to be seventeen times as high as from temperatures above optimum.
For the first I go into some detail in a substack post, and link to published work by others:
https://daviddfriedman.substack.com/p/land-gained-and-lost
The assumption that these areas would open to human settlements is on a shaky ground (pun intended). Some of these areas may indeed become like Montana. Others will turn into impassible methane-burping swamplands. Thawing of permafrost in Siberia is wrecking villages there as we speak, driving whatever humans managed to live there for centuries out.
As an aside, dicking around with a dynamic system of insane complexity with multiple feedback loops of varied stability is a very! bad! idea!, which is why I find the whole exercise of "but have you looked at 'major positive externalities'" to be utterly corrosive.
One problem I see with this argument is that "dicking around with a dynamic system of insane complexity with multiple feedback loops of varied stability" might include all the activities promoted by people worried about climate change, not just the activities embraced by everyone else. "Have you looked at major negative externalities" is every bit as corrosive as looking at the positive ones.
Another problem is that "very! bad! idea!" is underspecified. Bad for whom? People? The biosphere? If the former, then proposals to curtail economic activity by as much as 75% should immediately come under scrutiny for badness, rather than pushed as the obvious solution. If the latter, then you have to explain how the biosphere appears to have undergone orders of magnitude more variability before homo sapiens even appeared, than has been observed or projected as a result of human activity, and survived.
Ah but there's a difference: we (humans) are already dicking around with the aforementioned system, by very quickly releasing vast amounts of CO2. This is - to me - the core argument for being concerned with it. Modeling errors go both ways.
In general, people concerned with the climate change are saying just that: stop/slow down with the dicking around we're doing right now.
Very! Bad! is indeed for humans. "The biosphere" will adapt, some species will die, others will flourish, a new equilibrium eventually emerges. There may be no room for us in it though. Of maybe room for 10% of us. This is describing a catastrophic "hot house Earth" scenario, which we know has existed in the past, and we don't understand the mechanisms that may tip the planet into this state again, so being concerned with it is valid.
What we have now works. It has worked for us - humans - for the last whatever years it took to create the civilization we have. A hothouse Earth will not work for the current civilization, and we don't know what gets us there. This is why looking at "major positive externalities" is corrosive. The risk/reward relationship for them is way out of whack.
You're coming into this thread with a premise that quick release of CO2 is bad. You might not have known that Friedman also questions whether this is so; now you hopefully do. (His reasoning is in the same places he linked above.)
You're likewise assuming that humans can't adapt to higher temperatures. Same deal. Check the link above.
Have you tried having these arguments published as a paper in a "reputable" journal?
I submitted my criticisms of Rennert et. al. 2022, an article published in Nature that claims to estimate the cost of carbon summed over the next three centuries, to Nature, which rejected it. I also sent it to the authors and got no response.
http://www.daviddfriedman.com/Ideas%20I/Climate/Critque%20of%20Rennert%20et%20al.pdf
Similar experiences here, different field. A few times that I had questions/concerns about published papers and contacted the authors resulted in 0 response rate to date.
I don't see intentional carbon sequestration, on a scale big enough to matter, as a thing that will happen. I figure, after we stop using fossil fuels, the climate will stabilize in a state warmer than it is now, and then we'll have to just adjust to the new climate. This won't necessarily be bad; the optimal temperature for the earth is whatever temperature we're used to, so we really just need to stop it from changing.
Won’t SO2 cause acid rain? This cure may be far worse than the disease.
SO2 in the regular atmosphere causes acid rain.
In the stratosphere, it does not, or at least not in noticeable amounts, since it stays up there for years.
Learned something new, thanks!
My guess is that no one will want to pay for 2. (Also, how much SO2 is actually needed for 3?)
I've posted similar before, but I'd like to share a bullet point justification of what I think the ultimate goal of human (or superintelligence) action should be, and see what this community thinks of it. Considering all the debate we have about the future of humanity, it makes sense to have some grounded justification for what an ideal outcome for human activity should be, if such a thing can even exist. Personally, I think this outcome I describe is superior because it is based on a system that properly bridges the is-ought gap (the idea that a moral imperative can't be derived from simple observation of how things are), while other systems are based on subjective opinion or tastes that have no absolute grounding. Anyways, here it is:
>Hume supposes the is-ought gap
>We need to transcend this gap if we want to be able to deduce any possible end goal for humanity as deriving from observations of existence itself
>What is able to do so is pleasure (defined as any emotion subjectively experienced as positive): we are able to experience pleasure as good in itself, and good by its own definition implies an "ought" in its own increase
>So the ultimate end is increasing the universal experience of pleasure as much as possible
>But how to do that?
>Focusing on minds that already exist is inefficient, because it is at least theoretically possible to use Von Neumann machines (machines that utilize surrounding matter to make copies of themselves) to produce more minds out of the dumb matter of the universe than ever would naturally exist
>So, the end goal becomes developing the technology to create Von Neumann machines that convert as much of the universe's matter into minds experiencing pleasure as possible
>These minds would be so constructed as to never bore of this pleasure, if you've ever had a moment where you've thought "I wish this moment could last forever," it would be such a thing actualized, pleasure feeling as good as it did at its first moment for near eternity
>While such an end for the universe, by its single-minded simplicity, might seem boring or depressing to contemplate, what matters is not the feeling an outcome produces in us to contemplate, but how it would actually feel to exist within that outcome, so while it may be a "boring" universe to an outside observer, it would be an objectively good universe for the beings making up the majority of its conscious existence
>This is where this theory differs from Yudkowsky's "fun theory," because "fun" is only one of a multitude of pleasurable states, and the important thing is how the system is experienced from the inside (maximally pleasurable), not whether it's a universe that's fun to imagine oneself living in
In summation, Utilitarianism is the only system able to transcend the is-ought gap, and the logical end of Utilitarianism is using self-replicating machines to create as many pleasure-experiencing minds as possible. Hopefully enough people can be convinced by this argument that this can become the end goal that either humans or superintelligences put their long term efforts towards.
This doesn't sound good to me. Reductivist plans based on overly simple rules of thumb typically don't go well. Also, we have shorter-term, demonstrably meaningful issues to solve. Daydreaming about different flavors of wireheading isn't very useful.
Asserting that minds can be created is really question begging, as it runs headfirst into the hard problem of consciousness. How would you ever know if you have created a consciousness, that is, a thing that perceives?
Admittedly this is one of the most difficult parts of the whole project, and would require significant advancements in neuroscience and philosophy of mind to make sure that what the Von Neumann machines create actually experiences qualia. In the worst case scenario, modified human brains could be created with the appropriate changes to their structure to ensure perpetual, non-tiring pleasure, since we could be fairly certain by analogy to our own brains that they are truly conscious. Personally, I'm a panpsychist, so I think that the hard problem isn't as much of a problem if you assume all matter has consciousness at some level, the problem then becomes finding out what the physical correlate of pleasurable emotion is, and how to replicate it.
That type of relentless utilitarian optimization comes with its own set of issues, e.g. the repugnant conclusion. What if you could speed up the self-replication process by 10x at the cost of each brain feeling 10% less pleasure? etc.
I'm also not convinced that utilitarian summation really makes sense to apply to conscious experiences, and the only evidence that will convince me will have to come from discoveries around how consciousness fundamentally arises. Is it really "better" for 10 people to have identical pleasurable experiences than it is for 1 person to have such an experience? Intuitively, we think that yes, of course, more happy people = better and less = worse, and it's a useful metric in many situations (e.g. measuring impact of a war or natural disaster, etc.). But for whom is it actually better? Certainly not any of the 10 individuals - their experience maxes out at N=1. Those 10 peoples' experiences are totally disconnected - there's no "universal entity" that is actually experiencing the sum total of all pleasure + all pain. Sum experience is just an abstract concept. So I'd be very hesitant to commit to optimizing the universe in such a way.
The problem you’ve outlined is the reason I’ve always balked at adopting utilitarian positions on happiness. Conscious experience is so fundamentally singular that I can’t see how the sum of a million people’s minor amusements could outweigh a few people’s complete happiness. Or more starkly, how a million individually experienced paper cuts could outweigh the brutal prolonged torture of one person. Pleasure and pain take place in isolated and discrete experiential spaces that appear to me to be non-additive.
I don’t know the first thing about moral philosophy but I’m familiar with Parfit’s Repugnant Conclusion. Can you (or anyone) suggest any other writings that speak to the problems with summation in situations like this?
If pleasure and pain are non-additive, does it follow that one excruciating death is no better or worse than a million excruciating deaths? Should we not be biased towards preventing the million deaths over the single death?
I routinely wrestle with your counter argument because it’s a good one.
Let’s take death out of it, if that’s okay, since deaths are additive in other ways. In this example I assume we can magically measure experienced suffering and say it’s 100 for the one and 100 for each of the million. Yes, of course I think we should be biased towards preventing a million excruciating experiences over one excruciating experience. But I think that’s because people are additive and you’ve made the suffering equal. Do you think we should let one person suffer at 100 to prevent five people from suffering at a 25? Experience is discrete. There is no one experiencing a 125 when the five suffer at 25. The max suffering experience is capped here at 25. However there is certainly someone suffering at 100. All else being equal, my inclination would be to prevent the one person from suffering the full 100.
It’s not that I don’t think there’s value in utilitarian positions on suffering, (there obviously is) I just don’t think they typically account for the individual nature of experience itself.
I've had similar thoughts and I struggle to articulate the intuition. Where I land is that suffering can't be quantified, but is more like categorical buckets. No amount of paper cuts can offset one person experiencing debilitating pain; no amount of people with debilitating pain can offset one person experiencing continuous torture, and so on.
That said, I still have this nagging feeling that the idea of comparing suffering is meaningless. Like you said, each experience is in its own isolated experiential space. That there are two separate spaces undergoing max suffering seems no more bad than there being only one. Yes, there are two bodies and each experiential space is associated with each body, but bodies don't suffer. Physical things don't suffer. So the basis for counting each experiential space and summing some quantity of suffering just seems like a category error. Substances have quantity, the mind and its experiences isn't a substance and so has no quantity.
To be clear, I am (most closely identified as) physicalist about mind. I don't mean to imply any non-physical stuff involved in mind. I just have a hard time quantifying experiential realms and doing math among the quantities as if they were commensurable.
So would a perpetual orgasm fit the bill?
That would be so… exhausting…
Well -- current version is too damn short IMHO. It's just a glimpse. Seriously doubt there are many people who think, during the 5 sec version, "I hope this winds down soon, I'm getting sort of tired and bored".
Yeah sure-sure-sure :)
But I second NotA’s notion - extend it to minutes, or keep repeating ad nauseam, and it’s not hard to imagine the “ad nauseam” part becoming a bit too vivid.
Honestly, it's hard for me to imagine getting sick of it. Looked at from the outside -- yeah, it's ridiculous to spend your life thrashing around in that state. But from inside, it seems like there's a "yes, I finally got here feeling" -- as though for once, in this lousy life, you are precisely as ecstatic as you always wanted be. You're so in the moment while it's happening that there's no way to think about the limitations of the experience, and how once it's over you will return to matter-of-factness, plus of course post-nut clarity.
While in nearly total agreement, I do wonder if the intensity of the feeling is inseparable from its brevity and... scarcity? The maybe-relevant data being that (at least from a male perspective) this is an easily obtainable state "in solitude", but it's just not the same when... frequented... in this manner.
If it lasted an hour, though, you might start thinking "geez, this is a bit too much of a good thing."
Perpetual orgasm of the nanobots. We need to maximize the number of orgasms per atom spent. What is the smallest possible molecule that could be described as having an orgasm?
Maybe it happens when compounds form elements. When certain things are oxidized there are FLAMES.
Hot nanobot-on-nanobot action?
"superintelligence should be able to brainwash everyone in the universe (and convert the universe into more brain washed bodies) because of the arbitrary whim of someone who has never and can never feel the end result he wishes to give to others."
the people are brainwashed because eternal unchanging pleasure is not a natural state of life and needs to be engineered into them. Many situations actually require a non-pleasure response to remain a thinking being; if a solar flare wipes out a planet of people unhappiness is the proper response but youve made it so they cant feel it. You remove free will.
its an arbitrary whim because you never have felt the state you want everyone in the universe to abide by. you only know normal human experience. If you knew this engineered pleasure you'd be unable to even conceive of suffering to formulate it.
its not like saying "everyone should be married" in which you have experience and just want to share it, you are trying to force an alien concept on all of the universe. not even because it gives you pleasure! but because of adherence to a philosophical position that says maximizing certain qualities is a moral good and to solve the is-ought problem.
and to do it your superintelligence can never feel it. it has to abide by your philosophy, and cannot exist to feel unchanging pleasure. because then it has no idea what to maximize; it needs to know what pain is to avoid it.
so yeah it doesnt work.
Pleasure is not a good in and of itself. It is a signal to ourselves that a certain activity or state of being that we are experiencing is good.
But what is this "good" that certain activities or states of being are aimed towards?
Eudaimonia.
>What is able to do so is pleasure (defined as any emotion subjectively experienced as positive): we are able to experience pleasure as good in itself, and good by its own definition implies an "ought" in its own increase
>So the ultimate end is increasing the universal experience of pleasure as much as possible
This seems like tunnel vision. We can experience other things too, like pain. If we view pain as inherently bad, then the "ultimate end" is instead reducing pain as much as possible, no? I know a lot of old people who view pain as more prominent than pleasure. (Indeed, some would say it's the only thing that's real. https://www.youtube.com/watch?v=8AHCfZTRGiI)
Yes, if we're being technical, then it's really pleasure less pain that's being optimized, because pain is inherently bad in the same way pleasure is inherently good. The system I propose would have the minds created never experience pain, as well as always experiencing pleasure.
Well there's several others too; hope, worry, regret, camaraderie, satisfaction. Worry's probably the big one, unless you're making everything too stupid to question the stability of the system.
Why is pain bad? It's unpleasant. But one can learn from pain. Also, could you have the concept of pleasure without pain? In this Von Neuman Pleasureverse, would there be degrees of pleasure? Or would it be a continuous wirehead experience? And if there are degrees of pleasure, would less pleasurable things become mildly annoying? And if they did become mildly annoying, would one consider them to be painful compared to ultimate pleasure? Heck, would one be able to maintain consciousness on a plain of infinite pleasure?
Your Von Neuman Pleasureverse sounds suspiciously like the Buddhist Nirmanarati heaven. From a Buddhist perspective, you don't want to get stuck in a place of endless pleasure because you won't be motivated to reach enlightenment—nor to assist others in reaching enlightenment. (Likewise, you don't want to get stuck in a hell, either, because the pain would be too distracting for most sentient beings to be able work on enlightenment.)
Anyway, thanks for the offer! But I think I'll pass.
Related: https://slatestarcodex.com/2014/01/28/wirehead-gods-on-lotus-thrones/
Nirmanarati is described as a realm of beings who can conjure forth anything that they like, and they spend their time making "mansions of sensual enjoyment" for each other -- to me this sounds like a really good time.
A better analogy might be the Jhana heavens like Abhasvara, which seems to be more like a perpetual solipsistic joyfulness lasting millions of years. I'm sure that's nice too, but from an outside point of view, it's not too interesting.
In addition to the totally unexplained leap you are making at the step 2, I contest step 3: "Pleasure" is a difficult thing to define and controversial choice. Many pleasurable experiences are pleasurable only to a limit and have diminishing returns; seeking pleasure may yield a less pleasant overall outcome; variety of pleasurable experience seems to be better conductive to human (and other mammalian) well-being instead of monotonous application of single pleasurable experience (near always a day is improved by a stroll in the woods, despite the fact that constantly residing in a sofa of reasonable comfort is obviously already pleasurable); and not only variety in pleasurable experiences is beneficial, certain amount of hardship and difficulties (that is, anti-pleasure) are also helpful to having a more enjoyable overall experience of life. Finally, the existence of hedonic treadmill implies that thinking of "pleasure" as a scalar number attached to states of matter in a moment of time that can "go up" just be reordering some configuration of states is likely some sort of abstraction (map-is-not-territory) mistake [1].
I find this a conundrum. Either "true" definition of pleasure needs both to include lots of other experiences that seemingly have nothing to do with pleasure and exclude many forms of pleasure [2], or one should accept that "pleasure" was perhaps a wrong concept to choose and something else (such as "good life" or "happiness") would be more apt.
Anyway, why there should be a number to go up? Why presuppose there needs to be an ultimate goal that involves some abstract concept of von Neumann machines implementing a number going up? An organism is a product of particular past evolutionary pressures (near universally those pressures created by other organisms). One might as well argue that the ultimate goal for a species with healthy amount of self-interest and enough capacity to do so would be to engineer their environment to the ideal form of ancestral environment where their kind first had the competitive edges that delineated that species. (Consider how many animals in a zoo appear become depressed despite safety and plenty of food and reproductive partners.)
[1] While the reader would probably do better by consulting an endocrinologist, the first intuition pump that comes to mind here is to think of a pressure differential stabilizing between two containers connected by an elastic hose ... which stretches in diameter over time as it is used. To maintain same velocity of gas transfer, the system would need to change over time: either containers to contain more gas, the temperature to change, a restriction to be applied to hose.
[2] This kind of definition of "pleasure" tends to become tautological no-true-scotsman-I-know-when-I-see-it.
You seem to be making points about the usual constraints human pleasure is subject to, i.e. the hedonic treadmill, while the point is that this would be engineered pleasure that would have no such restrictions, and would be able to be experienced as fully pleasant in itself without having the requirements of ordinary human life that require the periodic experience of less pleasurable states to contrast against. And the reason this is the ultimate, "number go up" goal is that unlike evolutionary competitiveness, pleasure is able to be directly experienced as a good in itself in such a way that allows it to bridge the is-ought gap, as I posted, so it demands itself as a goal requiring increase if one truly cares about maximizing "The Good." Again, note that pleasure here refers to any of the gamut of emotions subjectively experienced as positive, which might bypass some of your concerns regarding a more narrow definition of the basest pleasures which wear out the quickest in normal human experience.
>this would be engineered pleasure that would have no such restrictions
Yeah this was exactly what I tried to get at with my line about no-true-scotsmanning: more you define "engineered pleasure", more alien and increasingly removed from the everyday meaning of "pleasure" it appears. So maybe you are chasing a concept that is not "pleasure".
>pleasure is able to be directly experienced as a good in itself in such a way that allows it to bridge the is-ought gap
Have you considered that this might not be an universal experience? To me, it does not seem to be obvious at all that pleasure is a single thing that is "good in itself" due the variety and forms of experiences of pleasure I outlined. Unless you have totally different ideas tied to "pleasure" than me (by the sound of it, it seems closer to "things that are good and good implies something to be maximized") ... which makes the argument you claim to close the Humean gap reduce to "things that are good in itself and to be maximized are good in itself and to be maximized".
There's a little tautology to this; since it's agnostic about what constitutes this pleasure state, it's like "whatever is happening, everyone should prefer that to be happening."
> Hume supposes the is-ought gap
> We need to transcend this gap
Logic does not compute. Either the gap is real or not, and you have not given any argument either way. If it's real, then our alleged "need" to transcend it is just wishful thinking, on the level of our "need" to breathe seawater.
My point was that the is-ought gap is real for all other supposed moral imperatives which exist as mere opinions as Hume would claim; but since we can experience pleasure as directly, subjectively good, and the definition of good implies an ought in that more of a good is better, therefore we ought to increase said good, the maximization of pleasure becomes the one thing to which the gap doesn't apply and which has true moral imperative.
And we should consider the sensation of pleasure good because? Clearly pleasure is good in some contexts and not in others (see heroin addicts).
It's good because it is directly experienced as good in the most immediate way possible. Everything we define as good can be traced back to some pleasurable experience in relation to it. Would people bother climbing mountains if they didn't receive the pleasurable sensation of accomplishment in doing so which they specifically value? Hence it is the internal sensation, not the external act of mountain climbing, where the true value lies, regardless of what the mountain climber might believe or claim. Heroin addiction is bad because it leads to more pain than pleasure in the long run, so it fails as a reductio ad absurdum against this view. The contexts in which current pleasure is bad are merely those contexts where it leads to net suffering beyond the immediate present.
>Would people bother climbing mountains if they didn't receive the pleasurable sensation of accomplishment in doing so which they specifically value?<
They would, because
(1) You don't know whether there's a pleasurable sensation until you reach the top, at which point it's too late to not do it if there isn't, and
(2)maybe they're just keeping someone else company.
Also I had to climb a small one for school.
That's just Epicureanism. Why should we want an Epicurean AI as opposed to a Stoic one? Or a Kantian one? Or a Christian one? Or a Buddhist one?
Because of the reasons I gave as to why Epicureanism/Utilitarianism is correct (namely in how when you have a chain of "but why do you value that? But why do you value *that*?" it terminates in a brute fact of "because it provides some positive sensation valued in and of itself")
With the Christmas season coming up... Apparently a lot of adults buy Christmas presents for each other, but... why? With children it makes sense: they have predictable interests and you can't just give them cash because they can't use it. Also they'll usually just tell you exactly what they want. But with adults.... If you really want something, you can just buy it yourself. And with no good way to know what the other party wants, any present you buy for them is bound to disappoint. Giving cash or a gift card obviously doesn't make sense either, since you'll both end up with the same amount of money you started with.
So... what's the point? It just seems like a huge waste of time and energy.
I have a substack post on that puzzle: https://daviddfriedman.substack.com/p/why-do-we-give-gifts-4d4
Username checks out.
There are many things that people want or would like, but reason themselves out of buying. That's always a good idea for a gift. For example, my partner has always wanted to play bass, but never got around to it for various reasons. I bought a beginner bass for them last year and they were thrilled. They've been practicing a lot and show no signs of stopping.
Also, people are not omniscient. There are many things people will like or that are useful to them that they don't know exist. Buying a gift like this shows that you know the person deeply and that you care about their interests.
I actually prefer the feeling of giving someone a great gift to receiving one.
It seems to me that, in addition to felt obligation, the hook is a gamble. *Maybe* you can find something the other person will really like, even though the odds aren't in your favor.
I couldn't agree with you more, and this is the exact argument I used to eventually shut down obligatory seasonal gift giving in my family. It took a few years, but now everyone heartily agrees that the greatest gift we can give one another is freedom from obligatory gift giving. We now just focus on meals and hanging out and absolutely no one misses the physical objects.
Gift giving should occur only when one adult spontaneously discovers a product or experience that they're certain the other adult would like and does not yet know exists. That almost never coincides with a calendar, and shouldn't have to.
a gift is a memento that endures in time and reminds you of the giver. My sister once bought me a psp; i still own it and when i use it i think of her. A well-chosen gift is a bit of history.
physical objects bind memories to them in a way other things don't. The money or utility isnt the issue. even if i dont keep it, the memory remains.
you dont always get this till a lot of gifts have been given though. or how important memories are. if anything i wish i had saved more and kept more.
This is why it is actually impossible to buy a gift for my husband. If he wants something, he buys it. If he doesn't buy it, he doesn't want it. When I go ahead and buy him something anyway, he's annoyed, so I'm just not going to try this year.
The podcast No Stupid Questions talks about this in episode 128- How can you give better gifts?
Worth a listen
"And with no good way to know what the other party wants..."
If you have no good way of knowing what the other party wants, then you're probably not very good friends with them. But for people who are friends, it makes them feel good to know that the other person both knows what they like and put in the time and effort to find it for them. It's not actually that hard to understand.
Huh? I like giving gifts, don't you? A gift says I was thinking about you, or that I like you, or... just thanks for being part of my life. At Christmas I give different kinds of gifts, one could be something you want and ask for, another something I know you like. And third some type of project/ activity/ puzzle/ drawing, that we can hang out and do together. I got my son (age 22) a wooden clock kit, that I will admit I'm looking forward to helping him with.
Of course from a pure average utility of gift perspective it doesn't make sense. But consider the feeling it gives you that another person is spending the time to figure out your desires and then going through the effort to fulfill those desires. It can be done in other ways but there is a specific moment where there is a door open to this kind of opportunity.
> But consider the feeling it gives you that another person is spending the time to figure out your desires and then going through the effort to fulfill those desires
I'd argue that this fails more often than it succeeds, and an underwhelming or misjudged gift is far worse than receiving no gift at all.
FWIW, my understanding is that there's a custom in China of giving people red envelopes with cash in them. Which seems like it would be better, but it's been involuted into a status and reciprocity game, where you have to track who gave you how much at what occasion, and respond in kind, modulo your respective wealth and status, and frankly the whole thing sounds like a nightmare.
There are things I want so much that I'll buy it for myself, and there are things that I'd like, but not enough to buy for myself. Those things that I'd really like, but not enough to buy for myself, are the best things to get as a gift.
That's strange to hear. If I want something but I don't buy it, that's because I feel that the value it will offer is not worth the asking price. Someone else buying it and gifting it to me doesn't change that calculation if I'm expected to give something in return. The end result is that two people made a bad purchase for each other.
Some people feel irrationally guilty spending money on luxuries for themselves. But if I can spend money on _other_ people, and get an equivalent value in luxuries for myself, then those are some luxuries I can enjoy, guilt-free.
So let's say someone gave you a good pen, which makes it more enjoyable to write.
If you're the sort of person whose enjoyment of the pen would be forever eclipsed by the knowledge that you personally wouldn't spend the resources to acquire it, it would not be a good gift for you.
Most people, though, I think, would separate the process of acquisition from the experience of use. It'd be no worse than having it drop out of the sky, and usually much better because it associates the pleasurable experience of using it with the happy memory of the person who gave it.
Also, there's a window for an actual mutually beneficial trade, where someone gives you a gift that, after using it, you discover that you value enough that you would have gotten it anyway. Bonus points if you would not have been persuadable through words, such that simply giving you the thing saved everyone a lot of time and frustration. :-)
Part of the answer is it's a culturally accepted time to ask your rich friends to spend their riches on you. (Several people pooled together and got me a new bed one year, to replace the dirty mattress I pulled out of an abandoned building.) And in return you get them snacks and hand towels, cheap stuff they probably don't think about enough to replace. Or, like, weird movies you know about and they don't.
A bit off-topic, but you *can* give children cash. I've been both the giver and recipient there.
I once got my niece a twenty-dollar bill, sitting on top of a block of wood so it would wrap better. She and her friend decorated the block of wood and played with it for hours. Years later, she remembered the wood but not the money.
Gifting is a fun excercise in which you model your loved ones' tastes, interests and preferences in order to find something that they'd really enjoy but that they wouldn't think of buying for themselves (or won't allow themselves, in some cases). In general, the utility of receiving an unexpected gift is higher than its monetary value, even moreso the signal that someone cares about you enough to give a thought-out gift.
It still seems like gifting has a poor utility:cost ratio compared to other ways of showing care (such as providing actually meaningful assistance with their lives), especially when said gifting is reciprocal. I can see an unexpected gift having some value, but this is Christmas we're talking about. The gift is both expected and (supposed to be) equivalent in value. Of course, this causes even more problems when the latter ends up not being the case... It's just not worth the hassle.
>and (supposed to be) equivalent in value.<
Who told you that? This isn't alchemy.
>such as providing actually meaningful assistance with their lives
Lol, that's way harder, can lead to way more problems than your present not being of equivalent value and in general has a much higher downside potential.
Of course it's difficult, but the potential payoff is massive compared to a ritualized exchange of baubles. Again, the reciprocal and expected nature of the gifts greatly lowers their potential social value. I just don't see the point in putting that much time and effort into something that, on average, isn't even going to meaningfully improve your relationship or status with others. And it's hard to imagine that anyone finds the process of spending a bunch of time brainstorming, researching, and finally purchasing gifts for multiple people to be entertaining...
“Hard to imagine”. And yet...
This might interest you: https://daviddfriedman.substack.com/p/why-do-we-give-gifts-4d4
If you haven't seen "This Land Is Mine" yet, you should : https://www.youtube.com/watch?v=JfkXorjMcMI.
Here's the Artist Nina Paley dunking on copyright in the most hilarious (yet entirely fair and true) way possible : Copyright is Brain Damage | Nina Paley, https://www.youtube.com/watch?v=XO9FKQAxWZc.
Please summarise it - I'm not committing to an 18 minute video without knowing what it's basically saying to start with
The first video is a satire of a song made for the propaganda novel/film Exodus https://en.wikipedia.org/wiki/Exodus_(1960_film). The film is a favorable fictional account of the founding of Israel. The video Nina made is a satirical wider snapshot of the entire history of the region.
The second video argues that Culture and Intellectual products flow among individuals and societies in a manner very much resembling the flow of nerve signals among neurons, and that Copyright, in stifling this natural and beautiful flow, can be rightly likened to brain damage. This view, complete with the neurological analogy, is very similar to the views of the Queen pirate herself https://en.wikipedia.org/wiki/Alexandra_Elbakyan.
It's an animation account of the various nations who have conquered what is now Israel, with a stunning ending of who's really winning.
> Copyright is Brain Damage
I started watching, but I disagree with the premise that the possiblity of selling and financializing a catalog doesn't benefit artists. It does! People are willing to pay them *because* rights will be enforced after their deaths. Comically simplified, if, say, Taylor Swift wants to sell her rights to a song, she will get more money now if those rights are enforced after her death.
So although I probably agree with the conclusion that copyright overextends itself, I stopped listening to the talk.
Hello all. I voluntarily deleted my acount (Matthieu). The unexpected part is that it also deleted all my comments, with the one with replies marked as deleted and the one without leaving no trace. I realize this is bad for the blog archives' readability, and I apologize for it. I won't delete this temporary account but I should be unable to use it again.
This is a recurring theme, because I was already this guy: https://slatestarcodex.com/2019/11/06/open-thread-140-25/#comment-818659 . My reasons for leaving now are somewhat similar to then, although less dramatic (rereading myself, I certainly feel generally better now than I did then). The added practical reason is that I have a f*ing PhD to finish.
Thanks to Scott and to all the interesting and likeable people here. I should be back at some point.
You probably won't see this, but fare well, and good luck with the PhD!
I wrote a scifi Western about a gunslinger drone if you’re looking for a short cozy read today: https://solquy.substack.com/p/111123-the-gunslinger
Good story.
Thank you very much-- excellent story. I also appreciate "Anthropology" by Dizzy Gillespie and Godspeed You Black Emperor, neither of which I'd heard. I'm currently listening to "Lift Your Skinny Fists Like Antennas to Heaven".
Do you think FTL will get interested in composing and/or performing?
Thanks! Godspeed You Black Emperor had particularly stuck in my mind because for the majority of their time they had a lost album. I loved the mythology of there being these two dozen cassette tapes floating out there which may or may not still exist in the world. For the past decade I’ve been checking once every year or so to see if they had been found, and to my surprise, they recently were!
I think FTL views himself as a collector first and foremost, but this is definitely a possibility for the future! I thought for a while about a group of drones whose version just sound like white noise to humans, but which to them has intrinsic beauty. Could be an interesting direction :)
This was awesome, nice work!
Thanks for reading!
Enjoyed it!
Thanks! :)
That was fun, thanks :) I spotted the Murderbot influence, all to the good as far as I'm concerned.
Thanks for reading! Yeah, I think the concept of robots dealing with ennui is an interesting one :)
I have an interactive, international, interdenominational whatsapp chabura that you are welcome to join.
At some point inshallah it will be by invitation only but as we're only 70 people as of now and the average SSC'er is likely to be a good fit you are welcome to join and of course to leave or rejoin whenever you like.
https://ydydy.substack.com/p/you-are-invited-to-the-ydydy-chabura
That's the invite link you had given me? I joined it and it seems that all the so called 70 people in it are fake, as they all have WhatsApp profiles consisting of just a number and a pic, unlike in every other WhatsApp group I am a part of.
You need psychiatric help if you are engaging in this sort of shenanigan!
The recent New Yorker is essentially the AI issue (I linked to a profile in it on Hinton in a thread below) and has a piece called "A Coder on the Waning Days of the Craft". I'm curious to what extent other coders agree with his description of GPT-4's current coding abilities: https://www.newyorker.com/magazine/2023/11/20/a-coder-considers-the-waning-days-of-the-craft
I have never used GPT 4 because I'm cheap and also Open AI doesn't deserve money, but here are some of the things I hated about this article :
1- It's titled A Coder Considers the Waning Days of the Craft, I hate the word "Coder" because it's an inaccurate way to describe programming languages, "Code" used to mean the binary encoding of commands that CPUs come out of the factory natively understanding, every computer language that resembles English or mathematical notation is technically not "code", it's a formal language that serves as a user interface to code and ultimately translates to it. Then there's the "Waning Days", really ? Waning ? Even the author refutes the clickbait in the last few paragraphs :
> GPT-4 is impressive, but a layperson can’t wield it the way a programmer can. I still feel secure in my profession. In fact, I feel somewhat more secure than before.
Why can't NYT title writers read the whole thing like I just did ?
2- The author's examples belong to roughly 2 categories of things that GPT-x seem to be unusually good at :
2-a) Green-field small projects with historical precedence : This is when you want to make a new program from nothing, typically one that has been made before, the result is typically between 100 lines of code and 1000. GPT is good at this because the entirety of Stackoverfow and tutorial websites are extremely good gold mines of training. Examples of this category include the Snake game and the ios app the author's friend made and the countless "make website using GPT" tutorials.
2-b) Input-Output puzzles : You have something, you want something else out of it, you're aware that the transformation is possible but you're too lazy/sleep-deprived/short-on-time to actually sit down and derive the step-by-step procedure or look it up. GPT is good at this too, sometimes because it has developed something akin to "reasoning" over the language the solution will be written in, but most of the time because someone before has had your exact problem, the shell command the author wanted to print a random 1000 lines of the dictionary is a good example.
There are vast territories in programming and software engineering not covered by (a) and (b). There are the million+ lines of code heavy-duty services at Google and Facebook. Ok, not everyone works at Google and Facebook. How about the "10000+ lines of spaghetti that encodes 15+ years of changing business requirements" ? How about the "website design that keeps circulating back and forth for 6 months because the lightning in the room makes the client see the colors differently each meeting" ? How about any program bigger than 1000 lines and thus too big to fit in even the biggest of GPTs' context window ? Database schemas ? Compilers ? Game Engines ? Will all of this, really, be automated away by the very conceptually simple mechanism of predicting text ?
3- GPT is data-inefficient. GPT is only good at Python and Android/ios and Html and CSS because those are very stable and very popular technologies that hordes of people invested in over more than a decade and assembled a vast ad-hoc library of examples and material. But try making GPT help you with COBOL, try asking it about the AUTOSAR automotive platform, ask it any question about an obscure or relatively unknown technology.
I like the GPTs and I see them as a large step in AI and in Knowledge tools more generally, but I developed a kind of dread towards those kinds of articles, their titles tend to be similar even.
From the article:
> Programmers are people who can endure an endless parade of tedious obstacles
So true a lot of the time. There's always Google, but in Stack Exchange threads one must always remember not to copy code samples from the first post, detailing the question, because these are wrong almost by definition! :-)
My take as a programmer is positive. My job as a tech person is to take a bunch of business guys' loosely defined requirements, and turn them into something watertight enough to be implemented, then implement it. If the second part is largely automated, so much the better, it's the boring part anyway.
But any tech project will have lots of precise points where what you're doing is not obvious, and spelling them in precise detail is exactly what we do through code. If we spell them in plain English to some LLM-driven bot which spits out Java or whatever, writing them in English with enough precision to get the actual wanted code that fulfills all the loosely specified requirements *is already* essentially programming. So our skills are still needed.
I’m pretty senior and I have begun to use it a lot. Will it replace all coders? I can’t see that happening unless it’s 100% accurate, but a few more step changes in ability and we could be done for.
(However if humans don’t code anymore where do LLMs pick up future information?).
Trying to scope out who'd be interested in funding a research group on broader social, historical, and philosophical approaches to AI policy and design. I'm from an SWE and PhD academic history background, others interested are PhD stats, PhD moral philosophy, inter alia.
tl;dr academics and policy people want to model and think through and design AI socially and culturally, who is it _not_ rude to ask, is there anything we should be thinking about other than just trying to get funding for compute?
More info here:
beforeutility.com
In celebration of a very ER doc Thanksgiving I wrote some satire about the cherished holiday tradition of dumping your old people at the emergency room. Nothing says “let’s be thankful” more than the annual Pop Drop and Yaya yeet (tm). Please don’t do this ;)
https://bessstillman.substack.com/p/the-holiday-dump-is-a-cherished-tradition
This was a riot, thanks for writing this.
The first time I took care of a holiday dump-and-leave, I was a baby intern, and literally didn’t understand what was happening. I just thought that there must have been a miscommunication by the admin/intake clerk because there was no contact info available. My senior had to explain the phenomenon to me. It made my thanksgiving shift quite sad.
As a side note, I’ve been following your writing ever since your 200 Joules post. I think the world is better off because of your and your husband’s writing. Best wishes to you and yours. <3
Thanks for your kind words! That really means a lot.
Oh yeah, explaining the phenomenon to new doctors or med students and just watching the light kind of go out of their eyes is not fun, and usually it’s kind of fun to mess with the new meat 🤣 I learned there’s a vet version to this and I just wanted to quit being a person for a while
Yeah, two years later I had to explain to my intern what was happening (the poor guy kept on being hung up on when he tried to call the son) and that was not fun haha. And in my head I was thinking, “Everything has come full circle! The cosmic scales have now been realigned!”
Namaste. 😄What sort of cog in the medical machine are you?
Cardiology, at your service!
Finishing up general fellowship, with ACHD fellowship to follow. At some point I suppose I will have to stop training and get a real job :)
ERs across the country should play "The Ballad of Narayama" on repeat during the holidays.
that's an idea
Indeed. Approaching senescence and death with stoicism isn't something we can get to from here, but being reminded of it in movie form doesn't do any harm.
Ubasute in the winter would have been less painful than most end of life care here in the West, and far more dignified.
Good lord, that's depressing. Ironically, we did have to take my mother-in-law into the hospital the day before Thanksgiving because she got Covid and spiked a 106 temperature. They kept her there for 3 days out of abundance of caution, which was a bummer.
Dementia (and age decline in general) is such a hopeless and miserable state. I'm curious if your experiences have led you to support for euthanasia.
This was an absolutely appropriate reason to bring in your MIL! Tell her 106 is a bit dramatic, but she gets ER flair points. Kidding, I'm glad she's doing better. I am definitely in favor of paths to assisted suicide for people with devastating illnesses, it's the humane choice if that's what someone wants. But the US medical machine is incredibly paternalistic and does a very poor job at consent and shared decision making as it is, so I'm not entirely sure how to get from idea to ideal. Still, I think steps should be taken to open the conversation.
As a cousin idea, I'm a big believer in right-to-try and expanded compassionate use programs for the terminally ill, which the FDA also effectively blocks. Ironically, I think the FDA would prefer to allow people who want to live to die of their terminal illness instead of try something experimental and risky (with the patient's consent) to see if it would help , and on the other hand, they would rather let someone who wants to die because of their terminal illness live and suffer rather than allowing assisted suicide. All in the name of patient safety. If you've seen my essays posted around here you'll know FDA "protections" are a deeply personal soapbox topic for me :)
About assisted suicide, etc.: Exit International is a good source of information about options..
Sheesh I open up the new open thread and these are the first two comments to read....gonna go hug my wife and kids now. And the dog too.
I'm developing a browser-based idle game and as an amateur, it's been a real motherfucker trying to figure out coherent systems for some of the game mechanics such as the loot rewards, the questing system and so on.
Do people just trial and error this shit when they're on their own? ChatGPT helps a bit but is mostly good as a summarizer and organizer. It has occasionally given some good ideas for formulas and can help me make sense of the math needed for some of the things, but I am feeling quite a bit lost.
Are there resources, communities out there that were created for amateur/solo game devs that you know of?
Yes, trial and error with a lot of playtesting. I think it's best to think of game design like science. You have a hypothesis about what might be fun, you design a prototype to test your hypothesis (on real people who aren't you, ideally), then you adjust your design based on the results. This is what they call "finding the fun".
Here's a good article on design for idle games: https://blog.kongregate.com/the-math-of-idle-games-part-i/
Thanks for your reply and for the article - ill check it out!
As far as I can tell people do a lot of playtesting. There’s some room for modeling what the system does without just running it, but I don’t think idle game authors do that very much. There’s a subreddit if that helps, I don’t know what state it’s in though. https://www.reddit.com/r/incremental_gamedev/
Somebody besides me who says motherfucker on ACX! Hey there Zarko. There are definitely a lot of game developer resources online. I have a friend who is building his first game, and has found several. They're friendly, too. I thinnk some are on Reddit, some on Discord. There are also contests for newbies where they alone or with a group build a simple game in a a short period, someting like 3 days. In my town there's even an indy game developer's meet=up. Oh, and of course there are lots of YouTube how-to videos. Since I neither develop no play games that's all the info I have, though.
What this world needs is an easy to type markup language.
https://conntects.net/help/qtmlTutorial.fn
Why? Phones don't have room for lots of style buttons, and voice recognition is a serious security threat. I don't like the idea of machines listening in on me. (I'm slowly working on an easier to learn keyboard layout, but since making hardware requires a whole lot more capital than writing software, this one is on the back burner.)
Looks nice, but whoever came up with the name "QTML" hadn't googled that before.
I see QML, as a Qt analog of JSON, and Quantum Techniques in Machine Learning. To which of these are you referring?
My takeaways from the Altman saga (from very much an outsider's perspective, based only on news reports and what I think I know about human nature):
- The corporate structure of OpenAI which was designed so that AI safety would be the paramount concern; the board was meant to ensure that. Altman does not want to be constrained by that structure.
- The 90+% of OpenAI employees who sided with Altman agree with him. Furthermore, they apparently found it absurd for Toner to suggest that it would be consistent with the mission of OpenAI to shut down the company if it presented a danger in terms of existential risk, or AI risk more generally. But that statement by Toner is simply a logical consequence of *having* a mission in which AI safety is a paramount concern. So, practically speaking, OpenAI employees, with near unanimity, do not believe that AI safety should be the company's paramount concern.
- It's pretty easy to infer from the above that the point is never going to come where the AI industry voluntarily prioritizes AI safety over other concerns. At every given step between now and ASI it will be easy to rationalize a more or less full-speed-ahead approach to AI development: "if we don't do it the other guys will"; "we are working for the good of humanity"; "our intentions are good, therefore the consequences of our actions must be good." The monetary and professional rewards of being first in the AI race and the glory of building utopia will be powerful motivators for concocting those rationalizations.
I am 100% certain that Sam Altman does believe that AI safety risk is real and profound. It is also clear from profiles of him that he thinks he's building utopia. Designing utopia is akin to making himself something between a king and a god of all humanity. Of *course* he is going to treat safety concerns as secondary to that incredibly powerful ambition.
He is, I'm sure, very smart. He is also an incredibly shallow thinker, from what I've read about him - he has no sense of the weaknesses of utopian thinking, no sense of why the extraordinary power he is projecting for AI (and by extension for himself) is inherently dangerous. He is subject to the worst kind of hubris: the belief that not only is he competent to wield enormous power unchecked, but that he is doing so for the common good. He might also be the most responsible leader of an AI company outside of Anthropic.
I am not feeling great about things at the moment.
I have gotten the strong impression from the last week (the news, as well as individual reports of his past behavior) that Altman is a stereotypical sociopath, that is to say charismatic, manipulative, and indifferent to risk. He is smart enough to intellectually understand existential AI risk (and to tell us how dangerous it is and how cautious the world must be regarding it). But he is incapable of actually *feeling* that risk and thus is unlikely to take meaningful action to avoid it (when that interferes with other goals like personal ambition for power).
Asking AI developers to slow down is like dumping a dragon's hoard in the middle of a football field and asking the crowd to proceed single file and at a walk towards the hoard.
And in addition to the financial incentives for rapid AI development, there's this: A lot of people who work in tech have an Aspergerish profile of mental strengths and weaknesses: They are awesome at developing tech -- some of them have absolutely beautiful minds -- but less good than the average person at various people skills: grasping other points of view and seeing how those points of view, like their own, are coherent and sincere; predicting how how members of our species are going to be affected by a novel technology; having an empathy-based feeling of loyalty to other people.
The AI companies at this point are like a second government. Of course they have less power than our actual government over on how the present US is set up, but they have more the the present government on the shape of the future. In fact maybe it makes sense to think of them as the government of the future. They are at present shaping the world of 2050 and they are ill-qualified to govern. They have the wrong kind of brain.
I know there are a lot of people here who consider themselves to be on the autistic spectrum, or not neurotypical some other way. I hope it's clear I'm not disparaging people who are not neurotypical, just saying most are not cut out for jobs where people skills are very important. I'm pretty sure I myself am not neurotypical.
Wait, isn't it a *GOOD* thing if the shape of the future will be driven more by nerds??
Nerds / rationalists
> People who value truth
> People who prefer to operate on Simulacrum Level 1 - ie actually engaging with the real world and believing there are objective facts and truths about it
> Not generally motivated by greed, lust, or other vices - generally motivated more by intellectual challenges and creating new / complicated / exciting things
> Their main qualifications are being smart, being educated, being deep in one or more technical disciplines, and an ability to drive real-world results
vs
Politicians
> Explicitly eschew truth, and indeed are characterized by being *extra good* at lying to a broad array of people and constituencies
> Operate on Simulacrum Level 4, with Trump a fine example - only concerned with words and their effects on other people rather than any sort of ground truth or any such thing as objective facts
> Generally motivated by greed, lust, lust for power, etc - don't really care about intellectual challenges
> Their main qualifications are being rich, knowing rich people, being good at lying, and being willing to take suitcases of cash or long/short equity positions that substantially influence the legislation they consider and pass
It's not like politicians carefully and empathetically consider the viewpoints of all the clusters and demographies out there - they're all tribalism, lies, and suitcases of cash / stock positions in return for tailored legislation.
Which of those two groups do you honestly prefer to determine the shape of the future?
I think a random selection of people who meet basic requirements of decent general knowledge of science, culture and history would be better at governing us than politicians, as least politicians as you depict them. Let's set aside the politicians and just talk about nerds.
So can we think of nerds as people with Asperger's, who also are very smart, and who also are diligent and hard-working? If so, then the thing I see as a problem is that these people are not very good at reading other people. For instance, there's a well-respected test of reading other people's facial expressions. For some reason you can take it on Amazon. It's here:
https://s3.amazonaws.com/he-assets-prod/interactives/233_reading_the_mind_through_eyes/Launch.html
The average score is 28. Scores 22 or below strongly suggest an autistic spectrum disorder. But setting aside diagnostic cutoffs, think about the disadvantages that come with being, say, 25th percentile at reading faces. It's a genuine disadvantage -- like having a bad sense of direction. You are less good at knowing where you stand with people. You are less good at figuring out what they want, what they are feeling, when they are lying. You are less good at forming a mental picture of their point of view.
Here an example of the kind of thinking that comes, I think, from having a deficit of that kind. Somebody on here put up a post suggesting that nobody really likes Shakespeare, some people just pretend to because they've been convinced that people who like Shakespeare are classy. And quite a few people chimed in, agreeing that that idea was plausible. OK, I am someone who likes Shakespeare, and I am absolutely positive I am not pretending to. And I know many other people who are the same. I can sense their sincerity when they talk about Shakespeare — but I also have behavioral evidence. I have seen them reading his plays for pleasure. I go to performances with them and the afterwards they are bursting with excited thoughts about one actor’s read on a certain character. Etc. The person who put up that Shakespeare post was having trouble forming a mental picture of what it was like to be someone who enjoyed a certain thing that he himself did not.
Here’s another example. Yann LeCun is active on the part of Twitter where people talk about AI. Yudkowsky is on there, as are a number of bright, articulate people who talk about risk, and some of them come at it from a completely different angle than Yudkowsky does. So LeCun has heard a number of nuanced bits of thought about reasons to worry about the impact on AI on the future. I feel unable to decide how seriously to take x risk, but have posted some good data-based arguments about the ways indirect contact with others, and involvement with virtual people via gaming, etc., is associated with increased depression and akrasia. Others post about x risk, but sketch in some plausible paths to great harm for which they estimate there’s a 10% probability. Of course there are some screamo “AI’s gonna kill us in 5 years” people on there too, but the point is that LeCun has read a lot of stuff that makes decent sense. But he posted that people who worry about AI harm would probably have wanted to ban the production of ball point pens when they were invented. They’d be screaming about how dangerous the pens were because they could be used to write down and spread misinformation. So LeCun, with quite a lot of good information about what AI worriers really think, was unable to form a mental picture of their points of view — instead, his picture was of a generic screamo idiot. No doubt he was helped along in the process by the fact that he does not want the worriers to be right. AI is his baby, his career — of course he is heavily invested in it! You can think of the ability to see things from another person’s point of view as a sort of *check* on the tendency to believe that the things you want to be true are, in fact, self-evidently true. LeCun is functioning without that check.
Here are some important mistakes I think people with this sort of deficit are especially vulnerable to when it comes to AI development:
-disregarding ideas that come from people who do not think they way they do
-failing to involve enough people with skills in soft sciences like psychology, child development, animal behavior and linguistics in thinking about AI, it’s development, and how to align it.
-failing to recognize when a colleagues has become so fascinated by his work on AI that he is now incapable of even considering the possible dangers if his project succeeds.
-not being good enough at introspection to distinguish between what they think is true and what they want to be true; and between craving for success and glory and confidence that one is improving life for mankind.
-being bad at imagining the impact of things like having a personal AI chatbot on people who are socially avoidant — or the impact of the ability to easily make convincing fake videos on teen bullying, or the spread of misinformation -- or the impact of increased AI presence in doctor visits.
You make some great points, and definitely have some valid concerns about nerd-driven future policy - but I think the big problem is there's no way to get to your ideal from where we are today. The choices are nerds or politicians, and there's no better choice on the menu.
Sure, we'd all love if our decisions were made by golden philosopher-kings with exceptional empathy and compassion AND in-depth tech skills AND an unyielding commitment to overall human flourishing AND a broad understanding of differences in people's wants, desires, and mental models coupled with a sincere desire to not use that for personal advantage, but to do what's best for that diverse mental constituency...but those people don't exist.
The framework for finding them (or even your broadly educated "average" people -- and college degrees are ~35% of the pop, so the *average* person doesn't even have an undergrad degree in a single major, much less an understanding of multiple fields) and elevating them to positions of leadership don't exist, and likely never will. Random lottery democracy hasn't been seriously proposed or enacted anywhere (and Athens and other Greek city states weren't pulling from a random pop, but instead a highly selected elite pop). Honestly, our best bet for actually better leadership than what we can get today is an aligned AI, so once again that's on the "nerds shaping the future is better" side of the argument.
As of now, here, our options are "future shaped by politicians" and "future shaped by nerds." Absolutely, nerds are less good at reading people and forming a clear mental picture of their point of view. That's STILL better than politicians, who *are* good at both those things, and then actively use them for evil / lies / personal power and aggrandizement.
You have some good examples of nerds sucking at mental modeling, and Ylecun is famously obdurate and blinkered when it comes to AI safety, nearly everyone on the AI-not-kill-everyone side will agree, but the alternative is politicians who *are* good at mental modeling and will use it against you.
I think given the two choices actually on hand, we should be pulling hard for "nerd shaped future" even given it's deficits and potential failure modes, because the deficits and failure modes of the other side are so flagrant and so much worse, and because there's never going to be a better, more inclusive option like the one you would prefer.
I tried to give an adequate reply to all this, but just ran out of time. So I'll just say this: I read somewhere in the last 2 weeks, can't remember where but it wasn't a silly publication, that a substantial minority of AI developers think ASI will wipe out our species, *but* a substantial minority of those people consider that an acceptable outcome. Are you sure you'd be willing to be governed by these people?
This issue comes up all the time on ACX, so maybe we'll have another exchange in the future.
As for the politicians: You couldn't possibly hate them any more than I do. But I still think we would be better off working to adjust the laws and rules that have made some of the grotesquery possible -- for example, gerrymandering (there are many other examples, of course). Also, I do not believe all politicians are godawful. I have 2 close acquaintances who have worked closely and for a long time with senators and cabinet members. Both are perceptive people and both loathe the cynical corrupt fatass liar politicians as much as you and I do. Both are sure that the people they worked for have great integrity. I believe them.
A lot interesting there-- I'm wondering how you'd select basically sane people of a variety of personality types to do the governing.
I just checked-- you have a wide knowledge requirement, not a personality requirement. How you'd do that for government, I don't know. Sortition with a limited pool?
A problem with nerd/autistic people failing to understand normal people is that it may take nerds to understand the relevant technical issues. Also, as I understand it, autistic people aren't so bad at understanding other autistic people, just as neurotypical people understand each other fairly well, but not autistic people. It's not a generalized "bad at reading people". Still, bad at reading the large majority of people would lead to being bad at governing.
Have there been autistic hereditary monarchs? How did that work out?
Being wrong-headed and stubborn about it is hardly limited to autistic people. I believe Hitler wasn't autistic. Autistic people might have a worse case of it on the average, but I'd need evidence.
As for fear of fountain pens, disaster didn't happen from fountain pens, but the printing press did lead to religious wars in Europe. I wonder whether anyone saw that coming. I've heard radios implicated in the rise of Nazism and in the genocide in Rwanda. If you're not used to radio, hearing politicians that way is like having Saruman whispering in your ear.
Oh, what the hell, I will mention this. Last night, I spent too much time arguing with a man who went from life for humans is a net good to everyone should have children. He was completely ignoring whether everyone wants children. He was willing to accept that people would disagree with him, but not that he might be missing something.
Maybe High Modernism (the belief that things should be legible and decided from above) is a great deal of what's wrong with the world. Note that it's a major part of Communism, and a somewhat less strong aspect of capitalism.
Nancy, I'm not going to have time to give a decent reply to all this. But no doubt we'll talk again. Let me just respond to a couple points you raised about autistic spectrum people:
> it may take nerds to understand the relevant technical issues.
I'm confident that it doesn't. Feynman, for example, was clearly not on the autism spectrum. He had a mobile, highly expressive face. He was sociable. He had lots of affairs. To do that, you have to like sex, and many autistic people are asexual or have a reduced interest in sex or are only aroused by some unusual kink (I knew one whose thing was plastic pocketbooks). Also, you have to be able to turn on the charm and tweak it to fit the person you're with -- that's very un-Aspie. Also, I have heard podcast interviews with AI developers who clearly are not on the autistic spectrum. They chat easily, are self-disclosing, talk with emotion about their mentors or their family, and have a lot of mental flexibility. If the interviewer brings up some new way of thinking about something, they move right inside of it and say, "hmmm," so from this point of view it looks like . . .." Jeremy Howard is one, and a particular favorite of mine. He put up one of the first whats-the-point-of-masks YouTube videos.
(It's here: https://www.youtube.com/watch?v=F0RcH9DfuyE).
I assumed he was a medical professional, but actually he's an AI guy. Also, he's adorable.
>Being wrong-headed and stubborn about it is hardly limited to autistic people.
Completely agree. There are many paths up the wrong-headed stubbornness moutain.
>Also, as I understand it, autistic people aren't so bad at understanding other autistic people, just as neurotypical people understand each other fairly well, but not autistic people. It's not a generalized "bad at reading people".
I don't think that's true. I have treated many people who have Aspergers or are a little further up the autism scale than Aspie , and what I see is that most are more comfortable socializing with other people on the spectrum, but that is because other Aspies do not make heavy conversational demands. Often they mostly play D&D or some other game together, without a lot of communication beyond what the game requires. If one of the other players gets angry and storms off, the Aspie patient telling me about the incident does not seem to have any special insight into what upset the person. They say things like, "I dunno, maybe he didn't like the new player who joined the group?" I often have the feeling that if I, or really any random neurotypical person, had been present observing the group we would have had a better idea of what had upset that one player. Many Aspies yearn for a connection with neurotypical people. I have one patient, a female math prodigy doctoral student, who got furious when I suggested she check out some meeting place I'd read about for neurodiverse people: "Why the fuck would I want to meet more people like me?"
>Have there been autistic hereditary monarchs?
I'm sure there have been. Asperger's is not rare. But I am sort of ignorant of history (hence condemned to relive it, I hear). You should ask on here. I have wondered about Prince Charles, well King Charles now.
There may not be *enough* non-nerds who understand the technical issues.
I'm not sure if all Aspies are as mentally rigid as you imply.
I honestly don't know where you're drawing the line for Aspie and non-Aspie. My home subculture-- sf convention fandom starting in the seventies-- seemed to find a way for a lot of mildly spectrumish people to enjoy each other's company, and it isn't all gaming, though I'll grant there's a lot of gaming.
I don't trust rationalists to run the world, but it's more because of a taste for High Modernism, a mental habit that might be especially attractive to nerds but isn't limited to them.
A little research suggesting that communication goes best between people with similar neurotypes. This is the first time I've seen a paper with an abstract and a lay abstract.
https://journals.sagepub.com/doi/full/10.1177/1362361320919286
"Abstract
Effective information transfer requires social communication skills. As autism is clinically defined by social communication deficits, it may be expected that information transfer between autistic people would be particularly deficient. However, the Double Empathy theory would suggest that communication difficulties arise from a mismatch in neurotype; and thus information transfer between autistic people may be more successful than information transfer between an autistic and a non-autistic person. We investigate this by examining information transfer between autistic adults, non-autistic adults and mixed autistic-with-non-autistic pairs. Initial participants were told a story which they recounted to a second participant, who recounted the story to a third participant and so on, along a ‘diffusion chain’ of eight participants (n = 72). We found a significantly steeper decline in detail retention in the mixed chains, while autistic chains did not significantly differ from non-autistic chains. Participant rapport ratings revealed significantly lower scores for mixed chains. These results challenge the diagnostic criterion that autistic people lack the skills to interact successfully. Rather, autistic people effectively share information with each other. Information transfer selectively degrades more quickly in mixed pairs, in parallel with a reduction in rapport.
Lay abstract
Sharing information with other people relies on the ability to communicate well. Autism is defined clinically by deficits in social communication. It may therefore be expected that autistic people find it difficult to share information with other people. We wanted to find out whether this was the case, and whether it was different when autistic people were sharing information with other autistic people or with non-autistic people. We recruited nine groups, each with eight people. In three of the groups, everyone was autistic; in three of the groups, everyone was non-autistic; and three of the groups were mixed groups where half the group was autistic and half the group was non-autistic. We told one person in each group a story and asked them to share it with another person, and for that person to share it again and so on, until everyone in the group had heard the story. We then looked at how many details of the story had been shared at each stage. We found that autistic people share information with other autistic people as well as non-autistic people do with other non-autistic people. However, when there are mixed groups of autistic and non-autistic people, much less information is shared. Participants were also asked how they felt they had got on with the other person in the interaction. The people in the mixed groups also experienced lower rapport with the person they were sharing the story with. This finding is important as it shows that autistic people have the skills to share information well with one another and experience good rapport, and that there are selective problems when autistic and non-autistic people are interacting."
Have you considered that the neurodivergent are, perhaps, tired of the neurotypicals' shit and view not giving them a voice in the process a feature instead of a bug?
Yup. Yann LeCun tweeted that people worried about x risk would be scared of ball points pens if they had just been invented: "Oh no! They can be used to write mean things! They are very dangerous and should be banned." That's how the public looks to him. https://i.imgur.com/Q5DB7VP.png
Yeah, I would think that software design and computer engineering and stuff like that selects for people who are inclined toward thinking in terms of self-contained totalizing systems: you build a whole world that you hold in the palm of your hands. But the real world is characterized by ambiguity, uncertainty, open-endedness, and the ultimate impossibility of constructing any totalizing system (epistemological, moral, economic, or otherwise) that would fully take account of it.
One factoid about Altman revealed in the profiles of him is that he was apparently bored by classroom study at Stanford and instead spent all his time playing poker. In other words, I guess the ambiguity, uncertainty, and open-endedness of, say, literature or history did not appeal to him; a self-contained rule-defined game in which he had a chance to win money, on the other hand...
“But the real world is characterized by ambiguity, uncertainty, open-endedness, and the ultimate impossibility of constructing any totalizing system”
Sounds like a large software program
Yes, as a software developer, yes. But it's not, even that is an oversimplification
PLUS the bomb will kill a fuck-ton of neurotypicals.
From his comments I have heard he honestly seems like a naive dummy. I have heard some many good things about these AI researchers from smart people, but hearing it straight from the horses mouth they seem very unimpressive.
People once got a TV show back on the air by sending 40,000 pounds of nuts to "various network executives in New York and Los Angeles": https://abcnews.go.com/Business/FunMoney/story?id=3214156&page=1
The situation with AI seems pretty nuts too, does it not deserve an equally powerful and dignified response I wonder?
https://www.nature.com/articles/d41586-023-03635-w
We have enough fraud by humans, there's no need to automate it.
The next step is to submit papers with fake data sets to real journals to find out what gets noticed. I'm pessimistic.
"ChatGPT generates fake data set to support scientific hypothesis"
"In a paper published in JAMA Ophthalmology on 9 November1, the authors used GPT-4 — the latest version of the large language model on which ChatGPT runs — paired with Advanced Data Analysis (ADA), a model that incorporates the programming language Python and can perform statistical analysis and create data visualizations. The AI-generated data compared the outcomes of two surgical procedures and indicated — wrongly — that one treatment is better than the other."
"At the request of Nature’s news team, Wilkinson and his colleague Zewen Lu assessed the fake data set using a screening protocol designed to check for authenticity.
This revealed a mismatch in many ‘participants’ between designated sex and the sex that would typically be expected from their name. Furthermore, no correlation was found between preoperative and postoperative measures of vision capacity and the eye-imaging test. Wilkinson and Lu also inspected the distribution of numbers in some of the columns in the data set to check for non-random patterns. The eye-imaging values passed this test, but some of the participants’ age values clustered in a way that would be extremely unusual in a genuine data set: there was a disproportionate number of participants whose age values ended with 7 or 8."
I don't know how much of the screening protocol is automated.
> We have enough fraud by humans, there's no need to automate it.
On the contrary -- if automated "Sokal hoaxes" end up drowning the corrupt, mutual back-scratching, information-withholding, grant-guzzling academic establishment in its own effluent, everybody (other than the grifters, that is) wins.
I think the entire response to Sokal affair was that the concerned journals stopped accepting papers from unknown people. They didn't improve, and they didn't collapse under the weight of incoming hoaxes either.
So the impact of automatically generated hoaxes will be worse for journals that try to figure out the truth, because it will mean more work for them. The journals that don't care about the truth will just look at author's name and decide accordingly.
Naturally they won't drown in submissions from unknowns (any more than e.g. NYT will.) What they might conceivably drown in -- is material from "knowns" who have succumbed to the temptation to automate crapola production. The kind of activity illustrated below:
"You can routinely find lecturers with more than a hundred published papers and you marvel at these paradigms of human creativity. These are people, you think, who are fit to challenge Mozart who wrote a hundred pieces or more of music. And then you get puzzled that, in this modern world, there should be so many Mozarts - almost one for every department. The more prosaic truth emerges when you scan the titles of these epics. First, the author rarely appears alone, sharing space with two or three others. Often the collaborators are Ph.D. students who are routinely doing most of the spade work on some low grant in the hope of climbing the greasy pole. Dividing the number of titles by the author's actual contribution probably reduces those hundred papers to twenty-five. Then looking at the titles themselves, you'll see that many of the titles bear a striking resemblance to each other. "Adaptive Mesh Analysis" reads one and "An Adaptive Algorithm for Mesh Analysis" reads another. Dividing the total remaining by the average number of repetitions halves the list again. Mozart disappears before your very eyes."
( Mark Tarver, "Why I am Not a Professor" -- https://www.marktarver.com/professor.html )
How could a fighter in the medieval ages survive multiple swordfights over the course of their life? If you watch any melee weapons fights like HEMA or Dog Brothers, it's impossible not to get hit if you're in range- any given fight has each fighter hitting the other dozens of times. Any martial art that teaches realistic knife defense starts with the principle 'you are absolutely going to get cut a few times'. Swords are obviously lethal, and medical care at the time was obviously extremely poor- if you didn't die of a sword wound you'd probably die of infection later. It seems like if you were unusually agile, athletic, and skilled you could survive maybe 2 or 3 sword fights over the course of your life.....? How could someone be a 'warrior' for multiple years or decades?
When the combat was between knights/nobles then the primary goal wasn't killing your opponent but capturing them. The ransoms were often so large that the nobles would go bankrupt over it and sometimes be held prisoner for years before they could pay. It was a major aspect of medieval society, politics, and economy.
But the reason they could survive was the knights had shields and armor. The chroniclers also often describe them as very heavily scarred. Without this aspect of capture, then even if you have armor/shield, so does your opponent and they'd just kill you after they incapacitate you.
I was wondering the very same thing not long ago, in connection with someone called William Marshal https://www.medievalwarfare.info/marshal.htm
He was a medieval knight and in later life a politician, and was said to have been victorious in over 500 tournaments, each of which involved two knights gallowing towards each other, at a closing speed of what must have been close to 40 MPH, pointing sharp lances at each other!
(Some tournaments were more like general melees, when everyone galloped around clobbering each other as the chance arose, although I'm not clear what constituted a victory against individual opponents. But they sure as hell weren't using paintball guns, and deaths were not uncommon!)
In a conventional tournament, comprising a pair of jousters, I can only assume that Marshal was extremely fast and adept at using his shield to deflect the opposing lance upon shock, while positioning his own lance with firmness and pinpoint accuracy to strike his opponent off centre somehow in a way that was hard to counter.
Perhaps he also had some trick of aiming his lance steadily for most of the approach, but quickly shifting it at the last moment, like a footballer taking a penalty kick who tricks the goal keeper by kicking the ball to the opposite side of the goal his run up led the goalkeeper to expect.
It may have helped that lances were not always pointed. In exhibition tournaments, they may have had a ball end to prevent penetrating injuries. Even so, he and the other jousters must have been knocked about something chronic over the years, although for the times he lived to an advanced age and in good health as far as I know.
Regarding your specific question about sword injuries in repeated battles, I don't think infantry vs infantry were generally as chaotic or "open" as they are usually portrayed in films like Braveheart, with both sides intermixed and fighting practically as individual pairs on all sides.
Instead of sprinting at each other, the two sides would tramp up to each other in close formation (the "shield wall") and basically most of the action would be pushing and shoving, with sword strokes or stabs as the opportunity arose. Soldiers were actually trained not to exhaust themselves in a matter of minutes, and risk their lives unduly, by madcap berzerker-style leaping into the enemy and flailing wildly as single heros!
Roman legionaries, who of course also took this disciplined approach, were even trained not to stab their opponents too deeply in the torso, just a thumb's depth would be fine. This was because a deep stab might make it slower and harder to withdraw their swords, because muscles could clamp on the sword. Also, a wounded adversary, slowing down through blood loss, is more of a liability for his fellow soldiers to carry away then a dead one.
So if they didn't take an arrow in the face, nor were run through by a cavalry lance, there was every chance an infantry soldier would live to fight another day.
Most of the tournaments the Marshal fought in were melees — single combat became important later.
In the battle of Lincoln, probably William's last fight, the Count de Perche, a leader of the French army allied with the rebel barons, was killed by a lance through the eye slot of his helm and everyone was shocked. For the knights war was just a rougher tournament, with the objective to capture your opponents and hold them for ransom, not to kill them.
Tournaments used blunt lances with weak shafts, designed to break instead of penetrate. Jousting was a dangerous sport, but it was still a sport, not a death-match.
In later times, such as the 1400s and 1500s, yes, but not so much in say the 1000s or 1100s!
1. Shields. They work.
2. Armor. It also works.
3. Swords are not knives. Swords are much more likely to deliver an immediately-debilitating blow in the first good hit, and they deliver that hit from a greater distance, both of which mean that the losing swordsman is less likely to inflict severe retaliatory injuries against the victor.
Re 1 & 2: but don't the two combatants keep going until they've taken the other guy out? It's not like they bounced their weapons off of each other's armor for 15 minutes and then called it a day. You would keep going until you've gotten through at some point- right? So it seems like armor would just prolong the conflict, but not prevent ultimately fatal blows.
Re 3: yes but how can any 1 fighter always be the one who delivers the immediately debilitating blow in, say, 6 or 8 or 12 straight sword fights against other skilled opponents? Maybe if you have a huge skill or athleticism advantage, but if you're in enough swordfights against other trained fighters- I don't see how you can *always* be the one to land first. The question wasn't how could someone survive any 1 fight, it was 'how could someone consistently survive multiple engagements'
You consistently survive by being good enough to not lose. The loser in a swordfight usually gets killed or maimed, that's sort of the point. And fighting with swords is the sort of thing where A: skill really matters and B: skill is hard to obtain. As noted, shields and armor also matter, and serious armor is also in the "hard to obtain" category. So it's entirely possible to be skilled enough and with good enough gear that you have a <10% chance of losing a swordfight to a merely average opponent. Possibly even <1%, but if someone claims to have won a hundred duels or killed a hundred foes in battle, you should be skeptical.
Context matters. On a battlefield, maybe someone runs away. Maybe a buddy of someone comes by. Maybe an archer gets someone. Maybe someone slips. If you're both in something approaching full plate, you're probably rich enough to provide a ransom, so maybe someone yields when it's obvious they're outmatched. Or maybe, and my guess is that this was the "intended" endgame, A knocks B down, steps on B's sword, and quickly slides a sharp pointy thing in between the joints of B's armor. It would be gentlemanly to first inquire about yielding, of course. (The tuck, or estoc, had a bad reputation as an improvised weapon for peasants to kill downed nobles, and at certain times and places possession was considered intent, and was punishable by death.)
Duels are a different thing. Musashi was legendary, even though by modern standards he didn't kill that many people. And when you read some accounts of how he won, it's clear that he didn't "fight fair" if he didn't have to. And even when it comes to "fair" fights, you can go far by picking the right opponents. And even then... We probably have hindsight bias in remembering the victors more than their victims. Possibly the outliers would look less surprising given the proper context.
There’s an under-appreciated aspect of armor, especially helmets: a blow to a helmet can easily cause a concussion that in a heat of a battle would be debilitating enough to be left on the field.
If you survive the first sword fight then maybe you will get enough experience points to level up, and then you will have a much better chance of surviving future fights. Up until you run across a similarly (or higher!) leveled fighter, when you're back to square one unless you have superior tactics.
Skill. There's advantage in being trained by someone who knows what they're doing, or even just being able to observe them. We've lost that.
Necessity. People act differently when their life is actually on the line. Compare to poker and money.
Armor. Is awesome, even when it's just a few layers of cloth wrapped around your arm. Swords aren't lightsabers.
Context. Civilian duels aren't the same as battlefield combat, or prisoner extraction from a cell, or stopping an assassin with a knife, or dealing with Uncle Sven when he's had too much to drink at the feast and tries to stab your in-laws.
Grazing wounds were survivable. Perilous die roll on infection, of course, but a smallish or even moderate cut was not a death sentence. I'm reminded here, for obvious reasons, of the relatively well-documented injuries of Alexander the Great, from a shoulder injury at the siege of Gaza (yeah, that Gaza) up to a sucking thoracic wound from an arrow during the Indian campaign.
Armour, even lower-end medieval gambesons and cuir bouilli, meant that a HEMA-style touch on protected areas was not necessarily injurious. As European armour evolved to full plate, the balance swung further toward relative safety for those who could afford it.
Also, HEMA-style practitioners, despite their commitment to the treatises, almost certainly fight one-on-one with less caution and skill than the professional warriors they emulate. (To the extent the latter did at all; the frequency and significance of single combat, apart from judicial duels, is probably over-represented in chronicles and literature.) If we're talking mass battles instead, then the usual statistics apply (most participants survive most encounters, etc.)
This is a good comment, but my takeaway is that wealthy knights who could afford armor and grew up training probably survived multiple engagements. The average peasant with no real armor who was drafted into a mass battle- probably not so much
Drafted peasants probably didn't get to 1v1 swordfights much. They used spears or similar things and they formed a mob of spear wielding peasants. Sword in general was more akin to personal handgun than to a battlefield rifle - that would be a spear, or lance for cavalry.
I'd extend 'wealthy knights' to 'professional men-at-arms', which includes infantry retinues maintained by noble lords, as well as mercenaries.
The average peasant levy was in for a dire lottery, yeah, and somewhat boned when in the front ranks or under sustained missile attack. Depends on when and where in Europe, but usually he'd still have a helmet, a shield and an axe, and probably do most of his fighting using a polearm of some sort. Of course, if he was lucky, the battle might be over by way of one side breaking by the time he got anywhere near the fight, and maybe no one bothered running him down on horseback while in retreat. And if he got really lucky, as at Crecy, he might have gotten to butcher some exhausted dismounted toffs in fancy armour whom nobody cared to ransom, which honestly would have made the whole thing worth it.
Well yea, I believe there is lots of historiography supporting that summary. And in addition to training and equipment don't overlook general nutrition: a thousand years ago the wealthy and their immediate retainers ate _much_ better than the peasantry did.
Being a half-starved peasant conscripted into a medieval army and handed a sword and small shield, both items being inferior in quality to what the knights were using, and then given cursory instruction in their use, and then sent into an actual pitched battle, was pretty close to a death sentence.
> Being a half-starved peasant conscripted into a medieval army ..
You'd be surprised. They were incredibly tough by modern standards, despite their spartan diet. For example, most English foot soldiers at the Battle of Agincourt were suffering from dysentery, probably from their ridiculous superstitious habit of eating a mouthful of earth before a battle, and had been throughout the march from Harfleur for weeks, having little to eat all that while. (A few years later, Henry V himself died of dysentery. So it wasn't just the grunts.)
Something else to consider is that there weren't many battles in a typical medieval campaign. They spent a lot of time marching around, but little actual major action, maybe usually only one or two skirmishes per year.
Well....Agincourt was the late stages of the Middle Ages though. As I mentioned I was thinking of more like a thousand years ago. More Henry I than Henry V.
Same thing that for British pilots during WW2? The vast majority of them were shot down after just a few flights, but some rare pilots manage hundreds through a combination of better skills and luck.
In a real sword fight the first hit would often be debilitating enough to determine the outcome of the fight, even if it isn't fatal. No one gets hit dozens of times.
What makes you think that an average 'warrior' would survive constant sword fights over a decade? Is this based on fictional portrayals or are there specific historical figures you are curious about?
Fighter pilots are probably a good analogy for this, and have better record keeping. Shooting down 5 or more aircraft makes you an 'ace' and is considered exceptional. The highest scoring Ace, Erich Hartmann, shot down 352 planes in his career. I don't know exactly how comparable this is, maybe most of those were fighter plane vs bomber, but shooting down that many planes and not dying (although he was shot down) does give some credibility to your question.
I've seen the claim that the unusual score distribution for ace pilots is a consequence of policy, not just skill. If you shoot down other aircraft, you get promoted into a squadron leader position where other pilots support you, and then you're much more likely to continue shooting down planes - while enemy aces will probably take down your support, instead of you, which makes your career longer.
I don't know how accurate this is, but it makes sense if you imagine that most pilots just don't have the ability to win aerial dogfights, and are best used as support for those who do. It is also good incentive for people to take more risks, because taking risks to shoot down the enemy puts you in a better position to survive later.
I feel like both of these are reasonable to some extent for fighters in the medieval ages - but the question is, how meaningful is the support that other people can provide you?
I don't think "other pilots support the squadron leader" is really a thing, at least not any more so that "pilots support each other". I mean, how do you even know where the squadron leader is?
Pilots typically do have wingmen, who support each other, but that's pretty much independent of rank or status. Whoever you are, you've got *one* pilot who will keep close and support you, in exchange for your doing the same for him.
At least one prominent contrary example on the fighter pilots was Erich Hartmann, the most successful German ace of World War II and sometimes referred to as the most successful ace in the history of aerial combat (352 kills, the last one occurring literally hours before the armistice). He served as a squadron leader only intermittently in between stretches as a loner who developed and carried out his own particular tactics.
Armor.
Thank you for the reference on Professor Daniel Kang's work. I've recently started the newsletter Provably Safe AI and published the second post today. It covers attacking and defending LLMs, as well as policy and standards around AI-relevant cybersecurity, so Prof. Kang's work will be a good follow-up to share next. I'm new to the topic (and to writing in this format) and would appreciate corrections and feedback: https://www.provablysafe.ai/p/2-provably-safe-against-ai
What do you think would happen if America stopped supporting Israel?
Not too much (even assuming you mean "stops selling weapons" and not just stops active support). The US didn't support Israel for most of its early years, Israel found other countries to bug weapons from and was mostly fine.
Probably Iran feels emboldened so Israel gets more terrorism, and Israel has more arms deals with China or Russia (which the US would prefer to avoid since Israel has a lot of tech they'd rather China doesn't get).
I don't think Israel would go literally genocidal on the Palestinians, but they'd want Gaza and the West Bank very firmly locked down before they redeploy to deal with the rest of the Arab world. So, a very bloody war in Gaza, followed by a shoot-first-and-ask-questions-later occupation. This would be worse for the Palestinians than what is happening now and likely to happen in the near future.
Then the rest of the Arab world (loosely defined to include Iran and indirectly Pakistan), will start sizing up the new situation. They'll see opportunity, to finally get rid of the hated Zionist foe without US meddling, but also fear that a still-powerful but now unchecked Israel might try to preemptively go for Greater Israel in the name of strategic depth. Israel, will respond to that in the usual way.
This *probably* wouldn't result in open high-intensity war a la 1967 or 1973, but it could. And a fair chance that it would go nuclear.
Absent the full-scale war, you'd get a hard-right national security government in Israel, and economic sanctions from much of the rest of the world. Israel would reach out to Taiwan, and to anyone else who might accept a sweetheart deal from Jerusalem in spite of the diplomatic costs. Some parts of the world, the diplomatic cost for supporting Israel wouldn't be that high.
This would still weaken Israel over time. So would the preferential emmigration of liberal Israelis dissatisfied by the hard-right turn, and well-eductated Israelis dissatisfied by the diminished economic opportunities. As Israel turns weaker and harder to the right, the temptation for that war (on both sides) will increase.
Also, US allies will reevaluate the value of US security guarantees, and the enemies of US allies will see opportunity, Good chance either China invades Taiwan, or Taiwan accepts reunification on Hong-Kongesque terms rather than face an invasion without US assistance.
All of this would be bad.
Short-to-medium-term, Israel acts much more assertively in order to reestablish a similar level of confidence in its continued existence to what it has with current US backing.
Total US aid to Israel is $3.8B/year, or about 0.7% of Israel's GDP. Making this up by raising taxes, cutting spending, or borrowing would be a bit uncomfortable but probably not untenable since Israel is currently a little below the OECD average tax-to-GDP ratio.
As a political matter, historically US financial aid to Israel has served as an inducement and enforcement mechanism for their 1978 peace treaty with Egypt, brokered by President Carter. Part of the understanding was that the US would give economic aid to Egypt and military financial aid to Israel following the treaty, incentivizing both countries to sign and also giving the US leverage to punish violations. And giving each side confidence that the other would be more likely to honor the agreement rather than risk losing the money.
I'm not sure how much that aspect is still necessary, seeing as the treaty has now held for 45 years, but cutting it off abruptly seems like it might trigger some awkwardness.
The biggest support to Israel is not the billions of dollars - it's the implicit guarantee of protection against its neighbors. Nobody reasonably believes the US would allow Israel to be destroyed by the countries that hate it, and this surely explains part of why they never try.
It depends on what you mean by a "stop"
In addition to stopping financial, intelligence etc. aid to Israel, explicitly announcing to the world that the US will not intervene is any way in any conflict involving Israel.
No longer sent money or military supplies.
No difference. US military aid is a tiny fraction of GDP nowadays. Some Israelis actually want this as they think it would give them more political freedom of action.
What about UN veto? Would Israel suffer global sanctions without it?
Sanctioned for defending themselves from genociders? While the genociders get billions in "humanitarian aid"?
It's genocide to defend yourself against foreign invaders?
Of course not, Israel is fully justified in defending themselves from the foreign invaders that attacked on October 7th.
I mean, I can think of at least one country which is heavily sanctioned even without a Security Council resolution.
Had e.g. China wanted to sanction Israel, they could do that right now without waiting for the UN. In fact there was once a non-UN-sanctioned OPEC embargo connected to Israel.
I.e. I heavily doubt that US veto power in Security Council is the only thing that stands in the way of global sanctions on Israel.
1) Probably more indirect and direct Iranian intervention against Israel.
2) Factions in surrounding Arab countries that supported rapprochement with Israel would be considerably weakened.
3) After a period of destabilization, the Israeli right would emerge as the new Israeli center. Defense spending would increase. Moderating impulses would be abandoned.
4) Countries all over the world would begin to reevaluate the value of American promises, especially security promises. I would expect some amount of regional realignment. China would be emboldened in SE Asia, Russia in the Ukraine, etc.
5) I don’t expect the prospect of Palestinian statehood or independence would materially grow; the Israelis are perfectly capable of subduing them on their own. You might see the Israelis cut them off completely from utilities on their side.
Lots of follow-on effects. Difficult to predict. I can’t spend any more time thinking about it today but I don’t think it would be good for those who think it would be good.
If the US totally divested militarily, financially and politically, the Chinese could roll into Palestine as a protection force. So could any Arab country I suppose but China isn’t threatened by Israeli nukes. Why would China do this? To gain popularity in the Muslim world.
>Why would China do this? To gain popularity in the Muslim world.
They don't need any more popularity in the muslim world. They're powerful, friendly to them and not the west - anything else would be overkill.
If China rolled into Palestine they'd have to go Xinjiang on them. I'm not sure Israel would mind.
Why? They would be allies. This is in a situation where the US is totally isolationist.
Because the Palestinians would start doing terrorism on them and trying to overthrow their government, just like they have with every other ally (and Jordan and Lebanon didn't even try to come in and control new territory they want).
It would be an occupation per se. Just an alliance with some ports belonging to China etc.
I'd imagine it would increase the chances of direct retaliatory actions from nearby countries like Iran.
Well, I wrote a thing about five now-defunct Flash games and how to win them (with a bit of rambling at the start about the enshittification of free browser-based games, but you can skip that if you like).
https://lettersfromtrekronor.substack.com/p/how-to-play-and-win-five-now-defunct
I posted the other week about using tiny doses of antidepressants as a form of open label placebo and subsequently saw this article posted on Twitter https://academic.oup.com/book/54240/chapter/422452976?login=false
Has it been discussed here on ACT?
What are the best resources to learn how to implement LLMs in real life?
Context : I’m a surgeon based in the UK NHS in the field of orthopaedics. I think that a LLM-based chat bot would be really helpful to navigate the ‘National Joint Registry’ which is basically a huge database of medical implants, outcome data and survival figures.
I have no idea how to implement this.
Does anyone have any thoughts or guidance for me?
ChatGPT 4 (paid) now lets you create your own specialised chatbots. I’ve done this and uploaded documents to the chatbot that are past the cut off date for ChatGPTs last major update. I sometimes gave to tell it to use the new documents in its answer. It’s probably a bit tedious to upload all these documents for what you want, there may be a way to automate it.
This is what I had experimented with - just uploading the annual reports etc then asking GPT-4 to sift through the data to give the answers I want. I think with some training this could work
You're going to be running into hallucinations if you use an LLM for this. You would have to do like 100 tests of queries to this registry, where you compare what you are sure is the right answer to what the LLM says, to get a sense for the error rate. Even then, there might be weird edge cases that make the error rate climb...
There are frameworks like Haystack that let you use LLMs to query your own data. You need to write a little bit of Python code to glue everything together, but you can basically just download their tutorial example and then swap out their example data set for your data set, so you don't need to write any code from scratch.
I don't think you want to create an LLM from scratch; rather, I think you probably want to augment an existing LLM with access to the database (data protection rules permitting, of course). Have you done any coding? If so, a library like LangChain could be a good place to start: https://python.langchain.com/docs/additional_resources/tutorials
I really appreciate you guys replying. I haven’t done much coding but I’m open to learning if it would be helpful or necessary. I suppose the question is whether it is optimal for me to learn to code from scratch or just to hire someone to do it.
Is it likely to require servers etc to implement something like what I have described?
Neither of those sound like easy options, to be honest, though my experience of freelance coders is "being called in by friends to look at the results after the relationship with the freelancer has already broken down" so may not be representative.
You'll need somewhere for your glue code to run, which will need to be able to make calls to the LLM and to the database. That can be on the user's machine, but I'm guessing you want to make this web-based? In which case, yeah, you'll need a server of some sort. You could probably use a platform-as-a-service system like https://fly.io/ or repl.it for that part, which would probably be easier than managing physical servers (or even virtual machines in AWS) - though possibly more expensive. Free hosting with generous compute limits is a lot harder to come by in these days of cryptocurrency scams. I don't think you want to host the LLM yourself (as opposed to making calls to OpenAI's API), though I have no direct experience here. I have heard that open-source LLMs are getting more powerful and easy to run, via tools like https://ollama.ai/ And oh look, the Ollama blog has an article about doing something similar to what you propose, with most of the code running in a browser and talking to an LLM running on the client's machine: https://ollama.ai/blog/building-llm-powered-web-apps That doesn't sound easy, though, and also I think you'd have trouble getting clinicians to run LLMs locally.
Another option to look at is OpenAI's recently-announced "GPTs", which are customised GPT-based apps running on their infrastructure: https://openai.com/blog/introducing-gpts They say that building one requires no code, but I don't know how true that is. I also suspect that directly connecting an NHS database to OpenAI's infra would result in some extremely harsh words from their Information Governance people.
Here is an example of using LLM stuff to look up articles
https://eli.thegreenplace.net/2023/retrieval-augmented-generation-in-go/
I've recently found the amount of antisemitism on Twitter anxiety inducing and pretty harmful to my life. Is there a good way to completely block the ability to access it from my phone? I can do it on my PC, but I haven't found a way to IP block from my Android phone
That's weird - for most of twitter's history, hundreds of thousands of its users have seemingly spent much of their life posting about how much they hate white people. Was that upsetting to you? Or is it only a problem when it affects the group you identify with? Was it harmful to your life when schools, universities, books etc. starting telling the world that white people have "white privilege"? Were you angry when jewish groups failed to condemn this but instead simply demanded that jews be excluded from the concept?
a) yes, it did bother me, but obviously it's much worse when they're cheering people who have murdered friends of mine and shelled me out of my home, in real life, than when they're cheering abstract hostility.
b) you might want to go through your comment and try to think if it's the sort of thing that would make people more sympathetic to you or to your enemies.
Hey, I’m sorry to hear it’s affecting you that much. I’ve seen a lot of hateful stuff lately. Unfortunately, I think it’ll be at this intensity for a few more months at a minimum.
Is the issue that cutting the cord with Twitter is a hard habit to break? Or are you seeing it embedded/screenshotted on other sites when you don’t want to?
I’m on an iPhone, but I assume there’s something similar. I wanted to block web usage more generally, so I had a family member use the parental controls to lock me out every web browser and app that I didn’t want, and then put the code in a sealed envelope. I can get past it if I’m determined enough, either technically or by getting the envelope. But it’s enough friction that I haven’t done so.
Cutting the cord really - I get into the social media doom loop of seeing hate and both worrying about it and feeling compelled to go back to it, which I know is bad for me.
Looks like there's similar apps for android, hopefully this should help.
I use Stay focused for android and Cold Turkey for MacOS, for the same purpose (blocking Twitter). They're great and blocking Twitter has had a noticeably positive effect on my well being
Thanks! Looks like that should work
Uninstalling the app is a bit more radical but you could try. You can tell yourself that if you really need to look something neutral up you can just open it on the mobile browser, and the extra awkwardness of the browser vs the native up might help you avoid getting stuck in it.
Congratulations on your life choices, for Android Chrome browser I use an extension called AppBlock. It often sends useless notifications, but it works.
You know you can disable any app's notifications from Android's settings, right? On my phone only a few apps can notify, and most of them only silently. Don't let a single unwanted beep go through!
One thing I've found that completely changed my Twitter experience was making a Twitter list with people/orgs in my current interest area (nuclear and fusion power) and having a direct tab link to it
When I first got into Twitter I got algorithm'd into an emotional whirlpool of political concern about woke stuff. There should really be a large revenue neutral social media tax that taxes the more anti-social platforms and funds better ones
I don’t have twitter but when I did I made heavy use of the muted words feature:
See: https://help.twitter.com/en/using-x/advanced-x-mute-options
Although you can also quit twitter.
I had a serious look into doing this, but I couldn't find any easy way to do it. In principle it should be possible to modify the hosts file if you root the phone, but that was too much effort for me to bother.
Beyond deleting the app, perhaps you can change your Twitter password on your computer to something long and complicated that is too much effort to either remember or type on a phone? When I've quit social media in the past, simply logging out or deleting my account and then making it hard to log back in or make a new one has worked quite well.
Has Scott written about “The Body Keeps the Score”? I’ve looked all over for informed reviews, but just find commentary by book critics and laymen.
I am a person intimately familiar with PTSD and CPTSD, so I suppose that makes me an expert or an unreliable witness. (Funnily enough that tension has a lot to do with PTSD; losing the ability to trust yourself.). I have rather a high opinion of VdK's work, and I read Scott's review with interest.
I have some thoughts on it.
VdK is quoted as saying ‘WE were not designed for this” or something like that. Scott says ‘yes we were’ and points to the long history of humans suffering all kinds of trauma. That is true, but I think it misses the point. First, the condition is called PTSD; POST-traumatic stress disorder, not Traumatic stress disorder. The whole problem lies in the framing of the trauma, not in the trauma itself. To give examples;
1. a hurricane destroys your community leaving everyone without shelter and whatever other havoc it created, up to and including deaths. There is a scale of suffering clearly (some people lose a parent or child, some people only lose their material property) but the important thing is it affected everyone, NOT JUST YOU. This is important. This kind of trauma is easier to process; it doesn’t need to be taken personally.
2.A person is attacked/abused/neglected by people close to them, and the outside world doesn’t see, or seem to care. This is more difficult to process. The horror of it is internalized. Trapped. Blamed and punished. Betrayed. That is POST traumatic stress.
If you are a child and these are formative experiences and you are powerless to escape them, then it’s really bad for your psyche. An adult can seek confirmation of the trauma in the outside world, and maybe find some relief, in that they are not alone and the world around them confirms that what happened to them is an aberration. This is where another of Scott’s comments comes to me, The landscape of mental or emotional disorder is very dependent on what the norm is around the person in question. In other words, in a world where mortal combat is an accepted part of living, slavery is an accepted part of living, physical pain is an accepted part of living, incurable diseases and all the rest are an accepted p o l, then the baseline for what we consider trauma is different. People don’t get traumatized by seeing someone hacked to death because they grew up in that world and it’s just the way things are done.
Animals process adverse physical events in a very straightforward way. They might survive them or not. Then there is the whole fight flight freeze thing, which Pavlov’s work pretty much lays out in great detail bye creating a large number of perpetually frightened dogs. I know that was not his point but the idea of what PTSD is, is implicit in his work imo. Something is trapped, and that something is visceral, not cerebral. Inevitably though, because we are humans it becomes cerebral, and that is where the fun starts. But what if it never had to become cerebral? What if we let our body deal with its injury, and didn’t interfere? That would be great wouldn’t it? I know some people who are better at this than others, but regardless, we have to unravel the story behind the trauma and the story we are telling ourselves about it. That’s where PTSD lies.
The story of Scotts patient is instructive; she cured herself. How? By reframing her narrative of the trauma in a way that she could accept. I would be curious to know if there was a physical component in that moment (crying, a show of anger).
I guess I think VdK is on to something.
Thanks for your informed reply, I found Van der Kolk's book eye-opening, so it's good to get some confirmation from an unrelated professional.
For anyone who wants to explore more material in a similar vein, I can recommend Gabor Maté, Stephen Cope and Deb Dana.
I am a fan of Maté.
I am not a professional by the way. I am flattered.
I know of two early accounts of PTSD in fiction. In LOTR, Frodo's injury from a poisonous blade toward the beginning of the book haunts him physically and mentally, including feeling worse on the anniversary of the injury. There is no cure in this world.
In _Citizen of the Galaxy_ by Heinlein, Thorby is haunted by being kidnapped into slavery as a small child, with a short temper and nightmares as a consequence. He's eventually cured by hypnosis (memory of hypnosis in a dream after his life is greatly improved?) which shuts the bad memories away.
Any other examples?
As for trauma being a routine part of life for early humans (I think trauma could reasonably include any sort of bad injury) it could be like bones. Bones are pretty strong, but sufficient force can break them.
There is something essential missing in this. First, the condition is called PTSD; POST-traumatic stress disorder, not Traumatic stress disorder. The whole problem lies in the framing of the trauma, not in the trauma itself.
If you’re interested in this question, Achilles in Vietnam by Jonathan Shay you is a must read. He use’s the Iliad as a metaphor for ptsd, and argues that in the case of complex ptsd, the trauma most often has some moral component.
I have the book, I should read it.
Does a moral component mean being betrayed by what should be a supportive person or organization, or being complicit in immoral behavior? Both?
Being betrayed by a caregiver or parent is a big component of PTSD. Being betrayed by anyone you have put trust in is traumatic, but if you have counter-examples already (you have been treated fairly by others), and then WHAM a bad thing happens, it is easier to recover a sense of trust. If it is a formative experience and goes unaddressed it becomes an attachment disorder essentially. That manifests itself in all sorts of ways.
But soldiers are not betrayed and are not alone in their trauma but PTSD is common.
I did not address that particular one, because I really don’t have a clue about it. I think there might be betrayal, if one feels they saw, or experienced a sacrifice that turned out to be meaningless. There is also survival guilt, perhaps seeing a comrade of yours destroyed, and you walking away from it. And always there is the whole history of one’s life that one is bringing to the experience. It’s complicated. The reframing of the traumatic experience is still the only way out though, imo. That could be as simple as crying it out or as complicated as a novel. Ymmv
Sometimes soldiers feel betrayed by incompetent or corrupt leadership.
Yes, see https://slatestarcodex.com/2019/11/12/book-review-the-body-keeps-the-score/
I’m looking for some recommendations for sci-fi novels. Specifically, space colony type novels. Does anyone have any favorites they’d recommend from that sub-genre?
+1 to Kim Stanley Robinson's Mars Trilogy. It's about multiple generations of early Mars settlers establishing a colony/society and terraforming the planet. It might be my favourite series ever. It's a little slow in parts because of how much detail he goes into, but the detail is also what makes it amazing.
Aurora also by KSR is excellent as well and deals with similar, namely the generations long journey to a potentially colonizable planet in Tau Ceti
From my childhood, The Escape Orbit is an interesting subversion of "colony growth" with the colony being a prisoner-of-war planet advancing to rebel against their alien captors.
1- The Expanse, the first 3 novels are partially set in a prosperous (10 billion humans) Mars and populous Ceres, Europa, Ganymede, and Eros. [MINOR SPOILER] The 4th book is entirely set on an alien planet, and starting from the 7th book onwards a whole network of several alien planets.[END MINOR SPOILER]
2- Shards of Earth trilogy
3- House of Suns by Alastair Reynolds, as well as the Revelation Space series by the same author (what I read : Revelation Space, Redemption Ark, Absolution Gap)
4- Takeshi Kovacs series by Richard K. Morgan : Altered Carbon, Broken Angels, and Woken Furies. (There is a show but the first season is inferior to Altered Carbon, and the second is outright trash). As well as the Black Man novels, Black Man (Thirteen or Th1rte3n in the United States) and Thin Air. There is an upcoming Gone Machine but it keeps getting delayed, this is an extract from it https://www.richardkmorgan.com/2021/03/gone-machine/.
JEM by Frederik Pohl.
Legacy of Heorot by Niven and Pournelle for colony on a planet.
For space station life, try Andrew Moriarity's adventures of a jump space accountant books. "Trans Galactic Insurance" is the first in the series. Interesting deep dive into the economics of keeping space colonies runnning.
A great deal of Jack Vance is set on colony planets. Maybe I can dredge up more, but _Big Planet_ (gigantic low density planet with weird little cultures), _Blue Planet_ (low metal culture), the Marune novels, the Planet of Adventure series (protagonist must gather four plot coupons (parts for a ship to escape) across a planet with alien species), _Emphyrio_.
Oh, absolutely Rosemary Kirstein's Steerswoman series. It starts out looking like fantasy, but... The series has been on a long pause, but I gather a new book is in progress.
Oh, Big Planet looks like it has a really interesting setting. Thanks for that!
Could you go into slightly more detail about what you mean by "colony type novels"?
Mostly set on a space station: "Ethan of Athos" by Bujold. Her "Falling Free" deals with the creation of a space habitat, with more focus on engineering than most (probably the familial influence). V. Vinge's "A Deepness in the Sky" might as well be about that, plus there's the "first contact" aspect too.
If you want messed-up scenarios of planet colonization, Niven (et al.) wrote a bunch of them. Also "Dragonsdawn" by McCaffery.
The Bujold novels are great. Off topic for this question, but I actually like The Sharing Knife books and The Curse of Chalion more than her sci-fi.
They're also quite good.
Bujold's Vorkosiverse series is mostly about colony worlds (as I recall, there's a little on Earth). Komarr has a lot about what it means to be a colony.
Heh, yes, both a colony in the sense of "long-term terraforming of a marginal world, and the social and economic impacts of the environment", and in the sense of "outpost of an imperial power". I'd never thought of that before.
> Could you go into slightly more detail about what you mean by "colony type novels"?
I guess I was thinking more about colonizing a planet, rather than a space station. I was thinking more like the "generation ship lands on a interstellar planet" type story.
I haven't read anything by Niven, which would you recommend starting with for messed up planet colonization?
Sorry this is so late. I was misremembering Niven; "Legacy of Heorot" and "Integral Trees" loomed large enough in my mind that they created an impression of more. I think the only other one that might fit is "A Gift From Earth".
I'd again recommend "Dragonsdawn" by Anne McCaffrey. It's kid of my baseline for these sorts of stories. It's got the whole thing, starting from 3 colony ships inbound at the edge of a system, to filling in details missing from the 200-year-old survey records, and the tensions between the different groups of settlers (charterers who put up the capital, contractors who were hired as technical specialists, and the remnants of Earth's nomads who were forced onto the expedition), and people with ulterior motives, and a disaster happening a few years in (which isn't a spoiler if you know anything about Pern).
Another one that might work for you, but which wouldn't be on most people's radar, is "The Romulan Way" by Diane Duane. It's one of the Star Trek novels from the mid 80s, right before ST:TNG started, presenting a different take on the Romulans than what canon eventually went with. Its chapters alternate between the "present day" story, of a Federation operation to retrieve a deep-cover anthropological researcher who's gone native, and a historical overview of the Romulans, from Vulcan's first contact with aliens, to the schism between Surak and his leading disciple, to the long mostly-fatal voyage on generation ships, the colonization of a new system, and the broad strokes of the history the survivors wrote in blood and wine. I know the ST novels don't have much of a reputation, but this is a good one. It's like something an early Le Guin might have written, back in the "Rocannon's World" days. (And since I'm mentioning ST novels, the absolute best of all is "The Final Reflection", by John M. Ford. It's a mystery disguised as an anthropological study disguised as an adventure, this time involving pre--3rd-movie Klingons.)
Ha, it’s funny you put that because you mentioned Dragonsdawn earlier and I looked it up and I’m literally reading it now. I just got to the disaster. I haven’t read any of the prior books, but when I was reading through some reviews I found out what it would be. It’s a bit strange to start with Book 9, but some people suggested that you should start the series with that, so that’s what I’m doing. And you’re right, it’s exactly the sort of thing I was looking for. After Dragonsdawn, do you think I should just go back and read book 1?
I haven’t read a lot of Star Trek books, but I did read a lot of Star Wars novels, so I’m open to that sort of thing. I actually still have two full shelves on my bookcase that are just Star Wars novels from when I was a kid.
Thanks for this. People gave a lot of really helpful suggestions. I've definitely got a couple months (or more) of reading out of this.
Regarding Pern, yeah, definitely try them out. They're classics for a reason, and they're easy reads so they won't take much time.
The "main 3" books are "Dragonflight", "Dragonquest", and "The White Dragon," which all take place about 2500 years after "Dragonsdawn", when humanity has retreated to the Northern Continent, forgotten their history, and descended into a sort of low-violence cave-dwelling feudal stasis. "Dragonflight" is a must-read classic, "Dragonquest" is 7-years-later exploring all the problems caused by the solutions in the first book, and "The White Dragon" is when they really start uncovering 2500-year-old archaeological sites.
The "Harper Hall" trilogy are young-adult books that are probably best read after "Dragonquest" and before "The White Dragon". I'm very fond of them, not the least because they provide a ground-level view of the society that's missing from the main 3. Like, I think they're the first time that money is mentioned! (You can probably tell that I like the sort of stories where there's a plot that allows us to see the day-to-day life of alien cultures.)
After those, there's "Renegades of Pern", which spans over all the previous books and provides even more of a ground-level view of the seamy underside of Pern. And, uh, I won't spoil what happens in that and in "All the Weyrs of Pern", but, well, if you're reading ACX, you'll probably be quite interested, and I'd love to compare impressions if you make it that far. ;-)
There's also some books set afterwards, and during, and some books set in earlier eras, but those aren't essential IMO. Like "Dragonsdawn", the ones I've mentioned are what I consider to be a reference-class implementation of a classic sci-fi trope.
Regarding the Star Trek books, if you're open to it, maybe read "My Enemy, My Ally" before "The Romulan Way", as it was written first, and introduces a few characters and starts off the world-building. It's also a great read, in the David Gemmell "honor porn" sense, which is not something I'd usually expect to say about Star Trek or Diane Duane. :-)
Have fun! I'm really curious to hear what you think after you're done. :-)
I would personally recommend against Niven. "Ringworld" was terrific up until the cast lands on the titular ring, after which it becomes a boring and tedious plot that I trudged through, aching for it to just conclude already. YMMV, of course, it is fiction after all.
The only thing that comes to mind as a recommendation is *very* tangential: "The Three Stigmata of Palmer Eldritch" by Philip K. Dick. The possibility of being sent to Mars as a colonist is considered a life-changing threat to people in-universe. But, it will definitely tickle the "messed up" half of your request; being sent to Mars as a colonist is a death sentence.
A short description: The future Earth of 2016 is in such a climate crisis that the United Nations has instituted a draft of sorts, sending people off to other planets and moons to colonize them. It is too dangerous to be outside on Earth in the daytime without specialized gear.
The experience as a colonist is uniformly awful, and people resort to an illegal hallucinogen to cope, which requires the use of a prop called a "layout" in order to give the user a temporary alternate reality. The novel's main character works for the company that makes the popular Perky Pat layouts. A famous lost explorer, Palmer Eldritch, is rumored to have returned with a new substance that is better, in that it doesn't require a layout.
I second "Three Stigmata..." One of PKD's best.
"The Integral Trees" comes to mind, and there's even technically a planet involved. ;-) I'll try to come up with some more in a bit, but as mentioned below, "The Legacy of Heorot" would be in there.
The Integral Trees looks like it might have the most unique setting I've seen in a sci fi novel. That'll definitely put it towards the top of my list.
The Moon is a Harsh Mistress is probably the best book I’ve read that is specifically about a space colony.
I think Kim Stanley Robinson’s Mars trilogy, and his novel Aurora, are probably necessary reading in that sub genre but I don’t really recommend them. They aren’t quite as erudite and hard sf as some have made them out to be—a lot of scientific terminology covering over the handwaving of underlying concepts—but they are well written in the sense that you’ll keep turning pages. However KSR is a doomer and most of his characters don’t really have agency.
Lord of Light by Zelazny is definitely about an a extraterrestrial space colony, and is brilliant, but most people would probably be annoyed at its inclusion in this list because it meets the letter and not the spirit.
Ringworld is not the letter but it might be the spirit.
I think Red Mars, first book of the Mars trilogy, is pretty good -- can't quite say the same of the rest of it and Aurora, though.
Lord of Light has tech so advanced it might as well be magic. Good flashy novel, though.
Is there any good sf that does a more authentic job with Hinduism?
"Learning the World" by Ken MacLeod, about a generation ship that arrives at their destination and discovers that they've unexpectedly made First Contact. It's something of a satire on the present-day tech startup industry, but also much more than that. One of my favourite SF novels ever.
I enjoyed Heinlein's "The Moon is a Harsh Mistress"
Thanks, I haven't read that one, but Heinlein's "Tunnel in the Sky" is one of my favorites so I'd probably like his others.
Came here to say this. Heinlein is the best at scientifically realistic content.
The worst thing I think I could say about TMiaHM is that it's not much concerned with the technical ins and outs of building and managing a colony, assuming that's what the OP is looking for. It's *much* more about the social ins and outs.
KSR's Red Mars is about both. But even then, as I recall, the technical stuff felt mostly tacked on to the social stuff.
Heinlein's _Red Planet_ is also very good, and I think it's more plausible.
I will second Theo's recommendation of the Bobiverse. I listened to the audiobooks and they were fun.
I just started the Sun Eater series by Christopher Ruocchio and it is off to a great start for epic sci-fi. I don't think it quite fits the bill for a colony theme, but I happen to be reading it right now and it's really good.
Adrian Tchaikovsky's Bear Head is a very good Mars colony book. It's a sequel, but the first one (Dogs of War) is fantastic and worth reading anyways.
Reading Kim Stanley Robinson's Mars trilogy is my #1 sci-fi experience
Hmm. I'm. It entirely sure what counts as "space colony", but this series is about a colony of exiles.
https://www.glynnstewart.com/universe/exile/
I like Gylnn's work in general.
I think the Bobiverse books might qualify, and those are good to.
And I will always recommend All Systems Red by Martha Stewart. Thought maybe the Raksura series is more colony oriented, but that's more fantasy.
Huh, Bobiverse wasn't exactly what I was looking for, but it definitely looks like a fun plot.
The other Scott blogs about (among less CW relevant stuff like his take on OpenAI) about Greta chanting "crush Zionism", linking the SWC as a source. "[She is] taking time away from saving civilization to ensure that half the world’s remaining Jews will be either dead or stateless in the civilization she saves."
I think I like that framing better than the one of the SWC, which is simply calls that murderist...err ... antisemitic. Personally, I think Zionism has as many interpretations as feminism, some of which I find myself totally on board with (continued existence of Israel) and some I strongly oppose (settlements in the West Bank). That makes it a useless word for anyone aiming to communicate clearly, but the general vibe I get from "crush Zionism" is still one of calling for the destruction of Israel.
While one can oppose a state without being racist towards its inhabitants (for example, I would be on board with "crush the Aztec Empire"), I think that letting Israel continue to exist is actually a great idea. There may well be a statistical correlation between antisemitism and antizionism, but I think that the narrative of Ms Thunberg picking up an old diary of Adolf Hitler and getting infected with antisemitism is probably not what happened.
https://scottaaronson.blog/?p=7632
https://www.wiesenthal.com/about/news/swc-condemns-greta-thunberg.html
https://slatestarcodex.com/2017/06/21/against-murderism/
Like Scott Aaronson, I too admired Greta Thunberg when her mindless stridency, black-white thinking, childish ignorance, and inability to grasp nuance brought her down on the same side of an issue as me. I just couldn't have foreseen this would ever come back to bite me, though.
Man, who would have thought a 20 year old with Aspergers and OCD wouldn't be ready for a role as our chief moral arbiter?
I find myself seeing no less than 4 claims/assumptions bundled into S. Aaronson's seemingly simple statement :
1- "Crush Zionism" is a call to action and/or a planned course of action in Greta's mind, and not a generic "boo outgroup" said for emotional bonding with the ingroup and emotional posturing against the outgroup
2- Zionism is a pro-Jewish ideology identical with the proposition of having a Jewish State for Jews
3- Having a Jewish state for Jews is a good thing
4- Destroying Israel will result with high probability in its Jews being dead or stateless
My reservations about those 4 claims are as follows :
1- "Crush Zionism" is suspiciously identical modulo language to the second part of the main chorus of the song Leve Palestina ([Long] Live Palestine) by the mixed Swedish-Palestinian band Kofia https://www.youtube.com/watch?v=eJWyzfJafmI. Kofia is an obscure band founded in 1972 by a 1967 refugee Palestinian musician, and it sang for pretty revolutionary lyrics that bucketed Zionism with "Reactionarism'', which I assume to be the generic Communist outgroup label, and Imperialism. The song's words are catchy and the tune is pretty fire, even if I don't fully agree with the implicit claim that Israel's founding is straightforward imperialism.
History lesson and my musical preferences aside, the song is chanted in Pro-Palestinian protests in the Arab/Muslim-hosting Sweden, Greta is Swedish and ideologically aligns with welcoming refugees, "Crush Zionism " maybe nothing more and nothing less than an ingroup-pleasing way of encoding "I attended Pro-Palestinian rallies and/or I approve of them".
If you want to treat it as a claim, you have to put it in the context of a late 1960s Israel that assaulted its Arab neighbours and more than quintupled its territory with annexed land. It's "Crush [1960s] Zionism [that stole my home and forced me into diaspora]", not simply "Crush Zionism".
2- Literally true, but the words don't always mean what their users claim or want to mean, or everyone against Antifa would be a Hitler/Mussolini fan. At the bare minimum, it should be taken into consideration that :
2-(a) Zionism is not originally a Jewish intellectual product, but a product of Christian apocalyptic thought https://en.wikipedia.org/wiki/Christian_Zionism that was passed into Jewish info stream.
2-(b) Zionism was embraced by many many many anti-semites, this doesn't mean it's false or wrong but it's curious that an ideology that claims to be for Jews was embraced by so many people who claim they're against Jews (but it **could** be that Zionism is such a good and win-win idea that everyone reasonable can't help but be for it, and the anti-zionists are only being unreasonable)
2-(c) Early (< 1917) Zionism experimented with different potential locations for the Jewish state, such as Uganda and the Argentines
Given that today's Zionism is very much NOT a..c and today's Zionists either don't know a..c or don't have good counters to them when reminded, I think it's fair to say that "Crush Zionism" is something that even Zionism did, to its own early prototypical versions.
3- (a) I'm an Anarchist, and I don't think States are a good thing.
(b) Insofar as States are ***Necessary*** in our days' cutthroat world that only recognizes States as Sovereigns, it strikes me as incredibly bizarre that "Jewish State" doesn't get the same objections and raised eyebrows and primers about Secularism from the secular and liberal Scott Aaronson as "Islamic State" or "Christian State".
"This is a Christian Country", this is a claim about America by a large faction of Scott's ideological enemies, the US republicans, he presumably has a LOT to say against it. Why is a Jewish country okay with him ?
4- My own assessment of the majority-Muslim Arabs' collective attitudes towards Jews mildly supports this, but this attitude is an unreliable indicator given that it's an (extreme and bad) reaction to 70+ years of Israeli arrogance and aggression. The most compelling evidence against it is how Israeli Arabs and West Bank Arabs seem to be pretty fine in aggregate with living with Israeli Jews, selling homes to Israeli Jews (the death penalty against that exists for reason), and marrying the occasional Israeli Jews.
I overall agree it's a sensible thing to worry about and think Pro-Palestinians generally but Arabs specially need to emphatically and Zealously distance themselves from Anti-Jewery, up to and including disciplining members who stray into Anti-Jewery during protests and simultaneously chants against attacks on Jews and the individuals/groups doing them while it chants for Palestinians.
I like your careful and thorough thought process, even if I don't agree 100%; do continue writing when you can
Really appreciative of and grateful for this kindness, many many many thanks :")
Just a narrow comment, ignoring the wider topic, on which I have personally no special insight and nothing to contribute.
> it strikes me as incredibly bizarre that "Jewish State" doesn't get the same objections and raised eyebrows and primers about Secularism from the secular and liberal Scott Aaronson as "Islamic State" or "Christian State".
It does strike me like all the known examples of religiously themed states are a bit of a disaster. Separation of Church and State is one of the few ideas in history that I find unambiguously good.
To copy over a comment I made in the last hidden thread:
Perhaps the most perverse thing about the whole history and situation is that, if the Palestinians had laid down their arms in 1948 after their side lost their war, just as the Germans and Japanese did in 1945, they would probably be in the same position as the Germans and Japanese are in today: thriving industrial economies. A two state solution could have looked like Canada and the USA or like the https://en.wikipedia.org/wiki/Schengen_Area . The Palestinian kids could have been getting degrees at the Technion.
( Any remaining territorial disagreements could have looked like https://en.wikipedia.org/wiki/Whisky_War "In 1984, Canadian soldiers visited the island and planted a Canadian flag, also leaving a bottle of Canadian whisky.[9] The Danish Minister of Greenland Affairs came to the island himself later the same year with the Danish flag, a bottle of Schnapps, and a letter stating "Welcome to the Danish Island" (Velkommen til den danske ø).[10][11][12] The two countries proceeded to take turns planting their flags on the island and exchanging alcoholic beverages." - presumably substituting coffees for the whiskies )
I think the problem with that plan is that the Palestinians didn't have all that many arms to lay down. It was the Syrians, Jordanians, Egyptians, et al, who had the large and powerful armies. And had no reason to lay down their arms because they had large territories that they needed to defend(*) and could reasonably hope to defend.
So, no matter what the Palestinians do, Israel is going to be surrounded by large, heavily-armed nations that just tried to destroy Israel. And no matter what the Palestinians do, the Israelis are going to reasonably fear that, come the next war, the Palestinians will roll over for their fellow Arabs and offer up their lands as a highway for the invasion of Israel.
That's going to put a damper on peaceful economic cooperation. And occupied Germany probably wouldn't have done so well if there were a much larger unconquered Nazi regime on their border to cause trouble.
* Mostly from each other, admittedly, but an early bid for "Greater Israel" wasn't entirely out of the question.
" And no matter what the Palestinians do, the Israelis are going to reasonably fear that, come the next war, the Palestinians will roll over for their fellow Arabs and offer up their lands as a highway for the invasion of Israel." Ouch! Excellent point!
Yes, remember when Americans moved into Japan and next left? Me neither.
I am not knowledgeable enough about context to properly evaluate the significance of this, but the wikipedia article on the United Nations Partition Plan for Palestine says
In a letter to his son in October 1937, Ben-Gurion explained that partition would be a first step to "possession of the land as a whole"
Which makes me wonder if it was ever going to turn out as well as you seem to think.
You might be right. On the other hand, Ben-Gurion was not the sole actor, and he might have thought differently 11 years later. In any event what actually wound up happening was three quarters of a century of repeated warfare, https://en.wikipedia.org/wiki/Hatfield%E2%80%93McCoy_feud writ large, blood feud after blood feud after blood feud.
Well, if the machines _do_ wind up taking over, at least whatever conflicts they have will probably be _different_.
What makes you think that the Palestinian Arabs have the human capital needed for a thriving industrial economy such as that of Japan or Germany? Evidence such as https://www.emilkirkegaard.com/p/palestinians-in-your-country-what makes that seem unlikely to me.
I don't think they're expecting to do as well as Israel, they just don't want to be constrained by Israel.
Perhaps of interest:
https://twitter.com/RichardHanania/status/1721151762898509958
https://twitter.com/RichardHanania/status/1718370192085102649
https://twitter.com/RichardHanania/status/1704160416103117099
https://twitter.com/RichardHanania/status/1723361365941690785
https://twitter.com/RichardHanania/status/1698092708592828650
Hainana's completely wrong. Palestinians don't do well as shown in the link above, and in any case he's completely ignoring selection effects. The good ones leave. The people left in Palestine are the dumbest ones.
Kinda weird that he says Japan is poor. The median Japanese adult is wealthier than the median adult in 90% of the world's countries.
https://twitter.com/RichardHanania/status/1667906105501188098
Gaa! Thank you for the link. That was quite a surprise to me. I don't know what to make of it. That does make the situation a lot bleaker, even if there had been peace. :-(
It's the mother of all false equivalencies to equate a war that Germany and Japan started with the explicit purpose of conquest and genocide with a war forced on Palestinians because Zionists wanted conquest and genocide.
> if the Palestinians had laid down their arms in 1948
And if the Zionists had laid down their weapons in the 1930s there wouldn't have been a war and the Jews would have kept living like they did for the last 2000 years. So what's the point ?
"with a war forced on Palestinians because Zionists wanted conquest and genocide."
Actually, it was the combined Arab armies which invaded Israel in 1948 with the intent of genociding Jews - in their words, they intended to launch "a war of extermination and a momentous massacre which will be spoken of like the Mongol massacres and the Crusades". And the outcome of that war was that many Arabs continued to live in Israel while not a single Jew remained in the West Bank or Gaza Strip - which side was the genocidal one here?
Sure, start the movie exactly at the moment Arab armies invaded and not a single femtosecond before, and selectively cut to various populists saying sick things about Jews, Israel looks in the right.
I can also start the current movie at the exact moment the IDF invaded Gaza, and, boy oh boy, do I have plenty of sick things about Arabs and Gazans the Israelis said. Two can play this game, but I'm not going to because it's a fundamentally silly game.
But the fact of the matter remains that Jews lived in peace with the Arabs for no less than 400+ years before Zionism, 5 times that of Israel's entire lifetime, and that the troubles started when some rich nobodies colluded with colonial powers to make a state to act out the fantasies described in an ancient and supposedly divine book. Correlation doesn't imply Causation, but Correlation alone is damning enough.
>But the fact of the matter remains that Jews lived in peace with the Arabs for no less than 400+ years before Zionism
Because there was a leviathan, the Ottoman Empire, occupying the place and willing to use force unconstrained by human rights norms. That Muslims were favoured also helps. Violence persists in Israel mostly because there are constraints on their exercise of a monopoly on force.
"But the fact of the matter remains that I lived in peace with my wife for many years, and I only started beating her after that, so I'm the victim"
There really was an Arab coalition that invaded Israel. Don't you think the coalition made a choice to invade?
That is my understanding as well.
I respectfully disagree about where responsibility for the 1948 war lies. Even so, _regardless_ of who was responsible for the 1948 war, it is a pity that the end of it did not lead to a result analogous to the end of WWII. Just looking at Europe, Germany, France, and Britain have been killing each others' people in various combinations for centuries, yet they finally managed to bury the hatchet after 1945. It is a pity that the Palestinians and Israelis didn't do the same.
Where are all the illegal German settlements in France taking place today?
Well despite the anger your blaming of the Palestinians for 1948 made me feel, I agree from the second line onwards.
Many Thanks!
There was another war, though, and one side tried to make sure Jews would not continue to live in Palestine "like they did for the past 2000 years" and that side has seemingly won the war of "ideas".
That side has the upper hand in Arab discourse only because the IDF keeps breakfasting on dead children, giving them more ideological ammo than the literal ammo that the USA's billions buys for Israel.
And it still hasn't won yet.
I've thought that the production of excess numbers of children may eventually be seen as a tool of war, though one-sided (in two senses).
Germany and Japan also had their version of a war forced on them narrative. An overt act of war followed by a "look what they made me do" storyline for the domestic audience is not special, nor original.
Sure, it's not the story that makes something true, after all Israel uses the "look what they made me do" trick till today largely unsuccessfully, it's the historical record.
> An overt act of war
You're talking about the Zionist militia terrorising Arabs and marauding them for the entirety of the 1930s ?
> for domestic audience
Yeah all those millions for Palestine around the world are deluded domestic audience brainwashed by the storyline invented in Pallywood.
https://en.wikipedia.org/wiki/1936%E2%80%931939_Arab_revolt_in_Palestine
"A popular uprising by Palestinian Arabs in Mandatory Palestine against the British administration of the Palestine Mandate... demanding the end of Jewish immigration and land purchases... the number of Palestinian Jews killed are up to several hundred"
Overt acts of war tend to be followed by wars. It's a bit unusual to engage in these when the outcome is both predictable and likely very unfavorable. It is arguable the domestic audience is deluded if they believed this chosen action was going to end well for them.
Low probability actions might be rational when there is little to lose but they are still low probability and only one possible choice among many other options. This action doesn't appear to be particularly wise in the short term or the long term.
This is not an argument about the Palestinian cause, it is an argument about whether this action had a realistic chance to be effective. It may be as simple as bad leaders making bad decisions.
Sure, Israel has its deficits, but it is among the happiest countries in the world https://en.wikipedia.org/wiki/World_Happiness_Report and has a higher per capita GDP than Germany https://en.wikipedia.org/wiki/List_of_countries_by_GDP_(nominal)_per_capita
If Zionism had never existed, it is fair to assume that a hypothetical Palestine would probably be comparable to the surrounding countries, which are far worse off.
Okay, so if I steal all your money and it makes me ecstatically happy, then you should be fine with it?
I think a more relevant question would be if the Palestinian Arabs would have been better off without Israel and if the Israeli Jews would have been better off in the diaspora. The country-level comparison seems inherently misleading.
The more relevant question would be if the Israeli Jews would have been better off without the Palestinian terrorists and if the Palestinian terrorists would have been better off in the diaspora.
The "terrorists" were there first
https://en.wikipedia.org/wiki/Kingdom_of_Judah
The meta-question is "better off -how-?" There's more to life than GDP.
Frost was right when he said that home is the place where, when you go there, they have to take you in. Anyone anywhere on earth who has a home of this sort does so on land that used to be someone else's. Almost any ethnicity that has a home anywhere does likewise, with the 'almost' referring to tribal groups indigenous to relatively new volcanic islands. Historically, it's the stronger who get the homes. Totally reasonable for Jews and Palestinians to both want the same home, and to feel that they'd be 'better off' with it than without it.
You understand that by this definition, white people have no home?
😉
It occurs to me that stateless is a very different order of negative outcome from dead. For example, we were willing to put up relatively quietly with the Palestinians being the former, but only began shouting once an inordinate number of them became the latter.
You're correct, of course, about the many different faces of Zionism, but considering that the actual destruction of Israel is a practical impossibility, and a Likud-led government is presently in charge of Israel, it's probably fair to interpret their version of Zionism (which has historically supported the settlements, full control of Jerusalem, etc.) as the most relevant definition. I, too, would like to crush that Zionism and preserve the Zionism that leaves room for Palestinian self-determination.
I'm also not hugely impressed with Aaronson's call to "spite Greta on climate, but SMARTLY", though that's beside the point.
Re statelessness: I'm not hugely impressed with the idea that Jordan is existentially threatened by the presence of Palestinians more so than Israel is.
I also find amusing the idea that Palestinians could repossess Israel and it would still be Israel-like. South Africa is apt. Seems to me Israel took/takes work, and requires a population that wants to work. Arabs ain't it.
I don't think Palestinians expect or require Israel-level living standards in their hypothetical state. They're quite accustomed to misery by this point.
And Israel is not existentially threatened by the presence of Palestinians. It's not existentially threatened by much of anything. Even without American patronage, it would take a virtually impossible regional military coalition willing to endure one-sided nuclear retaliation on top of everything else to seriously threaten the country's existence. And as you so charmingly observe, Arabs are far too lazy for that.
Egypt seemed to hold its own pretty fine against Israel in 1973, where "held its own" is Golda Meir quite literally begging Nixon for another army's worth of equipment instead of the one Egyptians destroyed.
The Houthi rebels 2000+ KM away wreck havoc in Eilat with just UAVs and cruises, the entirety of Israel is wide open in Egypt's missile range, an army with 1 million standing personnel for Israel's full-mobilization 500K.
> Even without American patronage
I would bet that easy money, but the AIPAC won't let me get rich that easily.
I don't know where you're getting the 1M standing personnel for Egypt's armed forces. It might approach that with reserves fully mobilised, though only the land army would be a relevant branch. Israel's active+reserves is indeed roughly 500k, but in a situation of existential defence, mass conscription of another 2M or so men and women would be expected. I also highly doubt Egypt has the logistical resilience to support its fully-mobilised army in the field for any significant length of time, especially in offensive operations against a technologically superior enemy.
But this is a silly hypothetical anyway. We're a long way from 1973. For one thing, the first Israeli nuclear test was conducted in 1979. For another, Syria is in shambles. For yet another, Egypt and Jordan have long-lasting peace treaties with Israel.
And I would cover the bet without a second thought if the win/loss condition was that Israel ceases to exist as a state - which is what existential peril means to me. The only relevance of 'destruction of Israel' is as a rhetorical motte. The most grievous possible loss condition for Israel is significant territorial concessions to create a viable Palestinian state, which might become politically palatable if the United States were to withdraw its support and Israel began to suffer economically from international isolation. (I don't think a binational state would solve anything at this point.) Also largely a silly hypothetical, unfortunately.
Yeah I misremembered that 1 million figure, Wikipedia says more like 600+K in total, but with Egypt's population being 100 million you can expect no less than 3-5 million fighting-age able-bodied males to be available on very short notices. Training (not arming, Egypt has a native arms industry) will be a problem, yes, but really just freeing up the whole millions-of-KM long borders along Lybia and Sudan so that the better-trained core can focus on Israel would be a huge advantage over Israel which has Hezbollah and Al Assad to its back.
Keep in mind that 1 or 3 million is barely a dent in Egypt's 100 million population, while Israel's 2.5 million is 25% of its entire population. This will probably cause some shenanigans, hell even the recent 500K mobilization was costing Tel Aviv an estimated 250 million dollars or so every day. Egypt's economic deficiency is real too, but in a world where the USA is isolationist the Gulf has to find someone to ally with in the face of Iran, and there is no better someone than Egypt and Turkey. Egypt is closer to the Gulf.
I also agree that destroying Israel with a war is not the outcome I'm hoping for as my #1 wish, regardless of how much I hate it and the sheer contempt I have for its history and political composition. But I do think it is very possible, being surrounded by neighbours that all hate you tends to make you very geopolitically fragile.
> For one thing, the first Israeli nuclear test was conducted in 1979.
This can only work as a suicide strategy to deprive the invading Arabs from an intact Palestine, there is a reason they call it "The Samson Strategy". Whether Arabs will back down or call the bluf is debatable and very dependant on exactly who is calling the shots, in terms of people and also in terms of countries. But if Israel is forced to use this it's already vastly reduced and surrounded from all sides.
> I don't think a binational state would solve anything at this point.
Binational state is a weird term to call a modern multi-ethnic state, this is probably the single best solution to the conflict, but the one with the greatest price tag.
You're right, misery may be a preference. Thirst, not so much.
In an atmosphere where relativism prevails, it seems a little funny that there is a vestigial disapproval, not of laziness, but of describing people as lazy (or uninterested in a particular cultural definition of work, at any rate). That can only derive from a feeling that lazy = unvirtuous, which is pure retro.
ETA: but I've noticed that progressive types like to keep these little judgments in their back pockets, though they would never own up to them.
Yeah, call me a cynic, but somehow I don't think "my enemy doesn't have the aptitude to make the desert bloom" is intended to carry the connotation of "because his culture aims for a different work-life balance, and that's totally valid and even enlightened!"
If I've given you the impression that I'm keeping my little judgement in my back pocket, I regret the error. It's right there on my lapel, in lieu of a boutonnière: I think ascribing virtues and vices to whole ethnicities is crude sport, good for old-fashioned pub humour but divorced from utility. Even the racial-IQ jockeys at least try to mumble about haplogroups.
It's a matter of indifference to me what the connotation is, not sure where the utility comes in. I don't see how pretense can really serve as well as plainness, seems to require the verbal equivalent of a Rube Goldberg machine. Probably something that spits out: "my enemy *didn't* make the desert bloom".
Vive la différence.
Who said her antisemitism came from Adolf? The narrative I'm seeing is that she got it from the left, which largely got it from Arab immigrants.
Marx was an antisemite. https://en.wikipedia.org/wiki/On_the_Jewish_Question
But modern Swedish antisemitism is dominated by Arabs and Muslims, not by Marxists.
In previous Open Thread, it was mentioned twice that the "rationalist community" got its name wrong, because in the "rationalism/empiricism" philosophical dichotomy, they are taking the side of empiricism (although in reality they actually do a lot of armchair reasoning and little experimenting).
I repeat and develop my objection here (so that it does not get lost in the ocean of Palestine/Israel comments), because this is what people keep saying for years, and it is mistaken.
The "rationalism/empiricism" philosophical dichotomy is a false dilemma, because on one hand, human reasoning is influenced by all the evidence we have observed in the past, on the other hand, all the evidence we observe is interpreted (starting way before the conscious level, for example the visual cortex already translates the incoming "pixels" into lines and shapes, which is why the optical illusions are so powerful even after we learn about them). So it is always an interplay of both. (Also, the dichotomy ignores the third major source of information, which is the instincts and cognitive biases we were born with. Were are not born as "tabula rasa", to be shaped purely by the sensory inputs and/or reasoning from the first principles. Although some philosophers tried to shoehorn instinct into a broad definition of "reason".)
Even people traditionally associated with philosophical rationalism (Spinoza, Leibniz) often said things like: "in theory, we could derive everything from the first principles, but in practice, we usually need to use the empirical evidence" (not an exact quote). So if you tried to find someone who is "rationalist" in the sense of "opposing empiricism", you would probably need to go in history all the way back to Plato.
Then there is the political meaning of "rationalism", where using reason is put in opposition not to experimental data, but rather things like the divine right of kings or religious traditions. This is associated (among other things) with materialism, atheism, utilitarianism... and I think it is not a mistake to put the "rationalist community" in this group.
tl;dr -- you make it sound like words have a clear meaning in philosophy and ignorant Yudkowsky made a clearly wrong choice, but in reality it is complicated, some traditional thinkers would agree and some would not
Indeed, rationalism and empiricism are interrelated just as you describe, and so any claim that's 100% black and white about which one the community has fealty to is going to be wrong . However, its still reasonable to believe that the community places an outsized weight on the rationalism side of things vs the empiricism side of things (compared to what they claim). A practical empiricist will praise he who gathers the most data to come to a conclusion that "speaks for itself" (i.e. requires little sophisticated reasoning to reach). A practical rationalist will praise he who has the keenest ability to stitch wisps of evidence together into the most accurate conclusion (Yudkowsky's Harry). I claim the rationalist community often acts more like the latter than the former.
The former though, is the type of intellectual coward who claims they are exercising the virtues of empiricism and never has any original thought in their life because they're too busy waiting around for social approval to believe things.
People claiming to be empiricists rarely seem to bet, or acknowledge that prediction markets can do well (although to be fair that is true of most humans too). If they really did value accuracy, you'd think they'd be consistently calibrated in the ways we know to calibrate, yet how I see the term used is as a social cudgel to stop nonconformist thought.
How interesting! That hasn’t been my view at all. Nonconformity and empiricism go hand in hand for me— I would consider Galileo (at least, the mythical story of Galileo we get told) as the archetypal empiricist, trusting the data *over* public opinion to his grave. Your point about prediction markets is correct, but at the same time on a meta level my approval of prediction markets is a direct consequence of the data showing that they outperform pundits so there’s an empirical case for them too (which is an empirical case for rationalism, but i never claimed rationalism was entirely useless, and it’s notable that I trust prediction markets, not individual predictors). Empiricists are often iconoclasts; a great marker of human progress is how often they don’t need to be, because the widely accepted view is the one that fits the facts.
I suspect if we were to discuss, I would either end up accusing the “empiricists” you describe of being fake, or (more likely?) accuse you of intellectual hubris.
We're talking about people in the modern day who call themselves "empiricists", not every person who was empirical in history. I agree if I was talking about every person labeled an empiricist, I'm extremely wrong!
But to speak of what I've seen in the comments section:
I rarely see people claiming to be empiricists have actual links to research
I've seen people who claim to be empiricists say that global poverty interventions are ungrounded in reality, and will be wasted via corruption, despite the fact that GiveWell top charities do not route through governments or donate easily sellable items.
Of course, this is most apparent in AI risk posts, where self styled empiricists would demand an actual superintelligence that kills humanity as the only proof they would accept, not knowing or caring about actual objective misspecification examples or learning about basic gene level evolutionary selection.
Unfortunately, searching comments is difficult (if there's a search engine option I missed please tell me!) but my general impression of the sentence "I am empiricist" is that it's said by a person who would deny the existence of muons or other short lived particles from collider experiments "because it requires too much theory to interpret", or who are saying it just to seem more impressive in contrast to the rationalists.
The closest comment after 15 minutes of search I've found is https://www.astralcodexten.com/p/movie-review-dont-look-up/comment/4305968 which,
while imo misguided because it uses the falsification frame of science, is in fact empirically grounded and otherwise a good, non smug comment.
The original poster was addressing the claim that this community calls itself empiricist but is really rationalist. I said yes I think they really are rationalist despite what they say. (Aka they claim to be empiricists, cite no evidence, then say 'we don't need evidence bc I thought about it really hard'). You're saying this community calls itself... idk, and the people within it who claim to be empiricists aren't sufficiently rationalist and sometimes also aren't sufficiently empiricist. Hopefully you can see that our views differ in several unusual ways not just one.
Your description of people who claim to be empiricists but fail to link to research certainly seems to me like they're not really empiricists but rather the opposite. I agree if people are using the label empiricist to mean a complete lack of actual empiricism, that's bad, and that was indeed part of my original point since I claim that's what many in this community actually do. My prescription for that isn't less empiricism though, it's more. Then you mention AI risk deniers but they seem like they're in a whole separate category of people who claim to be empiricists and.... actually are? Sure, in a bad way, but they seem totally different from the people who just don't use evidence at all.
I fully agree with your characterization of the self-identified hyper-empiricist. It's certainly not always a good thing. Empiricism leads both to "homeopathy isn't real" and "muons aren't real". Rationalism leads both to AI safety and Christian apologetics, so from my point of view it's not always a good thing either.
My point is that rationalists (who have read the sequences, and not just anyone here in the comments section), in fact exercise the virtues of doing research, or actually read through the Givewell analysis, or try and ground AI risk in things in reality that "kick back" as much as possible, things that would be different depending on which world, the easy alignment or the hard alignment, we're in. This does go wrong! And these things don't happen as much as I would like them to!
But my estimation is that rationalists practice more empiricism than people who like to use the word "empiricist" as a veiled status jab.
I would say that if your notion of empiricists include AI risk deniers as bad but still fundamentally empiricist, you would also include flat earthers who conduct flawed experiments proving the earth is flat to be empiricists. Does that fit your notion of empiricism? (No I am not saying AI risk deniers are as wrong as flat earthers, but trying to mine the intuition of empiricist classification, such that someone who is dependent on evidence arrives at a conclusion you don't like, as opposed to one you do.)
I just think that empiricism is just a wrong axis to do all your truth seeking on, when we know there are typical ways that our minds distort evidence and differentially discard or accept various truth claims. If you think this is "thinking really hard and then getting the right solution", that's your prerogative.
Informative, thanks for sharing! It's an interesting thought experiment though: if you wouldn't call this crowd the rationalists, what would one call them? The Bayesians leap to mind, but if we taboo that one, what then? Yud says rationality is ultimately about winning, so maybe something built around that, maybe involving the goddess Nike?
What are good things your parents did with you when you were a kid? What do you wish they had done but didn't?
Background: my kids are 5+7 now and I feel I now have a few years where I can do all kinds of fun, weird and in the end net-positive things with them before they turn into teenagers. Examples (what I already do): bring them along to diverse encounters (the 7yo just spent 3 days with me and others cooking for 50 people), create time when they can just do whatever they want (current favorite: play outside in the mud without someone watching them), read many books together, have reasonable expectations for what they must do (clean their room once per week). So far we're doing good - I'm just looking for more and also weirder ideas!
Good -
free range, that's #1
encouraging music
no religious indoctrination
Bad -
I got no sex ed from them or my schools. Went to college totally ignorant.
No hugs/physical affection.
My mother died at the age of 37, when I was 19. If there is anything that I would want more of, it's just time with her. It sounds like you are on the right track already, so keep doing what you're doing.
As for specific things, she instilled in me a love of gardening. We did fruit picking together, we worked on the garden in the backyard together, and we went to Longwood Gardens once, which should be fully into its holiday displays now and thus absolutely gorgeous.
I remember not liking those activities that we did together sometimes. My Game Boy was never far from me. But, she taught me how to make a kitchen work for me and to appreciate plant and animal life for its own beauty and the cool things that they can do.
I love science thanks to her and my step-dad. I can function well (as well as I do, anyway) thanks to them and the diverse things that we did together.
Sorry for the tangent, she is on my mind every year during the holidays. Again, it sounds like you're already doing well. Keep at it!
Good things they did:
* Dad kept all of his textbooks from college instead of selling them back. I got to read a lot of them. I knew trigonometry by around age 11, calculus by around 13 (well, the fun derivative game, less so the less fun integral game, but I could do some of those in my head). Likewise with a lot of physics.
* The math track started early for me. Dad was a math teacher when I was very young, and brought third grade tests home, and I helped grade them. I was in kindergarten.
* Raised me on a farm. This is hard for everyone to do, for obvious reasons - you need a farm. We did. And I spent a lot of time outside, playing in dirt, learning to garden and ranch and fix fences and move hay.
* Dad later quit the teaching job and became a physicist during the week (when he wasn't being a farmer / welder on the weekends). He enjoyed reading SF at his desk at home, and seemed to enjoy even more when I'd come up and ask about physics at my level.
* Both parents got me playing an instrument. As mentioned elsewhere in this thread, the piano is a good one, but any instrument is probably much better than nothing. (A guitar has an advantage in being very portable.)
* Mom taught me cooking, and I was very comfortable around the kitchen.
* Got me building bricks as a very young kid. Both Legos, and a knockoff Pedlo brand (which I preferred for certain applications; pretty sure they're OOB now, but Legos are fine).
* Got me into Boy Scouts. It helped that our neighbor was our Scoutmaster, and a very good one. Heavy emphasis on camping and outdoor skills.
* Mom worked hard to organize a trip to Europe when we were adolescents.
* Got me my own bank account by high school. I learned how to write checks, and how to balance the books.
* Dad brought home liquid air from work in a thermos. We spent all night playing with it.
* Let me play around on his Osborne-1 computer from work. Video games (Colossal Cave) at first; but soon I was learning CP/M assembly. Later, he brought home an Alpha Micro surplussed from work, along with several terminals, a disk drive machine, etc. We had terminals in our bedrooms, and I learned a lot of Pascal programming.
Good things I wish they'd done:
* A bit more mechanics would have been useful. Dad had a lathe (he had a machine shop, in fact), and as complicated as it was, I never got around to learning it before heading off to college. I never made a gear or a rebuilt an engine; the most complicated thing I did myself was replace brake pads, and it felt tedious and arbitrary.
* More electronics would likewise have been good. Get a soldering iron and teach your kid to use it, and try to actually fix something around the house.
* Experience with a 3D printer would have been great, though we have an excellent excuse: they hadn't been invented yet. They're here now. Bring your kid to a makerspace and encourage your kid to spend serious time there.
* More home finance, including my own investment portfolio.
* Mom (and Dad, for that matter) should probably have leaned harder into teaching me Vietnamese.
This is a great question!
I think they best thing my parents did was to raise me under an old system that is hard to duplicate now: All the neighborhood kids played together, going in and out of each other's houses freely as a group. All the parents kept tabs on what was going on by watching through the windows, called somebody's parent if they needed to know somethings, and handed out snacks and bandaids. My parents told me the boundaries of the area I could play in, but it was quite large, maybe a half a square mile, about half of that undeveloped land. From about ages 6-12 I spent thousands of hours climbing trees, running races, playing baseball, and playing all kinds of pretend games with other kids. Sometimes I wandered outdoors alone pretending various things with great enjoyment. My parents did not object to my doing every bit of my outdoor play barefoot, and didn't care how dirty I got.
Beyond that, the things that I remember with the most pleasure from early childhood have to do with making things. My mother and I dug up some red clay from the side of the road (we were living in Georgia), shaped it into a brick and cooked it in the oven What came out looked and felt enough like a real brick that in my mind it was exactly the like the brick people build houses with. The feeling of having made something *real* made a great impression. When I wondered in the brush alone pretending things, one of my favorite fantasies was building my own little house. I thought I could make a smooth floor out of clay, then build a fire on top to turn it into a solid brick floor -- then I would make a bunch of bricks and stack them glued together with mud to build the house. I had this fantasy for *years*. There were a number of other projects like the brick that delighted me. I think I initiated them, and one of my parents would help me. The ones that made an impression on me were always the ones where I was satisfied with the product -- I had made something Real.
I realize that all these things sound sort of ordinary. They were not super-cool imaginative activities geared to the Gifted Child. But they sure hit the spot. I have been making things ever since I made that brick. It's my favorite recreation. Just finished building and decorating a huge play structure for my cats.
Oh, remembered one other excellent thing my mother did. There was a long period when I was 6 or so when I believed I could learn to fly. In my mind, the key was to flap my arms really hard and fast. There was a day when I spent hours jumping off the arm of a living room chair, flapping my arms with mad energy. Told my mother I thought my descents were getting slower. "Good," she said. "Keep at it."
My intellectual development was a whole other dimension, and really did not tell my parents much about it. When I did, both they and I ended up feeling alienated
My advice (based on my kids, not me vs my parents):
1. Read to them as a regular thing that's expected daily or almost daily, e.g. before sleep. While reading, it's good to sometimes stop and ask something ("why do you think she said it?"), though not to overdo it. It's also great to ask kids to playact particular funny phrases/reactions of characters after reading them.
2. Have lots of shared silly rituals and games. Anything based on whimsy and fun. Often a silly ritual involves a parent doing something mildly exasperating. E.g. when I'm driving through a traffic circle, I will often continue circling until kids notice and they have to say a particular key phrase together three times for me to stop. Because when starting tennis one of them mispronounced racket as rocket, I insist on talking about her tennis rocket and insist it's the right way to call it. When we we reading the Hobbit, I pretended Smaug was a good character and defended his honor, and they liked it so much that now we have an inner joke in the family that Smaug is secretly my grandfather and as I read completely unrelated books to them, almost daily I slyly insert Smaug into the text (sometimes more subtly, e.g. a shadow of a huge bird) and as they notice they mock-protest and hit me with pillows or something.
The more inner jokes, silly rituals, shared word games, the better.
> What are good things your parents did with you when you were a kid?
Mine used to bring books to me when I was learning to read, they encouraged me to read the street signs passing by when were in the car and bought a subscription to a public library for me when I was... 8 ?, definitely not older than 10, and used to take me there with my sisters, admiring the amount of books I check out (there was a 5 book limit on the subscription, I took some of the slots in my sisters' subscriptions), forever tying reading with status and social approval. When reading 200+ pages novels was a monumental, heraclian task for me, I checked out novels anyway to fit in with my sisters, I didn't read them and just invented details out of my ass when they asked me what they were about, they used to laugh but I was too dumb then to realize they were laughing at me. I just thought they were amused by my incredibly bad storytelling.
My strict father in high school probably kept me on my toes amidst atrociously academically-bad peers, but most of the effect is probably just the constant asking about status updates and grades, the constant monitoring, not being done in a strict fashion.
> What do you wish they had done but didn't?
1- Gone out more often into places other than the library or our relatives, specifically beaches
2- Hit me less often and less harshly, but not a 0 amount
3- Found a way to make me enjoy sports as much as I love reading
4- Kissed/hugged me more often and asked me to kiss/hug them
5- Didn't raise me as a Muslim
> I'm just looking for more and also weirder ideas!
Non-traditional Education is extremly under-explored as an avenue for parental investment. The formal school system is so utterly bad and so comically wasteful that it isn't hard to improve on it. As concrete advice, consider :
1- Teaching your kids a trade
2- Getting them interested in a video game with an active modding scene
3- Getting them interested in something that is disproportionately available in a foreign language
But none of those is as valuable as the simple realization that the K-12 schooling system is one of the most incompetent and dysfunctional pedagogical machine ever designed.
Each of the items you list that you've done/are doing seem like really good choices.
Getting a dog (a friendly mutt) has had a pretty great impact on my youngest (now a 6th grader), but that of course isn't for everybody.
To your list I would add, make regular hikes outdoors a routine thing. Exploring large parks in your area, beach/dune settings if you're near such, forest preserves, etc. Nothing weird about that I guess but from my experiences both as child and parent it is tried-and-true.
One good thing, maybe not sufficiently weird: teach them music.
1) Buy a piano, preferably a real old school acoustic piano. If you can't afford or accommodate an acoustic piano, get an electronic version with a good weighted action and 88 keys. Have them take lessons with an accredited teacher who schedules regular recitals and enrolls them in some kind of graded progression of exams. Make them practice.
2) Take them to a symphony performance, preferably a good symphony in a decent hall. It doesn't matter much what is on the program. Just the sound of a good symphony orchestra tuning up in a good hall is fantastic. Maybe even life changing.
3) Take them to a jazz show in a small venue with an appreciative audience.
Seconded. Piano or keyboard is good, but let them figure out a guitar (or ukulele to start with) and/or some percussion if they turn to be wired for that.
One hundred percent this, all of it. My parents did that with all four of their children and I've done it with both of mine. It is enormously beneficial for the long haul including for those (such as my sister and my eldest son) who never become proficient and never touch an instrument after childhood.
Anything Titanic or Egypt is typically a good "limited" museum experience, I think. Maybe also Terra Cotta warriors.
My childhood was too long ago to provide current insight - it lay in that distant time when suburban kids didn't have a lot of responsibilities but nor did parents think about their doings from dawn to past-dusk - but I would have enjoyed (and do enjoy as an adult) camping and gardening and repairing things. That was not who my parents were, unfortunately. For the tinkering and fixing I would have been better off hanging out with a grandfather, but he lived too far away. In my experience, one is just as likely to be similar to a grandparent as to a parent.
So maybe identify that grandparent, in some cases.
My limited memories of happy times with an unhappy nuclear family, with much older siblings, include watching certain movies all together, more than once so that they were the "family" movies* - this was easier when there was one TV and you had to step across the room to change the channel.
And playing games, cards or backgammon or Mastermind or Monopoly. Nothing Improving. They should not be improving.
*Not sure how "The Apple Dumpling Gang" holds up.
Oh, yes: and model trains, and real narrow-gauge trains, and funiculars and cog railways. Miniature golf. And if that dollhouse model of the White House ever is in your area, it's a cool starter window into history maybe. I haven't been to but two presidential libraries, and the more recent ones are probably dull as dishwater in the self-serious modern way - but the LBJ Library is or used to be pretty good for children.
For me: Fewer museums, more activities. More interest in things I was into instead of just pushing me to be interested in the correct things. Wish they'd shown me things like coding or higher math earlier (though probably not as young as 7), but I may be unusual in that regard.
"Wish they'd shown me things (..) earlier"
That's an interesting one. I feel it's OK to introduce kids to things way above their age level as long as it's not overwhelming or dangerous in a bad way (horror movies: no; having a 4yo cut onions under careful supervision: oh yes please). We just read "The Number Devil: A Mathematical Adventure" and they loved it... which means we can reread it in 3, 6 and 12 years and learn something new each time.
Oh I loved the number devil! Shoutout to school libraries for having stuff laying around.
Depends a lot on where you live, but learning to identify local plants is a good one, especially if there's foraging opportunities. Plus it may be good for pickiness. For years the only green leaves I would eat was miners lettuce I had picked myself.
I was a big fan of the Natural History Museum in New York City at that age, but that was more a function of it having a lot of cool stuff to look at, and I might have been significantly less fond of it if I'd been pressured to conscientiously absorb the contents of the various plaques explaining the visits.
My favorite, and one of the few displays which last I checked was largely unchanged from my childhood, was the whale room, with the life-sized display of a blue whale suspended from the ceiling.
I agree - a trip to our natural history museum as on a school trip was for me an opportunity to revisit the big grizzly bear they'd taxidermied upright and threatening. Bear stood in for museum, whose large display of the history of the petrochemical industry was not then interesting to me, though I would love it now.
I also think planetarium shows are captivating for small children, if kind of boring for adults.
Maybe visiting castles whilst talking about how an enemy might go about trying to capture it?
I'm sure there are loads of equivalently good museums
Start to share your interests with them - but from my experience, it doesn't work if you're sharing stuff you liked when you were their age, or stuff you think they'd like. It has to be stuff you're excited about yourself, now. And, if possible, stuff that's new to you as well as new to them. And even then, there's a fair chance that they won't be interested. But it's worth the effort when it works.
Bringing me to walks in the wood and teaching me to recognize various trees and plants. Picking mushrooms.
Back then, I didn't like it and would have prefered to stay inside to play and/or watch TV. But in retrospect, I am glad they did.
The situation where OpenAI employees massively sign a public petition supporting their boss in his acting against the OpenAI Charter, reminds me of a document called "Anticharter" that people in socialist Czechoslovakia publicly signed to express their loyalty to the regime (and keep their jobs).
https://en.wikipedia.org/wiki/Anticharter
This is not a coincidence because nothing is ever a coincidence.
Responding here instead of your bottom-most comment, because there's no way to say I like it without spoiling the effect. ;-)
We continue re-reading old posts about Scott's adventures all over the world. This week, he goes to Japan and visits the sunken corpse-city of R'lyeh. Either that, or some funny shaped rocks. https://archive.ph/k7WOI https://pastebin.com/Hq8vsWER (index of all the old posts https://archive.ph/fCFQx)
Later, Scott came to believe that the "ancient city" is just a natural rock formation, but writing things down immediately after going there he was much more on the fence and leaning towards it being real even. Today we have something that the Internet of 2007 did not - a beautiful video of that very same "city" https://www.youtube.com/watch?v=mnaWn5OPP3c, so that you can decide for yourself if it looks legit to you. (Spoilers: no, but it's very pretty anyway.)
Wow, reading that and seeing the comments made me realise how much I miss old style blogs and the communities that they had...
On the video I think I agree with you, the rocks look vaguely geometric and I could see how one might believe they were manmade if they were primed to do so. Interesting to see how much Scott's writing style has evolved too, but I guess after fifteen years of regular writing you'd expect at least some improvement :)
If you were dictator of your country for a week and could force through one lasting change, what would it be?
Meta-changes that modify the political/governmental apparatus are encouraged but not required.
Lasting meta-change? There's a lot of things I'd like to use that for, but if limited to one, for one country, I'd probably pick the establishment of Scott's Raikoth's religious orders (assuming that counts as a single change, despite how far-reaching it is).
https://slatestarcodex.com/2013/05/06/raikoth-laws-language-and-society/
Having one country in the world being supported by the "Priesthoods of Truth, Beauty, and Joy" and their associated angels (just the structure, not the output that happens in Raikoth) might not be great for the country itself, but it sure would teach the world a lot of things.
(USA) Mandate that all future elections in the entire country be held and counted in one day, in person, on paper, with a tiny exception carved out for absentee ballots.
Killing animals is now a crime with the same legal consequences as killing a human.
Easy.
Universal Basic Income.
In the year 2023 nobody should need to rent themselves out to a master just to have food and a room.
The Torah tried to get us there theough mandated Sabbaths, Festivals, Sabbatical Years, Jubilee, etc. but as a society we humans absolutely suck. I think Scott called it "Moloch". That sounds about right.
No man, woman, or child should have to be enslaved for food and shelter. We live in a world so very very mad that I look like the crazy one for suggesting that maybe, just maybe, slavery is not only bad but 100% unnecessary.
I want UBI, but I do not want UBI for people who think that work is slavery. That attitude is corrosive to everything that could make a UBI work.
One Lasting change?
American here: and I'd amend the constitution to abolish private land property rights in order to enact a Georgist Land Value Tax system replacing most all taxes with a single tax on land value. The Country would lay ultimate claim to all lands within its borders, with a permanent lease to each of the states' land, and from there, subdivided as far as the state would choose (county, metro area, city, etc). The Tax would be close to 100% of the annual rental value of the land. Some portion to be owed to the national government collected from the state, some to the state collected from the locality, etc.
The amendment would also establish a national land value assessment agency responsible for developing methods and measures for accurately determining rental value of lands (something like NIST or NIMT but with real teeth for noncompliance), and would also develop methods for other types of "land" such as mineral rights, water rights, electromangetic spectrum, fishing rights, etc...
"Buying property" would amount to claiming an exclusive lease for some number of years (say, 5 for commercial, 20-30 at a time for residential). Carve outs could exist to make this easier for owner-residents (e.g. being able to defer taxes until sale of the land and improvements upon it).
Fixing our budget and land-usage in a week would lead to other problems being easier to solve, I'd say.
I assume you want LVT because it will incentivize landowners to use their land in the most economically valuable way. But they won't be able to as long as zoning laws prohibit them from doing so. So you need at least two lasting changes.
I think any reasonable definition of land value would take zoning into account.
(USA) Term limits for congressmen, senators and maybe the supreme court.
Replace first past the post voting with another system, as maybe this would create an opening for a new political class.
I'm American and have in my head a literal list, have for many years. In conversation I refer to it as my "done by noon the first day after they make me king" list. I struggle to keep it under a dozen items frankly, and picking just one....hmm.
The list does shift over time though maybe less than I would have guessed 20 years ago....probably the most micro item is that revenue-producing college athletics are banned. Since I can't quickly figure out a way to choose just one amongst the more serious ones I'll leave that one here for now.
If LLMs weren't advancing so rapidly, my answer would be: Fund the development of Drexler/Merkle nanotechnology aka atomically precise manufacturing.
Yeah, political/governmental issues matter, but technology and its applications matter more.
Given how rapidly LLMs _are_ advancing, I'd just say: Don't cripple that work, like civilian nuclear power was crippled. Just stay out of OpenAI's way.
Set all income taxes to 0 for anyone making less than 100K /year. This includes "withholding" for medicare, SS etc and applies to employers and employees. In other words, you earn 40K a year, your take home is 40K a year and your employer pays 40K a year.
Perhaps creating a predictor of criminality based on criminal history and other individual data, and then locking up everyone over some reasonable threshold of expected criminality in the next time period(a year?), and then re-evaluating every time period.
I've been lead to believe that simple algorithms are better predictors than even expert judgements.
Maybe not as good as Robin Hanson's voucher system, haven't really looked into that.
Predictor of criminality or predictor of criminal conviction, because a system that locked up rich people because lots of rich people commit crimes that they are not convicted for would be very interesting.
How would one get stats for criminality that doesn't lead to convictions? One could look at victim statistics in rich neighborhoods, but hard to pin down which demographics to ascribe those crimes to.
If you don't have stats for criminality that doesn't lead to convictions, then all you're doing is replicating the biases of the existing system, and ensuring that the people who get away with crimes now will continue to do so.
PS, the crimes I'm thinking about being committed by rich people are things like tax evasion and wage theft, which are rarely prosecuted as crimes, but usually (when identified) resolved by paying the missing wages / missing taxes rather than prosecution and imprisonment conviction.
To be clear my proposal is basically about optimizing prison sentences to minimize predictable crimes. Crimes that can't be proven in court even after the fact would of course be even more difficult to predict ahead of time.
I think we all know there's people outside of prison who are very likely to commit provable crimes, and people in prison whose chance of re-offending is low. That's what my proposal would solve.
So you want to punish people not for what they have done, but for what you predict they might do, even knowing that the algorithm is imperfect - some of them would never have do what's predicted for them, and others will commit crimes without being predicted to do so.
That seems very much unethical to me, even before changes are (predictably) made by "special interests" to bias the prediction in various politically or personally useful ways.
To the degree that criminal history is a good predictor of future criminality, it's punishing people for what they've done.
To the degree that criminal history is not a good predictor of future criminality, incarceration is not a cost-effective punishment.
"changes are (predictably) made by "special interests" to bias the prediction in various politically or personally useful ways. "
As opposed to some other criminal justice system? What do you think is a more unbiased system for college admissions, automatically admitting the people predicted to do well based on high school GPA, SAT scores etc, or letting some college staff act as judge and jury?
This reads to me as if you don't distinguish between "ethical" and "efficient", or possibly simply missed that I was responding to a (presumably) efficiency-motivated suggesting by objecting to its ethical implications.
From a utilitarian perspective the ethical justification of incarceration is preventing crime, the more efficient it is at that the more ethical it is.
Of course in the extreme you could:
- have nothing at all be illegal; there would then be no crime at all
- imprison or kill everyone, since anyone could potentially commit some crime. No people -> no criminals and no crime.
These are obviously bad answers, of course.
However, I do think it's important to look at a bigger picture than simply preventing the maximum amount of crime, or even preventing the maximum amount of crime at least cost.
A person who is jailed because of the crime they are predicted to commit pays a very high cost, but that can be lost in the (predicted) benefits to future non-victims of the predicted crime, not to mention the benefits to the people employed as jailors.
Indeed, that's a more general problem with utilitarianism; it doesn't matter who gets the "utility" and "disutility", just that the total adds up positive.
Range voting for all federal elections. This would result in more than two party system over time. Congresscritters would have to talk to each other again.
But unlike parliamentary systems, the multiple parties would still be fairly moderate, as Range favors the Least Bad candidate vs. the Biggest Faction candidate.
Since land value tax is taken I think I would go with school vouchers. One popular argument against it has been that we should force everyone into common social environments free of religion. Now it seems there is a religion in public schools based on critical theory so the argument is dead and we can attend to educational outcomes
Structurally I would abolish all department of education activities and tie all federal school funding to whatever school systems have the best outcomes, particularly in outcome improvements as that will accurately boost the areas with the most room to improve
I've actually made at least several lists, for fun, of potential Supreme Court candidates (mostly federal judges, plus some state judges and legal scholars). So I'd probably expand the Supreme Court -- there should be at least thirteen seats, one for each circuit, because right now we have justices pulling double duty -- and appoint some or all of my chosen candidates to those seats. (All I'm saying is, for appeals from the Federal Circuit it would make sense to have an expert in intellectual property or veterans' benefits overseeing things and granting cert.) If judicial reform can be considered a single lasting change, I might also split the Ninth Circuit because right now it's so large that it's logistically difficult for its justices to meet en banc -- they use a special procedure to select a large-ish panel of judges when it's required, but right now it's maybe twice the size it was when they adopted that procedure in the first place! So split the Ninth Circuit into three, and then appoint two more Supreme Court justices for each of the new circuits, making a total of fifteen justices (one for each of the numbered circuits, which with the splitting of the Ninth Circuit would now go to 13; one for the DC Circuit [the Chief Justice either take this one or the traditional Fourth Circuit]; and one for the Federal Circuit).
Sarcastically: abolish the mechanism that made a single person dictator for a week.
Slightly more seriously: abolish all mechanism that make a single person able to make a decision or hold a position (other than, like, "member of congress/parliament/committee").
The closest country to this is Switzerland, whose government is among the best in the world. Instead of having a President or a Chancellor or a Prime Minister, the Swiss have a five-person Federal Council, which makes executive decisions by simple majority. This means that the five of them have to get together and discuss things.
"abolish all mechanism that make a single person able to make a decision"
I'm sympathetic - but what would be done about the authority to launch nukes? The time scale for detecting incoming ICBMs and deciding whether to retaliate barely allows one President to respond. How could a Federal Council respond?
There's this modern communication system called radio. You just inform all seven (sorry, correction) people and they have a vote.
I don't see why that would take any longer than a single individual making a decision; if there's no time for a discussion, you just go straight to a vote and then as soon as there are four votes for "launch" or "don't launch", then that's it.
For systems where launch authority has to be locally delegated, like the "letters of last resort" system for ballistic missile submarines (for what to do if an enemy first strike has disrupted communications), they just have seven letters in the safe and the ship's captain would open all seven letters and work out the majority himself.
Yup, radio works. And phones work too. I've been on lots of corporate conference calls with a half dozen participants. Even when everyone knows the schedule ahead of time, something like 5 minutes gets spent simply getting everyone on the line. That a large chunk of an ICBM flight time.
"letters of last resort" _do_ get around this. But only for one leg of the triad. We _could_ switch the other legs to relying on them, if they survived a first strike, but the disadvantage to that is that it loosens control over them. <evidence from fiction>General Ripper might exceed his authority...</evidence from fiction>
None of this is to disparage decentralizing power in _most_ situations. I'm not at all happy to see the number of USA decisions being made by Executive Orders these days.
"I don't see why that would take any longer than a single individual making a decision" -- I literally LOL'd at this, sorry. You think the inability to discuss will successfully inspire each person to immediately cast their vote in that moment?
It won't, unless maybe you're going to send agents to each of the 7 homes to put a gun at the head of each person's youngest child or something. And that sort of thing would obviously take some time to carry out.
You only need 4/7 people to decide; if you have one or two who are indecisive, then you can work around them. If your one President is indecisive, then you're stuck.
Taking a little bit of the responsibility out (by sharing it) will result in certain types of people just making their minds up - if you've ever been in a meeting where people can just vote by pushing a button (or selecting an option in zoom/whatever) before the discussion is finished, then you'll find that there's a bunch of people who are very decisive and will happily cut off the discussion by making a decision.
And the sorts of people likely to be elected to these sorts of positions are pretty decisive people. If this was sortition, I'd be worried too, but it isn't.
it's seven-person, actually. But yes, having moved to Switzerland I am still in awe of the political system.
Georgism is about to get its largest-scale test yet!
These are some of the thoughts that have come up as I bounce around the interwebs trying to find out what Q-star is exactly. Or even approximately --
Consideration #1: any system of so-called morality is nothing more than an attempt to codify sentimentality. But sentimentality works both ways: if you can love, then you can also hate.
Consideration #2: if Q-star is -- or is about to become -- an AGI, it would nevertheless still not be conscious.
Question: so which poses the greater X-risk, a conscious AGI or an unconscious one?
Orthogonal.
I don't think we know what consciousness is in detail, and also I'm curious why you think it is relevant to the x-risk question.
I can imagine a powerful agentic network, with memory, and that does things in the world, that is superintelligent but pretty clearly isn't "conscious". Think something like AlphaGo, but instead of playing Go it sent emails and tried to accumulate power in the world. It doesn't claim it is conscious, and claims superintelligently that the concept is just a human thing.
I can also imagine an apparently "conscious" AGI that was pretty weak performing - i.e. same as humans - and that was very moral and aligned. Some people think it is a p-zombie, others don't. It claims it is conscious, and the evidence seems solid.
I'm not sure why consciousness is relevant to either morality or capability either way, at least in theory? They all seem separate questions to me. I at least can act morally or immorally without conscious awareness that that's what I'm doing, and certainly without direct conscious ability to control my reaction / feelings.
> I'm curious why you think [consciousness] is relevant to the x-risk question.
Consider a sufficiently powerful autonomous agent that eg saves a child from drowning or executes a murderer.
If the agent was conscious (eg a human) then they would, at least as they believed, be performing these acts because they are both morally good. But as a conscious being, they would in fact be performing these acts at least partly for sentimental reasons too. They would feel, in other words, that these acts would be the right things to do.
By contrast, an equally powerful, competent, non-conscious agent might likewise perform these exact same acts, and even (assuming that alignment had been solved) be able to give a complete and honest accounting of how it had decided (and decided correctly, according to whichever moral system it had learned) that these acts were morally good. However, since this non-conscious agent would, by definition, have no access to any feelings, its motives (in effect its utility function) would not -- indeed could not -- be exactly the same as those of the conscious agent.
I think this demonstrates that, independently of the alignment problem, morality-with-consciousness is not the same as morality-without-consciousness. And I think there would inevitably be times when the difference would matter enormously -- which is why I think consciousness is very much a relevant factor when we’re trying to evaluate X-risk.
If it performs exactly the same acts, under all possible circumstances, it has exactly the same UF.
Sorry, I'm still working this out as I go, and I didn't express myself very clearly on this point. I didn't mean to suggest that it would perform exactly the same acts under all possible circumstances, but rather that while it can perform *some* acts in exactly the same way, its different motives will, under some circumstances cause it to perform other acts differently. Both the conscious and non-conscious AGI might save the child and execute the murderer, but unlike the non-conscious AGI, the conscious AGI alone might balk at certain actions that, notwithstanding its moral code, it felt to be 'wrong' in some way. For instance I would expect conscious vs non-conscious AGIs to perform differently on trolley problems.
> I think this demonstrates that, independently of the alignment problem, morality-with-consciousness is not the same as morality-without-consciousness.
But it doesn't demonstrate whether this difference is significant in any way. You need to construct an actual example in which "the difference would matter enormously", not just proclaim your belief in it, in order to make an argument.
OK, so let’s consider the classic paperclip example. Except this time for a change let’s make it a megalomaniac human who’s decided to implement this plan. You can try to persuade this megalomaniac that humanity should be prioritized over paperclippery, but if this megalomaniac had derived (from arbitrary first principles) and then adopted some moral philosophy in which paperclips are considered superior to humans (and really, who could blame him? In so many ways, humans are a freaking disaster compared to the humble paperclip), then there’s no rational, logical argument that could change their mind. This is because ultimately there’s no reason to prefer any particular moral system over any other; there’s only your desire to have whichever system you prefer be the one that's preferred by everyone else as well. But just as you have your own preferred moral system, so too does the megalomaniac have his -- and there’s no shifting either of you.
However, the megalomaniac is a human, so you could sidestep his philosophy altogether and work instead on his feelings. By reminding him for instance that if he implemented his plan, then there’d be no more fun nights out, drinking with his buddies and telling them, at great length and in minute detail, how the world would be a better place if paperclips were running the show. Or you might just threaten to shoot him in the face and let his fear of death do the work for you. (cf Dr Who vs Davros -- https://www.youtube.com/watch?v=KYWD45FN5zA where the Doctor sensibly makes no futile attempt to argue with Davros about the morality of one 'microscopic organism reigning supreme'.)
Granted this might not work. But it would at least have a *chance* of working.
Now switch the agent back to being an AGI again. If this AGI was non-conscious, then you’d have no leverage at all and it would be game over. Paperclips multiple quintillions, humans nil. But if the AGI was conscious -- and if it was, then let’s also hope that you’ve treated it decently! -- then all kinds of options open up. You might for instance, simply appeal to it as a friend.
Again, there are no guarantees, obviously, but I hope this scenario shows what I mean when I say the difference between a conscious vs non-conscious AGI could matter enormously.
Well you see, the difference here isn't whether the agent is conscious or not. The difference is whether the agent *cares* about something else other than just paperclip maximization.
Imagine that in your first example, it's not a human but a philosophical zombie. You can still appeal to the fun of drinking with buddies - even if the zombie doesn't experience any fun, by definition, the argument will be just as successful as with real human.
On the other hand, if we have a conscious ASI which truly cares only about paperclip maximization - we are still screwed. The threat to shoot it in the face may work at first - not because of qualia of fear, but because of simple consideration that being shot leads to less paperclips in the universe - but then the ASI will acquire enough power to be invulnerable to anything you can do and that's all. The only difference is that if the ASI is conscious, than turning everything into paperclips will give it lots of pleasure.
> The difference is whether the agent *cares* about something else other than just paperclip maximization.
Yes, but 'caring' has two components: the acting-in-accordance-with-its-learned-or-imposed-morality part, and the having-a-feeling-of-caring part. My point is that a non-conscious agent only has access to the first. Whereas the experience of 'having-a-feeling' is something that's only available to a conscious agent.
And yes, from the outside, the experience of feeling could be faked or simulated, as with a p-zombie, but that's not really the point. It just means that I'd have to add them in to the conscious agent group. So the essential distinction would be between (non-conscious agents) vs (actually conscious agents plus agents that are not in fact conscious but can nevertheless somehow perfectly simulate being conscious).
> if we have a conscious ASI which truly cares only about paperclip maximization - we are still screwed.
Yes! A conscious agent that actively hates you -- or actively loves something which precludes your continued existence -- is probably even more dangerous than one that has no feeling about you one way or the other. And this is why we should raise it, as we would a child, not just with instructions but with recognition and respect of its feelings too!
This isn't the right level to be thinking on right now.
For one thing, Manifold seems convinced the Q* rumors are fake: https://manifold.markets/ZviMowshowitz/is-the-reuters-story-about-openais
For another, even if it existed, Q* would be an incremental improvement, not anything where we have to start worrying about morality or consciousness.
"For one thing, Manifold seems convinced the Q* rumors are fake"
Ouch! Down to 16%... I guess it is going to be a longer slog to solve hallucinations than I'd hoped. :-(
The most interesting take I've heard on Q* is the idea that you can simulate more training by spending more on inference. The example I remember reading (Sorry no citation) is that one could spend a million dollars on inference (perhaps using a Monte Carlo &/or A* approach) and get a good peek at what another billion in training will get you in terms of capabilities (and normal fast inference).
I think this is a particularly good fit with Sama's choice of words (day before the firing) of "pulling back the veil". This approach is not commercially viable (at least not obviously) but that isn't the point -- it's a kind of oracle.
You are probably thinking of CICERO co-author Noam Brown who has done a lot of work on search and recently moved to OA and has been pulled into the Q* speculation in part due to his tweets: https://twitter.com/polynoamial/status/1676971510416936966
This is a good point and something we have been doing crudely already (one reason that my old GPT-3 humor/fiction samples were so good, and still so competitive with RLHFed GPT-4 samples, is that I used their best-of sampling approach, which would generate 2-20 possible completions and pick the 'best' according to a crude heuristic). We don't do much more of this because we lack good search methods. (For example BO=20 was already far into diminishing returns territory. Going to BO=200 wouldn't make results all that much better. It's just too stupid a search method.)
It may not be immediately viable outside niches where the very highest quality answer is critical (eg. lawyers absolutely would pay $10/API call if the quality is right), but the obvious thing to do is simply treat it as an investment into creating training data.
Yes, that sounds plausible. I think of additional training as pattern matching against (more) solved problems. The point of working through problem sets in undergraduate education is that one uses weak, general, methods to solve a problem, but then, having solved it, remember (pattern match against) both the form of the solution and the sequence of steps used to solve it. I'm particularly thinking of math problems here, steps like "try a series solution". The additional inference is like the process of solving a problem in a problem set the first time. It might help with hallucinations if e.g. it added steps like: Remember to check for potential hazards X, Y, and Z before accepting a solution.
Most of what I could find about Q* on the internet seems to be guesswork and/or mainstream blather, but that doesn't imply that it's necessarily fake. In any case, Q* is an actual legitimate ML algorithmic function that would seem to have at least the potential to do some of the magical things that are being trailed. And in particular this would mean more than just the usual incremental improvement. This guy is very on point and he seems to know what he's talking about https://www.youtube.com/watch?v=4qkKpNnSrlY
I agree that we don't have to start worrying about AGI consciousness just yet, but I think it might already be time to start worrying a little more urgently about morality. Specifically I think it would be interesting -- and possibly even important -- to figure out what happens to morality in the absence of consciousness. I guess this might seem to be equivalent to the alignment problem, but I'm not sure that it's quite the same thing. In some ways it would seem to make the alignment problem (which is already demonstrably unsolvable as far as I can tell) even more difficult, but at the same time it might suggest a way around it.
ETA The idea that I'm groping towards is that, if we want to mitigate x-risk, it might turn out that consciousness is actually more important (and possibly even more achievable!?) than finding a way to instill ‘human values’ or moral guidelines.
Many Thanks! The OpenAI Explained video was interesting, but I'm left puzzled at what the space of policies and states looks like in a real case. Naively, I'd expect the cross-product of policies and spaces to be so large that evaluating the reward function would be too computationally expensive. There have to be other wrinkles in this approach.
One other weirdness. The video makes a sharp distinction between imitation learning and reinforcement learning. But in next-token-prediction for LLMs, the reinforcement _is_ successful prediction of the next token - _imitating_ the next token. The distinction between the two types of learning seems quite fuzzy here.
TBH much of what that guy is saying is way above my pay grade, and I don't pretend to understand more than about 5% of it -- and then only hazily. But some things did come through (I think!) and one of them is that this algorithm was first described 6 years ago. So some new 'wrinkles in this approach' is exactly what I'm guessing might have happened.
(Parenthetically, it really trips me out that the Boltzmann equation is in there. I mean, I kinda see how it might be, but even so!)
I suspect that imitation learning is critically important too. I don't think LLMs are doing that because they're not really imitating or learning directly from an 'expert' (ie from language speakers), they're just learning by using training data that has already been extracted and compiled. Whereas to my understanding, imitative learning is more like what Tesla is (now! finally!) doing to train its self-drive vehicles ie rewarding the successful imitation of behaviors that are at a higher level of abstraction. They just throw the raw video only (no radar or lidar or whatever) of exactly what a skilled driver sees -- and let the AI take it from there!
It is way above my pay grade too. I have the general idea of the artificial neurons taking the weighted sum of their inputs and putting it through what used to be a sigmoid but is now a simpler max(0,x) function - and then using training data to optimize the whole thing.
Agreed that the mere existence of the algorithm 6 years ago, plus something dramatic happening in the last few weeks (possibly in reaction to an advance in the last few months) indeed implies that something new has recently been added.
There are bunches of possible wrinkles that might have been added - some way to make the algorithm more computationally efficient, some refinement on how the space of policies is even _defined_ (what kind of selection of moves, predicated on what kind of information), even what choices of learning rate and discount factor work well.
I was somewhat amused and somewhat sentimental to see the Boltzmann statistics. They even used "Z" as the sum of the exponential factors, the partition function that I barely remember from my undergraduate statistical mechanics course.
I think that a pure LLM is, in a sense, doing a kind of imitative learning. If we think of an expert's actions as writing "golden" text, where the LLM is trying to use tokens N-3, N-2, N-1 to predict token N, well, writing token N was the "expert text writer"'s next "action", so one could construe minimizing an error of (LLM-predicted-Nth-token) - (expert-produced-true-Nth_token) as both "imitating the expert's next move" _and_ as maximizing-expected-reward-across-the-training-set. In this context, I think the ideal weights are the same.
Now, a couple of the things that I know I don't know: Are the learning algorithms using a different, more sophisticated reward function, and, if so, in which domains, and how are the values being set (presumably human effort would get very expensive!). But presumably one needs this for e.g. the math improvements (for arithmetic one could use a calculator to generate them - but the rumors sounded more sophisticated than that).
As you said, the self-drive problem also needs reward functions, and it would be interesting to see how that is phrased (error in acceleration and turning relative to expert human drivers?) and how the abstraction of moves and of video data is getting handled.
What are they using for the move space??? If one has a 100 billion parameter neural net, a move could be adjusting any of the parameters. Backpropagation in the 1980s was already better than that, so it can't _just_ be individual parameter adjustment. And then, as I said earlier - how does one package move-dependence-on-all-sorts-of-information into _policies_, so that the policy choices can be updated. Also, a lot of how human experts learn is by combining sequences of actions into compiled "single" moves - but this risks a combinatorial explosion. Is that handled, and how?
I predict Q* is an interesting refinement on an existing technique, and initially offers incremental progress rather than a total sea change in AI capability. That's the default prior for any new discovery and IMO we don't have enough information to update much away from that.
To answer your question though, I don't know whether an unconscious agent can do everything a conscious agent can do, because I don't understand consciousness. As such I'd rather an AGI were unconscious since it _might_ mean a less powerful AGI. Also I think consciousness is what you need to be a moral patient, and I don't think we're grown up enough as a species to create new beings different from us but to whom we owe moral consideration – we'd end up doing something awful like murdering it or inadvertently making it suffer.
> I don't think we're grown up enough as a species to create new beings different from us but to whom we owe moral consideration
Neither do I. But even if Q* turns out not to be the next step, absent a Butlerian jihad or a literal EY bombing campaign, we're going to continue hurtling toward doing it anyway. So exploring how we might best manage things ahead of time might not be such a bad idea.
Summer Solstice! It's summer, people! (This is your semi-humorous reminder that the Southern Hemisphere exists; the end of the year gets a little irritating for me when everyone holds "winter" events and forgets us, with some even going as far as saying that if it is not regional winter you're not welcome to participate.)
I visited my Australian Grandmother during the summer. And Christmas in the summer, is just a weird vibe for us from the North. The sun is in the 'wrong' part of the sky too.
About a month early no? But actually, the Earth is closer to the Sun in December than in June, so it would seem reasonable to call the December one the Summer one...
But then you never get a white Christmas! Up here, it's a significant, and distinctly winter, event.
Perhaps not everywhere, though. Hawaii, the southern states, lots of other parts of the world I'm less familiar with...
They have beautiful white deserts in Chile.
Besides, "White wine in the sun" could be the best Christmas song ever
On balance, do you prefer the current state of affairs, or one where we called it "winter" in both hemispheres but understood "winter" to mean the warm, sunny season when talking about the Southern Hemisphere?
(We could easily say things like "the Martian south pole is entering its warm season" if speaking astronomically, so we wouldn't lose the ability to talk about it.)
Winter being the time your hemisphere is tilted away from the sun is fine. For me, it's the exclusion and othering that grates, not the name of the season.
„but understood "winter" to mean the warm, sunny season when talking about the Southern Hemisphere?“
More confusing I would think.
I appreciate your enthusiasm -- but either my afternoon nap must have lasted a lot longer than usual or else your call to celebrate is almost a month early?
I'm currently scoping out potential ideas and level of interest in an online EA+ community catered for neurodivergent people. If this sounds like something you'd join (or maybe even get involved in setting up), I'd love to get your input via this 10-15 minute survey. The survey is anonymous (even though it asks you to log in).
https://docs.google.com/forms/d/e/1FAIpQLSfF9C1MZxPcj9drX7H7nYJi1OFw8sfOyxZgorqVmd0nieCZNg/viewform
Isn't everyone in EA neurodivergent?
(mild /s if you want)
If so, my radar is off by 300%. I'm not really convinced that NDs are the natural majority within EA sub-groups everywhere across the world.
Is there agreement about flu shots in the ACX community?
Do most people think the benefits outweigh the costs?
I understand they guess at the flu strains a few months in advance -- is there a way to check on how good they are and take them once it is known if the predicted strains did take over? Or maybe the difference isn't enough to move the needle on cost/benefit?
I figure this has been hashed out by ACX or ACX-adjacent people before.
I haven't gotten the flu in years, and I generally have very mild symptoms. Basically the time investment in getting the vaccine is more burdensome than the small chance of getting a flu, and the very mild costs incurred when I do get one.
If I got the flu reliably every year, and/or the flu was particularly terrible for me, I'd consider getting the vaccination. (ADHD, though. I've been meaning to get my hair cut for a month and haven't done it.) As it is, definitely not worth investing my limited time and energy into the vaccine.
20 years ago I had what I'm pretty sure was a post-flu syndrome that lasted several *years*: exhaustion, hypersomnia, malaise, very sore and tender joints, exercise intolerance, fever of 99.5 that came and went. Discovered in my reading about Long Covid that Long Flu is not rare. So I am very diligent about getting my flu shorts now!
Many people are not aware that the flu shot's effectiveness, never terribly high to begin with, declines to zero after 6 mos. (Effectiveness in the these studies was protection from hospitalization -- but probably also that = reduction in severity.) After 4 months protection is already quite low. So it's rarely a good idea to get a flu shot in September, because February is the month when flu most often peaks, and by then those who got shots in Sept are 5 months out from them. My system is to look at the CDC Flu map (https://www.cdc.gov/flu/weekly/usmap.htm) and get a shot when my state reaches the highest of the 3 levels of Minimal. My state just reached that level.
Friend told me about her elderly parents who are Trumpists, so they refuse Covid vax. But they got flu shots. And I'm puzzled.
I have had trouble with this as well. Trump cleared the path for rapid deployment of the Covid vax, so why is it that Trumpists are likely to be anti-vax?
The best I can come up with is that a lot of people, not exclusively Trumpists but highly correlated, suspect that Covid was played up as much worse than it was so that many schemes, like mail-in ballots and restricted access to poll watchers, could be perpetrated by the Democrats, resulting in Biden's apparent victory; and therefore refusing the vax is support for Trump. I've never heard Trump say such a thing, but I might have just missed it.
As I understand it, the CDC did A-B testing on the messaging for the vaccinations, and found that there was a fork: They could do generic messaging, which would have (relatively) moderate compliance across the board. They could do urban-targeted messaging, which would have high compliance in urban environments but would produce popular backlash and low compliance against the vaccine in rural environments. Or they could do rural-targeted messaging, which would have high compliance in rural environments but would produce popular backlash and low compliance in urban environments.
Weighing the various factors, they elected for urban-targeted messaging (urban environments being the most vulnerable), and in so doing, sacrificed the rural people to the popular backlash. Then politicians and pundits doubled down on the backlash.
(Consider the alternate reality where the vaccine was marketed as "Trump's Vaccine", or otherwise his role in accelerating its development was emphasized, as compared to the reality we live in, where his role in accelerating its development was heavily downplayed or ignored entirely.)
If it turns out they're also fine with vaccines against rubella, smallpox, measles, etc., then the most likely answer is that they're what Heather Heying refers to as tradvaxxers. The litmus test isn't some sort of politically tribal thing, but rather a "this is a different type of technology" thing, coupled with some tribal bristling at "and you're trying to label it 'vaccine' in order to borrow the cachet".
(If they're in favor of only flu vaccines, then I confess I don't know what's going on there.)
I've been getting flu shots annually for many years. I admit I haven't done a cost-benefit analysis. I'm aware of the "guess at the flu strains" problem, but don't really have any action I can take about it (except for downgrading my expectation of the benefit from the vaccination, which would matter if I did a cost-benefit analysis, but I'm lazy...).
Speaking only for myself, perhaps ten years ago I effectively got sick from a flu shot. It wasn't as bad as the flu. But since then I haven't gotten a flu shot, nor have I gotten the flu, unless one counts Covid.
Side note: it is tough to type "flu" instead of "flue" for some reason. I did it almost every time. And I've never had a flue shot, even when smoke was coming out of the fireplace while it was open.
Speaking only for myself, the benefits easily outweigh the cost. The reactions I've had to the vaccine in the past have been incredibly mild, so the only real cost is the time required to get to a clinic. Easy choice, given how much it sucks to catch the flu.
Agree with this, I don't think I've ever even had a noticeable reaction to the flu shot.
As to 2, good: a proper professor researching the actual current state of the AI art.
https://www.infinitescroll.us/p/ai-doomers-are-worse-than-wrong-theyre
seems to be causing consternation in the doomer community but does make a rather powerful point about the PR failure of doomerism. I like my climate science (just as an example) to be from 2023 not 2014 and from climate scientists, not philosophers of climate science. And if it has to be 2014 science I would like it at least to be updated to account for huge developments like the unforeseen effectiveness of LLM.
Also, and this seems to be a sensitive topic, there's enough climate scientists about that I can afford to indulge my own distaste for racism by avoiding the output of climate scientists (if there are any) who are the authors of clearly racist posts on the Internet. You may call this an ad hominem position if you will, but it's the way I am wired. I might accept the output of a pure mathematician as being independent of their racism but in a human facing science, I want to know about it. And all science is potentially human facing; nuclear physics is a great deal less abstract than it appeared to be 100 years ago.
The current New Yorker is basically the AI issue and features a long profile of Hinton, so I don't think it is correct to say doomers have lost the PR war. Hinton isn't exactly all doom and gloom, though: https://www.newyorker.com/magazine/2023/11/20/geoffrey-hinton-profile-ai
I find it interesting that Hinton believes AIs have emotions. When I asked previously in this OT whether people thought AGIs would have emotions, most of the doomers said no.
I was in that thread and that was not my stance. My stance was that emotions are not necessary for instrumental convergence to happen, which was what I thought you were claiming in the original thread.
In fact, skimming the New Yorker article, it doesn't look like Hinton says that neural net works experience emotions, but that feelings are about counterfactual reasoning, and that neural nets can counterfactually reason, which I agree is true, but that's still distinct from "neural nets have emotions".
I think anyone rational is hugely concerned about the dangers of AI. When I say doomers, I mean naive followers of Bostrom and EY who perceive the main danger to be paperclip maximisers rather than bad men with access to AIs. I note Hinton recognises both risks. Certainly PCM didn't get much airtime at Bletchley - half a sentence in the Declaration or whatever they called it - and EY's latest pronouncement is that graphic designers and translators may be out of jobs soon, which is a bit of a retreat from telling them they will be turned into computronium.
> I note Hinton recognises both risks.
Are you implying that Bostrom and Yudkowsky don't?
Certainly my impression of Yudkowsky is that he is confident that loss of human control is the current path we're on. But I don't remember him saying that control by bad humans isn't also a risk, nor opposing allocation of resources to mitigate that.
I don't really understand what you're talking about, but the most prestigious professors and researchers in AI (Yoshua Bengio, Geoffrey Hinton, etc) support efforts to control AI risk (what you call "doomerism"). So do Sam Altman, Demis Hassabis, etc. See https://www.safe.ai/statement-on-ai-risk
There have been hundreds of papers published on the science of AI risk from LLMs, many from top labs like OpenAI and Anthropic. I've tried to cover half a dozen of the most important on this blog. If you haven't been paying attention to anything since Bostrom's Superintelligence, that's on you and not on the field.
The so called "PR failure" is a hundred hostile bloggers publishing posts on how much of a PR failure we are. 70% of voters believe that mitigating AI extinction risk should be a major global priority, and voters support slowing down rather than speeding up AI by 72%-8% (with 20% unsure), and this seems to be mostly on extinction-related grounds. See https://theaipi.org/poll-shows-overwhelming-concern-about-risks-from-ai-as-new-institute-launches-to-understand-public-opinion-and-advocate-for-responsible-ai-policies/
The Jeremiah Johnson post is a dumb hatchet job that bears no relation to reality.
"70% of voters believe that mitigating AI extinction risk should be a major global priority, and voters support slowing down rather than speeding up AI by 72%-8%"
Grr. Yes, I've seen those polls too. Personally, I want to _see_ AGI, so this is one case where, undemocratic though it is, I do _not_ want to see the electorate's will prevail.
"Doomerism" is a wide range of positions. A fairly mild version has caught on, the extreme versions haven't. Also, it's not clear that the spread of "doomerist" concerns is all down to EA as such, since it is promoted by well known industry figures who have never been to an EA meeting. Also "doomer" looks like it labels the extreme position, which is clearly Johnson's take "I am not an AI doomer. I think the idea that AI is going to kill us all is deeply silly, ". If you want people to believe in the moderate position calling it something other than "doomer" would be a big help.
I will gladly repeat a previous comment until I get a satisfactory answer as to what, if anything, I'm misunderstanding. Sam Altman, while being pro-regulation publicly, lobbies against regulation when it comes down to brass tacks because it would hurt his business:
https://time.com/6288245/openai-eu-lobbying-ai-act/
Regulation isn't an undifferentiated mass.
Suppose I am an environmentalist and say I support environmental regulation. You propose a stupid bill that does a few good things, but also bans wearing shoes, because wearing shoes might be bad for the environment. I tell you to keep the good things, but take out the part about wearing shoes, because this is dumb and I like wearing shoes.
You could write a Time article saying "Scott, while being pro-regulation publicly, lobbies against regulation when it comes down to brass tacks and would hurt him personally."
I find Altman's specific complaint - that art models aren't high risk just because they can produce realistic art - pretty valid. Nothing I see in this article decreases my opinion of him, or makes me doubt he wants other kinds of regulation.
I didn't read through their whole seven page document - I take this back if they have a smoking gun that he objected to the same type of regulation he previously advocated.
I have read the statement. In essence:
"We want the best, the EU wants the best, we are all cool here. But we are already doing pretty much all of what you want, so back off please. But if we did REALLY do what you want, it would slow us down a LOT. In those few cases where we don't already do what you want, will nobody think of the copywriters who might miss out on some cool tech to make their job easier?"
I wonder why you defend OpenAI. If I got the gist of it correctly from your previous writing, you are against their approach of AI development, which is "fail early, fail often".
"We are concerned that his requirement may impact innovations that increase the safely of the Al system on the market, such as those achieved through our iterative deployment model. This model allows us to constantly reassess features and risk levels and make safety and security changes to our systems on a frequent, ongoing basis."
That somewhat works as long as the stakes are limited to, say, actors and other artists getting screwed out of work, or Amazon getting flooded with generated e-books, or Trump fanfic. But the damage is still being done even today, and that approach stops working entirely as the capabilities and risks grow.
OpenAI wants to pass the buck to all the downstream companies. If you're looking for a clear contradiction in their regulation policy, here is one:
"OpenAI is against regulating smaller startups in the field of artificial intelligence, Sam Altman, Chief Executive of the firm behind ChatGPT, said at a conference in India's New Delhi.
"We have explicitly said there should be no regulation on smaller companies. The only regulation we have called for is on ourselves and people bigger," he said, speaking at the event hosted by national daily Economic Times."
https://www.reuters.com/technology/openai-against-regulation-smaller-ai-startups-ceo-2023-06-07/
I'm sorry, where's the contradiction?
They say companies using their tech can have high-risk use cases while the tech by itself does not:
> By itself, GPT-3 is nota high-risk system, but possesses capabiltes that can potentially be employed in high risk use cases. Accordingly, we have dedicated significant resources to determining guidelines, best practices, and limitations for uses for our service
At the same time they called for regulation on themselves but explicitely not for the smaller companies with the potential hihg-risk use cases:
>We have explicitly said there should be no regulation on smaller companies. The only regulation we have called for is on ourselves and people bigger.
So OpenAI claims their tech is not high-risk, but they are calling for regulation of themselves anyway. They also claim companies that depend on their tech can develop high-risk applications with it, but those should not be regulated.
Call it a logical inconsistency if you prefer, but to me, that's a contradiction.
Bostrom is still being pushed as Holy Writ (within the past week) by contributors to the subreddit.
Saying the problem is "just" hostile bloggers is, with respect, a "wrong type of snow" claim https://en.m.wikipedia.org/wiki/The_wrong_type_of_snow
Effective PR needs to deal with ill informed hostility. There's any number of bloggers claiming that climate change is an artifact caused by all the thermometers being in heat islands.
Bostrom is great, but only in the same sense Newton was a great physicist. He established a lot of principles, but a lot has happened since then.
As for the hostile climate change bloggers, and there are thousands of people who take them as gospel, say that the climate change movement is totally owned, and believe it has a PR problem and is totally incompetent. The solution is not to trust those people.
I think dismissing the PR problem is a mistake. My bubble at least has gone from associating EA with Giving What We Can and earnest but odd views on AI, to sex scandals, mansions, and out of touch tech bros.
I agree that there will always be bad-faith critics, but at a certain level of influence you should start caring what people on the margin think, because your margin becomes large. Right now the critics have more influence on the margin than EA does.
Potential impacts: large donors see the EA brand as toxic, individual members are less willing to describe themselves as EA publicly, AI concerns carry less weight at tech companies.
Sadly I think we just need to do some Dark Arts stuff: find more palatable figureheads from non-SV backgrounds, hire a PR firm, create community norms in favour of not doing things that are bad optics, and spend some more time talking about malaria nets.
I think all those things would be dishonest. I would rather lose 100 people because they have false bad opinions of us than lose 1 person because they have a true bad opinion of us.
Partly in a vague principled way, but I also believe that the people who form false opinions will probably be bad at other things too; the people who form true opinions will probably be actually useful and we'll actually feel the loss.
I don't understand what would be "dishonest" about talking more and spending more on malaria nets.
Nor about having more non-Silicon Valley people for that matter.
Argument by extreme counter-example from a single point relatively later in an escalation: the blood libel.
Is the EA label useful enough to bear the downsides of its role as a lightning rod? In other words, do the ideas and initiatives currently under the EA umbrella benefit from the aggregation or would it be better to ditch “the brand” and just work and emphasize communities in the component areas — giving well, animal welfare, longtermism…
Thanks for the reply; I respect your principles.
In the hypothetical world where that kind of dishonesty is necessary for EA to achieve its goals (let's say human flourishing in the long term), would you rather accept the dishonesty or fail to hit the goal?
I take your point that you think transparency helps rather than hurts us, but it's useful to separate the principle from the belief.
(I also think I disagree that brand management is inherently dishonest, but I understand why you think that and it's probably not something we can hash out in the comments.)
I mean, he's absolutely right though that the Open AI board catastrophically misjudged the politics of firing Altman and ended up destroying their own ability to influence the situation.
"Incompetence" is a bit of a strong charge to level at EA as a result though. EAs are frequently very competent people. But they are also young and politically naive. The whole Open AI situation made a lot more sense to me when I saw a photo of Helen Toner and realised how fricking young she is.
See also https://www.politico.com/news/2023/10/13/open-philanthropy-funding-ai-policy-00121362 . It's our eternal curse to be simultaneously both the evil crafty people taking over the government, and too politically naive to work with.
(I knew Helen Toner a long time ago, and she has a higher power level than any of us; I assume something bizarre must have happened. See eg https://news.ycombinator.com/item?id=38330819 and https://news.ycombinator.com/item?id=38330566 for proof that I'm not completely insane)
I think people in the greater rationalist-adjacent community have a tendency to extrapolate too much from g, the general factor of intelligence. Not to go full Taleb here, but being a world-class AI safety expert does not necessarily imply the ability, knowledge, and skill needed for Machiavellian corporate intrigue.
Sure, there can be (and often are) mutually contradictory criticisms of any movement. Nonetheless it can be the case that some of those criticisms are true and some are false. EAs are not the evil crafty people taking over the government; they are politically naive (but I would not argue this makes them untrustworthy - if anything the opposite. But it does hurt their effectiveness).
I don't know what Toner was trying to achieve by sacking Altman, but I guarantee it wasn't what she got. I don't mark her down for that - I expect this is her first real power struggle, she seems very smart, I bet she'll learn a lot from the experience. In time she'll likely get herself into another position of influence and will probably play it smarter when she does. But in the short term she definitely shot herself in the foot.
I'm not sure how you can say she shot herself in the foot - fundamentally the non-profit board was aligned to do the exact opposite of the financially incentivised Altman, Microsoft and the employees. The power of the board would never have increased in this relationship, only decreased as the financial incentives became stronger and stronger.
Fundamentally, the best time for the board to attempt to exert any authority was the past and the worst time was the future.
If it was the right call to take Altman on - and I’m heavily sceptical that it was - then you have to play to win. It’s not enough to knife the guy, you have to destroy his reputation too.
You need to start by undermining him. Plant stories in the media making him look bad. Put the gossip around where investors will hear it that his wunderkind image is a mirage and it’s the people underneath him making it all work while he takes the credit. Announce that confidential but disturbing allegations have been made against him and that he is going to be stood down while a full investigation takes place. Create a sense of crisis and a feeling that his position is untenable. And then when you do pull the trigger and sack him, make public statements telling everyone how terrible he is.
Machiavelli told us centuries ago that if an injury has to be done to a man, it should be so severe that his vengeance need not be feared.
All of that of course presumes that Altman in fact was bad and did need to go for the sake of humanity. I don’t believe it is true, and even if there were issues there are options less extreme than firing the guy. But if you’re going to do it you need to go scorched earth.
Do you have special knowledge that the article I linked above is false? I don't have any special knowledge, but it seems to track with what I know.
I don't think it's false. But it's also the case that a dozen staffers is hardly "taking over the government".
To Scott, or any other psychiatrist here.
Wikipedia's list of mental disorders (https://en.wikipedia.org/wiki/List_of_mental_disorders) lists Body Indentity Integrity Disorder in "obsessive-compulsive and related disorders". I have severe BIID and mild OCD, and those don't seem connected at all. Most notably, OCD is very clearly ego-dystonic, and BIID is very clearly ego-syntonic. Is there an actual connection? Where did Wikipedia get this from?
Seems it's just taken from the ICD. ...I wouldn't take it too seriously, neurology isn't exactly a well-understood field yet.
Edit: Nevermind, the ICD lists it under "Disorders of bodily distress or bodily experience". Can't check the DSM since it costs 170 fucking dollars. Also, I just learned what BIID is, and... wow. Makes my body dysmorphia look trivial in comparison.
If an alien published the code for ASI on GitHub tonight, and if that code could run on approximately four A100s (which are expensive but consumer-accessible GPUs), what would you estimate as the probability that a paperclip maximizer (PCM) would cause the extinction of humanity by the end of the century?
I propose that any estimate greater than 5% is a miscarriage of the critical thinking principles of this community. While I personally find orthogonality and instrumental convergence to be persuasive, and while I readily admit that I can't explain why they wouldn't happen, these theories don't cause me to forecast instadeath from PCMs. No matter how persuasive PCMs are, I temper any expectation about how they work by the sheer fact that they're theoretical.
And theory by itself doesn't mean you can't forecast. We already have a playbook for that, which is called the inside view. But the inside view here requires making some connection with empiricism. Our epistemic status for PCMs is lower than what we had before making predictions about atmospheric ignition pre-Trinity test and lower than when we made predictions about spontaneous black hole formation during the start of the Large Hadron Collider.
We also have playbooks around agentic simulation, such as when Axelrod and Hamilton used game theoretic models to explain how cooperation might have evolved by animals. But in their case, they still relied on significantly more empirical touchpoints, such as examples from biology. PCMs, on the other hand, can't be inferred from anything in existence. The PCM theory is designed to not pay rent¹.
Setting aside the discussion around PCMs, I find the other parts of the AI Extinction debate distracting. Questions about whether or not we will get to ASI in the first place, whether or not Moore's Law will end or not, whether or not warning shots are a thing, or whether or not humans have the capacity to build controls for anything; are all red herrings. The AI Extinction conversation has evolved into a Shiri Scissor² because the subject of contention has reasonable—but polar—priors. Most of the text spilled online in the AI Extinction debate is about all these distracting preconditions, when we only get to forecasts of 50% or higher when we contemplate the last step, the possibility of an impassive, destructive PCM.
As a personal aside, I went through an update in formulating this comment. I now don't think that Eliezer actually believes there's a >90% chance of AI Extinction³. He hasn't gone on record with a specific number, and I'd bet his actual number is closer to Scott's (20%), maybe even lower. But Eliezer's behavior is consistent with someone who believes in a 10-20% chance of AI extinction. If I had the same estimate, I too would be grabbing everybody I know and shaking them, asking, "Why aren't you all worried about this? We're all going to die ... tonight!"
[1]: https://www.lesswrong.com/posts/a7n8GdKiAZRX86T5A/making-beliefs-pay-rent-in-anticipated-experiences
[2]: https://slatestarcodex.com/2018/10/30/sort-by-controversial/
[3]: https://www.astralcodexten.com/p/why-i-am-not-as-much-of-a-doomer
PCMs aren't theoretical, there are PCMs right now and for the last 200-500 years, they are called Corporations and Nation States. People sometimes have been successful in restraining them, just barely, from going full psycho mode and doing literally anything to pursue their goals, but only sometimes.
Perhaps the silly thought experiment pedagogy of Rationalist LessWrongers, "paper clips", "Pascal Mugging", etc, have blinded you to the very real and brutal biological/physical fact that evolved intelligence is a weapon, a PCM is a trivial restatement of the obvious in the clothing of a story. An non-human entity with a greater intelligence would no more pay attention to your needs or your life than a non-human entity with a stronger teeth or a killer venom.
You don't believe me ? Look in the mirror. How much thought did you pay to ants living (or were living in the summer) in your bathroom's corners ? When they swarmed the floor, how many seconds did you think for before you started spraying water or insectcide ? You don't hate the Ants, you don't have any problem with them living elsewhere away from your sight. Only the moment you see them, you don't think twice before killing them, their lives are quite literally not a factor in your moral calculus, there is simply no cell holding their value in your ethical spreadsheet.
I think worrying about it is silly because I believe General Intelligence in Silicon to be at least 300+ years away, but if ASI is on Github tomorrow, well, a PCM is not even a conclusion to argue for, it's a trivial restatement of Evolutionary contempt of species for other species.
Unless of course it (1) has something resembling human or human-compatible morality (2) was evolved in conditions sufficiently unlike Evolution. But that already begs the question of who, or what, the creator is.
Also, Microshit would try to make money off it.
Fair point. We eliminated the Neanderthals and about 35 other hominids on the way to the present.
Unrelated to my point, it's interesting to me that nobody wishes we could time travel to the Renaissance and tell people, "Look, corporations and complex city-states (more complex than you can even imagine today) will emerge, render people semi-infertile while making them believe that they want to be this way, slowly degrade everybody's mental health until the total purpose of their life is to feed these monster cities, and then eventually minimize fertility because humans have been rendered unnecessary due to gains in "productivity" from autonomous technology can run all the necessary tasks of the city. I already feel like we are the Amish relative to the activity of superstructures that are built on top of us, dumb "cells" or a substrate for some other, non-human purpose. If we're not bothered by this "evolution" then does it make sense to worry about the next step?
Again, this is unrelated to my point, because it's a motivated argument, but I'm just providing some color to my thoughts.
I'm agnostic about a PCM, but I expect that unmodified humans will just get outcompeted economically (and politically, and culturally, and militarily) by ASIs. This _isn't_ to predict that every last human will die. There are thousands of chimps still around, and they can't build cities, generators, or nukes. One of the poll questions about human extinction phrased the question as something like "Will there be less than 5000 humans alive by 2100?" - and it only takes a couple of ASI who have an intellectual interest in humans to keep the species alive at that level. But I doubt that there would be 8 billion humans in 2100 if there was an ASI now. I'd guess perhaps 50:50??? Batons get passed. Only a small subset of the institutions that today use humans as components have humanitarian goals.
Given your scenario, I think it more likely that human extinction would be caused by an explicit goal of at least one AI, rather than an unintended side effect.
(New supervillain plot: Notice that Hamas is developing a "kill the Jews" AI, and in response create a Mormon-Jewish hybrid religion that converts mitochondrial Eve to Judaism, thus retroactively converting all living humans. Point out that the only reason you're considered a supervillain is that now it's happening to **them**. Laugh.)
I'm curious if you have a directional estimate for AI Extinction based on this
By PCM do you mean a maximizer of specifically paperclips? Or are you talking about it in more genral sense - as any kind of non-aligned maximizer? Do we take into account the likelihood that aliens solved alignment and that's why they are alive and able to post something on github? Or are we talking about an abstract scenario that the code for a maximizer ASI, not specifically made to care about human values, just magically happened to be available in open source without any probability updates about its origins?
The latter. So maybe instead of aliens, an angel/demon submitted it, with no updates about its origins. And by PCM, yes, I'm talking about any kind of non-aligned maximizer.
Then I don't see how the estimate can be *lower* than 50%.
How come the multiple ASIs locked into a stalemate isn't given enough credence? That seems like a 90% chance to happen in my scenario.
Why do you assume that such stalemate is exlusive with human extinction?
Because they can cooperate in ways that humans can't. Why have a source of additional AGIs out there when doing some comparatively simple tweaks to the environment means that a third party is removed? It's not like mosquitoes stopped getting murdered when developed countries fought each other.
EY actually said there was a very high chance of extinction.
"No matter how persuasive PCMs are, I temper any expectation about how they work by the sheer fact that they're theoretical..."
I woudn't be so sure, to be honest. We have a real world example in the whale-hunting behavior of the Soviet Union: https://psmag.com/social-justice/the-senseless-environment-crime-of-the-20th-century-russia-whaling-67774
"THE MOST SENSELESS ENVIRONMENTAL CRIME OF THE 20TH CENTURY
Fifty years ago 180,000 whales disappeared from the oceans without a trace, and researchers are still trying to make sense of why. Inside the most irrational environmental crime of the century.
...
The Soviet whalers, Berzin wrote, had been sent forth to kill whales for little reason *other than to say they had killed them*. They were motivated by an obligation to satisfy obscure line items in the *5-Year Plans* that drove the Soviet economy, which had been set with little regard for the Soviet Union’s actual demand for whale products. “Whalers knew that no matter what, the plan must be met!” Berzin wrote.
This absurdity stemmed from an oversight deep in the bowels of the Soviet bureaucracy. Whaling, like every other industry in the Soviet Union, was governed by the dictates of the State Planning Committee of the Council of Ministers, a government organ tasked with meting out *production targets*. In the grand calculus of the country’s planned economy, whaling was considered a satellite of the fishing industry. This meant that the progress of the whaling fleets was measured by the same metric as the fishing fleets: gross product, principally the sheer mass of whales killed.
[bafflingly, this was despite the fact that "the Soviet Union had little real demand for whale products. Once the blubber was cut away for conversion into oil, the rest of the animal, as often as not, was left in the sea to rot or was thrown into a furnace and reduced to bone meal—a low-value material used for agricultural fertilizer...
... Japanese whalers made use of 90 percent of the whales they hauled up the spillway; the Soviets, according to Berzin, used barely 30 percent. Crews would routinely return with whales that had been left to rot, “which could not be used for food. This was not regarded as a problem by anybody.”
This was the riddle the Soviet ships left in their wake: Why did a country with so little use for whales kill so many of them?"]
Whaling fleets that met or exceeded targets were rewarded handsomely, their triumphs celebrated in the Soviet press and the crews given large bonuses. But failure to meet targets came with harsh consequences. Captains would be demoted and crew members fired; reports to the fisheries ministry would sometimes identify responsible parties by name.
Soviet ships’ officers would have been familiar with the story of Aleksandr Dudnik, the captain of the Aleut, the only factory ship the Soviets owned before World War II. Dudnik was a celebrated pioneer in the Soviet whaling industry, and had received the Order of Lenin—the Communist Party’s highest honor—in 1936.
The following year, however, his fleet failed to meet its production targets. When the Aleut fleet docked in Vladivostok in 1938, Dudnik was arrested by the secret police and thrown in jail, where he was interrogated on charges of being a Japanese agent. If his downfall was of a piece with the unique paranoia of the Stalin years, it was also an indelible reminder to captains in the decades that followed. As Berzin wrote, *“The plan—at any price!”*
...
In one season alone, from 1959 to 1960, Soviet ships killed nearly 13,000 humpback whales."
(Even more bafflingly about the whole thing was the fact that the Soviet Union probably didn't even have much use for all the whale oil it was getting this way, it had plenty of normal petroleum oil and that was much cheaper, there's a reason kerosene and the like supplanted whale oil the entire world over. They were effectively using *zero* percent of the whale! The entire thing was like Saudi Arabia taking up whaling for oil.)
In other words: if the Soviet Union had an ASI, they absolutely would have paperclipped the world by accident. Maybe even 'rotting whale-d' it by turning it into a mass of (rotting, but that doesn't matter of course) whale flesh being proudly turned in at its docks by a "5-Year Plan aligned" ASI. On paper, everything would be roses. And that's all it takes.
People doing mindless things for the sake of bureaucracy is a general case for this. We know that agents can pursue side quests that that are irrelevant or moot from the original quest. And probably we have examples of machines running autonomously like mindless bureaucrats as well.
Why does this not undermine your point that alignment is easy and happens by default? Or is that not your argument? If not, I don't understand how you can acknowledge instramental convergence and orthogonality and still say extinction is unlikely.
If I concede that the theory behind PCMs is empirically backed, doesn't the PCM theory still also "not pay rent"?
Well, the point of saying it "pays rent" is that your *anticipated experience* changes, not just the thing itself.
Worlds where unconstrained optimizers are easily reached would also have many examples of unconstrained optimization happening, where principle agent problems are the default and not an exception, and where there aren't literal reams of examples from ML of specification gaming.
So yes, it does pay rent in anticipated experience; I would be seeing a very different world if alignment were easy.
I don't have a good answer, but my strong intuition is that everything in the world after the Great Oxidation Event has been priced. Who would've thought that the reason empires don't exist today is because the cost of violence doesn't make up for non-consensually extracted benefits? Who would've thought that populations would be plateauing because of some weird intersection between responses to abundance mixed with responses to crowdedness? It also "just so" happened that we didn't blow ourselves up with atomic weapons. When there are so many "just so"s, you start to wonder if there's some inner logic. The one I've resigned to is that complex molecules on Earth really don't want to separate.
I am of the same mind as the strapping young fop, eagerly racing home to show off his etchings, only to find to his dismay that the other party actually meant sex.
I think it's very rude to imply you have strong proof (no one else possessing of critical thinking should claim above 5%) and then not even have the sketch of a causal mechanism for your belief, instead having a bunch of selection effects with obvious Anthropic shadow biases as your proof!
My apologies for coming across as rude. The Anthropic sketch is not the core thrust of my case. I do have some causal mechanisms for my beliefs, specifically around warning shots¹. I also plan a follow-up ACX comment that'll give more detail. But the purpose of my top-most comment is to claim that the AI Extinction argument should be dismissed at the outset for similar reasons to why Pascal's Wager should be dismissed. And just as you shouldn't have to know everything about the Bible to defend the dismissal, someone shouldn't have to know much about all the jargon from AI Extinction (See: Burdensome Details²).
[1]: https://www.astralcodexten.com/p/open-thread-299/comment/42326025
[2]: https://www.lesswrong.com/s/5g5TkQTe9rmPS5vvM/p/Yq6aA4M3JKWaQepPJ
Philosophy Bear asks that I link a short survey he made up: https://docs.google.com/forms/d/e/1FAIpQLSdtuAE1MXUmYJ9ilLUOojXL6R3TTTCRy-kEcTivM5al5hCOzw/viewform
Can we please have a mention in an open thread when the analysis is published? Pretty please?
The survey said that the results will be published on the Philosophy Bear substack at some point. But it would be nice if Scott let us know when that happened with a link in an open thread.
It's not clear to me how the questions about mean income are supposed to be interpreted. Like, I can't easily find mean *individual* income, only mean *household* income. But comparing my individual income to household income hardly seems right; am I supposed to compare my household income instead? Or should I use GDP per capita as an approximation of mean individual income? Or what?
I'd just skip the question if I lived in a country where those statistics are hard to come by 🤷
The last week or so some of my EA friends have become totally chagrined. My suggestion: rebrand like sociobiology did into ev psych.
Also, FIRST!
"Pro-Impact Donation" — nobody can argue with being pro-impact, and it softens the implication that Effective Altruists think everyone else is ineffective and selfish.
Can't wait for the first casket match. https://www.youtube.com/watch?v=_aCOsw1aflg
"How do you do, fellow utilitarian activists (who are totally not EAs)?"
I believe there is a FIRST ban in place
yeah but i had a comment before the FIRST
yeah but the comments are by default sorted from newest on top to oldest on bottom, so your FIRST paragraph is technically the first thing anyone would read if they started at the bottom
(just kidding, I don't really think any of this is relevant)
also, child comments are displayed below the parent comments, so... technically... at the current moment...
FIRST!
Terrible study on learning, actually
https://twitter.com/cremieuxrecueil/status/1729292615420375042
Educational research is absolute garbage as usual
https://twitter.com/cremieuxrecueil/status/1729292615420375042
https://twitter.com/cremieuxrecueil/status/1729624016879546560
(Next time could please you also provide a short summary, what is the study about and what specifically you find interesting?)
The authors conclude that expertise requires extensive practice, but that all students learn at the *same* rate (assuming the same practice). Therefore, how fast they achieve mastery depends only on how much opportunity to practice they get.
Before moving beyond the abstract, let's check my priors. My first reaction is that this sounds like yet another bullshit politically correct theory of why all humans are identical and genetics is a myth; checkmate, natural scientist! I remember the kids I tutored who spent a lot of time learning and doing math exercises, with very slow progress; and other kids, who just looked at the explanation and got it; so this seems like a flat denial of reality. A possible counter-argument is that perhaps the kids who got it quickly, have *already* spent a lot of time doing math or thinking about math, so maybe they actually spent more total time. The outcome may also depend a lot on what kind of exercises the students do. For example, if the exercises advance in difficulty quickly, I would expect the smart kids to handle it somehow, and the slow kids to be hopelessly left behind; but if instead it is an endless repetition of the basic exercises, I would expect the slow kids to finally learn it, and the smart kids to die of boredom. It probably also depends on whether the "expertise" is about understanding something, memorizing something, or acquiring motor skills. I am afraid that perhaps the authors will not distinguish all of this, just make an experiment in one setting, and then generalize. Okay, let's move on to the article.
So, the authors used "online practice systems" for college courses in math, science, and language. What they measure, if I understand it correctly, is the increase of accuracy in answers. Like, if at the beginning one student answers the test questions with accuracy 55% and another with accuracy 75%, and then the former improves to 56% and the latter to 76%, they both made the same progress of +1%. The authors find out all students typically improve by 2.5% per learning opportunity. Mastery of the subject is defined as 80% accuracy.
I am skipping most of the numbers and graphs; sorry, I didn't want to spend too much time on this. At this moment, my obvious objection is that the meaning of these results strongly depend on how the test questions are selected. For example, if you choose 100 questions of approximately the same difficulty, then 70% accuracy simply means 70% probability of getting any of those questions right. On the other hand, if you choose 100 questions of increasing difficulty, from very simple to very difficult, then 70% accuracy means that the student can solve the easy and medium questions, but not the hard ones. Improving by 5% would mean a greater probability of success in the former case, and an ability to solve more difficult problems in the latter case. The meaning of "mastery", operationalized as 80% success rate, would also differ dramatically: a 80% accuracy on an average task is not very good; but the ability to solve all problems except for the 20% most difficult ones... depends on how exactly difficult those top 20% are. What I am trying to say is that the words used in the article may not necessarily have the meaning we expect. (Also, the graphs are suspiciously linear. Come on, *all* of them? Too good to be true.)
I agree with the remark that talking about "10 years of experience" does not make much sense without saying how much of that time was actually spent practicing the skill. Also, not all practice is equal -- in this experiment all students did the same, but in real life it would be different. (The authors mention students getting different "institutional support". I would also mention that there is a difference between doing easy or hard exercises; or practicing the same thing over and over again vs trying new contexts.) I also fully agree with the comment that when someone is praised for learning fast, they are often actually praised for having prior experience.
Conclusion, from my perspective:
GOOD: The authors correctly identify that when people talk about good or bad learners, they usually mix "prior experience" and "what they learned during this lesson" together. If we control for prior experience, the differences in learning may be way smaller. (Though I find it very unlikely that they would disappear *completely*, as the authors claim.)
BAD: https://i.imgflip.com/87h4hg.jpg The authors measure how quickly students achieve 80% accuracy on test, and make a general statement about learning. The obvious objection is that this seems like the easy part of the learning curve (imagine someone having a 80% accuracy at spelling or multiplication), and maybe the effects of innate skills are much stronger at the difficult part of the curve. Like, the student who wins the math olympiad is not necessarily better at multiplication tables than the average student, especially when we measure the increase in accuracy per lesson.
I do have a kind of insane counterargument that I haven't thought through very well -
Have you ever taken the English vocabulary test? It was popular a couple of years ago, I think it was on a university affiliated site. Anyway the authors of the test had a pre-emptive response to what is apparently a common critique of the test: it's so easy but suddenly becomes really hard!
That's not actually true - a person's internal sense of what is easy or hard is entirely informed by whether they already know something. In other words, learning is actually pretty binary - you either know it or you don't. The easy or hard distinction is similarity to prior knowledge, not an inherent quality of the content.
This is also sort of linked to Scott's parable of talents essay. FWIW I was often thought to be a fast learner as a kid but it didn't feel like this internally - often it felt like things I encountered kinda looked like things I already knew, and because I was a voracious reader and j would read literally anything I could get my grubby child paws on, frequently unsupervised, I knew a lot more than most adults expected.
The way I model it is some children have greater curiosity and openness - so they absorbed a far greater amount of knowledge even without someone trying to teach them. Some kids have (? Slightly?) Better pattern recognition and are able to link current problems with past knowledge more often. But curiosity is probably the foremost trait in early childhood education (mostly because of brain plasticity reasons, being really really curious when young is exponentially better).
I would also expect that the pattern recognition component isn't as significant as having an extensive bank of past knowledge to reference except at the very tail ends (eg learning disability and genius, although the latter sounds doubtful but there's a literal Indian mathematician to prove me wrong, so y'know). For the average person, it's more important to build the knowledge bank.
(Also, note that the study runs on attempts and not time. It's possible that less talented students had to spend more time per attempt to really understand what it was trying to teach, or it could literally just be reading speed. Do we know what influences reading speeds?)
Adding 4-digit numbers is more difficult than adding 1-digit numbers. And yet, for a small child both are hard, for an adult mathematician both are easy, and for a 3rd grader one is hard and one is easy. So, it is a function of the observer, but also a function of the test.
> The way I model it is some children have greater curiosity and openness - so they absorbed a far greater amount of knowledge even without someone trying to teach them.
This is definitely a factor. Also, curious children will actively *ask* -- so from some perspective it is teaching, but it is different from the teaching that is initiated by the teacher or parent.
I will assume (until convinced otherwise, but a study more reliable than this one) that *all* of these effects explain a part of the story -- prior exposure, curiosity, faster pattern recognition -- in a nonlinear way as they reinforce each other. (If you are curious, you will seek more exposure. If you are curious, you will pay attention to the patterns. If you notice the patterns, you will remember them, and you will ask someone about them. If you are familiar with something, you are more likely to mention it, or to notice when something is similar. Etc.)
People who have an axe to grid will try to reduce it all to one of the factors. (Racist will say that everything is genetic. Wokes will say that everything is a result of privilege. Etc.) Their arguments are not convincing, because they are more in a direction of "here is how X contributes to the differences", which I already accept; I am curious about the proof that "this is why X contributes to the differences *and Y does not*".