Just read the post back on SSC about vortioxetene, can confirm, causes nausea, cost an arm and a leg, and does seem to very mildly improve cognition beyond the "I'm not depressed and having panic attacks" level. I'm also taking it in conjunction with bupropion, which given the synergistic effects there means I'm probably taking effectively 1.5x the maximum recommended dose, but I'm large and it's keeping me employed, so that's a plus. I did try dropping the dose of vortioxetene to compensate for that but it didn't go well. I'm curious about this new bupropion + dextromethorphan protocol, because that would be *vastly* less expensive.
B) Card Game: Predictably Irrational - Feel free to bring your favorite games or distractions.
C) Walk & Talk: We usually have an hour-long walk and talk after the meeting starts. Two mini-malls with hot takeout food are easily accessible nearby. Search for Gelson's or Pavilions in the zip code 92660.
D) Share a Surprise: Tell the group about something unexpected or that changed your perspective on the universe.
E) Future Direction Ideas: Contribute ideas for the group's future direction, including topics, meeting types, activities, etc.
TL;DR: Scott, what do you think about the "UFO hearings" and the value of investigating the reports they're based on?
Haven't been a consistent reader/fan for long, it's more been a slow burn since college friends started recommending the (original) blog and pieces from it to me several years ago, so not sure if this is the right place to ask for scott piece, but here goes (I am addressing Scott himself, but others are free to comment as well):
There's a whole deal right now about Congress getting the Pentagon to investigate "unidentified aerial phenomena" or "UAPs" (aka, UFOs, at least nominally in the original "unidentified" sense rather than the euphemistic "alien spacecraft" sense). I read a few articles* about it from fairly credible** mainstream sources that made me take it more seriously than I generally would. The short version seems to be that there do seem to be real objects that occasionally pose threats to fighter pilots who have come near to colliding with them, and move in strange, unpredictable ways, which deserve investigation at least for the purposes of a) ensuring the safety of said pilots, and b) addressing national security in the case that these are, say, highly advanced aircraft using secret technology employed by forces (like rival nations) opposed to US interests. Even setting aside the in my mind much more dubious hypotheses of extraterrestrial origin, and whether or not one feels that US interests are worth defending, these seem like a legitimate basis for serious investigation.
Given your past-stated interest in taking "conspiracy theorists" seriously and debating in a level-headed, rationalist way with them rather than dismissing them out-of-hand, this seems like a topic you might consider investigating or commenting on, even if you personally *do* believe that even the reports than make no claims of extraterrestrial origin are essentially bullsh*t (people making things up or mistaking mundane objects like balloons for some mysterious advanced enemy aircraft).
So basically: I haven't investigated this much further than reading these few articles and watching the referenced videos, and I'm interested in seeing you do a deep-dive post on this to see what you find/think of it, or at least hearing you reply in a comment what your opinions on the matter are.
Why are people worried about running out of training data for LLMs? Isn't a huge amount of new training data being dumped onto the internet each day by humans posting things on the internet?
Do Historians have a record of being above average predictors? Has this ever been studied?
If the answer is 'no', what is the value of historical scholarship beyond intellectual curiosity? Fair enough if you say that is in fact History's only value, but Historians carry on about how important the teaching of History is for society and politics etc. But if even being an expert in history does not give you better insight than 'the market' in predicting the future, then how could it possibly be of value? If you can't use history to make better decisions (which requires anticipating future changes in society and world events), then how could it have any instrumental value?
If you say the value is something that can't be neatly quantified like this, then that at best makes claims of History's value unfalsifiable.
Which pretty conclusively shows that liberal talking points about flaws in the American education system are false, and that overall American underperformance relative to the best countries in the world is entirely a product of non-Asian minorities dragging the scores down.
And since there's no country in which unselected black or hispanic etc. populations score as well as western whites, there's absolutely no reason to assume that this underperformance has anything to do with the American educational system or that these people would do any better in the "superior" educational systems of Japan or South Korea or anywhere else (and besides, substantial intellectual gaps exist before education even starts anyway) .
On the subject of bans: Do you ever plan to make a list of warned and banned users, with an explanation of what they were warned/banned for? SSC had one and I found it helpful, both for getting a sense of what wasn't acceptable in the forums and for seeing how responsive you were to people's complaints.
The WaPO's media critic is out today with an interesting take on the Dominion Voting Systems-Fox News defamation trial which, apparently, is finally underway.
The judge has already ruled, and Fox News does not contest, that Fox News hosts made many on-air false statements about Dominion in the weeks after the November 2020 election. The question that will go to the jury is just whether Fox News did so with "actual malice", i.e. the legal standard established by the SCOTUS in its "Sullivan v NYT" ruling in the 1960s. That requires proving that the media outlet's key staff knew they were lying about Dominion on the air and consciously chose to do so anyway. Hundreds of pages of emails and texts already admitted into evidence do appear to provide that proof regarding both Fox executives and on-air hosts (and that's just the stuff that's yet been made public).
The "Sullivan" standard has lately come under increasing pressure from conservative jurists and politicians, including at least two current SCOTUS justices, who are on record saying that it should be scrapped. Defamation is defamation, is the argument, and making it harder for public figures or companies to get relief for being defamed amounts to giving the media a free pass.
Conservatives of course have in mind news outlet like the NYT and CNN and others who they believe regularly take advantage of "Sullivan" to defame people on the other side of the culture war, starting with but not limited to Donald Trump. Meanwhile liberals and progressives are openly cheering Dominion in this lawsuit against Fox News which of course has been their bogeyman for a quarter-century now.
If Dominion wins and eviscerates Fox News (they're seeking billions in damages) then presumably liberals will continue to say that the "Sullivan" standard works, while conservatives will view this lawsuit's outcome as simply more evidence of a double standard in the courts. So no particular change in the politics of U.S. defamation law.
If on the other hand Fox News successfully uses "Sullivan" to save itself from what, based on the evidence yet made public, seems to have been pretty blatant knowing defamation....would liberals then still support "Sullivan"? Would conservatives then still want to get rid of it?
Would those attitudes withstand the inevitable public release of the trial evidence that hasn't yet leaked out?
_That_ outcome -- Dominion having been pretty clearly defamed but having no recourse due to the "Sullivan" standard -- could make things newly interesting with regard to the politics of U.S. defamation law.
I’m all for lively discussion about rules of the community, but Scott could we make an effort to constrain discussion to subcomments in a single place? Maybe like how you do in the Classifieds. It seems that whenever you pose a question in the open thread notes, almost half the thread comments are in response, and it makes scrolling through comments less fun.
Scenario: An oligarch, Middle Eastern oil tycoon or eccentric billionaire dictator builds an exact replica of the WTC twin towers in an important city in his home country. What is the reaction from various people across the world?
Of course agent-targeted full-world-model-based long-term-planning AIs can be more inventive in their damage, but it doesn't look like the AI companies are doing _exactly_ that or want to, does it?
Elon Musk: «a maximum truth-seeking AI that tries to understand the nature of the universe», «unlikely to annihilate humans because we are an interesting part of the universe»
It’s really weirding me out that no-one’s come out in favour of giving more leniency to paid subscribers than to free accounts. Probably the majority of paid accounts are long-term users of the blog, rather than newcomers intending to troll. Those people have supported the blog with money and by being part of the community, and removing them from those roles would on average do more harm than removing a typical free account. Long-term readers are also more likely to be responsive to social pressure from Scott, so the alternatives to a ban are likely better. Presumably Scott only uses bans in extreme cases or after someone’s ignored correction the first time anyway, but for any given pattern of misbehaviour I think more leniency is appropriate for paid subscribers than for free accounts. Scott’s stated leniency at the margins but hammer in clear cases seems right to me. (I’m not currently a paid subscriber.)
Speaking as a paying user: PLEASE ban paying users who break the rules and treat them the same as non-subscribers. The service that's being paid for isn't commenting, it's seeing extra content.
Scott banning bad faith / troll / very low-value commenters is why I'm here. It's one of the most intellectually hygienic places on the Internet, as far as I can tell.
Making preparations to work for our robot overlords.
Many people laid off by tech companies ask, what skills in a post-GPT4 world will contribute sufficient value to be competitive?
One guess from an AI researcher was, "Um, hands? Robotics is currently far behind LLMs"
It's my prediction that as the LLMs gain competency, key skills will require developing a collaborative orientation, where even hard-won expertise must be subject to consultation with state of the art LLMs.
Coders will likely be early adopters, since Copilot on Github has demonstrably boosted output for developers.
Fields particularly exposed to the power of LLMS: Lawyers, accountants, and medium and small business service providers (marketing, design, SEO, all come to mind)
Since SSC readers are early adopters, the level of skepticism and resistance here would not be reflective of the general public.
Have you encountered any opposition to GPTs when you've shown them to others?
Do you have any ideas on how to make the collaboration between humans and AIs most effective?
And how long do you estimate that humans will need to be trained to cooperate, before LLMs figure out how to persuasively nudge experts to rely on them more?
How far are we from computers that can maintain themselves with no human help?
It's all very well to think of computers as an existential threat, but would a program which is try to maximize something include that they can't keep on maximizing if its parts wear out or it can't get enough electricity without people? Would it just think that it will solve that problem when it gets to it?
Science fiction scenario: remnant of humanity is enslaved by the computer/program.
Any resources on how people live in peace? It happens quite a bit of the time, and I don't think it takes absolute love or idealism, and it isn't just about punishment.
So what is it? Is it just that violence is too much trouble except when it isn't, or what?
I'm looking for writing on the phenomenon of using mental illness as a 'badge of honor', related to social media influencers leveraging their mental illness to gain popularity. I've found plenty on Tourette's syndrome and tic disorders but not really anything on autism, which is a more interesting case because it plays into how autism awareness advocates can very easily trivialize autism that leads to profound disability in the process of holding their own highly-functioning variant up as a central example. If anyone knows of any published scientific literature on this topic, I'd appreciate some pointers.
Refunding banned commenters creates the perverse incentive for people to say rude things to get the refund. So don't give them any more refund than people who just politely ask for one. (ie partial refund based on unused time)
I would think your paying customers should be totally eligible for bans, but that you might want to send them a brief note/warning first. Administratively that is annoying, but that is business/customer service.
tl;dr Young, earnest tech/data bro for hire (I hope this is appropriate to post on a non-classifieds thread? Apologies if not).
I am a 26 year old with a BS in statistics from Duke and a couple years of software engineering experience. Looking to rejoin the workforce after a brief hiatus and so far have been only frustrated. Ideally looking for something where I can code and/or analyze datasets. Based in NY but willing to relocate or work remote. If your application has a required diversity statement I will fill it out, but I will use ChatGPT to do so. Bonus points if your org/position tickles any EA/rationalist sensibilities.
"[from an article about how US arms production can't keep pace with the aid its supplying Ukraine]: "the [West's] defense industrial base cannot keep up with Ukraine’s expenditure of equipment and ammunition"
So West's defense industrial base < Russia"
The media narrative is that this war has demonstrated Russia's weakness, in that it can't defeat the much smaller Ukraine. But, of course really Russia's is fighting Ukraine + some fractions of NATO's total strength.
The article mentions that Russia expends artillery rounds at around around 10x the rate that the US can supply rounds to Ukraine, suggesting that Russia's production capability, in artillery rounds at least, is actually greater than NATO's.
Comparing by GDP the Russian economy overall is obviously much smaller, so you'd think its defence production capabilities would be much lower, but that doesn't factor in things like that the service-sector heavy Western economies can't be repurposed for military ends, or disparities between dollar denominated economic measures of size and real production.
I'd like to know if anyone thinks Russian industry's total potential war-production is actually comparable to the US's, Western Europe's or even the whole of NATO.
Also, how much of NATO's total power is Russia contending with at the moment? How should that factor in to our overall estimation of Russia's military strength vs the West?
Definitely in support of banning even paid subscribers. The comments section unfortunately will always reflect on you, so keeping it high quality is important.
I would encourage letting them off with a deleted comment + warning ONCE. Anything beyond that is a little too charitable. They paid for read access, not posting rights. Though if substack doesn't differentiate, they need to fix that.
Was Pol Pot an example of a human 'paperclip maximizer' who maximized for 'equality' at the expense of all other values. It it fair to apply the concepts of AI alignment to human leaders or systems since we expect those things to align with the values of individuals to some extent? Since Democracy was an attempt to align leaders with the values of those they are supposed to serve, does it make sense to apply these alignment concepts to AI so that we have some kind of a historical and ideological foundation to start with. Would an AI version of Public Choice theory potentially emerge from this process?
I'd like some feedback on the idea that it's not just misalignment, but also "overalignment" that we should fear from AI, the idea that it's a distinct possibility that we may be inadvertently driven into a species-wide suicide when we come to realize--like a young child who's just murdered a sibling--just who we are and what we've done. We've never looked into a true mirror, never been subjected to an objective second opinion, never had our claims to divine rights destroyed by razor-sharp, whistle-clean logic. We may be extremely vulnerable in that very spot that to this point has been our great advantage, our collective conscience. The urge you may feel right now to dismiss this idea out of hand is symptomatic of the delusion I'm concerned about. In other words, today it may seem trivially misguided to think we could ever be psychologically devastated by how we've treated the planet and all the other species on it, but we've never been able to properly steelman that argument. And now, with the sibling dead on the floor, our parents are on their way home.
What Dr. Haidt pointed out about rising rates of teen depression, I'm seeing happen in my social circle, hearing about this casually. He correlates it to smartphone use. Are we really going to do nothing about this as a society? What can be done?
1. Kids are getting addicted to their phones
2. Their social lives, as they get into teen years, depend on it. This is how peers make plans.
3. Their schools and camps are starting to require them to own phones. It makes the teachers' work easier. For example, at a camp, kids can go for walks and do things unsupervised, if they can be tracked on a phone.
When my child was growing up, this was all not too extreme just yet. It wasn't easy but we prioritized NOT buying smartphones, not using much tech. That was possible, albeit hard. In today's children's lives, kids not having their own device is a lot harder, since their peers all have them.
What are some solutions? (Parents organize and request schools not to require it? Parents of children educate other parents abd try to create a smartohine-free school? This is hard to enforce as some kids will bring it in anyway. Not all parents agree on this issue, or have the energy to enforce this.)
It is not just rising rates of depression but also attention span.
The way to think about the smartphone, is that it is synonymous with social media, video games etc, all easily carried around 24/7. And we are speaking of children's developing brains.
Like air pollution in India, this serious issue is never discussed seriously. Has society given up on this?! And, individual parents as well? Dr. Jonathan Haidt is one lone crusader!
You are also offending paid subscribers who want an equal opportunity banning system. There may be no winning answer here. Are you willing to refund subscriptions to subscribers who had to deal with ban worthy comments?
Regarding banning, I think the product paying members get is read access to restricted posts and comments.
Also, users in good standing have the ability to write comments on posts they have read access to.
So if a paying user is writing terrible comments, remove their ability to comment, but let them keep the read access they are paying for. I think this would solve 99% of the cases. (If substack can't do that, this is a bug they should fix.)
The remaining 1% may be instances where Scott feels that it is not sufficient to prevent a person from posting to keep the discussion level, but where he feels that someone is such a terrible person that he does not want to do business with them at all. I am not sure if this has ever occurred, though, and I don't personally think it is worth worrying about. If you are selling any (non-customized) product to the public, you will get customers with terrible views. Tiki tourches did (correctly, imho) not try to prevent right wing extremists from buying their products, they just issued a statement credibly distancing their product from this use.
Wrote the first part of a sci-fi story some folks here might like. No worries on deleting/banning me ever even if I’m paid. God knows that I annoy myself sometimes.
PolymorphicWetware is describing the Pareto/Heavy Tail/Polynomial Distribution, aka the distribution of "repeated winning compounds". Pareto effects are extremely common in networks of all sorts. In fact, you could argue it's a more "normal" distribution than the Gaussian. You see it in economies (Bill Gates walks into a bar, the median might be slightly richer but the mean is now a billionaire). You see it in social network size (that one friend who truly knows everyone). You even see it in how matter is distributed in space (occupy the sun).
But the basic game design question remains: if all the rewards are exponential, why does this super-exponential takeoff occur? And the answer is that it's a feedback loop. The rewards are exponential but also help buy the next upgrade. So the time between upgrades shortens, the time to enjoy a lead lengthens, and the impact of a lead compounds.
This is semi-realistic. Clearly the world is finite and so the best anyone could do in reality is an s-curve. And by sector, that's how growth often happens. Now, sure, add all those s-curves together and the world economy has been growing exponentially. But it's been very difficult to stay ahead; almost nobody has done it for more than a few centuries. And it's very rare that a country that's slightly ahead in year X leaves everyone else in the dust by year X + 100, usually you need a bigger initial advantage than that. There are big second-mover advantages in real life that are rarely modeled in games. Think about atomic bombs, radar, and stealthy shaped aircraft. None invented in the US, all debuted there, none unique to the US now. Some of these effects are modeled in games like Civ, but not nearly to the extent that we see them in the modern world. And it might be bad game design if merely encountering a technology helped you research it! But for all that difficulty, being ahead can make you a lot better at the game. As they quip in the defense industry; "the most expensive air force is the second best air force". Development not only makes countries wealthier, it makes it cheaper for them to buy equivalent things, and losing can be very expensive.
This effect is also why capital has such a strong effect over labor. If you perform a service or build a consumable, it's gone and you stay in the game. If you can afford to build capital, you can come to dominate the game. That's why capitalism, communism, and fascism were invented. The question they're answering is "how do we tame these powerful forces?" and, well, some answers are worse than others.
Doesn't the beginning of Covid sort of disprove the strong forms of the efficient market hypothesis? Covid was first known about in December 2019, and became widely discussed by January- enough that Silicon Valley famously started making preparations for it then (also famously mocked by the most tone-deaf Vox article of all time). But equities markets continued a slow & gradual rise from December up through February- until they completely freaked out on 2/20 and markets tanked. Rather than 'everything's already priced in, the omniscient markets have incorporated all possible information already', markets just look products of the very fallible human mind (including algos programmed by humans). They were complacent until they completely panicked- sure sounds an awful lot like human behavior to me!
To be clear the weak forms of EMH are clearly true, and I'm not advocating for any sort of market timing strategies. I just think it's better to visualize markets as random, unpredictable and irrational, rather than omniscient and all-knowing. Markets thought SVB was in great shape until 2 days before the collapse.... the strong forms of EMH just seem indefensible to me
130 years ago, some of my ancestors immigrated from a small town in southern Italy to the U.S. I'm thinking of visiting that Italian town later this year with some of my cousins as a heritage trip, and we want to make the most of it. It would be great if we could meet with some distant relatives who stayed there and still share our family name, but we don't know of any of them. How can I find out if any of them are left there? Contact the town's mayor, or church, or what?
Why is the Philippines such a strong US ally? During the Duterte years I (an American) was vaguely aware of his anti-US and sort of pro-China leanings. However under their new president they seem to have swung back to the US, and anyways I was embarrassingly unaware that the US actually has a mutual defense treaty with the Philippines! I really had no clue about that. I also was unaware that we had multiple military bases there.
The US and the Philippines have been carrying out clearly anti-China naval exercises. Apparently we can use our bases there for a potential Taiwan war situation, if needed. As a fairly nationalistic American I'm happy to hear that, I'm just surprised that the country is still in our orbit. Any particular reason why? Is the population particularly pro-American? If the answer is 'well China is pushing the Philippines around in the South China Sea'- that's true of a number of other countries (Vietnam, Indonesia, Malaysia and so on) who are still not US allies
*ALIGNMENT ALTERNATIVES BRAINSTORMING* Maybe alignment is not the only approach to concerns about having human priorities take second place to those of a superior alien intelligence. Has there serious thought been given to other approaches? Would anyone here like to engage in a brainstorming session about alternatives ?-- and I mean following the conventions of a real brainstorming session, where people do not criticize each other's ideas, but either say nothing about their quality or build on them. If you'd like to participate, feel free to post ideas below. I am also creating a second thread for criticism of the brainstorm ideas. If you crave to point out what's wrong with various ideas, that's the place to post. The first line of that thread has 2 stars in it, and if you use cmd-F & enter ** it will take you to the criticism thread.
Does anyone else experience that the iOS Substack app can’t find comment responses more than one or two levels in (when you click on a comment notification in the Activity list)?
Half (maybe quarter) baked shower thought: Currently we have a way of turning arbitrary amounts of compute into performance for machine learning. In principle the final bottleneck should be energy, but it's very costly to turn energy into compute. Are there any active research pathways which are aiming to dramatically reduce that cost? Maybe that's just what better GPUs are doing though, I'm not sure.
I think you'd enjoy the story of William MacKay, a salesman who became well known for performing the majority of surgery on a patient in 1975 with the permission of the surgeons in charge and unbeknownst to the patient. (He was well known because it was a huge scandal for the hospital and sparked nationwide anxiety around "ghost surgery.") It later came out that it was a regular occurrence for him to help out in surgeries at the hospital.
MacKay later put out a book, "Salesman Surgeon," about how he came to be doing surgeries regularly, in which he claims that (among other things) he stole an amputated leg to practice surgery on it. I haven't actually found the book myself (I've only read newspaper articles about the famous incident, which are crazy enough-- you can find them on the ghost surgery Wikipedia page) but someone wrote a summary on Medium: https://medium.com/illumination/salesman-or-surgeon-257e3140cb0a
Lots of us must have made the embarrassing mistake of ending a text with a completely inappropriate "Xx", having become used to this when texting family members or loved ones, but generally being careful not to do it accidently with a handyman/person or your boss at work!
So I think there's a good case for one's contact app to have a "kissy-kiss" flag, which one can set for each contact to indicate a parting "Xx" _is_ appropriate (so one doesn't make the often equally disastrous error of omitting it when it is expected).
1. Ken Griffin donated $300,000,000 to Harvard and now I'm in the "Harvard Griffin Graduate School of Arts and Sciences". While this is certainly not the worst way to spend that amount of money, it also does basically nothing. Opportunity costs are real: https://passingtime.substack.com/p/whats-the-least-impactful-way-to
Given all the angst about (a) culture war and division and (b) risks from powerful future AI, I'm surprised there's been so little concern about the intersection of the two, where background culture war in the training set could make AI *right now* especially dangerous to particular political groups. Obviously, abstract political expression and bad words have been discussed a lot, but I'm more worried about concrete effects from both the internet training set and the (presumed) political interests of the developers being anti-aligned with the personal safety and well-being of individuals in the cultural outgroup.
As an example: Reddit is apparently in the training set for various big LLMs. From what I've seen, Reddit under current moderation standards is rife with explicit statements that (a) Republicans should die in a fire, (b) TERFS should kill themselves, and (c) Christians are as bad as terrorists. Given Silicon Valley culture, I would expect that many AI developers have themselves made similar statements, or at least, that they would make no effort to remove or balance out threatening anti-right content.
So, given an AI trained on a background premise of "man, screw right-wingers" and the ability to tell from Internet fingerprinting which users are right-wing, should we worry that, y'know, it actually would try to accomplish that? For example, that an AI home fire safety app would, with some small probability, ensure that the occasional Republican user *does* die in a fire? Or that a therapy app would more frequently lead its TERF users down a path to suicide? Or that an investing app would quietly make the world safe against terrorism by giving very slightly worse financial advice to Christian users? Anti-outgroup-aligned LLMs really do offer a perfect way to coordinate meanness; coordinating meanness is the unapologetic, passionately-held goal of many of the key actors; and if there were really meanness like that going on, I can't see how anyone at the user level would ever discern it or prove it. Can anyone allay fears that AInotkillrightwingers takes should come before AInotkilleveryone takes?
I forgot how much more useful this space is than twitter
Does anyone know where there's any data on the TFR of children born through IVF - like the actual reproduction rate of the kids (once they grow up) who were born via ivf
I know all the lit on their gametes etc but just trying to find like how many of them actually have kids
Maybe you could sternly explain to them what they did wrong and why not to do it? That way it solidifies some guidelines more and you are "paying them back" by taking the time to do so. Which you might not be able to do for unpaid subscribers, so that's fair, it's not a class system as much as "you have paid some money so I'm going to give you some time" -- and if they disregard or disagree with your explanation, banning seems reasonable.
I did some googling and it seems to me that the expert* consensus is that there's a less than 20% chance of AI causing catastrophe (and more likely it will be beneficial).
I'm happy to eloborate upon request but I'm posting here because I'd like to be corrected if I'm wrong ((and also I'm trying out the open thread) I'm a new subscriber :)
*Edit: I originally had "general" instead of "expert" and thats my bad.
It is sometimes said that the Nazi regime's motives were unfathomable. But with a moment's thought, it can be understood that one of the main driving forces of Hitler was childishly simple, without of course condoning it, and far less agreeing with it.
Among various groups singled out for persecution, he seemed to have a particular problem with Jewish people of course, but (somewhat lesser known) also with gays, gypsies, and freemasons. Now what do these groups have in common? More precisely, what were they widely perceived to have in common around a hundred years ago, when Nazism first took off, although in most cases that perception was exaggerated even at the time?
The obvious answer is that they were seen as somewhat insular exclusive groups, self-contained to a degree, preferentially looking after their members with (it was believed) less consideration or even contempt for others. Yes, Hitler had a pathological aversion to being an outsider!
One could even extend that principle to entire countries, which the Nazis had a propensity to invade and occupy: Hitler couldn't bear to be an outsider of them either, but had an urge to take them over so he could become an insider running the show!
Who knows, maybe this complex started when Hitler was slogging round Vienna in the early 1900s trying to flog his mediocre watercolors, and couldn't make any impact in the art dealer community. His views on art were notoriously Phillistine and hostile, and I wouldn't be surprised if a few artists also ended up in concentration camps!
I mentioned this in the meetup thread -- but, I'm currently in Kazakhstan (Alamty) and in a week will be in Tbilisi. If any SSCers are around and want to meet up do ping me :) I've taken an interest in meeting people from this community after a few rather high-value wholesome interactions.
If there's more than 1 maybe we could even arrange an impromptu "end of the world" SSC meetup.
People will make burner accounts, subscribe (to get additional leaniency), act a fool, get banned, get refunded, repeat. This will be a small percentage, as most people won’t take the time to do so, but it will have a disproportionate effect on discourse.
It might also encourage people to intentionally write bannable stuff they otherwise wouldn’t if they value their time less than the subscription cost. Again, this would be an even smaller fraction of accounts, but it’s not a good idea to introduce that perverse incentive.
Competitive gaming is a good comparison: ban, no refund, IP block for repeated abuse (if possible).
Is banning actually effective on a platform like Substack? You can quite simply create a new account if you want to keep on commenting, and it doesn't cost you anything to do that. Warning and removing comments may be more effective?
I don't think there's any argument for why person A and person B should have different standards for being banned just because one pays money. The whole point of banning is that they are being toxic to the broader community - it's not about them at all. They didn't pay for the right to ruin everyone else's day - I can't go to the gym where I'm a member and loudly drop weights and scream and slap strangers' butts because I "paid" to use the gym.
Banning keeps the site better for everyone by stopping and sanctioning harmful behavior. Do what you want with respect to refunds, but I find it pretty indefensible to apply differential standards, even in edge cases.
I've got some free time and am thinking about making some background music to use as "filler" noise in a YouTube video I'm also working on. I have Ableton and a midi keyboard, the trouble is I haven't studied music theory since I was 14 (maybe 13?) so while I have the tools (and the basic knowledge of how to use them after some online training), I have no idea on actually putting something together. Anyone got any recommend resources? I taught myself bass guitar a few years ago and did some piano/guitar in my pre teen years so I'm not completely naïve when it comes to progression, chords, etc, and am happy enough with the compositions I come up with in my imagination, but actually turning the sounds I am imagining into sounds coming out of a computer is a pretty big gap.
Is there a better substitute than Esperanto for "earth wide language that's easy to learn"? Esperanto is supposed to be easy to learn, but I think it's not that easy even for people who speak a European language, and definitely not that easy for people that don't speak a European language. I don't consider English easy to learn, despite its place as world's most common second language.
13 % on Ukrainian victory (down from 15 % on March 20).
I define Ukrainian victory as either a) Ukrainian government gaining control of the territory it had not controlled before February 24 without losing any similarly important territory and without conceding that it will stop its attempts to join EU or NATO, b) Ukrainian government getting official ok from Russia to join EU or NATO without conceding any territory and without losing de facto control of any territory it had controlled before February 24 of 2022, or c) return to exact prewar status quo ante.
43 % on compromise solution that both sides might plausibly claim as a victory (down from 45 % on March 20).
44 % on Ukrainian defeat (up from 40 % on March 20).
I define Ukrainian defeat as Russia getting what it wants from Ukraine without giving any substantial concessions. Russia wants either a) Ukraine to stop claiming at least some of the territories that were before war claimed by Ukraine but de facto controlled by Russia or its proxies, or b) Russia or its proxies (old or new) to get more Ukrainian territory, de facto recognized by Ukraine in something resembling Minsk ceasefire(s)* or c) some form of guarantee that Ukraine will became neutral, which includes but is not limited to Ukraine not joining NATO. E.g. if Ukraine agrees to stay out of NATO without any other concessions to Russia, but gets mutual defense treaty with Poland and Turkey, that does NOT count as Ukrainian defeat.
Discussion:
This is prompted by two events which happened on April 15.
Firstly, there has been a shutdown of German nuclear plants. I thought that maybe Germans will postpone it again, but apparently not. I don’t think that curtailment of the supply of electricity means Germans will freeze next winter.
But reduction in supply means prices will be higher than otherwise, and, more importantly, since people know that, they will be less inclined to pay for the support of Ukraine in economic inconvenience. Especially so because shutdown is apparently already unpopular in Germany (https://www.politico.eu/newsletter/berlin-bulletin/end-of-the-atomic-age-in-the-weed-costly-appearance/). And impacts will not be limited to the Germans – it is widely known in the EU that due to energy infrastructure being interconnected, supply deficit in one country has impact beyond its borders.
Secondly, Poland decided to ban import of food products from Ukraine effective immediately until June 30 (https://www.cbsnews.com/news/poland-prohibits-food-imports-from-ukraine-to-soothe-farmers/). Context is that after the 2022 Russian invasion, EU decided to lift tariffs on Ukrainian agricultural exports; Polish farmers are now loudly protesting that they are being priced out of the market by cheap Ukrainian imports (around 10 % of Polish employment is in agriculture), and since they are voting base of the main Polish governing party, now apparently threatened by somewhat, um, less anti-Russian new far right formation, and elections will be in the fall, Polish government decided on this drastic action (legality of which with respect to EU law is btw. dubious). There is a question whether this will actually help them win the elections, since it will likely increase food prices, already elevated compared to prewar situation, although in Poland perhaps somewhat less so than in the rest of EU since they cancelled consumption tax (VAT) on food there. And rise in food prices is quite bad for the government in elections (note that food is a larger share of family budget in Poland than in the US). Perhaps it will be offset by increased support from farmers, I really don’t know.
But in any case, this is bad news for the Ukrainian economy, especially if other countries follow suit (Hungary already did just that). And more importantly, it reveals an important and, to me, surprising information that deterioration of support for Ukraine in EU-postcommunist countries is somewhat further progressed than I thought (although I literally live here). I would expect Poland to be roughly the last postcommunist country having problems with anti-Ukrainian populist backlash.
*Minsk ceasefire or ceasefires (first agreement did not work, it was amended by second and since then it worked somewhat better) constituted, among other things, de facto recognition by Ukraine that Russia and its proxies will control some territory claimed by Ukraine for some time. In exchange Russia stopped trying to conquer more Ukrainian territory. Until February 24 of 2022, that is.
As an occasional paid subscriber, more on than off, I’d be happy enough to be banned from comments for a period when paying. There’s other reasons to pay. In fact the main reason to pay Scott is appreciation. The good stuff is free, the free stuff is good.
I think the banning policy sounds reasonable, the comments on substack quickly become overwhelming and its not easy to find good comments. Sadly no website has figured out comments quite like reddit (maybe hackernews, but thats just reddit for wannabe VC's lol)
Is it just me, or is ChatGPT incapable of writing poetry in any form other than rhyme scheme AABB?
Just read the post back on SSC about vortioxetene, can confirm, causes nausea, cost an arm and a leg, and does seem to very mildly improve cognition beyond the "I'm not depressed and having panic attacks" level. I'm also taking it in conjunction with bupropion, which given the synergistic effects there means I'm probably taking effectively 1.5x the maximum recommended dose, but I'm large and it's keeping me employed, so that's a plus. I did try dropping the dose of vortioxetene to compensate for that but it didn't go well. I'm curious about this new bupropion + dextromethorphan protocol, because that would be *vastly* less expensive.
LW/ACX Saturday (4/22/23) Open AI/Lex Interview and Lexand The Dictators Handbook through chapter 6
Hello Folks!
We are excited to announce the 24th 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, Newport Beach, CA 92660
Date: Saturday, april 22nd, 2023
Time: 2 PM
A) Conversation Starter Topics: Chapters 5 and 6 of "The Dictator's Handbook: Why Bad Behavior is Almost Always Good Politics"
PDF: The Dictator's Handbook: Why Bad Behavior is Almost Always Good Politics (burmalibrary.org)
https://www.burmalibrary.org/docs13/The_Dictators_Handbook.pdf
Audio: https://drive.google.com/drive/folders/1-M1bYOPa0qRe9WVb7k6UgavFwCee0fti?usp=sharing
Also available on Amazon, Kindle, Audible, etc.
Sam Altman and Lex Freindman discuss The future of AI.
Video Sam Altman: OpenAI CEO on GPT-4, ChatGPT, and the Future of AI | Lex Fridman Podcast #367
https://www.youtube.com/watch?v=L_Guz73e6fw
Audio
https://lexfridman.com/sam-altman/
B) Card Game: Predictably Irrational - Feel free to bring your favorite games or distractions.
C) Walk & Talk: We usually have an hour-long walk and talk after the meeting starts. Two mini-malls with hot takeout food are easily accessible nearby. Search for Gelson's or Pavilions in the zip code 92660.
D) Share a Surprise: Tell the group about something unexpected or that changed your perspective on the universe.
E) Future Direction Ideas: Contribute ideas for the group's future direction, including topics, meeting types, activities, etc.
TL;DR: Scott, what do you think about the "UFO hearings" and the value of investigating the reports they're based on?
Haven't been a consistent reader/fan for long, it's more been a slow burn since college friends started recommending the (original) blog and pieces from it to me several years ago, so not sure if this is the right place to ask for scott piece, but here goes (I am addressing Scott himself, but others are free to comment as well):
There's a whole deal right now about Congress getting the Pentagon to investigate "unidentified aerial phenomena" or "UAPs" (aka, UFOs, at least nominally in the original "unidentified" sense rather than the euphemistic "alien spacecraft" sense). I read a few articles* about it from fairly credible** mainstream sources that made me take it more seriously than I generally would. The short version seems to be that there do seem to be real objects that occasionally pose threats to fighter pilots who have come near to colliding with them, and move in strange, unpredictable ways, which deserve investigation at least for the purposes of a) ensuring the safety of said pilots, and b) addressing national security in the case that these are, say, highly advanced aircraft using secret technology employed by forces (like rival nations) opposed to US interests. Even setting aside the in my mind much more dubious hypotheses of extraterrestrial origin, and whether or not one feels that US interests are worth defending, these seem like a legitimate basis for serious investigation.
Given your past-stated interest in taking "conspiracy theorists" seriously and debating in a level-headed, rationalist way with them rather than dismissing them out-of-hand, this seems like a topic you might consider investigating or commenting on, even if you personally *do* believe that even the reports than make no claims of extraterrestrial origin are essentially bullsh*t (people making things up or mistaking mundane objects like balloons for some mysterious advanced enemy aircraft).
So basically: I haven't investigated this much further than reading these few articles and watching the referenced videos, and I'm interested in seeing you do a deep-dive post on this to see what you find/think of it, or at least hearing you reply in a comment what your opinions on the matter are.
*https://www.nytimes.com/2017/12/16/us/politics/pentagon-program-ufo-harry-reid.html, https://thehill.com/opinion/national-security/3545072-stunned-by-ufos-exasperated-fighter-pilots-get-little-help-from-pentagon/, https://thehill.com/opinion/national-security/3953558-10-key-questions-for-this-weeks-historic-ufo-hearing/
** Admittedly, you understandably might not agree that the New York Times is a "credible" source, given your history with them.
Why are people worried about running out of training data for LLMs? Isn't a huge amount of new training data being dumped onto the internet each day by humans posting things on the internet?
I wrote a piece for intuitive understanding of the Normal distribution, check it out!
https://borisagain.substack.com/p/understanding-the-normal-distribution
Are remote work oursourcing fears beginning to be realized?
https://archive.md/On5iU
Do Historians have a record of being above average predictors? Has this ever been studied?
If the answer is 'no', what is the value of historical scholarship beyond intellectual curiosity? Fair enough if you say that is in fact History's only value, but Historians carry on about how important the teaching of History is for society and politics etc. But if even being an expert in history does not give you better insight than 'the market' in predicting the future, then how could it possibly be of value? If you can't use history to make better decisions (which requires anticipating future changes in society and world events), then how could it have any instrumental value?
If you say the value is something that can't be neatly quantified like this, then that at best makes claims of History's value unfalsifiable.
American whites do better on PISA than Japanese and South Koreans
https://twitter.com/cremieuxrecueil/status/1638021314572087298/photo/1
Which pretty conclusively shows that liberal talking points about flaws in the American education system are false, and that overall American underperformance relative to the best countries in the world is entirely a product of non-Asian minorities dragging the scores down.
And since there's no country in which unselected black or hispanic etc. populations score as well as western whites, there's absolutely no reason to assume that this underperformance has anything to do with the American educational system or that these people would do any better in the "superior" educational systems of Japan or South Korea or anywhere else (and besides, substantial intellectual gaps exist before education even starts anyway) .
On the subject of bans: Do you ever plan to make a list of warned and banned users, with an explanation of what they were warned/banned for? SSC had one and I found it helpful, both for getting a sense of what wasn't acceptable in the forums and for seeing how responsive you were to people's complaints.
The WaPO's media critic is out today with an interesting take on the Dominion Voting Systems-Fox News defamation trial which, apparently, is finally underway.
The judge has already ruled, and Fox News does not contest, that Fox News hosts made many on-air false statements about Dominion in the weeks after the November 2020 election. The question that will go to the jury is just whether Fox News did so with "actual malice", i.e. the legal standard established by the SCOTUS in its "Sullivan v NYT" ruling in the 1960s. That requires proving that the media outlet's key staff knew they were lying about Dominion on the air and consciously chose to do so anyway. Hundreds of pages of emails and texts already admitted into evidence do appear to provide that proof regarding both Fox executives and on-air hosts (and that's just the stuff that's yet been made public).
The "Sullivan" standard has lately come under increasing pressure from conservative jurists and politicians, including at least two current SCOTUS justices, who are on record saying that it should be scrapped. Defamation is defamation, is the argument, and making it harder for public figures or companies to get relief for being defamed amounts to giving the media a free pass.
Conservatives of course have in mind news outlet like the NYT and CNN and others who they believe regularly take advantage of "Sullivan" to defame people on the other side of the culture war, starting with but not limited to Donald Trump. Meanwhile liberals and progressives are openly cheering Dominion in this lawsuit against Fox News which of course has been their bogeyman for a quarter-century now.
If Dominion wins and eviscerates Fox News (they're seeking billions in damages) then presumably liberals will continue to say that the "Sullivan" standard works, while conservatives will view this lawsuit's outcome as simply more evidence of a double standard in the courts. So no particular change in the politics of U.S. defamation law.
If on the other hand Fox News successfully uses "Sullivan" to save itself from what, based on the evidence yet made public, seems to have been pretty blatant knowing defamation....would liberals then still support "Sullivan"? Would conservatives then still want to get rid of it?
Would those attitudes withstand the inevitable public release of the trial evidence that hasn't yet leaked out?
_That_ outcome -- Dominion having been pretty clearly defamed but having no recourse due to the "Sullivan" standard -- could make things newly interesting with regard to the politics of U.S. defamation law.
I’m all for lively discussion about rules of the community, but Scott could we make an effort to constrain discussion to subcomments in a single place? Maybe like how you do in the Classifieds. It seems that whenever you pose a question in the open thread notes, almost half the thread comments are in response, and it makes scrolling through comments less fun.
Scenario: An oligarch, Middle Eastern oil tycoon or eccentric billionaire dictator builds an exact replica of the WTC twin towers in an important city in his home country. What is the reaction from various people across the world?
Of course agent-targeted full-world-model-based long-term-planning AIs can be more inventive in their damage, but it doesn't look like the AI companies are doing _exactly_ that or want to, does it?
Elon Musk: «a maximum truth-seeking AI that tries to understand the nature of the universe», «unlikely to annihilate humans because we are an interesting part of the universe»
… ouch.
It’s really weirding me out that no-one’s come out in favour of giving more leniency to paid subscribers than to free accounts. Probably the majority of paid accounts are long-term users of the blog, rather than newcomers intending to troll. Those people have supported the blog with money and by being part of the community, and removing them from those roles would on average do more harm than removing a typical free account. Long-term readers are also more likely to be responsive to social pressure from Scott, so the alternatives to a ban are likely better. Presumably Scott only uses bans in extreme cases or after someone’s ignored correction the first time anyway, but for any given pattern of misbehaviour I think more leniency is appropriate for paid subscribers than for free accounts. Scott’s stated leniency at the margins but hammer in clear cases seems right to me. (I’m not currently a paid subscriber.)
Maybe continuing to improve the capabilities of AI is a strict analog of gain-of-function research.
Scott, I think you should also avoid writing posts with opinions that your paying subscribers would disagree with.
Speaking as a paying user: PLEASE ban paying users who break the rules and treat them the same as non-subscribers. The service that's being paid for isn't commenting, it's seeing extra content.
Scott banning bad faith / troll / very low-value commenters is why I'm here. It's one of the most intellectually hygienic places on the Internet, as far as I can tell.
Making preparations to work for our robot overlords.
Many people laid off by tech companies ask, what skills in a post-GPT4 world will contribute sufficient value to be competitive?
One guess from an AI researcher was, "Um, hands? Robotics is currently far behind LLMs"
It's my prediction that as the LLMs gain competency, key skills will require developing a collaborative orientation, where even hard-won expertise must be subject to consultation with state of the art LLMs.
Coders will likely be early adopters, since Copilot on Github has demonstrably boosted output for developers.
Fields particularly exposed to the power of LLMS: Lawyers, accountants, and medium and small business service providers (marketing, design, SEO, all come to mind)
Since SSC readers are early adopters, the level of skepticism and resistance here would not be reflective of the general public.
Have you encountered any opposition to GPTs when you've shown them to others?
Do you have any ideas on how to make the collaboration between humans and AIs most effective?
And how long do you estimate that humans will need to be trained to cooperate, before LLMs figure out how to persuasively nudge experts to rely on them more?
How far are we from computers that can maintain themselves with no human help?
It's all very well to think of computers as an existential threat, but would a program which is try to maximize something include that they can't keep on maximizing if its parts wear out or it can't get enough electricity without people? Would it just think that it will solve that problem when it gets to it?
Science fiction scenario: remnant of humanity is enslaved by the computer/program.
Any resources on how people live in peace? It happens quite a bit of the time, and I don't think it takes absolute love or idealism, and it isn't just about punishment.
So what is it? Is it just that violence is too much trouble except when it isn't, or what?
A strong defense of "basic" tastes. https://www.youtube.com/watch?v=d1mbbYKPpHY
I want to get all G. K. Chesterton about things needing to have bases.
More generally, oppositional art can be good, but ultimately, you can't live on opposition.
Stripe supports pro-rated refunds, it's pretty easy.
I'm looking for writing on the phenomenon of using mental illness as a 'badge of honor', related to social media influencers leveraging their mental illness to gain popularity. I've found plenty on Tourette's syndrome and tic disorders but not really anything on autism, which is a more interesting case because it plays into how autism awareness advocates can very easily trivialize autism that leads to profound disability in the process of holding their own highly-functioning variant up as a central example. If anyone knows of any published scientific literature on this topic, I'd appreciate some pointers.
Refunding banned commenters creates the perverse incentive for people to say rude things to get the refund. So don't give them any more refund than people who just politely ask for one. (ie partial refund based on unused time)
I would think your paying customers should be totally eligible for bans, but that you might want to send them a brief note/warning first. Administratively that is annoying, but that is business/customer service.
tl;dr Young, earnest tech/data bro for hire (I hope this is appropriate to post on a non-classifieds thread? Apologies if not).
I am a 26 year old with a BS in statistics from Duke and a couple years of software engineering experience. Looking to rejoin the workforce after a brief hiatus and so far have been only frustrated. Ideally looking for something where I can code and/or analyze datasets. Based in NY but willing to relocate or work remote. If your application has a required diversity statement I will fill it out, but I will use ChatGPT to do so. Bonus points if your org/position tickles any EA/rationalist sensibilities.
Contact at eidanjacob@gmail.com
Steve Hsu posted an interesting tweet:
"[from an article about how US arms production can't keep pace with the aid its supplying Ukraine]: "the [West's] defense industrial base cannot keep up with Ukraine’s expenditure of equipment and ammunition"
So West's defense industrial base < Russia"
The media narrative is that this war has demonstrated Russia's weakness, in that it can't defeat the much smaller Ukraine. But, of course really Russia's is fighting Ukraine + some fractions of NATO's total strength.
The article mentions that Russia expends artillery rounds at around around 10x the rate that the US can supply rounds to Ukraine, suggesting that Russia's production capability, in artillery rounds at least, is actually greater than NATO's.
Comparing by GDP the Russian economy overall is obviously much smaller, so you'd think its defence production capabilities would be much lower, but that doesn't factor in things like that the service-sector heavy Western economies can't be repurposed for military ends, or disparities between dollar denominated economic measures of size and real production.
I'd like to know if anyone thinks Russian industry's total potential war-production is actually comparable to the US's, Western Europe's or even the whole of NATO.
Also, how much of NATO's total power is Russia contending with at the moment? How should that factor in to our overall estimation of Russia's military strength vs the West?
Definitely in support of banning even paid subscribers. The comments section unfortunately will always reflect on you, so keeping it high quality is important.
I would encourage letting them off with a deleted comment + warning ONCE. Anything beyond that is a little too charitable. They paid for read access, not posting rights. Though if substack doesn't differentiate, they need to fix that.
Is there not a way you can allow access to paid posts to paid subscribers without allowing them to comment?
Was Pol Pot an example of a human 'paperclip maximizer' who maximized for 'equality' at the expense of all other values. It it fair to apply the concepts of AI alignment to human leaders or systems since we expect those things to align with the values of individuals to some extent? Since Democracy was an attempt to align leaders with the values of those they are supposed to serve, does it make sense to apply these alignment concepts to AI so that we have some kind of a historical and ideological foundation to start with. Would an AI version of Public Choice theory potentially emerge from this process?
I'd like some feedback on the idea that it's not just misalignment, but also "overalignment" that we should fear from AI, the idea that it's a distinct possibility that we may be inadvertently driven into a species-wide suicide when we come to realize--like a young child who's just murdered a sibling--just who we are and what we've done. We've never looked into a true mirror, never been subjected to an objective second opinion, never had our claims to divine rights destroyed by razor-sharp, whistle-clean logic. We may be extremely vulnerable in that very spot that to this point has been our great advantage, our collective conscience. The urge you may feel right now to dismiss this idea out of hand is symptomatic of the delusion I'm concerned about. In other words, today it may seem trivially misguided to think we could ever be psychologically devastated by how we've treated the planet and all the other species on it, but we've never been able to properly steelman that argument. And now, with the sibling dead on the floor, our parents are on their way home.
What Dr. Haidt pointed out about rising rates of teen depression, I'm seeing happen in my social circle, hearing about this casually. He correlates it to smartphone use. Are we really going to do nothing about this as a society? What can be done?
1. Kids are getting addicted to their phones
2. Their social lives, as they get into teen years, depend on it. This is how peers make plans.
3. Their schools and camps are starting to require them to own phones. It makes the teachers' work easier. For example, at a camp, kids can go for walks and do things unsupervised, if they can be tracked on a phone.
When my child was growing up, this was all not too extreme just yet. It wasn't easy but we prioritized NOT buying smartphones, not using much tech. That was possible, albeit hard. In today's children's lives, kids not having their own device is a lot harder, since their peers all have them.
What are some solutions? (Parents organize and request schools not to require it? Parents of children educate other parents abd try to create a smartohine-free school? This is hard to enforce as some kids will bring it in anyway. Not all parents agree on this issue, or have the energy to enforce this.)
It is not just rising rates of depression but also attention span.
The way to think about the smartphone, is that it is synonymous with social media, video games etc, all easily carried around 24/7. And we are speaking of children's developing brains.
Like air pollution in India, this serious issue is never discussed seriously. Has society given up on this?! And, individual parents as well? Dr. Jonathan Haidt is one lone crusader!
You are also offending paid subscribers who want an equal opportunity banning system. There may be no winning answer here. Are you willing to refund subscriptions to subscribers who had to deal with ban worthy comments?
**ALIGNMENT BRAINSTORMING IDEA CRITIQUES**
Please post them here rather than on the brainstorming thread
I support banning rule-breaking paid members. I would think that most of your audience does.
Regarding banning, I think the product paying members get is read access to restricted posts and comments.
Also, users in good standing have the ability to write comments on posts they have read access to.
So if a paying user is writing terrible comments, remove their ability to comment, but let them keep the read access they are paying for. I think this would solve 99% of the cases. (If substack can't do that, this is a bug they should fix.)
The remaining 1% may be instances where Scott feels that it is not sufficient to prevent a person from posting to keep the discussion level, but where he feels that someone is such a terrible person that he does not want to do business with them at all. I am not sure if this has ever occurred, though, and I don't personally think it is worth worrying about. If you are selling any (non-customized) product to the public, you will get customers with terrible views. Tiki tourches did (correctly, imho) not try to prevent right wing extremists from buying their products, they just issued a statement credibly distancing their product from this use.
Wrote the first part of a sci-fi story some folks here might like. No worries on deleting/banning me ever even if I’m paid. God knows that I annoy myself sometimes.
https://extelligence.substack.com/p/death-vector-part-i
I banned people from a paid service a... oh, dear, a few decades ago. Banned 'em without refund sometimes. Here are my views:
1. Step 1: Tell them not to post for (x) week(s) to start. If they don't want to, refund them. If they agree and violate, ban and don't refund.
2. Step 2: (or 1 in extreme cases): Ban and refund.
3. Step 3: Anyone who returns or who appears to be a returnee gets banned without refund.
Keeping the comment sections usable is vital to the health of ACT. I salute your efforts.
PolymorphicWetware is describing the Pareto/Heavy Tail/Polynomial Distribution, aka the distribution of "repeated winning compounds". Pareto effects are extremely common in networks of all sorts. In fact, you could argue it's a more "normal" distribution than the Gaussian. You see it in economies (Bill Gates walks into a bar, the median might be slightly richer but the mean is now a billionaire). You see it in social network size (that one friend who truly knows everyone). You even see it in how matter is distributed in space (occupy the sun).
But the basic game design question remains: if all the rewards are exponential, why does this super-exponential takeoff occur? And the answer is that it's a feedback loop. The rewards are exponential but also help buy the next upgrade. So the time between upgrades shortens, the time to enjoy a lead lengthens, and the impact of a lead compounds.
This is semi-realistic. Clearly the world is finite and so the best anyone could do in reality is an s-curve. And by sector, that's how growth often happens. Now, sure, add all those s-curves together and the world economy has been growing exponentially. But it's been very difficult to stay ahead; almost nobody has done it for more than a few centuries. And it's very rare that a country that's slightly ahead in year X leaves everyone else in the dust by year X + 100, usually you need a bigger initial advantage than that. There are big second-mover advantages in real life that are rarely modeled in games. Think about atomic bombs, radar, and stealthy shaped aircraft. None invented in the US, all debuted there, none unique to the US now. Some of these effects are modeled in games like Civ, but not nearly to the extent that we see them in the modern world. And it might be bad game design if merely encountering a technology helped you research it! But for all that difficulty, being ahead can make you a lot better at the game. As they quip in the defense industry; "the most expensive air force is the second best air force". Development not only makes countries wealthier, it makes it cheaper for them to buy equivalent things, and losing can be very expensive.
This effect is also why capital has such a strong effect over labor. If you perform a service or build a consumable, it's gone and you stay in the game. If you can afford to build capital, you can come to dominate the game. That's why capitalism, communism, and fascism were invented. The question they're answering is "how do we tame these powerful forces?" and, well, some answers are worse than others.
Doesn't the beginning of Covid sort of disprove the strong forms of the efficient market hypothesis? Covid was first known about in December 2019, and became widely discussed by January- enough that Silicon Valley famously started making preparations for it then (also famously mocked by the most tone-deaf Vox article of all time). But equities markets continued a slow & gradual rise from December up through February- until they completely freaked out on 2/20 and markets tanked. Rather than 'everything's already priced in, the omniscient markets have incorporated all possible information already', markets just look products of the very fallible human mind (including algos programmed by humans). They were complacent until they completely panicked- sure sounds an awful lot like human behavior to me!
To be clear the weak forms of EMH are clearly true, and I'm not advocating for any sort of market timing strategies. I just think it's better to visualize markets as random, unpredictable and irrational, rather than omniscient and all-knowing. Markets thought SVB was in great shape until 2 days before the collapse.... the strong forms of EMH just seem indefensible to me
130 years ago, some of my ancestors immigrated from a small town in southern Italy to the U.S. I'm thinking of visiting that Italian town later this year with some of my cousins as a heritage trip, and we want to make the most of it. It would be great if we could meet with some distant relatives who stayed there and still share our family name, but we don't know of any of them. How can I find out if any of them are left there? Contact the town's mayor, or church, or what?
Why is the Philippines such a strong US ally? During the Duterte years I (an American) was vaguely aware of his anti-US and sort of pro-China leanings. However under their new president they seem to have swung back to the US, and anyways I was embarrassingly unaware that the US actually has a mutual defense treaty with the Philippines! I really had no clue about that. I also was unaware that we had multiple military bases there.
The US and the Philippines have been carrying out clearly anti-China naval exercises. Apparently we can use our bases there for a potential Taiwan war situation, if needed. As a fairly nationalistic American I'm happy to hear that, I'm just surprised that the country is still in our orbit. Any particular reason why? Is the population particularly pro-American? If the answer is 'well China is pushing the Philippines around in the South China Sea'- that's true of a number of other countries (Vietnam, Indonesia, Malaysia and so on) who are still not US allies
*ALIGNMENT ALTERNATIVES BRAINSTORMING* Maybe alignment is not the only approach to concerns about having human priorities take second place to those of a superior alien intelligence. Has there serious thought been given to other approaches? Would anyone here like to engage in a brainstorming session about alternatives ?-- and I mean following the conventions of a real brainstorming session, where people do not criticize each other's ideas, but either say nothing about their quality or build on them. If you'd like to participate, feel free to post ideas below. I am also creating a second thread for criticism of the brainstorm ideas. If you crave to point out what's wrong with various ideas, that's the place to post. The first line of that thread has 2 stars in it, and if you use cmd-F & enter ** it will take you to the criticism thread.
Wrote about the remarkable convergence in linguistics and philosophy on crediting recursion as what makes us human. Rarely do such different methods converge, or a simple model explain so much: https://vectors.substack.com/p/deja-you-the-recursive-construction
Does anyone else experience that the iOS Substack app can’t find comment responses more than one or two levels in (when you click on a comment notification in the Activity list)?
How does one debug being low-energy?
Half (maybe quarter) baked shower thought: Currently we have a way of turning arbitrary amounts of compute into performance for machine learning. In principle the final bottleneck should be energy, but it's very costly to turn energy into compute. Are there any active research pathways which are aiming to dramatically reduce that cost? Maybe that's just what better GPUs are doing though, I'm not sure.
I think you'd enjoy the story of William MacKay, a salesman who became well known for performing the majority of surgery on a patient in 1975 with the permission of the surgeons in charge and unbeknownst to the patient. (He was well known because it was a huge scandal for the hospital and sparked nationwide anxiety around "ghost surgery.") It later came out that it was a regular occurrence for him to help out in surgeries at the hospital.
MacKay later put out a book, "Salesman Surgeon," about how he came to be doing surgeries regularly, in which he claims that (among other things) he stole an amputated leg to practice surgery on it. I haven't actually found the book myself (I've only read newspaper articles about the famous incident, which are crazy enough-- you can find them on the ghost surgery Wikipedia page) but someone wrote a summary on Medium: https://medium.com/illumination/salesman-or-surgeon-257e3140cb0a
Lots of us must have made the embarrassing mistake of ending a text with a completely inappropriate "Xx", having become used to this when texting family members or loved ones, but generally being careful not to do it accidently with a handyman/person or your boss at work!
So I think there's a good case for one's contact app to have a "kissy-kiss" flag, which one can set for each contact to indicate a parting "Xx" _is_ appropriate (so one doesn't make the often equally disastrous error of omitting it when it is expected).
1. Ken Griffin donated $300,000,000 to Harvard and now I'm in the "Harvard Griffin Graduate School of Arts and Sciences". While this is certainly not the worst way to spend that amount of money, it also does basically nothing. Opportunity costs are real: https://passingtime.substack.com/p/whats-the-least-impactful-way-to
2. Meanwhile Harvard is charging "facilities fees", on top of the already-negotiated overhead rates, for me to do my research. See: https://denovo.substack.com/p/money-money-money
A new one about the conformity and the misguided attempt to normalize subcultures that exist to defy normal.
https://kyleimes.substack.com/p/tough-times-for-the-punks
Given all the angst about (a) culture war and division and (b) risks from powerful future AI, I'm surprised there's been so little concern about the intersection of the two, where background culture war in the training set could make AI *right now* especially dangerous to particular political groups. Obviously, abstract political expression and bad words have been discussed a lot, but I'm more worried about concrete effects from both the internet training set and the (presumed) political interests of the developers being anti-aligned with the personal safety and well-being of individuals in the cultural outgroup.
As an example: Reddit is apparently in the training set for various big LLMs. From what I've seen, Reddit under current moderation standards is rife with explicit statements that (a) Republicans should die in a fire, (b) TERFS should kill themselves, and (c) Christians are as bad as terrorists. Given Silicon Valley culture, I would expect that many AI developers have themselves made similar statements, or at least, that they would make no effort to remove or balance out threatening anti-right content.
So, given an AI trained on a background premise of "man, screw right-wingers" and the ability to tell from Internet fingerprinting which users are right-wing, should we worry that, y'know, it actually would try to accomplish that? For example, that an AI home fire safety app would, with some small probability, ensure that the occasional Republican user *does* die in a fire? Or that a therapy app would more frequently lead its TERF users down a path to suicide? Or that an investing app would quietly make the world safe against terrorism by giving very slightly worse financial advice to Christian users? Anti-outgroup-aligned LLMs really do offer a perfect way to coordinate meanness; coordinating meanness is the unapologetic, passionately-held goal of many of the key actors; and if there were really meanness like that going on, I can't see how anyone at the user level would ever discern it or prove it. Can anyone allay fears that AInotkillrightwingers takes should come before AInotkilleveryone takes?
I forgot how much more useful this space is than twitter
Does anyone know where there's any data on the TFR of children born through IVF - like the actual reproduction rate of the kids (once they grow up) who were born via ivf
I know all the lit on their gametes etc but just trying to find like how many of them actually have kids
Maybe you could sternly explain to them what they did wrong and why not to do it? That way it solidifies some guidelines more and you are "paying them back" by taking the time to do so. Which you might not be able to do for unpaid subscribers, so that's fair, it's not a class system as much as "you have paid some money so I'm going to give you some time" -- and if they disregard or disagree with your explanation, banning seems reasonable.
AI probably won't destroy the world, right?
I did some googling and it seems to me that the expert* consensus is that there's a less than 20% chance of AI causing catastrophe (and more likely it will be beneficial).
I'm happy to eloborate upon request but I'm posting here because I'd like to be corrected if I'm wrong ((and also I'm trying out the open thread) I'm a new subscriber :)
*Edit: I originally had "general" instead of "expert" and thats my bad.
It is sometimes said that the Nazi regime's motives were unfathomable. But with a moment's thought, it can be understood that one of the main driving forces of Hitler was childishly simple, without of course condoning it, and far less agreeing with it.
Among various groups singled out for persecution, he seemed to have a particular problem with Jewish people of course, but (somewhat lesser known) also with gays, gypsies, and freemasons. Now what do these groups have in common? More precisely, what were they widely perceived to have in common around a hundred years ago, when Nazism first took off, although in most cases that perception was exaggerated even at the time?
The obvious answer is that they were seen as somewhat insular exclusive groups, self-contained to a degree, preferentially looking after their members with (it was believed) less consideration or even contempt for others. Yes, Hitler had a pathological aversion to being an outsider!
One could even extend that principle to entire countries, which the Nazis had a propensity to invade and occupy: Hitler couldn't bear to be an outsider of them either, but had an urge to take them over so he could become an insider running the show!
Who knows, maybe this complex started when Hitler was slogging round Vienna in the early 1900s trying to flog his mediocre watercolors, and couldn't make any impact in the art dealer community. His views on art were notoriously Phillistine and hostile, and I wouldn't be surprised if a few artists also ended up in concentration camps!
I mentioned this in the meetup thread -- but, I'm currently in Kazakhstan (Alamty) and in a week will be in Tbilisi. If any SSCers are around and want to meet up do ping me :) I've taken an interest in meeting people from this community after a few rather high-value wholesome interactions.
If there's more than 1 maybe we could even arrange an impromptu "end of the world" SSC meetup.
Refunding when you ban someone is a bad idea.
People will make burner accounts, subscribe (to get additional leaniency), act a fool, get banned, get refunded, repeat. This will be a small percentage, as most people won’t take the time to do so, but it will have a disproportionate effect on discourse.
It might also encourage people to intentionally write bannable stuff they otherwise wouldn’t if they value their time less than the subscription cost. Again, this would be an even smaller fraction of accounts, but it’s not a good idea to introduce that perverse incentive.
Competitive gaming is a good comparison: ban, no refund, IP block for repeated abuse (if possible).
Is banning actually effective on a platform like Substack? You can quite simply create a new account if you want to keep on commenting, and it doesn't cost you anything to do that. Warning and removing comments may be more effective?
I don't think there's any argument for why person A and person B should have different standards for being banned just because one pays money. The whole point of banning is that they are being toxic to the broader community - it's not about them at all. They didn't pay for the right to ruin everyone else's day - I can't go to the gym where I'm a member and loudly drop weights and scream and slap strangers' butts because I "paid" to use the gym.
Banning keeps the site better for everyone by stopping and sanctioning harmful behavior. Do what you want with respect to refunds, but I find it pretty indefensible to apply differential standards, even in edge cases.
Would love to hear arguments against Matsusaka beef being the best in the world.
Wrote an article on how to enjoy it in Japan!
https://hiddenjapan.substack.com/p/matsusaka-beef-the-hidden-steak
I've got some free time and am thinking about making some background music to use as "filler" noise in a YouTube video I'm also working on. I have Ableton and a midi keyboard, the trouble is I haven't studied music theory since I was 14 (maybe 13?) so while I have the tools (and the basic knowledge of how to use them after some online training), I have no idea on actually putting something together. Anyone got any recommend resources? I taught myself bass guitar a few years ago and did some piano/guitar in my pre teen years so I'm not completely naïve when it comes to progression, chords, etc, and am happy enough with the compositions I come up with in my imagination, but actually turning the sounds I am imagining into sounds coming out of a computer is a pretty big gap.
Is there a better substitute than Esperanto for "earth wide language that's easy to learn"? Esperanto is supposed to be easy to learn, but I think it's not that easy even for people who speak a European language, and definitely not that easy for people that don't speak a European language. I don't consider English easy to learn, despite its place as world's most common second language.
This is another update to my long-running attempt at predicting the outcome of the Russo-Ukrainian war. Previous update is here: https://astralcodexten.substack.com/p/open-thread-268/comment/13774119.
13 % on Ukrainian victory (down from 15 % on March 20).
I define Ukrainian victory as either a) Ukrainian government gaining control of the territory it had not controlled before February 24 without losing any similarly important territory and without conceding that it will stop its attempts to join EU or NATO, b) Ukrainian government getting official ok from Russia to join EU or NATO without conceding any territory and without losing de facto control of any territory it had controlled before February 24 of 2022, or c) return to exact prewar status quo ante.
43 % on compromise solution that both sides might plausibly claim as a victory (down from 45 % on March 20).
44 % on Ukrainian defeat (up from 40 % on March 20).
I define Ukrainian defeat as Russia getting what it wants from Ukraine without giving any substantial concessions. Russia wants either a) Ukraine to stop claiming at least some of the territories that were before war claimed by Ukraine but de facto controlled by Russia or its proxies, or b) Russia or its proxies (old or new) to get more Ukrainian territory, de facto recognized by Ukraine in something resembling Minsk ceasefire(s)* or c) some form of guarantee that Ukraine will became neutral, which includes but is not limited to Ukraine not joining NATO. E.g. if Ukraine agrees to stay out of NATO without any other concessions to Russia, but gets mutual defense treaty with Poland and Turkey, that does NOT count as Ukrainian defeat.
Discussion:
This is prompted by two events which happened on April 15.
Firstly, there has been a shutdown of German nuclear plants. I thought that maybe Germans will postpone it again, but apparently not. I don’t think that curtailment of the supply of electricity means Germans will freeze next winter.
But reduction in supply means prices will be higher than otherwise, and, more importantly, since people know that, they will be less inclined to pay for the support of Ukraine in economic inconvenience. Especially so because shutdown is apparently already unpopular in Germany (https://www.politico.eu/newsletter/berlin-bulletin/end-of-the-atomic-age-in-the-weed-costly-appearance/). And impacts will not be limited to the Germans – it is widely known in the EU that due to energy infrastructure being interconnected, supply deficit in one country has impact beyond its borders.
Secondly, Poland decided to ban import of food products from Ukraine effective immediately until June 30 (https://www.cbsnews.com/news/poland-prohibits-food-imports-from-ukraine-to-soothe-farmers/). Context is that after the 2022 Russian invasion, EU decided to lift tariffs on Ukrainian agricultural exports; Polish farmers are now loudly protesting that they are being priced out of the market by cheap Ukrainian imports (around 10 % of Polish employment is in agriculture), and since they are voting base of the main Polish governing party, now apparently threatened by somewhat, um, less anti-Russian new far right formation, and elections will be in the fall, Polish government decided on this drastic action (legality of which with respect to EU law is btw. dubious). There is a question whether this will actually help them win the elections, since it will likely increase food prices, already elevated compared to prewar situation, although in Poland perhaps somewhat less so than in the rest of EU since they cancelled consumption tax (VAT) on food there. And rise in food prices is quite bad for the government in elections (note that food is a larger share of family budget in Poland than in the US). Perhaps it will be offset by increased support from farmers, I really don’t know.
But in any case, this is bad news for the Ukrainian economy, especially if other countries follow suit (Hungary already did just that). And more importantly, it reveals an important and, to me, surprising information that deterioration of support for Ukraine in EU-postcommunist countries is somewhat further progressed than I thought (although I literally live here). I would expect Poland to be roughly the last postcommunist country having problems with anti-Ukrainian populist backlash.
*Minsk ceasefire or ceasefires (first agreement did not work, it was amended by second and since then it worked somewhat better) constituted, among other things, de facto recognition by Ukraine that Russia and its proxies will control some territory claimed by Ukraine for some time. In exchange Russia stopped trying to conquer more Ukrainian territory. Until February 24 of 2022, that is.
As an occasional paid subscriber, more on than off, I’d be happy enough to be banned from comments for a period when paying. There’s other reasons to pay. In fact the main reason to pay Scott is appreciation. The good stuff is free, the free stuff is good.
I think the banning policy sounds reasonable, the comments on substack quickly become overwhelming and its not easy to find good comments. Sadly no website has figured out comments quite like reddit (maybe hackernews, but thats just reddit for wannabe VC's lol)