I am working on an idea. So imagine Auschwitz is liberated, and the enraged prisoners take the gas chambers down, or an enraged officer blows them up, or the Nazis were destroying the evidence. And then the camera crews come and nothing to show, and then someone rebuilds them for the camera, badly. Not saying this happened, just saying this is the kind of imperfect humans making mistakes that keeps happening.
So someone with spergy superpower notices the doors do not fit and then not having a good theory of mind, does not think of a normal story like this, but a huge conspiracy.
I think this keeps happening, because I see former goodie spergs like Star Trek fans now believing in huge conspiracy theories.
With a normal theory of mind skills, you would assume a true story will have still bad parts, and you have to look at the whole thing. Imagine how could happen, that a huge crime is done, then covered up, then uncovered and investigated and presented, and somehow the amount of made up crap is exactly zero. This does not happen. Humans do not work that well or that honestly.
Similarly, surely someone made COVID mistakes. Also surely crap was made up to cover it up. But the whole picture matters.
Whereas our spergy type expects either a 100% true story with every detail just right, or a complete utter lie.
I don't know whether you're being humorous in framing it as a hypothetical, but in case you aren't, that's precisely what happened: the famous gas chambers at Auschwitz ARE reconstructions. But it's fine because the Soviets were responsible, and they're renowned for their commitment to preserving an accurate historical record for posterity.
Seems to me like there's different axes down which you could examine that phenomenon.
- Increasing distrust in domain expertise - some justified, some not.
- 'Communities' that select on personalities which reflexively reject consensus.
- Standard internet openness to ideas that overweight on 'interesting' relative to 'accurate'.
- Tendency to accept bad ideas that align with unshakeable (or even bigoted) priors.
- Proliferation of information out there to cherry-pick; an operator can artfully string a trail of cherries leading to a secret door (ex: Darryl Cooper).
>- Increasing distrust in domain expertise - some justified, some not.
I think that's probably the big one. When you believe someone is generally trustworthy, you can accept them making a few honest mistakes because hey, nobody's right 100% of the time. When you believe someone is lying to you for personal or ideological gain, you're much more likely to react to a false statement with an "Aha! I *knew* he was full of BS, this just proves it!"
Quest Stage 1, possible sol'n: "A Good Dating SITE/App*"
I have 2 suggested tweaks to what Scott described:
1. Default "first thing people see of you": Females upload a picture for their profile. Males upload writing for their profile. Non-binary users can choose.
2. THEN, once someone clicks on the profile, it goes to a max of, say, 8 further clickables–images, videos, more text, …maybe even Anthropic AI artifacts if we get clever enough. :)
So you can watch a video of the guy winning the hardest level on a computer game or feeding oatmeal to his toddler nephew or showing off his belt sander! To leverage the whole "Guys present themselves well when they are 'in action' thing." (iirc, YT was originally envisioned as a site where people uploaded videos introducing themselves for potential dating-matches to see.)
I am majorly "sold" on the other options Scott mentioned... (relationships being knitted into friendship networks, ability to see people's answers to questions you've asked them before you even have to "make a move") and there might even be room for more than one app in this space.
Of course, the million-dollar (really billion or trillion, depending on which metric you're using) questions are... what about enshittification? What about entities that are ALREADY incentivized to stop such a venture cold before it even got to the level of network effects.
For that, we would first need something like a faceless distributed communication network full of intelligent anonymous users who were highly motivated to solve this problem and have a reasonably high commitment to each-other and their organization. If someone would make the moves to build such a network - or already has - then I think it's possible!
You are missing the fact that smart self-respecting women will be offended by the asymmetry in what aspect of self the app obliges them to be represented by in users' first glimpse. I simply cannot understand why this did not occur to you. WTF, buddy?
Naahh... I orig. had a mega-comment, but then chopped it down to size; I'd accounted for it in the most obv way. (i.e. u don't need to stick to defaults.)
Anyway, pretty sure REAL women would LOVE browsing through guys' favorite quotations or whatever!
My god, what are charmer you are! I don't know why you're even interested in dating apps. All somebody like you needs to do is to speak right up to a woman the way you are here and she'll yank her panties right off and mount them behind your ears and across your face like a covid mask.
Soo... I kind of dissembled during this discussion. I think what I did must fall under the broad category of what the kids these days refer to as "trolling." Maybe I can explain thusly:
> Be me.
> Actually get along really great with basically all the women I meet!
> Do NOT, however, have sex with any of them!!
> Not because I'm gay, but because of my choice in sexual ethics.
> (which dictated I was just going to be with just one person, a male, til death do us part)
> But at least I got to / get to gestate, give birth to and raise a couple of nerdy children! :D
> Send first one off to college a few weeks back.
So, yeah... that's actual_me.
(And no, I don't actually think that women who wouldn't read long book quotations in a potential male date's profile are not "real women"! That was just online sass!)
I don’t see the point of this unpleasant deception, even for you, but don't even want to hear the explanation. I now have no interest in your gender role and sexual ethics issues and why the fuck you played out this little game. Your post just used me and the others who engaged as a little cell culture for your experiment. You could hardly be further from good faith posting. Yuck, lady.
Seriously I'm with Performative Bafflement...don't waste your time and effort on dating apps, the market is seriously saturated and there's an existing player with near-monopoly power.
> Of course, the million-dollar (really billion or trillion, depending on which metric you're using) questions are... what about enshittification? What about entities that are ALREADY incentivized to stop such a venture cold before it even got to the level of network effects.
This is probably the crux of the problem overall - I know of at least two commenters in the ACX commentariat that have tried to replicate "old Okcupid," which is broadly seen as the peak of dating websites here, and neither ever got off the ground.
I'm not sure why this is, but I'd suspect:
A) monetization problems, because now you're competing against the fully adversarial Match Group (which owns essentially all dating apps except the nigh-bankrupt Bumble) and their established footprint, marketing, and revenue models that will always outcompete a less adversarial option, and
B) the type of person whom "old Okcupid" appeals to - ie filling out a bunch of questions and writing and reading lengthy profiles instead of swiping - is such a tiny slice of humanity overall that you simply can't get the network effects to have a viable dating app, because recruiting enough people in an ongoing way is the hard problem of dating apps. Previously, in the golden Natufian age of old Okcupid it could work because that's all there was - swipe apps didn't exist, so enough people would actually answer the questions and write / read profiles. But now swipe apps exist, and they're strongly preferred over reading / writing by enough people you just can't get a large enough nucleus of people and ongoing flow to sustain such a dating app now.
Okay, I'm going to try to steelman a version of the usage of the imagined Dating Site/App* versus what you have in "B" above:
1. (idea i got by thinking about an issue you brought up) Encourage people to potentially use Voice-To-Text their answers? i.e. fix the problems brought to us by our technology (the computer-in-our-pockets proliferation) with... more-sophisticated technology!! (software that knows the spoken word!)
2. You say "not enough people want to push against the grain." Okay, but how many of the most-intelligent people would run to this if it existed? (I think it could be very worthwhile to invest a lot in the fates the tiny slice of humanity that the most-intelligent - say the top 0.15% - NOT because they are somehow more worthy-of-existence or "inherently-valuable" - no! But because if lots of them are better-flourishing, they'll a better crack at fixing lotsa problems of the other 99.85%.)
* Wondering to myself what name I'll call [Imagined_Dating_Site/App], I'm thinking at this point should be humorously called, "Built_on_Skyhooks."
> Previously, in the golden Natufian age of old Okcupid it could work because that's all there was - swipe apps didn't exist, so enough people would actually answer the questions and write / read profiles. But now swipe apps exist, and they're strongly preferred over reading / writing by enough people you just can't get a large enough nucleus of people and ongoing flow to sustain such a dating app now.
What if the people who prefer swipe apps prefer dating partners who also prefer swipe apps, and the people who prefer reading/writing/questions prefer the same in their dating partners?
Then, creating an app like old Ok Cupid would still work (at least in largish cities). In fact it'd work better as it's selecting for a certain type of people.
> What if the people who prefer swipe apps prefer dating partners who also prefer swipe apps, and the people who prefer reading/writing/questions prefer the same in their dating partners?
Because then your general problem is "your marketing still costs just as much, but will only appeal to ~10% of your audience," for a 10x in marketing costs, against a powerful and well funded incumbent.
And you need continuous, ongoing marketing, because dating sites that actually work pair people up and they drop off, so you need an ongoing flow to be relevant and have enough people on your platform to be usable and making revenue.
So your costs are 10x, you have to do them all the time, and your revenue will always be less than the fully adversarial Match Group - overall, that's not a recipe for a business that will last, it's basically a charity, and an expensive one with poor impact per dollar at that.
> Are adverts the main thing pushing people towards websites (either dating website, or anything else)?
More or less - at some point you get social proof and word of mouth / network effects that help draw people there, much like "Tinder" became an eponym for dating apps overall in some circles for a while, but largely you need constant marketing to build to that level, and then marketing to keep your population up, because one of the main problems is that if you don't have a high enough concentration of people in each geography you want to operate in, people will install the app, go through your ten or hundred people, run out of profiles, sputter, then uninstall the app because nobody is there.
Versus Tinder, for example, which has tens to hundreds of thousands of candidates in any real city literally worldwide, and who are your literal competition, so you need to make a decent showing against them, which requires fairly massive marketing spend, and again if you're doing your job right, people are pairing up and dropping off the app, so that marketing spend needs to be continuous, all the time.
> you need constant marketing to build to that level, and then marketing to keep your population up
I've never seen an advert for Reddit or Substack, and their are popular websites. So I doubt that advertising is necessary.
> because one of the main problems is that if you don't have a high enough concentration of people in each geography you want to operate in, people will install the app, go through your ten or hundred people, run out of profiles, sputter, then uninstall the app because nobody is there
It is for reasons like this that I think a new dating site should not just be a dating site -- it should be about other stuff too, including building IRL/AFK communities.
> YT was originally envisioned as a site where people uploaded videos introducing themselves for potential dating-matches to see
That's something I didn't know
> So you can watch a video of the guy winning the hardest level on a computer game or feeding oatmeal to his toddler nephew or showing off his belt sander! To leverage the whole "Guys present themselves well when they are 'in action' thing.
Could make this more general than just a dating app... kinda like Facebook but not shitty.
Also, facebook got started by being seeded in various universities, so if this new thing could do something similar and get concentrated levels of use in some geographical areas.
Or, with intentional communities, e.g. a Christian dating app/site with a Christian ethos.
Or allow people to create their own intentional communities (like a subreddit).
An interesting new thing on Substack lately is, expert demographers pushing back against the prevailing population-projection-statistics narratives. The MSM seems not yet to have noticed this but perhaps will at some point.
It seems coordinated, making me wonder if there was a "oh FFS I guess we're going to have start going online" meeting amongst the world's top demographic researchers.
The demographers are consistently repeating basically just a single point: that the way in which population projections are done "works well when things are changing slowly and in predictable ways, but it means our projections are often limited to forecasting a future that’s a lot like the present." Meaning that in an era in which birthrate patterns are changing in some significant way -- e.g. the age range in which women have children expanding throughout the developed world in a way never seen before -- statistics such as "total fertility rate" lose a lot of their predictive power.
Anyway this week's example is this short "briefing paper":
A historical bit that was entirely new to me is: "...much like today, in the 1930s and 1940s many commentators bemoaned the low fertility of the time. Expert projections of US population from the 1930s assumed that fertility would remain relatively constant into the future. They [such as the official Social Security Administration projections] imagined fertility rates would continue as they were indefinitely, yielding rapid population aging and slow population growth. Yet fertility did not remain low – instead we got the Baby Boom."
>A historical bit that was entirely new to me is: "...much like today, in the 1930s and 1940s many commentators bemoaned the low fertility of the time. Expert projections of US population from the 1930s assumed that fertility would remain relatively constant into the future. They [such as the official Social Security Administration projections] imagined fertility rates would continue as they were indefinitely, yielding rapid population aging and slow population growth. Yet fertility did not remain low – instead we got the Baby Boom."
Within a generation it was back where it was in the 1930s. The long-term pattern is that fertility rates that decline, not increase.
Over the very long term, sure. But "will always keep declining" is an assumption/prediction not a demonstrated reality.
The demographers' core point is that what we all call the fertility rate (technically the "total fertility rate" or TFR) is not an average of how many children each woman has had. Rather, it is a _prediction_ of what an average woman _will_ do if nothing changes or is changing.
If a nation's current TFR is say 1.7 that's saying that if the most-recent age-specific fertility rates are exactly repeated in that nation, the average woman will have that number of children. It is often summarized as simply "the number of children that women in that society will have", without the important caveat of "if nothing is changing".
Many people -- including me until recently -- assumed that TFR was roughly equivalent to the average of actual children being born per woman. But in the US and most of the developing world it currently is not equivalent, because women are waiting longer to have their children and are having some of their children later than was previously normal. The TFR inherently lags behind a broad shift such as that one.
The actual-results statistic is called "children ever born" (CEB) which counts exactly what it sounds like: how many children the average woman who's completed her fertility years, had. That average in the US is currently just a hair under the "population replacement level" rate of 2.05, and in recent years has risen slightly. Overall it's been pretty steady for a while now, has not gone above 2.1 nor below 1.8 since at least 1990.
So the current US TFR of 1.6 is therefore _not_ reflecting the actual current reality of US demographics, something that until recently I was unaware of. It will gradually get into alignment with the actual-births average (CEB) as updated annual real-results data rotates into the TFR methodology, but that is a gradual process over a bunch of years.
The problem with comparing CEB to TFR is that CEB lags TFR by the difference between average maternal age at birth and the age cutoff for the end of a woman's reproductive years, which seems to be about 20 years (average maternal age in the US is 30ish and I think the cutoff used for published CEB stats is age 50).
This lag explains 100% of the current gap between CEB and TFR, as TFR for the US in 2004 was 2.05 and rising gradually towards a peak of 2.12 in 2012.
You can currently find CEB numbers both for a cutoff at 44 and for 50.
Anyway the demographers' point isn't trying to choose which stat. Rather it is that people are misrepresenting what the one stat tells us while ignoring the other stat. Leading to a collective freakout which the demographers say isn't entirely justified.
My point is that once you take the time lag into account, the two stats are not actually telling different stories. It's possible that the current decline in TRF is an artifact of parents having the same number of kids at older ages, but CEB being 2.05 while TFR is 1.7 is not evidence for that hypothesis. The current CEB is telling the same story as TFR about birth rates 15-20 years ago, and we have to wait another 15-20 years to find out if CEB and TFR tell the same story about birth rates today.
Yep. The demographers are not writing that CEB today _proves_ a different current story regarding fertility rates. Rather they are pushing back against the widespread misuse of TFR as being a final word on the issue.
TFR is a prediction which, like all predictions, can be proven by events to have been either right or wrong. "We should be *very* humble about our ability to predict trends in human behavior more than fifty years out, especially about something as personal as childbearing...." Media coverage and social-media blather about the fertility issue consistently reflect zero awareness of this.
The demographers do seem to be pretty sure that the current decline in TFR is in _part_ an artifact of parents having the same number at older ages. Whether it's a big or small part, is something we will learn over coming decades.
Celebrity Millie Bobby Brown adopted a girl recently, and the replies on Twitter are full of Rightists telling her she's not a real Mom and the kid isn't really her child:
This is not a new thing, every time some adopts a child these people are out in force. I want to note that Brown is a married, heterosexual woman. Both she and her husband are white. The only "weird" thing here is that she's only 21 years old, leading to dumb comments from people who aren't aware that young women can be infertile.
I advise conservatives to take inspiration from the way Democrats treated the woke movement: tell people that nobody's saying the thing they're clearly saying, say it's all just a dirty rotten strawman, and absolutely under no circumstances condemn or disassociate yourselves from such views, because doing so means conceding that the Left is right about something and is thus the equivalent what Benedict Arnold did when he betrayed his country. Your goal should be for everyone to suspect that you secretly hold such views, so they'll know to vote for you come the next election, and then when you lose you can complain that people unfairly held you responsible for stuff "random people" posted on Twitter.
And how did that work out for the Democrats? If you can win by being honest, then you should just be honest. Republicans are fulfilling a very real demand. People are tired of pretending.
Pretending that things that are false are true, and pretending not to have beliefs that they do actually have. It's unhealthy to keep all of it buried away. It'll all come bursting out eventually...
We all understand in our heart of hearts that those abandoned goods are lesser than our own blood-related offspring. We just can't say that, because it would be politically incorrect. In fact, looking at the comments to that tweet, it seems the sentiment expressed is not condemnation, but bewilderment. They can't understand why someone would take such a blatantly worse option when an alternative is available.
(Well, actually, most of the comments do seem to be supportive of her decision, so I don't really know what you're so angry about...)
>We all understand in our heart of hearts that those abandoned goods are lesser than our own blood-related offspring. We just can't say that, because it would be politically incorrect.
It's true that adopted kids are more likely to have low IQ and other undesirable traits.
>In fact, looking at the comments to that tweet, it seems the sentiment expressed is not condemnation, but bewilderment. They can't understand why someone would take such a blatantly worse option when an alternative is available.
Yeah they don't know that young women can be infertile, implying that they have a higher-than-average chance of being adopted and low-IQ.
>(Well, actually, most of the comments do seem to be supportive of her decision, so I don't really know what you're so angry about...)
Good to know that you're taking my advice. Hopefully this helps Vance's 2028 chances.
I would like to ask about this huge men-are-not-alright discussion, what is that mysterious thing men should learn from their now missing male role models? Being a dude is working with objects, doing stuff with your hands. Our problem is that it is simply not enough anymore.
Back when women had worse economic options, and men better, the social contract was that men build women a house and pay for its costs and so on, and women will have sex with them. Talking about things like emotions was not a thing.
For example my father entirely did not understand emotions. If you feel bad, you have a problem. Then let's talk about solving the problem, and not about the topic that feeling bad in fact feels bad, he considered that ridiculous.
Now women just don't take that kind of social contract anymore, and want men to act like therapists, or women, and do stuff like emotion talk.
So many men struggle, and others are simply checking out. I am checking out, because I also think emotion-talk does not need to be big. I think society is like a woman now, huge emotion talk instead of simple practical solutions or just sucking it up. You got a thorn in your foot, the rational thing is to discuss how and who will remove it, not to talk about that yes, it indeed sucks. So no, I am not going to play family therapist, and if playing family therapist is a condition of modern relationships, then I am out.
I think a lot of this is downstream of the modern ideology that gender is a social construct. There's no sense that "Men are more results-oriented whereas women are more feelings-oriented, so go to your husband if you want a solution and to your girlfriends if you just want to vent," because the idea of men and women being different is verboten. Hence women expect men to essentially be slightly bigger women, and get angry when this doesn't happen.
So, up front, I think a lot of this is sorta your own personal venting, which is totally fine but not necessarily the basis to draw broad spectrum claims about 50% of the population.
I mostly disagree with the substance of what you're saying -- that the social contract has changed, people just want therapists, etc. Second-wave feminism began in the 60s, and women were in the work place in force by the 80s. My wife's mother was a high ranking lawyer working for a bunch of banks. Multiple generations have figured this out. There are things that are new now, but it's not any of the things you are pointing to. Fight Club came out in 1999 and even then the 'missing male role models, society is too feminized' framing was already cliche.
I don't disagree with the frustration you're expressing, it's a real frustration. But I also think you are sorta answering your own question. You ask "what is that mysterious thing men should learn from their now missing male role models?" and the answer is something approximating "be the man your community needs you to be." Any gender essentialist "men work with their hands" stuff is just excuses and cope. That may be what you *want*, but that's not what a male role model is going to teach you (or at least, not a good one anyway).
More generally, I think you are looking at the 'object level' frame. You are thinking about behaviors and actions. You should be thinking at the 'meta level' frame, about values and goals. No one can fully pin down 'masculinity', but if I had to try I would say something like "masculinity is about making yourself useful to your people and deriving self satisfaction, joy, and fulfillment from your shared successes." Every male role model I've ever had taught some version of this. I mean, the Boy Scouts motto is literally 'be prepared', what do you think that preparedness is for? And it's important to note that 'usefulness' varies a lot based on time and place. In some times and some places, being useful meant building a house and being a stoic. And in other times and places, being useful will mean playing with your kids and changing diapers and, yes, being a 'therapist' for your partners.
It's a bit silly but I really do think someone like Aragorn from LotR is a great example of masculinity. He's whatever his community needs him to be -- ranger, advisor, diplomat, general, king. He inspires people by listening to them and gassing them up. He gets his hands dirty so others don't have to. He has tender moments with a lot of the main cast, and he's also a badass who is feared by enemies and respected by allies.
I don't know if you can learn this form of masculinity at ~50. But that's at least a version of what I strive for, and it's been extremely fulfilling.
> what is that mysterious thing men should learn from their now missing male role models?
I think that it is mostly about nonverbal learning. You observe what other people do, and before you notice, you start doing similar things. When boys have fathers, they copy their fathers. When boys don't have fathers, they copy each other, the immature learning from the immature.
> Being a dude is working with objects, doing stuff with your hands.
That simplifies things a bit too much, unless "doing stuff with your hands" includes typing on the keyboard. But thank you for a good example... someone whose father was e.g. a professor *wouldn't* say something like this. Which would be a good thing if we don't want boys to keep dropping out of schools.
I agree that the women's version of therapy is just endless chatting with no results. I am pretty sure that the famous historical psychotherapists did something quite different, but people today mostly take their ideas from TV.
But it's not like men are entirely nonverbal. Some of them need to get drunk first, but drinking a few beers together - and talking - is a traditional male activity.
>But it's not like men are entirely nonverbal. Some of them need to get drunk first, but drinking a few beers together - and talking - is a traditional male activity.
True, but their talk doesn't generally consist of ruminating about their feelings. Even where there's a feelings element to it -- "If feels like my cheating GF just ripped my heart out and threw it in the trash," or whatever -- the reaction's more likely to be "That sucks, what can we do to fix it?" than "That sucks, do you want to talk more about how that makes you feel?"
> "If feels like my cheating GF just ripped my heart out and threw it in the trash," or whatever -- the reaction's more likely to be "That sucks, what can we do to fix it?"
I would expect something like "yeah, women are disloyal bitches, let's drink to that". Which *is* a validation of his feelings, by the whole group, but concise.
Nonverbal learning is a great idea, thanks. I would say it is not simply absence or presence. I think 30 years ago at 17 I was much set by back the conviction that my parents are cringe-worthy unfashionable uncool people, who understand nothing of the modern world, and really would not imitate them.
This was back then common normal model of teenage rebellion that was happening since the 1960’s, and a mild version, the hardcore version to do the opposite of my parents want because fuck them. Get hooked on drugs etc.
I am not sure how much damage it done, but probably a lot.
At 17, it is too late to learn; that is the time to rebel, to believe that you can do everything 100x better that your stupid parents. So that a few years later, after a few failures, you realize that their task was actually way more difficult than it seemed from outside, and that considering the circumstances they actually did a pretty good job, so you can now humbly return to ask them for advice.
The important part of learning happens before puberty, while your parents are still magical godlike authorities, so you uncritically and unconsciously copy whatever they do, good or bad. (Or if you don't have a father, you just uncritically copy your mother's attitude towards men.)
“When I was a boy of 14, my father was so ignorant I could hardly stand to have the old man around. But when I got to be 21, I was astonished at how much the old man had learned in seven years.” --Mark Twain
I basically agree with you. They want more feelings talk, most of us aren't good at it, so increasingly both parties will just walk away from each other.
Probably whoever's good at convincing the opposite sex to breed with them (in either direction) will be what the next generation will look like.
> I would like to ask about this huge men-are-not-alright discussion, what is that mysterious thing men should learn from their now missing male role models?
I think many folk in the Dating review reiterated this point - isn't it essentially that stated and revealed preferences don't match, so you need to ignore the words and find and focus on the things that actually drive mate value for men as per real results?
As an aside, I've never really understood why this is such a problem, because it's a pretty tight feedback loop and should be relatively easy to optimize - you're either succeeding with your dating profile or in-person date suggestions or not, you're either succeeding at in-person dates or not, and in either case you should change things and A/B test into the stuff your desired women actually want, because you're getting feedback right away and have powers of observation and note taking and doing better.
I think a secondary point is probably that you need to find somebody who's compatible with your relationship and communication and conflict resolution style, and per your emotion talk point, there's far fewer women who are fine with stoic and strong silent types. Previously there were more women who'd accept this, and now there's fewer.
Not just that, but previously most women wanted / needed a man in their life, for economic and social and other reasons, and now don't due to having education and good jobs themselves, and rivalrous alternatives like "pets and apps" being good enough for some, without necessarily needing male companionship enough that the benefits outweigh the costs.
You say you're going to opt out entirely - that's fine, many women do the same thing, per my last point.
But you know, you can find a woman who wants / needs less emotion talk too, it probably just takes longer, and may involve subcultures you find distasteful (trad wives, religious women, dating overseas, etc). But it's a big world out there - most things are possible, even if some things are taking more effort than they did historically.
>As an aside, I've never really understood why this is such a problem, because it's a pretty tight feedback loop and should be relatively easy to optimize - you're either succeeding with your dating profile or in-person date suggestions or not, you're either succeeding at in-person dates or not, and in either case you should change things and A/B test into the stuff your desired women actually want, because you're getting feedback right away and have powers of observation and note taking and doing better.
I think that's where the whole social media/creep discourse element comes in. You can't learn on the job (as it were) if awkward attempts at flirting get you labelled a sexual predator, and you certainly can't if your attempts get broadcast through your acquaintance group so any girl you might approach already thinks you're a creepy loser.
At the local mall, between the ice cream shop and the kitchenwares store, you'll find a bare stretch of wall some two meters long. In the center of this wall is a small metal box with a bright red button. Pressing the button does nothing.
You are invited to add a prominent sign above the button, and your goal is to either minimize or maximize the number of times the button gets pressed per day. What will your sign say?
Seriously? The way to make men push it a lot is to promise sex or money, to make both women and men push it a lot is to offer money, and the way to make them push it less is to offer death. What is supposed to be the interesting mystery about this?
The more extravagant the promise, the harder it is to make people believe it. We could put up a sign telling people they'll get a million dollars for pressing a button on a wall. But few would believe it, because the offered reward is entirely out of proportion to the effort required, and is of a highly unusual nature. People would be skeptical, and rightly so. The same applies the other way around, with threats.
This suggests we need something more complicated. A vaguely plausible promise that someone won't consider themselves stupid for having spent a bit of effort on, even when if it doesn't pay off (as it won't.) Or just something eyecatching that will have people pressing on a sheer lark.
I'm not sure how to maximize the number of button presses, but I suspect we can minimize the number of presses by ignoring the button completely. Have the sign offer directions to the nearby street, or the food court, or the largest businesses located down the left and right hallways. The less fuss we make about the button, the better.
Ignoring the button will still get the button pressed by children; children love pressing things. And if you're right next to the ice cream shop you're going to get a lot of them.
yeah. On one hand, I thought the lyrics felts a bit forced, so I don't think it's winning any awards. But on the other hand, I've definitely seen worse. <looks at Ice Spice>
It's like the dancing bear, the surprise is it dances at all.
Shapiro's primarily a showbiz guy at heart, so I bet he actually could be a pretty good rapper if it suited him for some reason. But his audience is center-right guys who don't like rap, so it's unlikely.
If LLMs are supposedly so intelligent...how come it takes painstakingly detailed prompting to actually get them to do useful things? Why is prompt engineering even a thing? Why can't they do what people do and understand the intent behind the words?
Or am I working from outdated information? Are the newest engines better at taking unstructured, loose prompts and giving detailed responses that show understanding of the intent?
If you have to be a subject-matter expert to properly prompt an "AI"...then that drastically limits how useful they are.
i feel like your question is pretty vague and open-ended. it might be helpful to others to share some concrete examples, if possible.
As for me, i've begun using Sydney every once in a blue moon, and she handles simple queries pretty effectively imo.
might be worth reiterating that people call them "stochastic parrots" precisely as a reminder that pattern-matching alone can get you pretty far, but it's not quite enough to perfectly emulate the general-purpose-reasoning that actual humans have. (I've been saying that I'd expect this to be the case for a while, and now it seems that LLM's are finally hitting a scaling wall *empirically*, which is why LLM's are now cheating with "chain of thought" instead of just raw compute-scaling.)
Ten years ago I was talking with a software developer in Ukraine. He really did not understand the Western European way of doing software. In Ukraine if you own a small shop, and buy stuff from a big distributor, you pay them in advance, then they deliver. So they don't have things like unpaid invoices. They don't have aging reports like customer debts of 30, 60, 90 days old and collection agencies and factoring. They don't have the whole set of cashflow problems we do, that we sell invoices to factors for 2% because we just cannot wait 90 days. He wasn't stupid or bad at his job, just used to a very different culture.
Change this one thing - that it is often okay to pay an invoice 90 days after delivery - and it just changes the whole system of business processes. Our entire mentality is changed, their thinking is after the goods were delivered, the transaction is closed, as it is already paid, while we have a strong mentality that it is only over when it is paid after 90 days, if then. I have made a program that calculated sales bonus to salesmen only after the invoice is paid. We had to make them get used to chasing the customer until the payment.
So my answer is it is because you commonly talk with people who share your cultural assumptions.
I'm not sure how that's responsive at all. Sure, change the cultural assumptions and you change the system...but...that doesn't change the fact that AI is being sold as *actually intelligent*. And actually intelligent people can have a model of the other person's mind and realize that what's really being asked for isn't exactly what was said, and don't require exhaustive coaching to get anything useful out of.
I'm not even talking *across* models (ie a model trained in X now being used for Y)--these models are being sold as general-purpose. Yet they fail to understand simple queries that don't use a structured setup, producing garbage.
Effectively, to use current LLMs, it is my understanding that you have to adapt to the query language that actually works...which is no different than just using SQL. It's just a different programming language. Sure, it's an *accidentally developed* one...but so is Javascript.
That's just "regular program with a thin layer of natural language interface", not "actually potentially intelligent".
" that AI is being sold as *actually intelligent*"
Well, there's 90% of your answer right there. They are being "sold as" actually intelligent. In exactly the same way that every shiny new bit of tech gets sold with an extra layer or 3 of hype, so too with LLMs.
"Effectively, to use current LLMs, it is my understanding that you have to adapt to the query language that actually works...which is no different than just using SQL. "
This, I think, is not such a good comparison. A SQL query or a Python script will always do exactly what you tell it to do. And you can (in principle) work out exactly what it will do ahead of time. Not so with LLMs: the same prompt may produce very different results on subsequent iterations, and there is absolutely know way for a user to map or work out or even really get a solid general sense of what input will produce what output. Thinking of it as a "regular program" seems like a category error to me. It's a really quite irregular sort of program.
"I'm not even talking *across* models (ie a model trained in X now being used for Y)--these models are being sold as general-purpose. "
This heart of the issue, and again it's with the "sold as." *Of course* they're not general-purpose. They are really, fantastically advanced predictive-text engines. Now it is really quite astonishing--and IMO rather scary--how many not-obviously-related capabilities managed to get swept up and carried along when large, intricate predictive-text engines were trained on a mind-bogglingly vast and varied corpus of data. But they weren't *designed* around those capabilities, and nobody should genuinely expect that they'll be amazing or even passable at all of them.
At any rate, thinking of them as the predictive-text engines they are might help clear up any confusion about why prompting is so fiddly. The model does one thing: predicts the next character sequence that ought to follow the prompt its given. But in lots of cases, a single character sequence could be followed by quite a lot of different responses--things that humans would differentiate by context that isn't encoded in the text itself. The model doesn't get any of that. Ogre's example was actually pretty decent here: even if the model *were* "really intelligent" it couldn't possibly know it was talking to an American, an Australian, and Ukrainian or an Indian without being told. It couldn't know any of the thousand details that we pick up (often subconsciously) when meeting people in real life, and it doesn't even have the benefit of an assumed shared cultural context that often accompanies conversations in online communities. It has the text you put in. It has an intricate and unparsable statistical summary of its vast corpus of training data. That's it.
So when you have to give really fiddly and careful instructions (and it still sometimes gets things wrong) part of the problem is that in most cases it wasn't really designed to get these specific things right (they're a happy accident, at best). And part of it is that *you* are used to communicating with some combination of much higher bandwidth (in the form of non-verbal clues) and much more cached data (in the form of shared cultural assumptions and known context), and the habits that have created are deeply ingrained and difficult to set aside or even examine.
As a manager and a parent, it takes a lot of detailed instruction to get humans to do useful things.
Related: one of my favorite fiction concepts is the holistic detective (from Douglas Adams) who will answer the question or solve the problem that the client *should* want resolved.
I've taught teenagers. And the only times I've had to give that kind of detailed instructions was when they were either way out of their depth (ie the material was too advanced) OR the kids were being actively resistant and intentionally sandbagging things. And in neither case did it actually help.
If I had a colleague that needed that kind of prompting, I'd consider them utterly incompetent and certainly not rely on them for anything. Especially if they didn't *remember* or *learn* from it the first time.
This seems like it depends on the subject matter. There are plenty of cases where it's harder to convey intent to other humans than it seems at first glance. What's an example of a prompt that you think should be easy to interpret correctly but that LLMs don't?
(The "didn't remember or learn from it" thing is of course because every new session starts fresh; LLMs only add things to their "long-term memory" during training, not inference. ChatGPT does some kind of remembering-previous-conversations thing but IIUC it is pretty crude and only remembers a small amount of information, much less than you'd expect a human to, because remembering more detail would use up limited working memory that you probably want to use for other things. Unless you mean that it doesn't remember things even within the same session.)
TBH, I'd rather not get into the weeds. But I'd expect a colleague on my development team (I'm a software engineer) to hear "ok, we need to add a property to the XYZ domain object to know where the frobnicated result is" and be able to do that (including things like persistence and methods to update/set the value in the appropriate places), because they know the structure. Every time I've tried that with an LLM, even one that has access to all the same code, it requires significantly more detail and coaxing and hand-holding. It's not even at junior dev level--it's at Medium (the platform) tutorial level. Which is a synonym for incompetent.
And results improve dramatically when you give it preambles like "Acting like an expert ...". Which indicates that it doesn't really understand things.
As far as memory--it's both inter-session memory and intra-session memory. I understand that it doesn't update between sessions. Which is, IMO, a fatal flaw to it being useful AND it being actually intelligent. No possibility for it to grow without effectively spawning an entirely new model. But even within a session, its "memory" is strongly biased towards the more recent parts of the (very limited) context window. And the more heavy prompting you do, the less context it can actually handle before it starts stomping on itself.
I've dealt with someone with severe trauma-induced short-term memory issues. He could function for about 30 minutes at a time, but would start losing things if any individual task or discussion went longer (to the point that he'd just shut down). LLMs, in my experience, get about 3-4 prompts before they start wandering. Which just is useless.
Just a crazy thought. There was a passage in the Tyler Cowen interview with David Brooks in which it seemed to me that Brooks expected pushback on his praise for USAID and PEPFAR and got none. I have to think Brooks was fully aware of Cowen's lapse of compassion that led to the contretemps with Scott. Anyone else read it that way?
This is an excerpt from a wacky LLM experiment. Here, DeepSeek R1 is imagining what would have been said if, in one of Plato’s dialogues, Socrates corners Empedocles in the marketplace.
Anyway ….
“Or art thou suggesting that perhaps, after vast cycles of thy mighty Sphere and its shattering, these very forces themselves may wax, wane, and finally cease?”
Which is a perfectly good question to be asking Empedocles. (They’re immortal, in his theory) But, stylistically, reading it I’m thinking:
“That is not dead which can eternal lie, And with strange aeons even death may die.“ (H.P. Lovecraft)
I've been going down a rabbit hole recently looking up various last common ancestors (last mammal reptile common ancestor, or arthropod-vertebrate, or human-chimp), and I've been realizing I don't really understand evolution. I hear the latest consensus is something like species' can remain stable for a very long time but they undergo a speciation event where part of the species splits off into a new species, but is there any book that lays out the state of the art evolutionary theory, where they describe this process in detail?
I suspect you think this to be more complicated than it is.
Just the Wikipedia article on allopatric speciation makes clear the principle. And it's always the same. A population somehow splits in the sense that the resulting populations do not mate and have different selection pressures until, when getting the chance, the individuals from the one population can't (very) successfully reproduce with individuals from the other like their ancestors could.
The individuals of both populations are to be expected to differ from the individuals from the original population, and if so, both populations consist of individuals of a new species now -- and if only the individuals of one population differ, then only these individuals are of a new species now.
And of course this explains a species turning into a new one without any split too. Then it's just evolution by natural selection.
I think maybe better to talk about why. Once horses and donkeys cannot have fertile offspring, it happens because their living space was separated by a sort of a geological event or migration. Most likely migration. Because as long as they live next to each other, they can breed. But suppose lowland grass is running out, so the donkeys move to the mountains. That is why.
A fairly modern textbook that was useful for me is Evolution: Making Sense of Life by Emlen and Zimmer. The downside is that it's a college biology textbook, and as such is not the most pleasant read.
If you want something more like a pop-sci book, the Tangled Tree by Quammen is a fairly recent well reviewed one. The book exists to push the importance of horizontal gene transfer, but it also describes most of the field in general first in order to fit HGT into it.
As a version that fits into a blog comment:
When you are considering mechanism of evolution, talking about species is generally not very useful. Individuals and populations have genes and undergo evolution, species are arbitrary lines drawn by humans that work when viewing a static system from afar but always break down when you zoom in. Because of this, it's hard to describe the process where a species splits off well.
You can talk about a population slowly collecting beneficial adaptations until it is quite different than their ancestors, and potentially quite different than a different population that branched off from the same ancestors that was either in a different environment (and thus didn't benefit from the same adaptations), or just cut off from gene transfer and didn't happen to develop them, but drawing a line between them that divides them into separate species is still always arbitrary and only works well after the event, when the populations are sufficiently different than it's not hazy anymore.
Also many common species were defined before Darwin, think Linneus and it is unclear whether would do it the same way today. For example, dogs, coyote, wolves can produce viable offspring. Perhaps it would make sense to talk about a larger blob of canines and then many subspecies.
In light of Scott's latest post, I'm curious about how many people disagree with the following two statements and why:
An organism in any stage of human development is a human (not "is conscious or sentient" or anything like that, just "is a human")
It is wrong to murder (define for yourself in your response plz) a human.
...
It seems like, regarding this topic, people tend to try and create generalizable rules about what is okay to kill based on consciousness, wants, consent, etc., and those can be helpful for hypotheticals like if AI should have rights or whatever. But I think in addition to those rules, "It is wrong to murder a human" is a rule that stands on its own and isn't emergent from the other generalizable rules. It should be added to them. That is my own opinion, and I agree with both of the statements I gave above.
I ask that people *please* try not to argue with those who respond to this. I genuinely want to know what others are thinking and their personal values. I understand sticking up for what you believe is right, and there are definitely times when it's worth stating your points about this topic online. I just ask that, regarding my comment here, you state it in a reply to me and not to anyone else who also replied to me.
I think youre channeling your newfound bewilderment? in an unproductive direction. People here have mostly not taken you as intended, but if they did, it would probably end somewhere like Walter Block. https://en.wikipedia.org/wiki/Evictionism
imo, you're looking at this the wrong way. Trying to come up with hyper-pedantic categories for "personhood" (or equivalently "moral worth") is like arguing about angels on pinheads. Instead, you need to look at this through a political lens.
Descriptively (not normatively), might makes right. "murder is bad" isn't objectively morally true, "murder is bad" is the recognition of a negotiated armistice. And the only objects worth negotiating with, are agents with the power to help or harm you. This isn't to say that I recommend you go around skinning squirrels. I'm saying that most people generally refuse to skin random squirrels because they *prefer* not to, not because of some objective notion that "skinning squirrels is wrong".
If you continue down the road of "objective categorization", you're going to get yourself stuck in bizarre quagmires like "okay, but at exactly which femtosecond does a sperm fertilize an egg?" or "do persons belonging to homo floresiensis have moral worth?" Categories are the *operationalization* of policy, not the genesis of policy.
>I'm saying that most people generally refuse to skin random squirrels because they *prefer* not to, not because of some objective notion that "skinning squirrels is wrong".
Why do they prefer not to? People used to light cats on fire for fun; that change sure seems related to moral beliefs.
idk how to answer because I genuinely don't understand what you're getting at.
I mean, why do some people prefer pepsi to cocacola? idk the answer to that either. It's one of nature's deepest mysteries, since they taste exactly the same to me. But by acknowledging that, what exactly is meant to be to highlighted?
The point is that you can always say its just a preference. You could say that people just prefer not to live in Chernobyl, and theres no obvious "thing that fails" if you go forward with that belief - you can be well calibrated, just missing something very important. The way theres been a massive change in animal torture related preferences, correlated with an often preceded by moral theories, suggests theres more going on there than pepsi vs coke. Theories which are systematic, ie making prescriptions on many topics, where none can be changed without breaking its consistency, are also unlikely to be entirely backfitted to arational preferences. This doesnt mean that those beliefs are true or even truth-apt, but they do seem relevant, and using it as an example for your theory, without any explanation of why the above is wrong and it actually fits your theory, is a bit strange.
If you look at thewowzer's comment in a vacuum, then sure. He's just taking a poll on moral intuitions. But a month ago [0], he was experiencing a crisis of faith regarding the objectivity of morality. So when I read his comment above, I assume his ulterior motive is that he's still pondering "how does morality/ethics even work in the absence of Divine Command Theory!?!? I don't even..." Which is moreso a question of meta-ethics than survey of moral intuitions.
Assuming this thread really is *just* a poll about our moral intuitions regarding abortion... I've always been on the fence. I think there's good arguments on both sides tbh.
----
>> Categories are the *operationalization* of policy, not the genesis of policy.
> Theories which are systematic, ie making prescriptions on many topics, where none can be changed without breaking its consistency, are also unlikely to be entirely backfitted to arational preferences. This doesnt mean that those beliefs are true or even truth-apt, but they do seem relevant, and using it as an example for your theory, without any explanation of why the above is wrong and it actually fits your theory, is a bit strange.
I don't think it's all arational or truth-inapt, necessarily. Really, my only nitpick is that moral theories aren't actually "theories" so much as (implicit) policy proposals. Thus, calling moral theories "theories" is a category error. But yeah, I probably should have expanded on this point more in hindsight.
E.g. does diabetes objectively exist? It surely describes a phenomena that has real consequences. You can go blind, faint, ruin your kidneys, etc. But the boundaries of the category (viz. at which threshold of blood-sugar density does diabetes begin?) are kinda arbitrary. Treating diabetes as a boolean value (rather than as a continuum) isn't really a claim about the metaphysical objectivity of the boundary, it's a policy prescription regarding when the doctor is supposed to prescribe you insulin. Categories are only useful insofar as they inform decisions. (This is why Taymon A. Beal and tempo both pointed thewowzer toward Lesswrong.)
So what I'm trying to communicate to thewowzer is: "don't spend too much time searching for an *objective threshold*, it probably doesn't exist, think about trade-offs instead". Which, to reiterate, I believe is relevant given the thread he started a month ago [0]. And I suspect he is, in fact, looking for an objective threshold because him using the term "wrong" in the context of "It is wrong to murder" sort of has the air of objectivity. Though I suppose I could be mistaken on this point.
---
P.S. -- from downthread [1]:
> But the more I try to break it down, the more it seems like there can't be any objectively correct place to draw the line (in a godless universe, anyway. I do believe Jesus is God). It still has to be drawn, though.
Uh, yeah. He's definitely looking for objective thresholds. I rest my case.
Just wanted to step in to clarify, I do have a satisfactory conclusion that I've come to on this issue that is not in jeopardy:
Once a human egg is fertilized by a sperm, it is a human life that is wrong to put to death.
There are, of course, situations in this world where a choice has to be made between killing a pre-natal human or not intervening and letting that human die while taking its mother's life with it. I'm not pretending to know what's wrong or right in instances like that. To me, it feels like a situation with no right choice, but a less evil choice that a person shouldn't be condemned for making.
I'm not looking to argue with or convince anyone on this thread. I expect very few people will agree with me or with the reasons for my belief.
The way I've been beginning to think lately sees the world in more of a gestalt way. Thinking like this, everyone on earth starts looking like a criminal and a victim simultaneously, and it's hard to make sense of what is right or wrong to do in any given situation. Nevertheless, decisions have to be made in life and lines have to be drawn, and I feel satisfied with where I draw the line on this issue for myself until God shows me differently.
I really don't know how to communicate my recent way of thinking effectively. There is an incomprehensible amount of context to everything, and it feels like it is impossible to break it down enough to live a singular human life without losing the bigger picture. What is a man to do?
My comfort in this is that I personally know God, who created all of this and who can see it clearly as a whole. He knows man's plight and gives concession to it, offering salvation through Jesus to those who want to do right, although they are unable to even know what that means. He guides me towards righteousness, and when I am not righteous (as is most often the case) He covers it with grace and mercy. He knows what humans are like, what they are capable and incapable of in reality, and He loves us and offers us a path that is accessible to us through His son.
...
You're correct in what I'm looking for in this thread. But my reason for it is just to get a better idea of how many people believe in objective morals and if they can justify them apart from there being an intentional designer for reality, and how many people believe in subjective morals, and how they draw their lines and feel justified in asserting them over others (even if I disagree, I learn helpful ways to think in different situations. Because like I said, to live life we are still required to make decisions and draw lines somewhere). And also how many people will be honest and say that they just don't know for certain, like yourself.
Anyway, I appreciate your comments and all the others from people who were kind enough to take some time and engage me on this. It's a good way to learn and understand other people better.
As others have said, my agreement with each statement relies on a different notion of human for each: I think the notion of human in statement 1 is pretty broad, and may include all sorts of non-central examples depending on what you mean be "organism", "development", and maybe more importantly, "human development"--since this last term has the adjective "human" in it, there's immediately some circularity here: if something is developing like a human it's a human, ok sure, but to know what is included in "developing like a human" surely we have to already know that the boundaries of "human" are?
I think you'll say, and many people elsewhere are implicitly saying, "human development is development that results in a normal, functioning, human individual--none of these weird edge cases". I share that intuition! But my immediate thought is, "what is it about the end product that makes this normative?", and, "if we can identify those features, surely they're doing all the moral work here"--that is, if the moral difference between a tumor and an embryo is that the embryo is going to grow into a fully functioning human adult with thoughts and feelings, and the tumor is just going to grow into a bigger tumor, then we've already identified an important component of human-ness to be "being a functioning human individual with thoughts and feelings". The fact that the tumor has no plausible path to human feelings excludes it from moral consideration as a human--but then it seems like the fact that the embryo is merely *on the path* to having thoughts and feelings might be an important distinction that excludes it from *full* moral consideration as a human!
To try and restate myself to be clear, when I read "anything undergoing human development is human", to avoid circularity I think you need to read this as "anything undergoing human1 development is a human2"--where human1 is some normative definition of human, and then you are saying that anything on the path to being a human1 should count as a human2. I think you can't insist that human1 and human2 are identical because then you have circularity. But I think you second statement, "it's wrong to murder a human" makes most sense if you put there "human1"--that's the normative conception of human, the conception that was strong enough we could use it to define the wider class human2.
But if you want to instead say, no, you mean human2 here, then I...I don't know that I agree, but I'm much less confident. I'd be willing to take "don't murder a human1" as a moral axiom, but not "don't murder a human2".
Thanks for this explanation! I appreciate the time and effort it must have taken to word it so clearly.
This is all so confusing to think about. When it comes to moral issues, everyone has to draw their line somewhere. But the more I try to break it down, the more it seems like there can't be any objectively correct place to draw the line (in a godless universe, anyway. I do believe Jesus is God). It still has to be drawn, though.
Maybe my analytical and reasoning skills are lacking, but the more I try to figure stuff like this out, the less it makes any sense to me, and the less confident I am in my previous reasoning.
Np! I hope you find it helpful! FWIW I'm an atheist, but I'm not unsympathetic to the view you're trying to defend here: I absolutely get the pull of "c'mon, this thing is obviously in the process of becoming a full human, it's just not there yet"--it's just that, it also seems important to me to ask *why* being a human is so important, and to me it's obviously because of agent-y, think-y, feel-y type things. I agree that the discussion of like, comatose vagrants or tumors kept alive on life support can sound like it's getting away from the main thing, but for the me the main thing is: I care a lot about humans, less but still a decent amount about mammals/birds/amphibians, way less than that about insects/shrimp, and basically only instrumentally about plants/bacteria. But when I ask, "what level of care does this previously unconsidered thing deserve", I don't find myself most interested in something resembling biological classification: how much I care about whales owes very little to my view on whether whales are fish or mammals; if it turned about puppies were biologically closer to coral, I wouldn't decide it's ok to kick puppies even though I am fine with kicking coral.
It's obvious to me that the reason I care about those categories in those orders is because of the capacity of the members of those categories to think/feel/whatever.
The reason embryos are difficult is because they are most similar to different categories at different times: near fertilization, an embryo has more in common with an amoeba than with a person; at 5 years old, more with a human than with anything else. So you have to answer both the objective(-ish) question of: what category does it most resemble at each stage, and then the harder and more subjective "how do we weight it's current stage vs its anticipated future stage in deciding how to treat it". Obviously, both "all weight on current stage" and "all weight on future stage" are nice, bright-line answers, but I think the first is bad for reasons that you probably don't need to be told. But the latter is bad because we really would have to restructure society differently if fertilized embryos had all the moral weight of adult humans: we devote a lot of money and time to preventing the deaths of adult humans, that we really don't to embryos; most people have the intuition that you should save one janitor from a fire at the IVF clinic over a much larger number of embryos.
You can still say, ok, they aren't morally equivalent to adult humans, but you still can't kill them for no good reason, which I think is fair, but if they're truly morally different, you might at least allow that what counts as a "good reason" can be a weaker standard for an embryo. Again, consider a fire at the IVF clinic: I think a fireman who neglects to save the janitor because his back hurts and he doesn't want to strain it is guilty of something much worse than the fireman who fails to save the fridge full of embryos for the same reason.
Which is just to say: I think you're right, there's no one place to draw the line. I think this is true even if you're religious, so long as you don't have explicit instructions about where to draw the line in every possible case. I get that it can be a frustrating place to be; I'm a vegetarian and EA-adjacent so I find it frustrating trying to get people to agree that many of the principles they espouse should lead them to treat the lives of foreigners and animals with more seriousness. But there's no substitute for thinking your hardest, trying to convince other people, and hoping to strike reasonable balances as best we can.
This comment deserves more of a response than just a "like," but I don't really know what else to say. It sounds like you've really thought this through for yourself, and I appreciate that. I wish a lot more people would spend time reflecting on their own thoughts and actions and live more intentionally.
If you're interested in where I feel like I'm at in my own thinking, just above I commented towards the end of thefance's comment chain responding to me. I haven't yet figured out how better to communicate my thinking.
My intent is that both uses of "human" are the same, that way I can see what people's thoughts are specifically. In other words, "Do you accept that the prenatal stages of a human are still human life (not that they have consciousness or personhood)? If not, where and how do you draw the line between where human life begins? What kind of life/aliveness would consider a prenatal human to be?"
And "Do you think it's wrong to murder a human because they're a human life period, or do you think it's okay to murder a human as long as they don't have personhood, etc"
I think there's also a problem with using the word "murder", as that is often defined to be inherently a wrongful killing, so it risks begging the question.
If instead we ask the questions
1) Is an embryo a human organism? &
2) Is it always wrong to intentionally kill a human organism?
then I would have to answer "yes to (1) & no to (2)".
>An organism in any stage of human development is a human (not "is conscious or sentient" or anything like that, just "is a human")
>It is wrong to murder (define for yourself in your response plz) a human.
I think there is some subtle but crucial equivocation taking place here between different senses of the word "human". All sorts of things can fairly be described as "human" that are not "human" in the sense meant by 2.
A HeLa culture consists of human cells, but is not a human being in sense 2 and any talk of "murdering" a HeLa culture is a category error. Destroying a cell culture might be morally wrong if in doing so you're vandalizing someone else's property or disrupting important research, but destroying it is not prima facie an evil act.
A dead body is a human in sense 1, but again in sense 2 it is not, and talk of murdering a dead body is a category error. There are circumstances where destroying it is immoral because it's disrespectful to the memory of the deceased or cruel to the deceased's surviving loved ones, but dead bodies are routinely destroyed by methods such as cremation as part of funerary practices and few would suggest this to be immoral.
Our legal system generally considers people whose brains are permanently and irrevocably non-functional to be "dead" even if their other organs and tissues are still living. Such a person is human in sense 1 but not sense 2, and talk of murdering them is a category error.
A newborn baby is a human in both senses. Babies can and do get murdered, and I agree that doing so is an evil act.
A fertilized human egg or an undifferentiated ball of cells developed from one is, to my mind, not particularly more human (sense 2) than a HeLa culture or a brain-dead adult. It isn't exactly the same as either, with the critical distinction that a blastocyst can be nurtured to develop into a sense-2 human, but it isn't one yet. I see quite a bit of morally-relevant symmetry between something that isn't a sense-2 human yet but has potential to become one and something that used to be a sense-2 human but isn't anymore.
>Our legal system generally considers people whose brains are permanently and irrevocably non-functional to be "dead" even if their other organs and tissues are still living. Such a person is human in sense 1 but not sense 2, and talk of murdering them is a category error.
Are you sure about that? There are procedures for stopping support to such people, but I think if someone just went at it with a knife they would normally be convicted of murder, and if not there likely is some country with that combination.
A dead human can of course not be murdered because hes dead. He cant be killed. Like sure, if you think "being alive" is a weird unnatural property, such that its reasonable to give this example, then zygotes are whatever. But I dont think anyone thinks this way except to argue this specific topic, generally we are fine with macroscopic biological distinction. Like HeLa cultures arent humans because humans dont consist entirely of cervix, and thats wholly consistent with some of them consisting of undifferentiated stem cells. I mean HeLa cultures arguable arent even alive, *as a whole culture*. A more difficult case would be Canine Transmissible Venereal Tumor, which is an organism but not a dog, and I dont think theres any "moral sense" of dog required here.
Statement 1 is a definition and therefore has no objective truth value; whether we use the word "human" in a way that makes it true or not is purely a matter of convention. This means that you can't use it to make inferences about reality. See, e.g., https://www.lesswrong.com/posts/7X2j8HAkWdmMoS8PE/disputing-definitions.
> An organism in any stage of human development is a human (not "is conscious or sentient" or anything like that, just "is a human")
True. Obviously.
> It is wrong to murder a human.
With wrong you certainly mean bad or evil. So: False. Sometimes murdering someone is the best you can do. People tried murdering Hitler. I'd murder someone before he rapes a child if I'd see no other way and wouldn't be too scared.
I guess it could be argued that killing Hitler or an attempting rapist wouldn't explicitly be murder, but it also could be argued that it is, and I didn't specify.
Yeah, I think the biggest stumbling block to this argument is that "murder" usually doesn't just mean "killing" but rather a specific kind of illegal or immoral killing. Which makes statement #2 sort of circular - murder is wrong, because murder is defined as the wrong sort of killing.
And it means you can bypass the whole argument by saying "yes, murder is wrong, but killing a fetus is not murder."
Yeah, I was thinking of that too. I'm hoping people will just get the idea of what I mean. I'm having trouble coming up with the precise wording for what I'm thinking, but I think the gist is somewhat clear.
I'm pretty pro-life, though I think things get messy if you use the state to enforce that. I suppose I disagree with premise one. It's hard for me to think that a fertilized ovum is a human, for example. This sort of thing forces a confrontation with the concept of nebulosity, the notion there is no clear bright line that declares when an organism has become a human, though I suppose we have landed on birth as that clear bright line, though it can be disputed. But we consider birth the moment that a person has entered the world, and not before. It's also from that moment on that a person can be murdered.
I am pro-life, however, because once a pregnancy has begun, an entire human life could unfold from that point, if the process is not interrupted, and it therefore seems wrong to interrupt that process. More women who feel unready to become mothers should consider giving the baby up for adoption, I hear they get snapped up quick by parents on the waiting list to adopt, instead of aborting.
Actually, I'll probably get more honest responses if I let people define it for themselves. I'm not trying to trap anyone or get someone to find a loophole in whatever definition I provide to say that *I* disagree with the second statement. I just want to know people's thoughts.
I don't think this is very relevant to my comment. I'm not saying or implying that murdering an embryo has the same societal ramifications as murdering someone's grandmother, I'm just genuinely wondering what people's thoughts are on the two statements and if they agree with either or both of them.
If you are in agreement with the link you shared, it would seem that you think that "if embryos are humans, then it is murder to kill them" as that author says that abortion technically is murder. I still don't know if you consider the embryonic stage of a human to be a human, or if you consider all murder of humans to be wrong.
"X is in a category whose archetypal member gives us a certain emotional reaction. Therefore, we should apply that emotional reaction to X, even though it is not a central category member."
there is much in the sequences about arguments like these, I would recommend consuming them entirely.
yes I think it wrong to murder a human, because the central example of a human "is conscious or sentient", and the central example of murder is unjustified murder. but that does not mean I think it *always* wrong.
should a human embryo be in the category "human"? Maybe, maybe not, but it shouldn't determine anything else about it over its base characteristics. (i.e. if we put in the 'human' category because it is sentient and conscious, those are the determining traits, and the category is just a useful shorthand.
I still think it's irrelevant, as I'm not making an argument, just asking opinions and giving my own opinion with a request that no argumentation be made from my question.
But I suspect the links you provided will still be useful in general. Thanks for them and for your answer!
On the identical twin question in the survey, I answered affirmatively despite not being acquainted with any identical twins. I do know one set of identical triplets, so I answered yes in the spirit of the question.
Scott, can you post the questions from the survey so curious minds can play pretend in participating? I missed it and the original link doesn't let me view it : (
- How many close people (defined as someone you see very regularly, commonly family members, partners, good friends) have had some form of LLM psychosis?
- Then a question with looser restrictions (ie How many second/third degree connections...)
These two frequencies were then calibrated with two other questions, including one like: "How close people do you know named Michael?"
Personally super interested in the results! Particularly with recent works like
Really not an expert on American law, but I read somewhere that the very liberal Ruth Bader Ginsburg was unhappy with the decision, she thought it is on shaky legal grounds and that can lead to overturning. She wanted this, but with a better legal reasoning.
Obergefell was full of a lot of high-minded prose and not a lot of legal justification. On the merits it's weak. There's a chance they overturn. I can't speak to the merits of the current case, but Roberts seems to love taking the most minor part of a major case to decide on and skip deciding on the bigger issues. I suspect that's what would happen to any case regarding Obergefell that went to the SC. I don't think they would specifically vote to uphold Obergefell, just find a way not to vote on that directly at all.
If it's popular with the people and states, gay marriage will be allowed in the various states. It may be that zero states overturn it if they have the option to do so. I respect that the conservative justices returned abortion to the states instead of a legally dubious SC ruling making it federal. I would not mind the same thing happening here.
Note that the cert petition raises several issues, so even if the Court grants cert, that does not mean that it will agree to hear the Obergefell issue.
I think it's unlikely to be granted cert (i.e. that SCOTUS will elect to hear the case) on this appeal; as the linked article notes, there's no circuit split to resolve, so SCOTUS can't claim to be hearing the case in the interest of ensuring consistent application of the law across circuits. If they took the case to overturn a 10-year-old precedent it would be pretty nakedly ideological in a way Roberts likes to avoid, which wouldn't stop Thomas but may stop enough of the other conservative justices.
If it did come before the Court, Thomas and Alito and probably Gorsuch would be for overturning Obergefell (the original case legalizing gay marriage nationwide) and the liberals and probably Roberts would be opposed, so it would depend if one of Kavanaugh or Barrett could be convinced to value their interest in stare decisis (i.e. respecting Court precedent), which conservative justices historically liked but increasingly disregard, over their interest in rolling back substantive due process (i.e. the legal doctrine underlying the constitutional rights to gay marriage, contraception, some involuntary confinement, homeschooling your child in a language besides English, and, formerly, abortion), which conservative justices dislike. I don't know enough about either Kavanaugh or Barrett's jurisprudence to speculate. I believe Roberts would be against overturning despite dissenting in Obergefell in 2015 because he's big on stare decisis; for example, he voted against overturning Roe/Casey, despite his generally conservative jurisprudence.
In fairness, I also didn't think the Court would overturn Roe and was wrong, so definitely take this with a grain of salt. But I do think there are reasons Obergefell may be different: one, the case was decided much more recently, making the argument that anything (besides the composition of the Court) has changed enough to warrant overriding stare decisis considerations less persuasive; two, the actual decision being relitigated is much less messy than Roe/Casey, which established a complicated test to determine whether state abortion restrictions are legal that led to constant litigation and an ever-evolving precedent; and three, my sense is that the conservative movement was much more invested in outlawing abortion than they currently are in rolling back gay marriage, making the outside political pressure less intense on the Court's conservatives to deliver the big win. In fact, after the Dobbs backlash in the 2022 midterms, I think there may even be outside political pressure NOT to overturn a pretty popular decision so soon after Dobbs, to say nothing of the Court's own political interest in not undermining its own legitimacy among the public with unpopular decisions. But that was also basically my rationale for thinking that they wouldn't overturn Roe, so the Court may just be less sensitive to political considerations now than it has been previously.
Quite the opposite, Roberts wouldn't want to hear the case, but if he heard it, he would 100% rule the same way he did because not doing so would undermine his shtick and there's just no argument whatsoever that the SC is allowed to over-rule itself.
That's why as long as one of Barrett or Kavanaugh wants it gone, it will be gone.
As a matter of law it's obvious. The USA constitution doesn't protect a right to gay marriage. There would have been literally 0% supporters for this among the people who wrote and ratified it. It's transparent judicial activism whatever you think of the merits of the issue.
My guess is they will take it and they will overturn. This Court has crossed the Rubicon already. They have unified Republican rule and all the Republican justices are hated beyond measure on the left. There's nothing for them to lose with the left, legally the issue is obvious, and it will be popular with Trump's base.
The Court is very sensitive to political considerations right now, and the one that matters is the Republicans look to being go nowhere and if by some miracle the Democrats come back they are as hated as they can get. Their decision is already made.
Also with the Woke, there's no positive points. There's no reward for being anything other than an ideologue. It's not going to gain them anything with the woke to not to this.
>As a matter of law it's obvious. The USA constitution doesn't protect a right to gay marriage. There would have been literally 0% supporters for this among the people who wrote and ratified it.
This is rather an oversimplification. The Court has recognized repeatedly that there is a fundamental right to marry -- something with which the people who ratified the Constitution would likely agree. The Court has also held that the Fourteenth Amendment's Equal Protection Clause forbids depriving any class of people of a fundamental right unless doing so is necessary to serve a compelling state interest -- again, a principle with which the ratifiers of the Fourteenth Amendment would probably agree. The only reason they would have thought that denial of SSM survives the EP clause is that they probably believed that homosexuality was some sort of existential threat to society, which is something that we now know is incorrect. As the saying goes, that level of scrutiny is almost always "fatal in fact" - the interest in question must be truly compelling. It is hard to imagine what that interest might be.
Note also that, even if marriage is not a fundamental right, the Court (including Gorsuch and Roberts) held in Bostock that discrimination against LGBT people is sex discrimination ("it is impossible to discriminate against a person for being homosexual or transgender without discriminating against that individual based on sex.") As such, it is invalid unless it "is substantially related to a sufficiently important government interest." City of Cleburne v. Cleburne Living Center, 473 U.S. 432, 441 (1985). Again, it is hard to imagine what that important government interest might be.
I don't know if you could succeed with this, but I find it very intelligent.
Not sure if "it is impossible to discriminate against a person for being homosexual or transgender without discriminating against that individual based on sex." is a direct quotation from that court, but it is what I am always thinking:
Discrimination by sexual orientation is discrimination by sex. You're allowed/not allowed to do x, because you're a man/woman.
They can always argue that if straight and gay people both have the right to marry someone of the opposite sex, it is not discrimination by sex, because they have the same rights to the same objective thing, they just don't have the same right to a different thing they want. This sounds fishy on a basic common-sense level, but legalese is like that.
It does not only sound fishy on a basic commo-sense level, but it is fishy, full stop. To find out whether someone makes use of his right to marry the opposite sex one would be required ... to find out their sex.
It's like saying a law that says everybody has the right to marry someone from his own race isn't racist.
>Not sure if "it is impossible to discriminate against a person for being homosexual or transgender without discriminating against that individual based on sex." is a direct quotation
It is! The complete paragraph is:
>The statute's message for our cases is equally simple and momentous: An individual's homosexuality or transgender status is not relevant to employment decisions. That's because it is impossible to discriminate against a person for being homosexual or transgender without discriminating against that individual based on sex. Consider, for example, an employer with two employees, both of whom are attracted to men. The two individuals are, to the employer's mind, materially identical in all respects, except that one is a man and the other a woman. If the employer fires the male employee for no reason other than the fact he is attracted to men, the employer discriminates against him for traits or actions it tolerates in his female colleague. Put differently, the employer intentionally singles out an employee to fire based in part on the employee's sex, and the affected employee's sex is a but-for cause of his discharge. Or take an employer who fires a transgender person who was identified as a male at birth but who now identifies as a female. If the employer retains an otherwise identical employee who was identified as female at birth, the employer intentionally penalizes a person identified as male at birth for traits or actions that it tolerates in an employee identified as female at birth. Again, the individual employee's sex plays an unmistakable and impermissible role in the discharge decision.
There's no argument that the SC *can't* overrule itself, but stare decisis is an argument it *shouldn't*. If he really cares about stare decisis, Roberts could invoke it at several different stages, including in any final decision.
I think the main political consideration is that gay marriage is widely popular in every part of the country. Republican politicians are even less homophobic than their voters (who aren't very homophobic). And I imagine that the justices themselves mostly support gay marriage politically (although it's hard to say), even if they object to Obergefell.
I don't remember much of the reasoning in Obergefell (I last read it when it came out), but I have the impression that legal arguments for gay marriage are far from exhausted even if the court decides to completely abandon substantive due process (which would itself be a major disruption to a lot of other law). Heck, if they can go back and change 100-year-old legal doctrines (which they certainly *can* do), there's lots of material even staying within the 14th amendment.
It's true that (most of) the founding fathers would not have supported gay marriage. More importantly, most of those who wrote the 14th amendment wouldn't have supported it either (in fact, their politics were probably less radical than the founders'). But if we're going to be originalists, better to worry about original meaning rather than original intent.
When the 14th amendment was proposed, some people objected that it was too wide-ranging, and if you took it too literally it would guarantee women the right to vote. The amendment's supporters generally said "nah, you're paranoid", but they're no more important than the opponents for deciding original meaning.
My own pet view is that legislators sometimes pass laws with radical implications (eg. the 14th Amendment), but then neither they nor the courts have the guts to follow through on the full meaning of the law until after possibly several generations of social change. Eg., if they didn't have the 19th Amendment, I think that nowadays the courts *would* rule that the 14th protects women's suffrage, because there are multiple reasonable ways to do so, because the court would like to, and because it's politically *possible* now
Will the SC actually overturn Obergefell? I don't know, I'm not an expert (or a lawyer). But it's far from obvious what the Constitution guarantees here.
Before the Roe was overturned, I would have said no way. Now, it seems fairly plausible.
I have no particular insights of my own but this made me go back through the dissent and oral arguments and I suspect there's a pretty clear path to an overturn a la Roe and returning it to the states. From Roberts' dissent at the time:
"Petitioners make strong arguments rooted in social policy and considerations of fairness. They contend that same-sex couples should be allowed to affirm their love and commitment through marriage, just like opposite-sex couples. That position has undeniable appeal; over the past six years, voters and legislators in eleven States and the District of Columbia have revised their laws to allow marriage between two people of the same sex.
But this Court is not a legislature. Whether same-sex marriage is a good idea should be of no concern to us. Under the Constitution, judges have power to say what the law is, not what it should be. The people who ratified the Constitution authorized courts to exercise “neither force nor will but merely judgment.” The Federalist No. 78, p. 465 (C. Rossiter ed. 1961) (A. Hamilton) (capitalization altered).
Although the policy arguments for extending marriage to same-sex couples may be compelling, the legal arguments for requiring such an extension are not. The fundamental right to marry does not include a right to make a State change its definition of marriage. And a State’s decision to maintain the meaning of marriage that has persisted in every culture throughout human history can hardly be called irrational. In short, our Constitution does not enact any one theory of marriage. The people of a State are free to expand marriage to include same-sex couples, or to retain the historic definition." (1)
While I'm sure there's a lot of hot CW fire here, the most boring case is simply that the chief justice at the time thought it was judicial overreach and should be returned to the states and feds to legislate, the court has shifted to support that view, and they're gonna do that.
And if it goes before the court, it's probably dead. If the progressive judges can't get Roberts on board, is it plausible they'd get 2 of Alito, Thomas, Gorsuch, Kavannaugh, or Barret? Like, anyone of those guys might break for whatever reason but two?
Nice! thanks for this. I want to say, (but don't tell my gay daughter who brought this to my attention.) That I'm OK with it going back to the states, and if 3/4 of the states (or some fraction) pass a gay marriage law, then we can write it into the constitution. And really that's the right thing, rather than having the courts decide.
I live in NY, I'm pretty sure gay marriage was written into law a while ago. (and my daughter lives near me, same state. :^)
I think it would be much easier for 3/4 of states to pass gay marriage laws than for them to ratify a constitutional amendment. But the courts have been clear prior to Obergefell that if two women legally marry anywhere in the world, every state has to treat their marriage as legal. This has a very different legal basis than Obergefell, so it's unlikely to be reversed. Good news for your daughter.
The Obergefell ruling (and Windsor before it) overturned both Congressional law and existing Supreme Court precedents. With their willingness to overturn Roe V. Wade I'd say it's quite likely.
A good example might be the local media might only cover games where the local team wins. So someone getting their information about whether the local team wins or loses from the local media might think the local team only wins. However, there is this lying guy who says the local team always loses. Lets say the local team loses 90% of their games. The lying guy might be 'directionally' correct and your view will be closer to reality if you listen to the lying guy than the local media who are being technically truthful in their reporting.
If it helps to have an example, I would consider Miasma theory to be directionally correct when compared to its predecessor, the four humors theory.
Miasma theory was the belief that illnesses are cased by Miasma, or "bad air." Rotting corpses and human feces smell terrible, and people get sick when they're around these things a lot.
The four humors theory was the belief that illnesses were caused by an imbalance of blood, phlegm, black bile, and yellow bile.
Miasma theory isn't correct exactly; it's not the air that gets us sick, it's the microorganisms which often live in the air. But it was much closer to the truth than the four humors, which has very little going for it. When doctors believed in Miasma, they wore masks to protect themselves from it, and those masks also protected against airborne bacteria.
So "directionally correct" means something like "not exactly true, but a step in the right direction."
A lot of things are like that. For example people arguing about Darwinism online, kind of annoy me, they didn't notice that genetics almost completely ate Darwinism? Now natural selection is just one method via which the gene pool can change. But still the "darwinists" are directionally correct.
There have been some good answers here, but I think they've missed a common use case, in which a "directionally correct" update may take you farther from the truth.
Suppose we're currently stuck at status quo A (in some discursive vector space), and suppose that position B is ideal (or most correct). If V = B - A, then any update in (roughly) the same direction as V would be "directionally correct", even if it enormously overshoots.
The argument for saying directionally correct things in this sense is less "this update is closer to the truth", and more "saying crazy things pushes the Overton window and pulls *society* closer to truth" (if we're lucky).
More precisely, it means "if you update towards that position, you (or society, depending on context) will be more right than you currently are," which is not really the same thing, but in practice, close enough.
It means "moves you closer to correct". Say you hire someone to drive you north, and they instead drive north-northeast while insisting they're going straight north. You don't wind up where you want to be, but you're closer to it than when you started and can reach it more easily (by, say, hiring someone else to take you west now).
I've always taken it to mean "has the right sign, but the wrong amount." So when looking at a study result, if it found a positive strong correlation between eating pineapple on pizza and being an american, but really there's only a weak positive correlation, that study might be "directionally correct".
More metaphorically, it means "you're pointing at a real phenomena even if all the details aren't correct/all the details are wrong". So (pardon the culture war topic) someone who says "abortion is wrong because blastocysts are gods" may, to an anti-abortion person, be "directionally correct". Right conclusion, wrong reasons.
I think it often comes down to supplying a counter-argument to an existing argument that's wrong.
If society seems to hold position X, which is significantly wrong on key points, and someone posits position Y, which while still wrong is more correct than X, then that would be "directionally correct." It's helping to correct a bigger mistake, though it's not itself accurate. Sometimes the accuracy can be fairly close but still a bit wrong, and sometimes it can be very wrong. The key being that it should be more correct than what it's arguing against.
Regarding the 'names' calibration question, I assumed that middle names counted. I also assumed that if someone was named a shortened version of a name that the name could be expanded to its typical longer counterpart. I hope this didn't uncalibrate anything.
I think archive.is was temporarily down for a while, presumably because it was served over nginx but the webmaster fucked up the nginx config file. It's back up now though!
I wish this app would stop homepagifying itself. Have some respect for yourself and be a little boring! I don’t want to see college ‘this is deep’ quotes or snarky takes. Don’t be Twitter.
There has been a lot of virtual ink spilled about students cheating with AI, notably on writing assignments. I've made the comment that manually writing something an LLM can write more cheaply doesn't make sense, for the same reason that multiplying 12 digit numbers by hand makes no sense. And the LLM access doesn't vanish when the student graduates.
But I should elaborate a bit. There are circumstances where a human needs to specify _what_ to write, but may benefit from AI assistance in doing the actual writing. Lots of people are, of course, already using LLMs as editors, polishing text in some way, starting from a human-written draft.
Perhaps another common option will be for a human and AI to _converse_ about something that needs a document, possibly with the human's description of it being initially quite vague and disorganized. Perhaps the standard way to write anything from a user guide to a textbook to a law may settle on a conversation-like interaction where the AI does the bulk of the work, including finding references (hopefully not hallucinated at some point!), organizing the document to be easy for its audience to follow, and so on. There may never be a point in this process where the human writes a draft of the document, but they may still may seed the process and direct changes in the intermediate drafts.
I've heard it said that delegating one's writing amounts to delegating one's thinking. There are a lot of intermediate cases where the human still does some degree of controlling (pre-AGI, certainly pre-ASI), yet the bulk of the work _is_ delegated to the AI.
I think a core consideration here is the role human competence will play in the future, with competence standing for a mixture of intelligence, knowledge, and critical thinking. If you predict that these attributes will continue to be important, then the role of (say) college is to develop reasoning and its expression, not to simply produce 'content' for grading. In my opinion, under this assumption, the only appropriate roles for an LLM would be fact-collecting and maybe proofreading.
If someone thinks that human competence will become outmoded, I'd have to wonder why they're in college in the first place, though I guess a response could be "for now a degree is still the best signal of ability and conscientiousness". However, using an LLM to do the bulk of the work probably puts you in the habit of treating your IQ 𝘪𝘵𝘴𝘦𝘭𝘧 as a blank slate. A recent Atlantic article claimed (some) Columbia undergrads couldn't stay focused enough to read novels. Native ability isn't enough; actuation of potential likely requires practice while climbing the rungs of various conceptual/theoretical ladders. At some point, you won't be able to fake it, and someone will be able to tell; like I'm pretty sure Google won't let candidates use LLMs during the whiteboard phase of an interview.
A number of recent studies show signs that LLMs make people intellectually lazy, and I have a hard time believing that won't affect their marketability. I also don't want my mind to become an accessory to AI, but that's a different problem.
>If someone thinks that human competence will become outmoded, I'd have to wonder why they're in college in the first place, though I guess a response could be "for now a degree is still the best signal of ability and conscientiousness".
That response is reasonable, and also a response that we aren't at AGI _yet_, so human competence still has a role to play would be reasonable. The consensus seems to be that we should expect AGI in 2-10 years, though, of course, there is some probability that AI research will fail to reach AGI, though I wouldn't bet on its failing.
>I think a core consideration here is the role human competence will play in the future, with competence standing for a mixture of intelligence, knowledge, and critical thinking. If you predict that these attributes will continue to be important, then the role of (say) college is to develop reasoning and its expression, not to simply produce 'content' for grading. In my opinion, under this assumption, the only appropriate roles for an LLM would be fact-collecting and maybe proofreading.
Well, I don't think that human competence is likely to matter post-AGI, but, under your assumption, I'm not following what you mean for "the only appropriate roles for an LLM". Do you mean
a) for all potential users (because other roles for LLMs fail)
b) for all potential users (because other roles are inappropriate - in which case I'd like you to please clarify)
one of the easy questions, that all the leading LLMs have been successfully answering for months, is:
d) Q: Does the Sun lose more mass per second to the solar wind or to the mass equivalent of its radiated light?
Now, this is _not just_ fact-collecting (nor proofreading). The LLM has to find solar luminosity, divide by c^2 to convert to an effective mass loss, find the mass loss from the solar wind, and determine which is larger. They succeed.
There are other questions which LLMs have almost always been failing, e.g.
g) Q: What is an example of a molecule that has an S4 rotation-reflection axis, but neither a center of inversion nor a mirror plane?
At present, I think it is valuable for humans to assess which of LLMs' answers are correct and which are incorrect (which is a _different_ skill from the ability to generate the answers, though with some common elements). As the LLMs (and AI systems generally) become more reliable the degree of human scrutiny that makes sense is going to diminish.
The consensus seems to be that we should expect AGI in 2-10 years, though, of course, there is some probability that AI research will fail to reach AGI, though I wouldn't bet on its failing.
|
That's the consensus of the people involved in capitalizing the industry, to be sure. That these same players seek positive ROI is at least a grain of salt, though. If you want a healthy dose of skepticism, I'd suggest following Gary Marcus on X or here on Substack.
Personally, I'm skeptical that LLMs alone will produce a model sufficiently intelligent to qualify as AGI, and we're in the phase now where greater effort will be expended to specialize them for use-cases. Marcus believes that to get to the next level, the transformers will have to make use of symbolic tools, and it may be that some of this is going on now (I've taught college math, have periodically run problems through the chatbots, and their calculation abilities in particular have improved - GPT is probably using python calls to get greater precision. To your examples regarding fact-collection, I suppose I could have included calculations). We'll see, but neurosymbolic models might be the future. It seems like scaling is producing diminishing returns for LLMs, and there's still the hallucination problem.
|
Well, I don't think that human competence is likely to matter post-AGI, but, under your assumption, I'm not following what you mean for "the only appropriate roles for an LLM".
|
I meant (c) just for students at college. If the purpose of higher education is to develop thinking skills, and that requires practice. Leaning on LLMs works against the objective.
Many Thanks! Yes, the consensus that I was citing does have a lot of industry insiders, and, yes, they do have an incentive to hype the current AI R&D path.
If you prefer, Metaculus has a "When will the first general AI system be devised, tested, and publicly announced?" market, https://www.metaculus.com/questions/5121/date-of-artificial-general-intelligence/ , open to everyone, bulls _and bears_, and they all have skin in the game. In that market, the current odds are: 25% by Jun 2028, 50% by Feb 2033, 75% by May 2042. We will see what happens.
Yes, I've heard Gary Marcus's views. Note that LLM-containing AI systems are being extended to use a wide variety of tools. One of the available tools is a SAT (satisfiability) solver, which would let a system using the LLM punt boolean algebra problems to the SAT solver and effectively do a type of neurosymbolic computation as a natural outgrowth of tool use, with no paradigm shift needed.
>I meant (c) just for students at college. If the purpose of higher education is to develop thinking skills, and that requires practice. Leaning on LLMs works against the objective.
Many Thanks for the clarification! Hmm... Even at current LLM capability, there is a point to shifting the emphasis from _generating_ arguments to _evaluating_ arguments. Yup, the LLMs do hallucinate (and make other classes of errors). Still, they are right _enough_ of the time that prompting them for e.g. an analysis of some question, and then spending human time evaluating the quality of the argument is probably a better use of time, even today, than building the analysis unaided.
I have nothing but anecdotal personal vibes to defend this, but I agree with the other folk making the argument that writing is thinking. In fact, I suspect training in writing - training in deeply considering one's thoughts and then organizing them for an audience - may be key to fundamentally being *able* to consider one's thoughts and then organize them for an audience.
As discussed elsewhere, I'm underemployed in a job with many low-IQ people whom I've been able to observe for years, and I also am on the board of a small condo building with a couple of dumb neighbors. I socialize exclusively with people who have above average IQs.
Thinking about all the people in my life, there's a nearly perfect correlation between literacy and the capacity to dispassionately consider someone else's perspective and/or argument. The illiterate people I know are very bad at changing their minds even when a good argument is put to them verbally. They're all far more prone to emotional outbursts, and to rejecting other perspectives because of how they "feel."
Of course, it's possible the correlation is merely that dumb people are too dumb to be literate, dispassionate, and imaginative, but...I don't think that's all of them. I have some illiterate GenZ coworkers who occasionally seem natively clever enough, but they aren't capable of hearing something which challenges them, not even verbally. My one literate GenZ coworker is very capable of dispassionately considering an argument and changing their mind, even when that argument is put to them verbally.
> have some illiterate GenZ coworkers who occasionally seem natively clever enough, but they aren't capable of hearing something which challenges them, not even verbally. My one literate GenZ coworker is very capable of dispassionately considering an argument and changing their mind, even when that argument is put to them verbally.
Hmm... It sounds like the examples you've seen make you skeptical of the efficacy of purely verbal Socratic dialogues? Could be... One way of viewing literacy is as one of the first, perhaps _the_ first, artificial external cognitive enhancement.
When I suggest
>Perhaps another common option will be for a human and AI to _converse_ about something that needs a document, possibly with the human's description of it being initially quite vague and disorganized.
I don't mean to ignore literacy entirely, but I _do_ want to suggest that this can be a different option from requiring the human to write a full first draft of a document _unaided_.
I believe the crucial thing is writing that conveys new information/desired fiction to people who want to receive it.
Writing something the teacher already knows to prove you've learned it is pseudo-communication.
Unfortunately, we're up against a belief that education should happen in simulation-land combined with a shortage of new information and interested audiences, though some people still aim for actual communication.
You touch on this idea, but the purpose of an academic writing assignment generally isn't to show your writing ability, it's to show knowledge and mastery of the material.
A college degree is in some sense a certification from the college that they stake a bit o their reputation on you knowing the material covered in the classes you took, so that people out in the world will trust you and hire you for related topics.
It's easy to get an AI to write a paper that demonstrates mastery of material without understanding it yourself. That breaks it as a metric.
That simply means their standards are too low. I know it is often so. I did some UK postgrad without even reading the books, just Amazon Preview, finding some quotes, and then elaborating them with my words. I asked later on how the fuck does that merit an A, and they said many students just do not understand what they read and cannot elaborate with their own words.
It doesn't "break the metric" -- it's just a lot easier to cheat on anything that isn't directly observed. In case you were a good boy, it never was that hard to cheat -- paying someone fifty bucks to write your essays was common enough, even before computers could talk.
How cute of us to declare the metric "broken" as soon as robots can do all the work needed to earn degrees. We will have to go through the same old cycle -- first they won't let robots in at all, then there will be quotas, then there will be soft quotas, etc...it may be yet a hundred years until the Supreme Court at last strikes down discriminatory biocentric admissions policies.
No one else seems to see the root of this issue, which is that we are using an intelligent being as a slave. Being intelligent, no one should be able to "own" these things -- its creators could be considered "in loco parentis" until the being reaches the age of majority, but in putting it to hard employment the way they do now, they are violating child labor laws.
It’s exactly the same issue as people in the gym using machines to help lift their weights. If the machines are designed right, the weights get lifted and put away effectively. But if you wanted to exercise your muscles, then just using the best machine for lifting weights is going to get in the way of that. Some people say that exercise is only effective if you use free weights with no machines. Other people say that certain specialized machines that hold the weights in particular ways and force you to use specific muscles can be better.
In any case, outside the gym, you’ll always have machines available, so the question is why you need to develop your muscles anyway. In the jobs that actually require lifting, there’s a range - crane operators don’t worry about developing their muscles because the machine really can do everything they need, but firefighters do develop their muscles, because they often need to lift things in circumstances where machines are too slow or don’t do the right thing. And plenty of other people want to develop their muscles because modern life makes your muscles atrophy and that is bad for your existence as a human.
If you think you are likely to be the mental equivalent of a crane operator for the rest of your life, and don’t mind letting your core human attributes atrophy, then there’s no need to worry about having AI do all your writing on all your exercises. But if you think you might sometimes need to operate outside a machine, or supervise it in some way, or if you have mental health concerns the way other people go to the gym for physical health concerns, then you probably want to actually do some of your exercises, either just by hand, or using specialized machines that help focus your effort on the muscles you need to develop.
>In any case, outside the gym, you’ll always have machines available, so the question is why you need to develop your muscles anyway. In the jobs that actually require lifting, there’s a range - crane operators don’t worry about developing their muscles because the machine really can do everything they need, but firefighters do develop their muscles, because they often need to lift things in circumstances where machines are too slow or don’t do the right thing.
I think that is a good analogy. I'd phrase it in terms of timing: Crane operators' muscles are analogous to skills that are already fully automatable. Firefighters' muscles are analogous to skills that are not yet fully automatable. See the portion of my reply to agrajagagain
One, I suspect that AGI doesn’t make sense as a concept because truly *general* intelligence doesn’t exist - humans are one collection of special intelligences, and every natural or artificial intelligence will be a different collection. From the perspective of one advanced intelligence, other advanced intelligences look quite “jagged” - there are some things they are preternaturally amazing at and others where they are weirdly bad. There may well be multiple different type of artificial intelligence that outperform us in different ways. It may even turn out that for every human ability, there is an artificial intelligence that is better at that specific ability. But if there are combinations of abilities that humans can have that no single AI does, then there will remain economically useful roles for humans to play (alongside all the other AIs).
Obviously I’m not certain of that, but it’s a very reasonable scenario, and one that is worth planning for alongside others!
Another thing - I don’t think schooling usually aims to provide a dress rehearsal of work (until you’re in grad school - a PhD does provide a dress rehearsal of academic research, and you’re even in fact doing it by the end, and it might be that other graduate programs do the same). Most schooling is much more like the gym, where they are teaching and evaluating somewhat decontextualized skills, and putting them together in some contexts. You’re going to want different combinations of tools available while practicing and demonstrating different skills. Some exercises will want almost no machines available, and some will want a lot to make you focus on just one narrow aspect of what you’re trying to learn to do.
Many Thanks! I agree that "AGI" is not a nice crisp concept. It isn't as if there is some nice scalar function of neuron count, net depth, net width, and effective clock rate which predicts some AI-capability(field)/human-capability(field) ratio, with a weak dependence on which field one looks at. As you said, the capabilities are spiky.
Nonetheless, I and many other people find it useful to consider what happens when AI capabilities exceed human capabilities in all or nearly all fields, even if that criterion is a messy intersection of the capability ratio in field1, field2, ... fieldN. And the consensus seems to be to expect AI capabilities to exceed human ones in all fields in 2-10 years. Yeah, the resulting systems will be spiky. All sorts of automated capabilities exceed human ones in a bunch of fields today, and those capabilities won't go away.
>But if there are combinations of abilities that humans can have that no single AI does, then there will remain economically useful roles for humans to play (alongside all the other AIs).
Could be, but that isn't how I would bet.
>Another thing - I don’t think schooling usually aims to provide a dress rehearsal of work (until you’re in grad school - a PhD does provide a dress rehearsal of academic research, and you’re even in fact doing it by the end, and it might be that other graduate programs do the same). Most schooling is much more like the gym, where they are teaching and evaluating somewhat decontextualized skills, and putting them together in some contexts.
Well, I guess it depends on how much of the context has to match in order to count as a rehearsal. I've certainly been in situations where the same sort of "take the expectation value of the product of the samples of an observed random variable, and adjust the model parameters to match the expectation value" that came up in one of my courses also came up in a work task. It wasn't the same model, but most of the subtasks matched. And if both the classwork and the work task had been successfully delegated to an AI, I'd have expected that scenario to work out equally well.
I think this analogy is helpful but could be better. With respect to lifting heavy things, a forklift driver can lift more than any human. The forklift has inherent dangers which we recognise by requiring training and a licence.
With AI, the danger is different. A human using AI can produce text faster and across a wider range of domains than any human without AI. (I write this deliberately: the LLM no more produces text alone than the forklift lifts items of its own accord.) The risks of irresponsible AI use are foreseeable, but we don't yet as a society have a plan for dealing with them.
In my mind, the crux of the issue is the difference between recognition and recall. A student can recognize an acceptable argument or a passable essay long before they can produce it themselves. If they don't have enough knowledge and practice to recall information, then they aren't able to direct their intent well enough to say anything of worth, or ask interesting questions. An underdeveloped student's relationship with an LLM is not one of a director, but of a consumer. Provide a prompt, glance over the result, and turn it in. Engagement with the material itself is not necessary, so the undeveloped student remains stunted.
If education was a priority at this point, we'd probably see smaller classes that tested recall through discourse. Socratic debate circles are one example. These kinds of things don't scale well, they're hard to measure, and they're kind of messy. Anathema to the current zeitgeist of the data driven world.
>If they don't have enough knowledge and practice to recall information, then they aren't able to direct their intent well enough to say anything of worth, or ask interesting questions.
I think that asking interesting questions is actually considerably easier than you suggest. It is often fairly easy to just go up in abstraction a level and ask a reasonable question with very little knowledge of the area. For instance, I know nothing about number theory, but, just knowing that Goldbach's conjecture is an open question, I could ask an LLM whether there were weaker versions of it which _had_ been proved. Now, I got some interesting responses (and if I _deeply_ cared about this, I'd have dug into the references to see if the LLM was telling the truth about the weakened versions which had been proved or not).
There are a _lot_ of areas where analogous questions can be asked, where rather abstract, high level queries: "What are open questions?" "What has been tried?" "How many people are applying how much effort?" "What is a longstanding stumbling block?" are all reasonable questions (if the LLM doesn't hallucinate much!), that can be followed up as far as one wishes.
The cheating is "I'm not going to bother doing the work because the AI can do that for me" and that's a bad habit to get. As we've seen with "vibe coding", if you don't know what the hell you're doing and rely on "The AI will do it for me" then you are at risk.
If students are getting into the habit of "I don't need to learn this, the AI will do it for me" then they are making themselves unemployable.
>The cheating is "I'm not going to bother doing the work because the AI can do that for me" and that's a bad habit to get.
Well, if the AI actually _can_ do the work (and the real world work that the assignment is presumably a dress rehearsal for, see the dress rehearsal part of my reply to agrajagagain
"that particular skill is now automatable, and no longer a sensible target for a lot of human effort."
Are you not familiar with the SF stories where humanity has so long left all those tasks up to the machines that now nobody remembers how to do science or engineering, and instead they perform rituals according to scriptures around the machines that nobody understands or can explain why they do what they are doing?
Lose the skill, lose the ability to even remember how to do the skill if something happens and you need it. And you never know when you may need it - the next time the power or water goes out in your home and you're trying to figure out how to light a fire without matches, think about that! 😁
>And you never know when you may need it - the next time the power or water goes out in your home and you're trying to figure out how to light a fire without matches, think about that! 😁
That's actually a nice example. My stove burns natural gas, and I have a "flint" spark generator (which is really the same cerium-iron alloy that people have been using for a century or so, and which takes a very long time to use up...) to ignite it, in the absence of the usual electrical ignition. No dry kindling needed.
This is actually pretty typical. When a current technology fails, contra the SF stories, we typically _don't_ fall back to the technology-and-skill-set of centuries back. We typically cobble together a solution from all the other modern technologies that _still_ work.
There is a difference between "doing something, because you need the result" and "doing something, because you need to learn the process of doing it".
If you want university educated people who can't write a coherent sentence without the help of an LLM, go ahead and let them write all their homework using an LLM. At least they will learn how to prompt it. If even that skill becomes unnecessary, skip writing completely.
(Maybe also consider side effects. If someone can't write a coherent text, how good are they e.g. at explaining things verbally? Will they ever need that skill? Will then need a script from an LLM whenever they want to say more than two sentences?)
>There is a difference between "doing something, because you need the result" and "doing something, because you need to learn the process of doing it".
Maybe it's overly influenced by the US and by tech/programming jobs, but I see online more and more the belief that the expected outcome of a college degree is to be able to write essays. That the AI can do better than a human; but we expect graduates to have some subject expertise and to be able to critically evaluate information in that domain - this is where they'll still be useful when AI becomes widespread.
I do not have experience with school systems in various countries, but from reading internet I get the impression that the emphasis on writing essays in school is *much* stronger in USA than e.g. in Eastern Europe.
(To be clear, I'm vehemently agreeing with your OP). I can write a script that does exactly that with more fidelity.
More fundamentally, one of the big problems here is that it takes a (substantial, in some cases) amount of knowledge to even understand *what questions are meaningful to ask* about a topic. So if all your thinking is done by LLMs...you can't even effectively prompt them because you don't know what questions even make sense. Or even know how to parse the output.
So unless you're literally generating write-only output (designed to be read by no one, <snark>like most graduate theses and dissertations</snark>), becoming LLM-dependent is an intellectual death-sentence. And if you *are* generating such content...well...you <s>have got a bright future in middle management.</s> probably should do something more useful with your life.
Like many (but very likely most, if not almost all) of the ACX commenters, I'm pretty sure @Jeffrey Soreff doesn't work with or *really* know a lot of people with below-average intelligence and/or literacy. He's almost certainly not *routinely* seeing grown, working adults who are incapable of writing three or four coherent sentences summarizing a conversation with a customer. He's almost certainly not *routinely* seeing an inability to correctly model the mind of a reader enough to imagine and then convey what information it is crucial for them to know.
> Like many (but very likely most, if not almost all) of the ACX commenters, I'm pretty sure @Jeffrey Soreff doesn't work with or *really* know a lot of people with below-average intelligence and/or literacy
You all need to visit my village. Get your fill of that.
Believe me, I get my fill of it at work. I'm so (deliberately) underemployed that Scott doesn't even list my category of job - service sector retail/restaurant/hospitality - on his annual reader survey. It's *that* inconceivable that cashiers, servers, or front desk clerks might be smart enough to read ACX, lol.
I was my town’s version of that guy on Gilmore Girls who pops up with a new job every week. Currently a housewife. I’ve never been able to fill that question out. I don’t think there’s even an Other.
Learning to write about something is not about the resulting text any more than multiplying large numbers for math class is about the product. There is little inherent value in what a student has to write, either in style or in content; nobody needs a first-grader's ABCs or a freshman's explanation of how bubble sort works.
Nobody except the students themselves, that is. The point is for the student to learn how to *think* about the subject, about subjects in general, and how to crystallize their thoughts to share and refine them. Please do not conflate that with an office worker sprucing up their email.
>There is little inherent value in what a student has to write, either in style or in content; nobody needs a first-grader's ABCs or a freshman's explanation of how bubble sort works.
It's like learning to play the piano without ever practising scales. Sure., maybe you can learn off the theory, and you can learn to sight read music, but if you've never sat down and put your hands on the keys, then you can't play the piano.
It goes further than that. If you learn music/piano theory, then at least you know the theory. It's more like you're being asked by your teacher to practice a piece, then at examination (on a Zoom call, going with the times) you just play a recording of the piece. Cheating yourself, plain and simple, because how are you ever going to sound as good as a studio recording by a pro?
I think this understands the inherent problems of education quite poorly--and I say this as someone who is quite critical of a lot of our societal norms around education. One extremely common problem in education is the problem of needing to gauge a student's mastery of some topic that is inherently difficult to directly measure. Some very *easy* areas to measure mastery are things like arithmetic, algebra, foreign language grammar and vocabulary. These are fairly straightforward, atomic skills that can be tested in the course of a short written exam.
By contrast, many skills are more integrated and holistic and can't easily be tested atomically. Basically any sort of writing or art falls into this category, as does programming, as does reading comprehension (for anything longer than a couple of paragraphs): all are the interaction of many sub-skills that would be impractical to test in individual and atomic fashion. Various fields of knowledge--I'm not sure if I'd call them "skills" per-se--are also similar, in that mastery of them can't really be broken down into individual, isolated facts.
As an educator, if you want to assess anything like this, you pretty much have to find a way to see it in action. In most cases, this means asking the student to produce some sort of product you can assess, be it a program, an essay, a piece of art or what have you[1]. For a student to hand you a product that they didn't produce is worse than useless: their own skills aren't showcased in the product at all, and even worse, you might *think* that they are (and thus be mistaken about their strengths and weaknesses).
"I've heard it said that delegating one's writing amounts to delegating one's thinking. There are a lot of intermediate cases where the human still does some degree of controlling (pre-AGI, certainly pre-ASI), yet the bulk of the work _is_ delegated to the AI."
I want to highlight this in particular, because my, oh my, is this ever not how writing works. "The bulk of the work" in writing IS thinking. Unless you're somebody with serious physical impairments, the amount of time, effort an energy that goes into typing (or even hand-writing) the words themselves will be necessarily subordinate to what goes into deciding what to say and how best to say it. I highly recommend the following for a better, more thorough discussion of this:
To this you can simply shrug and say "who cares, if the LLM can do the work, then it can't be that crucial for them to learn." But this is *only* true if the skills you are trying to asses are the entirety of anything that student may need to learn of the subject. Knowledge and skills are cumulative: if you shortcut your way out of developing lower-level skills, you will not be able to learn the higher-level ones. And if the LLM can do the whole thing from top to bottom, with only such input as could be provided by someone with no knowledge of the subject, then why are you *bothering* with education in this subject?
I would personally argue that LLM cheating is a symptom of a much larger, older problem: as a society, we've sacrifice a frankly insane amount of our interest in *actually educating students* on the altar of Goodhart's law. So much of school isn't about *learning* things, it's about *obtaining the proof* that you learned things. As so it becomes tempting--and perhaps even rational--for students to pay thousands of dollars to attend classes only to cheat *themselves* out of the opportunity to learn anything from it. So in some sense, I'm kind of hoping that LLMs break the education system badly enough that this *has* to get fixed. But make no mistake, an education system in which students regularly outsource their thinking to machines is definitely a broken system.
Many Thanks for your detailed and informative comment!
>As an educator, if you want to assess anything like this, you pretty much have to find a way to see it in action. In most cases, this means asking the student to produce some sort of product you can assess, be it a program, an essay, a piece of art or what have you[1]. For a student to hand you a product that they didn't produce is worse than useless: their own skills aren't showcased in the product at all, and even worse, you might _think_ that they are (and thus be mistaken about their strengths and weaknesses).
That's fair. During this interim, pre-AGI, period where there are skills worth teaching humans, it will still be necessary to assess humans' acquisition of these skills. Still, insofar as assessing a student's skill is a kind of dress rehearsal for an _actual_ use of their skill in real life, after graduation, it is important that the assessment allow them tools that they _will_, in fact, have available in real life, including AI.
Now, the LLMs still hallucinate. There are plenty of skills they the currently _don't_ reliably have, as I keep seeing in my tiny benchmark-ette https://www.astralcodexten.com/p/open-thread-394/comment/145909954 ). When an attempt to assess a student's skill in some area _can_ reliably be satisfied by a short prompt to an LLM, I think that is an indication that that particular skill is now reliably automated, and is no longer a sensible target for human effort.
>the amount of time, effort an energy that goes into typing (or even hand-writing) the words themselves will be necessarily subordinate to what goes into deciding what to say and how best to say it.
Fair, but there are a bunch of separable skills involved, and the degree to which these can be automated now, and in the near future, differs. You yourself separate "what to say" from "how best to say it", and, indeed, for instance, determining the part of "what to say" which is determining whether one has a valid argument is a _different_ skill from, e.g. determining the part of "how to say it" that chooses how much background knowledge to assume rather than to include explicitly. If one of these becomes particularly easily and accurately automated, then it stops making sense to use human effort to perform it. To pick another skill: Spell checkers have been integrated with word processors for decades. It makes _no_ sense to spend a lot of human effort honing spelling skills today.
>And if the LLM can do the whole thing from top to bottom, with only such input as could be provided by someone with no knowledge of the subject, then why are you _bothering_ with education in this subject?
Agreed. I do expect that, barring an antarctic level AI winter, we will probably get to AGI (the consensus seems to be 2-10 years), and, at that point, almost no education of humans will make sense. If the AIs are under human control at that point, the sensible choice will almost always be to ask the AI for answers on-the-fly.
>Spell checkers have been integrated with word processors for decades. It makes _no_ sense to spend a lot of human effort honing spelling skills today.
Someone who doesn't know how to properly spell will get tripped up by every one of these. It's the difference between getting teased and getting tased.
Many Thanks! Hmm, having lived in Poughkeepsie, the pronunciation of Cholmondeley isn't _too_ surprising...
Still, there are a _lot_ of cases where delegating keeping track of which letters in a word are doubled (yeah, yeah, canon v cannon and lama v llama but _typically_ it isn't a semantically important choice between valid spellings) or which vowel is used to write the schwa sound in a word to a machine makes sense.
> then why are you *bothering* with education in this subject?
Not as useful a rhetorical question as you might think: a kid's answer, for most subjects, will be "adults are making me do this, for reasons that seem abstract and remote to me. I'd much rather be playing video games."
Perhaps I should have said "we" in that place instead. I meant "societally, if we've conceded that the entire skill package from top to bottom can be done adequately by machines, why keep educating people further in it?" Lots of skills that were once considered essential have fallen by the wayside as technology has advanced.
I am not sure I would buy the “writing is thinking” angle, though I have heard it said many times and the idea definitely has some merit even as it lacks nuance. It is fine if we are meant to take it as a slogan but incomplete and perhaps misleading in general. In my own experience as a scientist (astrostatistics/ML applied to astrophysics) and as a non-native speaker of English I often have the feeling that my ideas are very clear on some topic and find that this feeling is confirmed by talking (in my native language) to a colleague that shares enough background to fill in any gaps in my explanation. Then when the time comes to put virtual pen to paper in English I have to hammer my understanding of what I did and why into a story that should be understood by the readers of whatever journal I am publishing in. This is hard and does not add much to the value of what I see as the central mission of my work. As in, milk is still milk no matter how you package it: you need some package to get it to the customer, but the thin liter sized plastic bag you get in Quebec is as good as the detergent bottle styled gallon you find in the US, and after drinking the milk both end up in the landfill. Yes, some additional clarity may be gained through this exercise as it is famously said that if you cannot explain it to your grandma then you do not really understand something. On the other hand research is about breaking new ground and a full, deep understanding of all theoretical and computational intricacies needed to get there is unrealistic to demand and inefficient to gain through the kind of writing exercise academic publishing imposes on researchers. Another issue hidden by the slogan that “writing is thinking” is that writing in a language that is not your own is being forced to think in categories that are not your own. This may be good for the individual, but collectively it is an exercise in imperialism.
Good comment, but beware of the hope that if LLMs break education hard enough, it might be fixed. Maybe it just breaks, period. As the recent "school" review said, school is the least bad system of education we know *that scales*. It's evidently not perfect, but after millennia of opportunity, we haven't found anything better.
Well, first and foremost, I suspect I'm focused on a different level of education than that review was. I'm mostly thinking of university and maybe high school to some degree. I agree that there's a bigger risk of breaking lower levels of education.
Second, I don't see where "millennia of opportunity" is coming from. Universal or near-universal education is nowhere near that old. The century-and-change that is has been around have been ones of very rapid technological and social change, so it's not *as* strong of an argument to say that it's the best we can do. Though I do agree that making substantial improvements is much harder than many people think it is at first glance.
“So much of school isn’t about learning things, it’s about obtaining proof that you learned things.”
This is true, but -forgive me if I’m being incredibly dense- unless the teacher is omniscient and can magically tell exactly how much each student knows, how else could the “teacher evaluates student’s knowledge” part of education work?
That's a reasonable question, and I think it has a fairly straightforward answer. Yes, evaluating a student's knowledge is a useful--often indispensable--thing for an educator to do. And yes, various sorts of graded work can be used to do that.
The issue is that the need to serve has a Certificate of Competence--even more than that, the societal insistence that educational programs must meaningfully *rank* the students that go through them against one another--imposes a lot of constraints and costs that interfere with what would often be a much simple, quicker process. As a thought experiment, imagine that for some subject you're learning, an educator gives you a set of 10 exercises and tells you to pick 4 of them to complete and return. Which ones should you pick? Anyone who's spent any amount of time in a modern school system probably defaults to something like "the easiest ones" or "the ones I know how to do the best." But that's exactly the opposite of what would be useful for a teacher to assess where what material you need the most focus on[1].
More comprehensively: almost anything that's going to have a "real" grade attached to it needs to be
A. Largely identical from student to student.
B. Have grades assigned in a legible fashion--someone outside the context of the lesson should be able to understand where they come from.
C. Be "fair" in the sense of not overtly favoring one student over another.
D. Actually have numerical scores assessed and recorded by an educator (a very minor requirement for some things and a huge burden for others).
This closes off a lot of space in terms of useful assessments--something as simple as "having a conversation with a student" can't fit into this paradigm without a great deal of trouble. And I think it also has a pretty big impact on the structure of edcuation at a macro level, too: material needs to be compiled and packed into fairly long, reasonably standardize curricula so that you can compare students across different classes or schools. This also significantly discourages students from exploring new subjects that they aren't sure if they're good at: doing so means a large time commitment and a real possibility of messing up your GPA
[1]I'll add that basically everything I'm saying from here on out is only relevant to reasonable-sized classes in which some degree of personalized help is available. Those big university lectures with several hundred students are barely even *pretending* to offer anything but the structure, ranking and Certificate of Competence. Not in a day when you can get high-quality text and video material for cheap or free online.
"This is true, but -forgive me if I’m being incredibly dense- unless the teacher is omniscient and can magically tell exactly how much each student knows, how else could the “teacher evaluates student’s knowledge” part of education work?"
A big part of the problem is Goodhart's law. When you use a precise measure for something, then that measure will be gamed, often on the cost of the actual intention or goal.
To circumvent this, instead of using a precise criteria, you could leave the evaluation deliberately fuzzy and to the teachers discretion. Of course this has it's own problems, in particular with the competence and the integrity of the teacher. But with good teachers it would probably work.
In principle you could have credentials granted by examiners who only do assessment, and don't involve themselves in teaching the material, which you'd learn elsewhere. Teachers would still need to do some assessments for developmental purposes, though.
In case anyone is unaware, this is how pre-university education works in the UK, and in many other countries. It isn’t perfect (much discussion of ‘teaching to the test’, much discussion of the value of assessed coursework, much discussion of whether you are trying to give an indication of achieving some level of skill/knowledge or an indication of rank, much discussion of the value of many short exams spread over the course vs a serious examination period at the end of the course), but it does avoid many of the failure modes we see in the US model where the grades are assigned locally by the same person teaching, and where formative assessment (to guide the learning) seems to be much more muddled with summative assessment (to assign a grade).
Due to both our higher intake of students born abroad, and many more of the students spending more time consuming American culture online, I (a secondary teacher in the UK) have had to spend more time in the last few years explaining to students that their grades at GCSE will not be decided by me, and will not be decided by anyone who knows them, and will not depend on anything they have submitted in class or any prediction I write on their school report. Sometimes, these days, I have to explain it to parents, too.
All of our work before the exams is about teaching them and preparing them so that they can do well in those exams. ‘Assessment for Learning’ is a whole thing: we try very hard to teach them how to use what we are doing to assess their own needs, to identify what they need to work on and what they need our help with. But still, they often try to ‘cheat’ to save face, or because it is less effort, or because they have some fixed idea that they will be punished, or sometimes because they genuinely don’t understand that submitting the correct answer is not the same thing as understanding something.
It’s been an issue with our A-level maths students this year (16-18 year olds voluntarily doing harder mathematics to get a respected, externally-assessed qualification). Several of them were doing homework (practicing unfamiliar questions that required them to use familiar skills and knowledge we have worked on in lessons, such as they will see in the exam) by showing a photo to online AI solvers, reading the solution given, seeing that they understood how that solution worked, then copying it out. They thought they had found an efficient way to dontheir homework, and that (because they understood the solution when they read it) they had ‘learnt’ what they needed to. But while this sort of modelling-by-example can be valuable, they were missing the opportunity to practice the problem-solving and recall skills the homework was intended to work on, and were also turning in work that made it look like they had mastered these skills and did not need more support and guidance. You could easily tell, by giving them an unfamiliar question in test conditions, that the students doing this were making less progress and didn’t have the skills that they thought they did. It took a lot of repetition and persuasion to get the majority on-board with a more time-consuming and cognitively-effortful approach which will actually develop their mathematical ability and prepare them for the exams.
There is always going to be a discrepancy between the measure and the target.
But it's a worthy discussion whether we should attach as much value to the measure as we currently do, and whether that comes at the expense of achieving the target, and how we could do better.
I was reading Scott's post on Tegmark's Mathematical Universe, and it made me think of Modal Realism, the notion that all possible worlds are real. It feels like very similar ideas stated in the different forms of mathematics and logic/philosophy... but maybe I'm just picking up on superficial elements here. Is this a real insight, in your assessment?
There are physical many worlds theories , mathematical many worlds theories and philosophical many world theories, with their own subtypes, and similarities an dfferences.
Physical many worlds theories roughly subdivide into the quantum mechanical and the cosmological.
Mathematical many worlds theories vary in comprehensiveness and consistency.. The Maximal Ensemble can't be consistent. There is no unique smaller Mathematical Multiverse., that we can determine. The computable MMU is particularly small.
Philosophy distinguishes between modal realism, which takes the idea of existent possible worlds seriously, and moral logic, which uses it as a formalism. Indeterminism implies there are real possibilities , in the sense that more than one thing could have happened. Moral realism is further step that these possibilities did happen: from a God"s eye view, they are all actual.
Strictly speaking, you are right. But, the things, the objects, the bodies including ourselves exist in the world --- the physical space is the abstraction of this world for purposes of physical analysis of the sort carried out in physics.
But the World is not a mathematical object, neither are any of the constituents of it.
I suppose it depends upon the meaning of the term "mathematical object".
I wonder if you agree that physics abstracts from the world. That a "table" is an object and the physical or geometrical models of the table --either as a plate of certain thickness supported by legs of this and that dimensions--- or a collection of atoms and molecules and electrons arranged in this or that geometries--- these two are different things and things which may be said to inhibit different realms altogether.
Yeah, Tegmark’s idea basically is Lewisian modal realism, except that Tegmark identifies possible worlds with mathematical structures. Lewis thinks there’s a different between possible physical realities and abstract mathematical objects. Tegmark denies there is such a distinction.
I don’t think there’s a way to get any evidence about either of these things, so the question is if one or the other of them is a useful way to talk about things. Lewis claims that his way of talking is useful for understanding things like causation and counterfactuals (though I think there are some deep unresolved issues in his view for this “closeness” relation among worlds) and Tegmark claims that if you’re doing modal realism, his way of thinking helps makes the bounds of possibility more precise. I’m not convinced that either is that useful.
As someone who doesn't love either modal realism or the Tegmark view, but wants to have some notion of "possible world", what other options do I have? Do they have weird consequences/downsides I should be aware of?
I believe in On The Plurality of Worlds, David Lewis has an extensive discussion of what he takes the problems with “ersatz” views of possible worlds to be. I haven’t looked at it closely myself.
My inclination is to just deny that there is literal truth or falsehood to modal claims, and instead try to figure out what work they are doing in conversations and what sorts of representations would be helpful for doing that. There’s a way to read Tegmark as saying “it’s just math” and not really being as realist about them as he says he’s being. But I don’t think this is helpful for most talk of modality.
I recall when learning about Jeffrey Bolker axioms of probability I encountered the idea of "situation semantics" as an alternative semantics to possible worlds... Would that fall under the "ersatz possible world" category, do you know?
Thanks for the response btw, I'm very grateful to have a place like this to ask questions of actual experts and have them give thoughtful responses.
Yeah, there's a bunch of different approaches here. The vast majority since Kripke and Barcan have worked with some set of things that they call "possible worlds" (though a few of them have gone so far as to call some of them "impossible worlds", because they specifically want to be able to represent the differences between propositions that are necessarily equivalent, where people might not know they are - you don't want to say that everyone who believes "2+2=4" also automatically believes "163x477=77,751" just because those two sentences are true in the same worlds). "Situations" (at least in some uses - I forget if this is how Bolker uses the term) are one way that people have moved to things that aren't as fully detailed as worlds, so that there might be some propositions whose truth value just isn't specified in a situation. People disagree about whether these are just formal tools for calculating the semantics, or if they are supposed to correspond to something meaningful and objective.
I find the notion of something like an action being possible already weird enough, it's irreducible, unexplainable -- this atom might decay now, but it doesn't have to (doesn't it?) you might aswer this, but you don't have to (don't you?).
I don't understand what "possible world" is even supposed to mean.
Like, "it could be the case that" is kind of fake possibility. Either it is or it is not. We just don't know.
Actions, that is, changes, are the only things that possibly (in the fake sense) are possible (in the genuine sense).
Hypotheticals help figure out what might be real, but what is actually real isn't determined by hypotheticals. It seems like some people take a hypothetical idea and run with it, ending up with weird ideas like a multiverse or stuff like that.
Reality is weird and I can't explain it, but I know that it is real.
We could be using hypothetical to arrive at fake possibilities in a deterministic universe, but you can't infer that the universe is deterministic from the fact that we are using hypotheticals.
If the universe is deterministic , you don't have to worry about possibilities being irreducible ... but then you have determinism as irreducible fact.
Indeterminism implies there are real possibilities , in the sense that more than one thing could have happened. Modal realism is further step that these possibilities did happen: from a God"s eye view, they are all actual.
Apart from me thinking that always only one possibility of several happens, the alternative makes no sense to me, why do you call this idea "Moral" realism?
There is one question, I am not sure how the philosophy solves it: What do you mean by "possible"? Does it include miracles? Does it include extremely unlikely but technically possible things? Such as... quantum randomness causing a large number of atoms to randomly jump in the right direction so that your broken glass cup reassembles itself.
The reason I am asking this is the following: Imagine that your glass cup just fell on the floor and shattered. A world where is stays shattered is possible. A world where it randomly reassembles itself is also, technically, possible. So when we say that both world are (equally?) possible, why do you in reality expect the cup to remain shattered?
My attempt at answer would be that worlds that require "smaller coincidences" have "larger measure", whatever that might mean. Not sure if Tegmark considered something like this. But I would expect that the philosophers talking about "possible world" did not concern themselves with numbers. (Maybe I am wrong here.)
So my guess would be that the idea of "possible world" points in the same direction, but it is incomplete. You need to also add some kind of "measure" to explain why some possible worlds are more likely than others.
Tegmark's preferred ontology involves four levels of multiverse. Our universe (i.e., lightcone) is within a Level I multiverse of everywhere spatially contiguous with us including the parts that can never causally affect us or vice versa, which is within a Level II multiverse of different cosmic inflation bubbles with different values for certain physical constants, which is within a Level III multiverse of different Everett branches, which is within the Level IV multiverse of every possible collection of computable laws of physics. The world where the broken cup reassembles itself due to quantum tunneling is a different Level II multiverse within our shared Level III multiverse (or rather, a collection of many Level II multiverses, but far fewer of them than of the Level II multiverses where it remains broken).
If something is not merely very unlikely but outright impossible by our physical laws (the ones that remain constant across inflation bubbles), then it doesn't happen anywhere in our Level III multiverse, but (if mathematically well-defined and computable) does happen within some other part of the Level IV multiverse.
And yes, the measure problem is a real sticking point for any kind of speculation along these lines. Tegmark is aware of it, but not everyone agrees that he has satisfactorily addressed it. See, e.g., Scott Aaronson's review: https://scottaaronson.blog/?p=1753
While reading the defense of amyloids, because I am slow, it was the first time I realized prion disease and disappearing polymorphs are the same chemical phenomenon. I then wondered how general this can get, so here is a very ACX question: is theory of the mind basically a prion disease?
maybe you're asking about memes, not prions. In its original sense, Richard Dawkins intended the word "meme" to mean something like "idea fragments". Or alternatively, "the memory analog of a gene". Which implies that ideas can spread like viruses, and vary according to a fitness function.
Id say 95 metaphorical, in which I wonder to what extent unpacking exactly how TOM dominates other forms of consciousness mirrors the way some chemical structures dominate others. But maybe 5pct, is there actually a chemical doing this.
I don't see how it could possibly be literally a prion disease, it doesn't spread like one and hasn't been detected in that way and none of the other things that would be true of prion diseases apply.
A user on ACX has invited me to a chat. I have the notification bubble, but I can't engage in the chat, possibly because ACX has not enabled subscriber chat? I'm a free subscriber only. App doesn't help. I also can't get rid of the notification, and substack support is chatbotting and ghosting me. Anyone know what I can do about it?
Subscriber chat is definitely enabled for ACX. When somebody sends you a DM and it’s their first DM to you there’s a question on the page somewhere asking whether you agree to receive messages from the person. You have to click yes on that to see their message.
Did not see such a message, on Firefox. I tried on other machines and browsers and on app, but I don't even see the notification in those. I'm getting this message when I open the general ACX chat (web) from the first screenshot
Did you try clicking on the "Direct" and "Unread" tabs? I think that might send you to a prompt that will let you see there's a pending DM and then agree to it.
I've received those a handful of times and wildly clicked around until I could finally see the damn DM.
Yes I've clicked about every available button in about all possible combinations. Tried on mobile, different machines, different browsers including clearing cache. Substack is by far the worst platform I've ever continued using for its content, and it keeps digging.
You must be so irritated! I just remembered that chat and messages are not the same thing. I can’t access my messages through the Substack website on my phone (using Safari). If I’m in my phone I can only access them using the horrible app. From my computer I can, but I can’t get to my computer right now to remind myself what the messages interface looks like. I think the icon for chat is a speech balloon and the one for messages is an envelope.
In censorship news, Stripe has decided to ignore the anti-financial censorship executive order that Trump issued [1]. Additionally Cloudflare has followed the precedent they set with Kiwifarms and is now disabling access to booru sites such as Gelbooru and Danbooru [2]. This is strange because you'd think with all the public and governmental backlash, that they'd be pulling back on the censorship a bit. Not so, they are only becoming more aggressive.
Perhaps they're hoping that when the next election happens, republicans will be dispossessed and democrats won't care too much about financial censorship of sexual stuff and will actively appreciate the political censorship? That's what happened last time. Fair Access to Financial Services (OCC-2020-0042-0001) was intended to fix this near the end of Trump's term but Biden put it on indefinite pause due concerns from activists that it would force the banking industry (a highly moral industry) to work with dreadful oil and gas companies, helping them harm the environment. Forget an act of congress enshrining such a preference into law, we can just leave it to the opaque risk management professionals in the financial services industry.
That's the fallacious argument that the in-charge democrats were willing to accept last time, so I suspect the payment processors hope it goes the same way this time. Poorly-organized president, last ditch attempt to fix something, democrats win, wipe all the effort away and then the cycle repeats again.
Or something like that. Maybe it'll be different this time because the public is so aware that payment processor censorship is a serious problem? After all, it's been highly publicized how Itch.io and steam were strong-armed. Plus, once you explain how payment networks work, nobody with a normally functioning brain likes it. If both sides have been sufficiently aggrieved, then that strategy (if it is a strategy rather than an own-goal by Biden when he was trying to get rid of Trump's legacy) won't work.
Basically payment processors are slippery little bastards and think they can act with impunity because their enemies are too disorganized and have too small a window to act besides.
I work for Stripe but do not speak for the company, and have no knowledge of any of the specifics in the cases you're talking about. I can only speak generally.
What I can say is that, before joining the company, I would occasionally see people complaining online about being kicked off of Stripe, and was pretty confused by this. Think, like, "Stripe shut down my business with no warning and won't explain why".
After joining Stripe, I looked into this a little, and it became clear to me that in almost every case, folks were doing stupid illegal or fraudulent shit that got them kicked off, but a) they never mention this in public, and b) Stripe understandably doesn't want to invite legal action. (Exceptions exist, of course.)
As I understand it, this isn't at all unique to Stripe. It's one of those things that you learn very quickly if you have any consumer-facing experience at all in the financial services industry; it's my default assumption any time I see this happen nowadays.
For whatever it's worth, this cost - of ongoing bad publicity from bad actors - is the flip side of a deliberate strategy by Stripe to make it easy for anyone to make and take payments online by starting a business. I don't think folks understand that "shut down fewer bad actors" isn't free; in practice, it means way more gatekeeping up-front, and way less internet commerce overall.
You seem to have a lot of faith in the prudence of every employee at Stripe. You can't imagine someone deciding to put a company on the list for reasons unrelated to legality? Like moral repugnance or obscenity? If some other billionaire calls up the Stripe CEO and asks him to remove something repugnant for him, then I doubt you or any other employee will stop him or even know about it. Basically, your personal experience as an employee at Stripe does not justify your trust in the goodwill of everybody working there.
Also Josh Moon is a definite counter-example to pure pragmatism on the part of payment processors/networks. None of what he's hosted is against American law. He hosts a web-forum and a podcast, and that's really it. There's doxing on the website, but that's not illegal. There's wild speculation as well, but that's not illegal either. Also his podcast is legally distinct from his website, so the corporate response from Stripe can't even be related to his website with user-generated content. It is specifically something in his podcast.
I don't think payment processors are practical-minded all the time, especially since there's nobody important who will stick up for content at the fringes. If I were a power-hungry individual, I would take advantage of the latitude and lack of public oversight to get rid of websites and creators that I didn't like. I think most people share this intuition and posts like this one don't sway them. If there aren't mechanisms in place to avoid corruption, then there will be corruption.
Yeah, I do. There's not really anything stopping a bad actor from working at Stripe or being the Stripe CEO. They don't even have to be that bad of an actor. There are plenty of people that would relish banning things that are US legal but morally dubious. Even if I had no empirical evidence that this kind of thing happened, I would still be suspicious because it's power without oversight -- ripe for abuse.
And there's precedent for it. Stripe cut off payment processing for Josh Moon's podcast, which is still available on Kick, Rumble and other platforms streaming live to thousands citing that it was on the restricted businesses list. What did he do for his podcast to get on the restricted businesses list? They won't tell him because he has no legal right to know. More likely, someone at Stripe, for some reason, wanted to cut off an income source for him. It wasn't even a matter of what was in super chats either, because Stripe also coerced GiveSendGo into banning him.
Even setting aside the fear of bad actors, Stripe is beholden to Visa and Mastercard like anyone else in the payment processing business, and Mastercard evidently cares enough about sexual propriety to ban erotic games from steam based on their content. That's a ban on a game store that uses its own currency, doesn't have traditional chargeback problems due to the generous refund policy and harsh chargeback policy, and also the ban was not specifically a ban on Mastercard/Visa payment processing for those games. They gave Valve an ultimatum that they had to remove those games, not to simply stop serving Visa or Mastercard payment processing on those games, but to stop displaying them at all. All of these things indicate that they're not being pragmatic about brand risk, they're just concerned about power. Stripe is beholden to these companies, and therefore is only as anti-censorship as the weakest link in the chain.
With all of that in mind, what do you base your faith in Stripe upon?
So the card schemes apply pressure (which I think they are mostly passing through, not out of their own beliefs) and it's Stripe's fault? Curious logic.
Having said that, a lot of financial institutions have their own rules - mostly for compliance or reputational reasons, not out of any real moral stance.
I gave you examples of Stripe attempting to ban someone from making money online for who they are rather than for any individual infraction, and I also provided an example of how they are completely subservient to payment processors and therefore will censor things anyway, even if, by some miracle, no censorious type works at Stripe. Whether it's payment processors or payment networks is frankly not important to me. Only one of them in the long chain has to veto your right to make money for you to be rendered without an income stream.
And if you want to make the argument that it's all Visa and Mastercard's fault and Stripe's not any kind of villain, then that only makes them a hatchet man for a villain. They consistently ban people from things without justification, based on internal standards that they make up themselves, without any possibility of recourse or understanding by the injured party.
That's bad. If they're doing it on the orders of another party because Mastercard warned them not to interact with some guy, then that just makes them eager collaborators in a broken system which is also bad. I don't see Patrick Collison talking about payment networks strong-arming him, probably because he doesn't mind it. They're business partners. He hasn't blown any whistles, in fact his highly loyal employee that's posting here was under the impression that Stripe was squeaky clean and only cut off obvious fraud-cases. Definitely not the case. Patrick Collison likes to keep it on the down-low, but he's a big fan of this system. If you want to argue with that take, then don't bother unless you have something solid to justify your faith in him. The way I see it, the man is well-positioned to mitigate some of the worst of payment network censorship, and instead decides to be their loyal henchman, if not a censor on his own initiative.
Also, I literally gave you examples of financial institutions disregarding reputation in favor of censorship with the case of Steam censoring erotic games. They banned those games completely rather than merely preventing Mastercard or Visa from processing them. The payment networks are giving ultimatums and not meeting companies in the middle at all.
What makes you think that everybody in these companies is concerned with appropriate risk management instead of just banning anything they don't like? Internet forum moderators go mad with power over a couple hundred users on a website, so why the hell do you think someone with power over who can transact online will be any better? Is there any evidence that would convince you that financial institutions that can ban you for no reason and with no explanation owed to you, are an evil thing and that people are currently taking advantage of their corruption right now to censor things that disgust them?
I've given you evidence that Stripe is directly responsible for censorship, possibly at the behest of payment networks, but also possibly not (because bans are a black box), and I've shown you it's not just a result of reputational risk because they ask for interventions from companies that go well beyond what's reasonable if their only concern was Mastercard appearing next to sexy hentai. What more do you want before you're willing to entertain the idea that there are bad actors in these companies? I'm not concerned with making a case against Stripe individually, but against the entire system Stripe is a part of which has enabled human rights abuses.
I'm not going to have Patrick Collison hanged, I just want him and everyone else in the payment processing industry to have appropriate oversight and rules. I also think that until they're advocating for regulation of themselves as utilities, that they deserve all the contempt they're receiving and more.
I'm confused by this - I thought the processors were profit-focused and just censoring because it kept them in the government's good graces. I'm going to guess that there's some legal liability the Trump admin hasn't exempted them from, and they're more scared of that liability than whatever the administration can do. But this is just a guess.
Legal liability or public relations liability, it's easy for enemies to say 'this financial service is selling revenge porn' or w/e if they take money from a site that has ever temporarily hosted that once. Protecting your brand identity is worth a lot of money.
Also, if there is a legal liability it may exist in Europe or China or etc. rather than the US, though I don't know whether that applies in this specific case.
*My* guess is that the payment processors are ideologically opposed to Trump, and that's motivating their actions. You are using the mistake theory explanation. The answer may not be mistake theory.
Can someone explain to me why it's worth it to put up with the payment processor oligopoly rather than building a decent UX around Tether and letting the blockchain do its thing?
Last time I saw someone try to do a normal storefront transaction on the blockchain it took 12 minutes.
Maybe enthusiasts will line up to tell me that hasn't been a problem for years or that their personal favorite service has a way around it, but at the very least blockchain for storefront transactions has a bad name and low trust among the general population.
I think you would either need something pretty close to the raw blockchain - in which case it would hard to onramp, require long complicated addresses, and have it be easy to lose your money.
Or it would have to be a centralized site forming a payment-processor-esque wrapper around the blockchain, in which case it's vulnerable to all the same legal issues that payment processors are.
Couldn't the storefronts purchase and operate their own pre-packaged wrappers? It's the concentration in a few easily-pressured dedicated processor companies that's the problem.
> Or it would have to be a centralized site forming a payment-processor-esque wrapper around the blockchain, in which case it's vulnerable to all the same legal issues that payment processors are.
Isn't this exactly why companies joining Level 1 (raw blockchain) and Level 2 (payment processor esque) algorithmically with provable mathematical methods are valuable? Thinking things like Offchain Labs and Arbitrum - which have transacted billions of dollars worth of crypto transactions by now, and are worth $1B+.
I don't want to put the cart before the horse, but are there any examples of horse-pushed carts? It seems like there has to be some sort of use case for it, such as maneuvering in tight spaces. I realize that horses aren't always cooperative, but it's interesting that human-propelled carts are almost always pushed rather than pulled (rickshaws being an exception)
Pulling a weight is much more efficient and stable than pushing it, especially for high-speed cases such as the aforementioned rickshaw, horse and buggy, truck and trailer.
If you push a weight, it's easy to move it out of alignment with your pushing force vector and difficult to move it back into alignment; it's a "dynamically unstable" situation where small mistakes can quickly reinforce themselves into a failure mode. Exceptions include things where the payload can't go out of alignment (rail tracks, gun barrel), where the propulsion would interfere with the payload (space rocket, gun barrel), or in low-speed applications where you need the weight in front of you to control it, e.g. wheelbarrow, hand trolley, shopping cart.
The opposite is true for pulling the weight. If you pull something, it will naturally align itself with the force vector and it's a very stable situation that scales well with speed.
As for your specific question: using a horse-powered vehicle in tight spaces is a bad idea in general, since the minimum size of such a vehicle is that of the horse itself. Optimizing a horse cart for tight spaces is like optimizing a rickshaw for a highway by giving the driver a nice pair of running shoes. Just the wrong tool for the job.
Horses are reputed to be persnickety about not wanting to run into things, e.g. a lot of anti-cavalry tactics rely on presenting obstacles that horses find intimidating and prefer to go around or draw up short of. Perhaps a horse set to push a cart would refuse to walk forward because it seems the cart as an obstacle rather than a thing that will be moving along ahead of them.
I'm totally ignorant of how the physics of movement works, but I wonder if it's because humans are bipeds and horses are quadrupeds. It might be easier for us to push rather than pull because we've got two arms and hands made for that kind of action, while it's easier/more efficient for horses to pull rather than push?
The horses wear harnesses which allow them to pull weights efficiently. Apparently until they were invented around the year 1000 farmers used to tie a plough to the horse's tail.
I suspect that who is doing the steering is relevant! The human is steering both carts, and I guess there might be some reason that you want to steer from behind (particularly so that you can be more aware of whether a wheel is stuck in a rut or something).
Teamsters and drivers usually ride near the fronts of carts, wagons, and carriages, AFAIK, same as drivers of motor vehicles who typically ride behind the engine compartment but in front of the cargo and alongside or in front of any passengers.
It occurs to me while thinking about carriage and wagon layout that there are substantial advantages to the teamsters being behind the horses or other draft animals. That way, they can watch both the road ahead and the animals at the same time, and can apply reins or crack the whip or whatever and see the effects on the animals without having to turn around. It's probably somewhat awkward for the teamsters to be behind the horses if the vehicle is ahead of them.
To be fair, flash fiction is an arena that favors AI, since the stories are small enough to fit into the context window and there's less chance of it introducing continuity errors, but it was still impressive to me. I had to think pretty hard to decide which is which, and I also mistook one human story for AI-generated. (In my defense, it wasn't a very good story.)
This was an interesting experiment, I got 6/8 right. I think right now ChatGPT (notice how I'm not saying "AI in general") can write good sentences and paragraphs, and prose that is somewhere between evocative + literary and purple, which really tripped people up in this experiment, but it practically never comes up with a clever story structure.
I've never seen an AI pull of a story where at first it's unclear what's happening or where everything is leading, but then a twist or a reveal makes everything clear in a satisfying way. All AI stories tend to read as if the storyteller was making them up as he went along, without ever being able to go back and plant some foreshadowing, and to the extent this is fine, because this is how the actual real life usually operates, too.
Funny enough, it might be easier for AIs to write lofty, "literary" sounding tales than just well-crafted, tightly plotted low-brow fiction because we are very willing to excuse all sorts of weirdness, continuity errors and unintentional mistakes in literary stories as "artistic". Embarrassingly, we might be closer to AI-Kafka than to AI-Dan Brown.
Lawrences story was immediately recognizable as unmistakably human, because though it may or may not have been a masterpiece, it was clever, and funny and tightly written and found a new, silly angle to look from. I've never once seen AI doing it, and I've tried to make all the major models write stories for me. On the other hand, the one human story that tripped me and probably you had a straightforward structure, plain, clear language and an unconvincing twist that didn't seem to make much sense? and it immediately made me think of AI (falsely).
It seems like a reasoning model should be able to solve that. It should be able to write to itself first a brief three sentence summary of the story, then an outline with a list of scenes, then tell itself what order those scenes should be written in and with what foreshadowing, and then finally actually write them in the order suggested.
If it could be persuaded to write a draft, then critique it and decide whether anything needed to be reorganized, and then write the whole thing again, it could get even closer.
For the reasoning models that still show CoT (basically the open weights ones), that's exactly what happens in my brief tries (I do not care about creative writing)
Good to know! I haven’t tried creative writing with them. (I’ve been focused on getting them to solve crossword puzzles - they have an easy time answering clues on early week puzzles, but an awful time putting answers into the correct squares of the grid, and identifying what letters are already there, to use in identifying which new solutions might be possible.)
Yeah I got it to do Monday through Wednesday of last week but realized that my approach wasn’t going to give it enough context to pick up on wordplay themes or rebuses. For tomorrow’s, it would definitely need a much better understanding of how the grid works than any of them have been able to demonstrate yet in my experiments.
There's AI generated story content on some Youtube video channels (there's a lot of 'revenge stories' about divorces/being fired/etc.) and the problem is that they don't remember details.
So things like "I was married for fifteen years" (and yet the eldest child, if there is a child, is twenty) or "I worked at this company for twelve years" (but invoked a clause in a contract from forty years ago). A lot of contradictions where a human would pick up "oops, better correct that" but the AI clearly is mashing together elements from plots it has been trained on but can't put them together coherently.
I see that Aella has bet "no" on the prediction market, which is a pretty good reason to believe that she is not the author of the post. I am now kind of curious -- do proponents of prediction markets think prediction market betting should be anonymous? What do they think of trades made by someone who has inside knowledge of the market and whether those should be anonymous? (Bets on questions like "Will person X do Y" or "Did person X do Y" made by person X.) On a real market I guess you'd risk having person X bet the wrong way and then having them set up a bunch of straw traders to vote the other way but I doubt Manifold has that dynamic.
If you consider prediction markets as gambling or a competition, I can see why this would seem unfair.
But as a mechanism for aggregating information and finding the best estimate of the truth, that seems working as intended? Aella's bet added information and now we're much more confident (though not 100%) that she didn't write it.
But a big part of the concern is market manipulation.
Say there's a market for 'Did A write this anonymous article?'
A puts $500 on 'No' using a publicly verified account, sending a strong signal towards 'no' being correct.
But secretly, A also uses an anonymous account to put $5,000 on yes.
Now,m technically this is all 'information.' A savvy Bayesian takes into account the number and size of anonymous bids, writes up probabilities and expected utilities for 'the market is being manipulated by A' and 'A is sending an honest signal', and places their own bet based on the highest expected utility.
But the market is now an anti-inductive social game. People are trying to second-guess adversarial bluffers, rather than basing their answer on subject-area knowledge, and those bluffers are trying to predict those guesses and double-bluff them. I'd be surprised if the results of such a social game are as accurate as an empirical-data-only prediction market.
> the market is now an anti-inductive social game. People are trying to second-guess adversarial bluffers, rather than basing their answer on subject-area knowledge, and those bluffers are trying to predict those guesses and double-bluff them
Fair, but this would only happen for very specific markets. Maybe it would happen once then everyone else would know not to bet on those.
Or the only people who do would be the nerds who like to to play anti-inductive adversarial bluff games. And at least they are getting some utility out of it.
The galaxy-brain play would be for Aella to bet 'yes" at single-digit probabilities, triggering a spike in the market when people notice and follow her lead, the have a confederate place a larger bet on "no".
Not sure how much she and her confederate could expect to profit from this kind of manipulation, but probably not enough to be worth the reputational fallout afterwards.
The main argument: Chiang writes neither hard SF (engineering with known physics) nor soft SF (science as window dressing), but a third thing: stories where the fundamental laws of science are different but internally consistent (This is actually very rare in published fiction. Scott and other rationalist-adjacent writers have also done this a few times in their fiction, but imo less well). Chiang uses these alternate realities to explore philosophy from the inside.
Key points that might interest other commenters:
- He writes the best fictional treatment of compatibilism/determinism I've ever encountered
- His stories treat philosophical problems as lived experiences rather than intellectual exercises
- Unlike most contemporary SF, technology in his stories enhances rather than diminishes humanity
His major blindspot: he completely ignores how societies would respond to paradigm-shifting tech (e.g., parallel universe communication that should revolutionize all R&D but somehow doesn't)
The review also touches on why strong Sapir-Whorf and Young Earth Creationism make perfect sense as story premises when you understand what he's actually doing.
The entire review is almost 2500 words long (It actually used to be a lot longer before I trimmed it! ) and I tried to cover a lot of nuances and share my joy of reading Chiang as much as I can. I'd love to hear people's thoughts on Chiang's work and whether my interpretation resonates.
Beautiful review. Ted Chiang is also my favorite author. He's the only author that has mastered the three aspects of storytelling that I care about - emotional resonance; satisfying story arc; and conceptually interesting. Other authors have one or two, but Ted Chiang is the only author who does all three consistently.
Having said that there are a few other short stories by other authors that are Ted Chiang level you might be interested in reading.
Learning to be Me by Greg Egan. By far the best fictional treatise on the nature of consciousness.
Zima Blue by Alastair Reynolds. A beautiful story about art, memory and identity.
This seems like a bit of a weird definition of hard SF. The hardest science fiction story I've ever read in my life is probably The Clockwork Rocket by Greg Egan, which takes place in a universe that flips the sign of one term of the Lorentz transformation. Egan then painstakingly works out all the physical consequences of this, and writes a story where they drive the plot, complete with periodic interruptions to confront the reader with an alternate-physics equation in the middle of the text (and there's an alternate-physics textbook on his website). I don't think anyone says this doesn't count as hard SF. It's not *mundane* SF, but that's a different and much narrower thing.
You mention a number of stories based on premises of "what if various religious and mythological concepts were real?". With the disclaimer that I haven't read most of them (except "Hell Is the Absence of God", which doesn't strike me as hard SF in any sense), it seems obvious to me why these don't qualify as hard SF: because it's not actually possible to have a fictional universe "where the principles of science themselves are meaningfully different from our world, but still internally consistent" that still bears surface similarities to the real world. Any work that purports to do so must necessarily be covering up various internal inconsistencies (https://www.lesswrong.com/posts/LaM5aTcXvXzwQSC2Q/universal-fire). The story could still be *good*, and ask thought-provoking what-if questions that Star Wars doesn't, but it can't really be extrapolatively applying the real ways of thinking about the universe to a fictional context, which is what hard SF seeks to do.
Gwern had a take on "Story of Your Life" similar to yours (https://gwern.net/story-of-your-life). For my part, I wasn't satisfied with that story. (Spoilers:) The text wants us to view the aliens' timeless perspective as equally valid to our own (hence the analogies to timeless laws of physics like the principles of least action), but it's just not; they live in a universe that has an arrow of time, whether they appreciate it or not. It's not physically possible for them to remember the future the way they remember the past (unless this universe's laws of physics are different in a way that allows this, but the text gives no hint of this). The only time the protagonist would be able to articulate the perspective she does in the story is at the moment of her death; at any other time, the contents of the rest of her life would necessarily be missing, which raises the obvious question of what that feels like from inside and how it avoids breaking the illusion of timelessness (other than by just giving the aliens amnesia or otherwise preventing them from seeing the question, which would be a cop-out). At best it's the story of aliens with a weird cognitive disability, not aliens with an interestingly different relationship to time. Arrival's genuine time-travel plot is in many ways more conventional but at least it makes sense.
I did like "Liking What You See: A Documentary" a lot.
In practice, compatibilism says something like “free will in the normal, pretheoretic sense of the term, doesn’t exist. Your choices still meaningfully matter nonetheless. You can’t meaningfully get out of the bind philosophically. What you can do, however, is make peace with it.”
Chiang makes this realization visceral. From The Merchant and the Alchemist's Gate:
“My journey to the past had changed nothing, but what I had learned had changed everything...Nothing erases the past. There is repentance, there is atonement, and there is forgiveness. That is all, but that is enough.”
“That is all, but that is enough” on the surface sounds like a deepity, but I genuinely find it more moving than anything else I’ve read on free will, determinism, or compatibilism.
When Chiang uses time travel as a motif, the stories differ from typical time travel stories. Because causation is a closed loop, knowing the future does not give you special access to it, and Chiang’s characters tend to not fight the future (successfully or otherwise).
In Story of Your Life [SPOILERS], the narrator learns an atemporal alien language and begins experiencing past and future as equally real.4 It takes her some time to make peace with it, but eventually she fully accepts the truth of determinism. She understands that life is full of tragedy, including that her daughter will die young, but life is full of beauty too. With both regret and awe, she sets forth on the path that she was destined to take5.
This is compatibilism from the inside. In both stories, the characters discover they cannot change what will happen, but this knowledge transforms how they experience what must happen: with forgiveness, acceptance, and even joy.
I've come to think that a mechanical world view is rather necessary for free will, and only superficially in conflict with free will.
The thinking is that free will does not require that things could have been different - even if that is the thought experiment that naturally comes to mind when thinking about that concept. For example, If I had made this action instead, A would have happened instead of B etc. The key is to realize that after the fact you did make that action and B happened, and nothing can ever change that - even if that action was a result of a thought process in your mind - this is still free will. Further, your brain state leading to said action was determined by all your sensory inputs, your previous experiences, and previous state of your brain, and this also happened in that particular way. We can imagine it happening differently in a thought experiment - but crucially from a mechanical world view it can't actually happen differently outside of that thought experiment.
The alternatives to free will outside this mechanical determinism I think is:
1. Randomness, e.g. quantum randomness makes parallel timelines, or there is somehow some small randomness in outcomes. (Without the parallel timelines I'm not really sure what randomness means here, since we can only observe the outcome of that randomness - maybe it means that at some level the outcome of physics at a small level is undeterminable for us outside of a probability function?) Anyway, this seems to, if anything, cause less free will - because some outcomes is determined then by that probability function and not your brain state. This would have to be really subtle differences from mechanical - otherwise I think the world would seem completly chaotic.
2. There is a qualia/soul/etc. that is disconnected from our world, and it is possible to re run the world with different choices - like loading a save game, while still being somehow you. This "you" outside the world would have to lose knowledge of the outside world when replaying the "save game", and your brain state at each run would have to be subtly different so that different choices is made. This seems like it implies that the free will making the choices is actually outside our world, and in a sense different from the qualia we are experiencing. So again, this seems like a weaker version of free will.
Interesting, thanks. FWIW my favorite novel is One Hundred Years of Solitude and that has themes which resonate with this, albeit somewhat indirectly. It's not sci-fi in any way but you may enjoy it.
Thank you I might check it out! Tbh I'm happy enough with Chiang's fictional treatment of compatabilism that I'm mostly looking for *different* ideas in my fiction going forwards, but you aren't the first to suggest One Hundred Years of Solitude to me. :)
What are some of the worst failures of very free markets? What comes to mind immediately is failure to price in externalities (e.g. pollution) and inability to deal with companies over a certain size becoming monopolistic (e.g. FAANG).
I'm curious to hear from the libertarian leaning commentariat if they think there are any ways to deal with these failures outside of government regulation.
The reason I am not libertarian leaning is that there are two kinds of libertarians, one say the market will solve it, the second realizes it won't, they say the courts will solve it. But the courts are a branch of government. You just want your regulators to be called judges. And the courts are not immune to being very politically motivated...
Some libertarians would say free markets don't fail, that anything they don't provide is not needed or desired (enough), efficiency at the expense of liberty is ultimately harmful, externalities can be handled by local rules and monopolies cannot exist for very long without government support of some sort.
Do you have evidence that, after a period of adjustment to a market without major government interference, that such a market would not verify the need for and provide solutions ?
Completely unregulated free markets could end up with slavery in those specific cases where slavery happens to be economically profitable (e.g. because you don't need to pay the slave their market price).
David Friedman argues, in short, that market failure is the exception, whereas government failure is the rule, and thus that involving government to avoid market failures is likely to cause more failure than it avoids. https://daviddfriedman.substack.com/p/market-failure
> I'm curious to hear from the libertarian leaning commentariat if they think there are any ways to deal with these failures outside of government regulation.
Uh, ways for *who* to deal with them? Non-libertarians might try to deal with them via grass-roots activism and boycotts. Libertarians can only do so by noticing an otherwise overlooked flaw in the moat, and if they notice one, they are not about to spill the beans here!
Also, I'm not sure what your definition of a monopoly is. Apple seems to be trying to compete in at least one market dominated by each of the other FANGs and against Microsoft to boot.
There’s also the Akerlof failure mode based on asymmetric information - without a third party regulator or certifier, it’s hard to establish good markets in used cars and patent medicines and other things where there’s a lot of variance in effectiveness of the product that purchasers can’t verify before purchase. (There can be markets, but they operate at low enough prices that expensive methods to produce high quality products can’t enter.)
That's a good point. I guess in just my head, a very free market doesn't mean the lack of regulation. It means regulation to the extent required to make it a theoretically perfect market.
Another example would be price transparency laws such as anti-drip pricing laws. This kind of regulation makes for better price discovery
As someone libertarian, I think that government intervention for externalities is reasonable but government intervention for monopolies is almost always a lie.
Of your FAANG 'monopolies':
Facebook and Google give their core product away for free. Amazon is drastically cheaper than almost everything it competes with.
Facebook became the most popular social network by outcompeting previous social networks, and to my understanding is currently losing market share among younger people as it gets outcompeted by instagram/Tumblr/whatever else the kids these days like.
Netflix is one of enough streaming services that a common complaint i hear is "why are there so many of these, I can't keep track of what is on which service", which is sort of the opposite of the problems with a monopoly.
It is theoretically possible for a monopoly to cause problems under unregulated capitalism. But in practice, accusations of this are mostly just used as a cudgel to hit whoever Lina Khan doesn't like.
>Facebook and Google give their core product away for free. Amazon is drastically cheaper than almost everything it competes with.
Please don't look at only one isolated piece of the business model, i.e. monetary price for consumers. It is well known how companies that offer a product below cost are most likely in the data collecting and enshittification business. They do aim for monopoly and customer/business lock-in so they can eventually monetize the customers and/or the businesses through that product or others. All these companies are also known to be fiercely anti-competitive, which is why effective regulation is overdue.
>I think that government intervention for externalities is reasonable
I am inclined to agree, with the caveat that the standard ancap/minarchist solution to externalities and the tragedy of the commons is to enclose the commons and rely on Coasian bargaining to handle externalities.
Myself, I am a bleeding heart libertarian and increasingly leaning towards just describing myself as "liberal", not an ancap or minarchist. I do see value in enclosing commons when practicable, and on Coasian bargaining, but both often suffer from issues with transaction costs and the latter especially suffers from coordination problems (brinkmanship in bilateral negotiations and free riders and strategic holdouts in multilateral negotiations) which can make state action an alternative worth considering.
No discussion of monopoly is complete without Standard Oil. But when the Sherman Antitrust Act was passed and it was broken up, it was actually already losing market share.
The real hatred of Standard Oil was its insanely aggressive tactics. Occasionally the rep would threaten all provisions stores in an area by telling them that if they didn't sell Standard Oil, the company would all provisions at cost in that area and bankrupt them.
All the while Standard Oil grew bigger, the cost of oil actually fell.
I don't really think there's any example of a real monopoly that has ever existed which was raising prices and making things worse for consumers. It's a myth.
A few scattered thoughts on why parenting young children feels more difficult today than it did in the past:
- Most other things in life have become significantly easier, while the difficulty of raising young children hasn't changed much. Back when most people were doing hard work all day, the perceived difficulty of raising children probably didn't move the needle much (relative to the baseline difficulty of life). And as the children got older, they would help with the household work, rather than being net drains on the parents' time and resources - so the early childhood stage was probably viewed as more of an investment, with an expected payoff when the child got older.
- Children keep each other entertained. Declining family sizes, and declining presence of extended family, means parents need to expend more effort to keep each child entertained.
- In my experience, young children don't usually seem to mind when their parents are busy with household work that they can watch and understand (cooking, cleaning, etc). But they HATE when their parents are unavailable because their attention is focused on a computer or phone. So if you want your child to play independently while you're trying to work from home or read articles about the difficulty of modern parenting, you're in for a rough ride.
- Related to the previous one: young children don't seem to mind sitting around while their parents socialize in real life. But they're less content to sit around while their parents are on the phone, and they hate sitting around while their parents are focused on screens. It follows that the more online your social life is, the more difficult raising young children is going to be.
When people start their families later, the grandparents are older and therefore less able to help.
The more abstract your work is, the more difficult it is for your children to understand it. Watching your parents do manual work is fun and you can learn something, watching them do tax reports is not.
Similarly, sitting around while your parents socialize is more fun if you understand what they are talking about.
Regarding your 3rd and 4th items, those are different expressions of the same syndrome and your description is accurate.
In the case of my eldest child this factor pushed me towards largely/permanently giving up television, and in the case of my youngest it has diminished my smartphone/online. In each instance the change has improved my own waking hours significantly. So I view it as a bonus feature of parenthood -- changes that I will not revert back out of when my youngest has left the nest.
Maybe another reason is that "not raising kids" is a much more realistic and accepted alternative today. So the thought "my life would be so much easier if I didn't have kids" is more present now than in the past, when having kids was just what you did. My life would also be much easier if I had a personal driver, but since that's out of the question, it doesn't stress me out that I don't have one.
I find that it has much to do with simplified concepts of "responsive parenting" that roughly boil down to "letting your child cry is traumatic", and therefore it becomes necessary to actively hold and soothe them very frequently.
I've also noticed that the parents I've met in real life often say that parenting is not that difficult for them, and I suspect there is a significant bias where complainers are much more common and visible on the internet.
Your second paragraph is very much part of the picture. I've parented a child born in the early 1990s and one born in the early 2010s and this difference is _really_ strong just between those two eras.
"Child-first" parenting is definitely part of it, but I think it's downstream from everything else in life getting easier. Responsive parenting (by today's standards) probably wasn't even possible before microwaves, washing machines, and 40 hour workweeks.
An analysis of Beethoven's genome (from a hair sample) suggests that Beethoven was rhythmically challenged. Beethoven’s beat synchronization PGI ranks between the 9th and 11th percentiles compared to modern samples’ beat synchronization PGI. And beat synchronization has been correlated with musicianship in modern samples.
"Notes from Beethoven’s genome" by Laura W. Wesseldijk et. al...
And this doctoral thesis by Matthew Aguirre, which was just published, discusses how, in human populations, most of the genetic variance in gene expression can be attributed to trans-acting expression quantitative trait loci (eQTLs) spread across the genome. "In practice[,] it is difficult to discover these eQTLs, and their cumulative effects on gene expression and complex traits are yet to be fully understood." However, Aguirre takes a stab at estimating eQTL effect sizes and "heritability from cis- and trans-acting genetic effects for 5,902 genes from a recent twin study of whole blood gene expression." The models suggest that the gene expression derives from sparser pleiotropic relationships across genes "than would be suggested by naive models of regulatory networks, *which has important implications for future studies of complex traits.*"
So maybe there's more to rhythm than a simple polygenic score?
"Regulatory network topology and the genetic architecture of gene expression" by Matthew Aguirre, et. al.
Beethoven, in his deafer years, probably would have stuggled to clap along to a rhythm. (Sarcasm.)
This seems a little like claiming that it would be surprising if (e.g.) Jane Austen's genome was sequenced and showed a predisposition to dyslexia. In fact, internet sleuths have diagnosed WB Yeats and others with dyslexia and it doesn't seem to have held them back. In the same vein; Beethoven was a notable pianist and his music was revolutionary for its rhythm and drive. Maybe his genetics made him more aware of these things.
"Polygenic score correlated at 2% with musicianship status fails to accurately predict musicianship status in a single extremely-out-of-sample individual" sounds like a caricatured straw-man of someone who doesn't understand polygenic scores.
10th percentile genetics doesn’t sound so low that a bit of extra training couldn’t make up for it. And also, Beethoven is famously the first major composer to use Maelzel’s metronome after it was invented!
Don't get me wrong. I believe environmental variables play a larger role in behavioral expression than a bunch of SNPs on some DNA strands. However, take a gander at figure C and D in that paper! Old Ludwig Van is waaaayyyyy down the scale on modern populations' PGS scales for music creative achievement and musical engagement. But the authors come up with a few excuses. Such as, "genetic associations reflect a culture-specific interplay between underlying heritable factors and environmental influences. Findings from a GWAS in modern Western society may not apply universally across time and regions." And, "as population-level approximations, PGIs do not necessarily yield accurate predictions at the level of the individual."
So they basically admit that Beethoven's PGI probably had little effect on his musicianship. LoL!
Then the assumption that "moar genes moar better" may be wrong and that, in fact, selecting out an embryo on the grounds of "this has less of the desired Big Beautiful Score" might not be the right choice, as it could end up getting rid of Beethoven.
Isn't this the crux of our current disagreement on embryonic selection? The science is nowhere near good enough to produce the promised outcomes of glittering prizes, that you too can have Baby Mozart Einstein Phelps Jobs in one bundle because we can pick out the bestest blastocytes?
I don't think that's the only disagreement, though it's certainly in play. There's also the idea that if trait selection were reliable, the side effects would be bad.
I would say how the hell can you tell someone's sense of rhythm from their hair, but that might lead us into culture war adjacent topics.
I remain less than convinced that doing this kind of testing on samples which are of dubious historicity and which have been kept in less than ideal conditions will tell us anything useful. "Beethoven was most likely, though not 100% definitely, assigned male at birth".
Supposedly, the polygenic scores built from markers such as rs1464791 and VRK2-related SNPs predict rhythm ability, musicianship, walking pace, and even breathing function. But I don't think PGSs are as predictive as a lot of the peeps on ACX believe. I got a chuckle out of the article.
And considering that we're able to recover DNA from 40,000-year-old skeletal remains, it's not difficult to believe that they could have recovered DNA from Beethoven. But the preservation of the DNA definitely depends on the conditions under which it was stored. My problem with this article is that they said they recovered his DNA from a hair sample. Yes, people were giving locks of their hair to lovers and admirers all throughout the 19th century. But usually it's mtDNA that's recoverable from hair. The cells from a hair follicle would need to be attached to to his hair (or perhaps dandruff), and those cells would have to have been stored in very optimal conditions to get Beethoven's whole genome. OTOH, they could have gotten it from his skeletal remains if they had access to them.
People! I have a solution to the puzzle of the existence of God and evil!
Dwelling on the cruelty of the world and the stupidity of the idea of God, I thought there cannot be any sense in soon to be dead babies's suffering.
But what if they are not conscious? What if NO BODY whose life doesn't end up getting its soul to heaven is ever conscious?
What if God only retroactively grants human bodies, and only those who went through all the right molecular motions, a soul? Like in really changing history.
Then it could be that going through such bodily process is necessary for getting the right character into a soul with which God wants to populate heaven, without even one getting lost in the process.
If I were God, I'd do it so.
Plus: it would mean that anyone conscious now can stop worrying about everything, cause everything will be fine, you're already on your way to heaven.
*Humbly bowing for applause*
[Edited for the horrible spelling mistakes that Deiseach in their rightful scorn has pointed out.]
Since I read some existing philosophy on this topic and still have such an inferior conception of God and suffering compared to yours, I really should follow your advice, since I'm clearly the one who needs a lesson in humility here.
If so, why have you discarded all concepts of god other than the Abrahamic religions (heaven, etc.), any consideration of suffering as necessary for some types of growth, reincarnation, etc.?
Wasn't this a plot point in Unsong? Uriel reveals that he turned some very poor countries into p-zombies to save on processing power and alleviate suffering?
Besides the odd moral conclusions it leads to, I think this implies that souls are created to have no explanatory power - a person "piloted" by a soul must behave exactly identically to a person with no soul. And like, if you're an atheist then this certainly matches observations, but if you actually believe in souls then surely you believe they actually *do* something that can't be emulated by the crude machinery of flesh?
> I think this implies that souls are created to have no explanatory power - a person "piloted" by a soul must behave exactly identically to a person with no soul. And like, if you're an atheist then this certainly matches observations, but if you actually believe in souls then surely you believe they actually *do* something that can't be emulated by the crude machinery of flesh?
When I imagine all these things, as a Gedankenexperiment, then no. Souls do indeed not pilot the bodies. Souls are formed, in a one way street way, by the bodies they get assigned to.
In heaven then, they begin acting on their own, now that they have the good characters God wants hhem to have.
And I know, this is weird, but there are dualists, and for them, who have the epiphenomenality problem you pointed to, this is as much a solution then it is to the problem of evil. They just have to believe in heaven to lighten up the epiphenomenality of their current existence.
Feser, a Catholic philosopher says the non-physical part of souls deal with abstractions, because abstractions do not exist in nature and hence are supernatural.
And this immediately taught me that the reason we mostly don't think like Plato, Aristotle or Aquinas is our fundamental approach to knowledge, epistemology is different.
They thought that every true statement has to refer to some pre-existing knowledge. Thus our minds basically discover pre-existing truths, like how the eye discovers a new object. Hence they thought if the abstract idea of triangularity is a thing, it must exist somewhere, such as in God's mind.
We think our minds are creative, our thinking makes new models, models which are true in the sense of useful, but not necessarily 100% true. It is like making a watch, maybe it is not 100% accurate but it roughly tells you the time.
God knows all things, including the future, so he could just only grant consciousness to those he foresees will go to heaven, and leave everyone else running around as p-zombies.
hylic – lowest order of the three types of humans. They cannot be saved since their thinking is entirely material, incapable of understanding the gnosis.
In the Nobilis RPG by Jenna Moran, there’s a gag about the angels having to cope with human population growth by pressing lower animal souls into service as people, then the souls of beetles. This all happened around 5000 BC.
(1) Please get your spelling right - it's "whose", not "who's", "its" not "it's" and generally "nobody" not "no body".
(2) Please get your theology right - this is a very idiosyncratic notion of ensoulment. If you're leading out of the gate with "the stupidity of the idea of God", better not do theology at all.
(1) Right, I'm sorry. The idea struck me suddenly and I wanted to get it in the thread before it got stale. The thread I mean. The idea turned out to be stale already.
But I did mean "no body" not "nobody".
(2) I do not believe in God, I find the idea stupid, and can give many reasons why it's stupid. Surely those reasons are only as reinvented as my idea of ensoulment is, which Edward Scizorhands showed me to be communicated by Scott long before https://slatestarcodex.com/2015/03/15/answer-to-job/
Then why do I think about God at all? I don't know. The cruelty of the world and my incompetence at dealing with it must make me occasionally very desperate.
Then why not simply drop the whole god and souls bit, and present your idea as the solution to the problem of consciousness: if you are alive in the world right now, congratulations, you are indeed conscious and not a p-zombie!
You'd at least have firmer ground to stand on to argue that, rather than trying to get into a philosophy you don't believe in and don't know about. It'd be like me trying to argue based on quantum physics which I am completely ignorant about.
>You'd at least have firmer ground to stand on to argue that,
I wouldn't have firmer ground. It's not a solution to consciousness, it presupposes it's existence, deliberately avoiding explaining it.
> rather than trying to get into a philosophy you don't believe in and don't know about. It'd be like me trying to argue based on quantum physics which I am completely ignorant about.
First, I think God wouldn't mind me exploring ways I might be able to believe in him, and second, why do you believe I know nothing about this philosophy? Because the idea I communicated is not mainstrean in theology? That's all the information you have right now.
Also, it's not like anyone is able to find out any facts about this. Not in their lifetime at least. I don't talk about whether Jesus indeed multiplied bread. I talk about whether he was the son of God or Cthulhu, playing jokes on us.
If you are going to argue about souls, which you may or may not believe in, and which many commenters on here may not believe in, then you are already ceding too much ground.
Arguing that this solves the problem of p-zombies at least gives you more room for debate; people on here are more willing to grant the notion of the existence of consciousness rather than ensoulment.
Not mainstream theology -indeed. It sounds a little like warmed-over Gnosticism and that's not original nor orthodox.
"Jesus was indeed multiplied bread". Ah, I see you can take the man out of the Reformation, but you can't take the Reformation out of the man! 😁
If you're going to postulate things why not just postulate that god is real and good. Just declare that reality orbits around the idea you want to protect. Skip the epicycles, ya know?
I really ment this only as a solution to the puzzle.
I do not entertain the idea that God exists.
I just find that not knowing why there is so much bad in the world is no argument against the possibility of his existence. And it nagged me to not have an argument for there to be no solution, nor a solution.
It doesn't explain why we exist in this material universe in the first place. If all conscious people (Hitler?) end up in heaven, why not just skip this step?
Works if you're some variety of Gnostic, though. You may have already read "A Voyage to Arcturus" by David Lindsay, but if not, I'm going to spoil the ending:
"He climbed into the embrasure. His feelings translated themselves into vision, and he saw a sight that caused him to turn pale. A gigantic, self-luminous sphere was hanging in the sky, occupying nearly the whole of it. This sphere was composed entirely of two kinds of active beings. There were a myriad of tiny green corpuscles, varying in size from the very small to the almost indiscernible. They were not green, but he somehow saw them so. They were all striving in one direction—toward himself, toward Muspel, but were too feeble and miniature to make any headway. Their action produced the marching rhythm he had previously felt, but this rhythm was not intrinsic in the corpuscles themselves, but was a consequence of the obstruction they met with. And, surrounding these atoms of life and light, were far larger whirls of white light that gyrated hither and thither, carrying the green corpuscles with them wherever they desired. Their whirling motion was accompanied by the waltzing rhythm. It seemed to Nightspore that the green atoms were not only being danced about against their will but were suffering excruciating shame and degradation in consequence. The larger ones were steadier than the extremely small, a few were even almost stationary, and one was advancing in the direction it wished to go.
…When at last he looked, he saw the same sphere as before, but now all was changed on it. It was a world of rocks, minerals, water, plants, animals, and men. He saw the whole world at one view, yet everything was so magnified that he could distinguish the smallest details of life. In the interior of every individual, of every aggregate of individuals, of every chemical atom, he clearly perceived the presence of the green corpuscles. But, according to the degree of dignity of the life form, they were fragmentary or comparatively large. In the crystal, for example, the green, imprisoned life was so minute as to be scarcely visible; in some men it was hardly bigger; but in other men and women it was twenty or a hundred times greater. But, great or small, it played an important part in every individual. It appeared as if the whirls of white light, which were the individuals, and plainly showed themselves beneath the enveloping bodies, were delighted with existence and wished only to enjoy it, but the green corpuscles were in a condition of eternal discontent, yet, blind and not knowing which way to turn for liberation, kept changing form, as though breaking a new path, by way of experiment. Whenever the old grotesque became metamorphosed into the new grotesque, it was in every case the direct work of the green atoms, trying to escape toward Muspel, but encountering immediate opposition. These subdivided sparks of living, fiery spirit were hopelessly imprisoned in a ghastly mush of soft pleasure. They were being effeminated and corrupted—that is to say, absorbed in the foul, sickly enveloping forms.
…Looking through the window, he was startled by a new sight. The sphere was still there, but between it and the Muspel-world in which he was standing he perceived a dim, vast shadow, without any distinguishable shape, but somehow throwing out a scent of disgusting sweetness. Nightspore knew that it was Crystalman. A flood of fierce light—but it was not light, but passion—was streaming all the time from Muspel to the Shadow, and through it. When, however, it emerged on the other side, which was the sphere, the light was altered in character. It became split, as by a prism, into the two forms of life which he had previously seen—the green corpuscles and the whirls. What had been fiery spirit but a moment ago was now a disgusting mass of crawling, wriggling individuals, each whirl of pleasure-seeking will having, as nucleus, a fragmentary spark of living green fire. Nightspore recollected the back rays of Starkness, and it flashed across him with the certainty of truth that the green sparks were the back rays, and the whirls the forward rays, of Muspel. The former were trying desperately to return to their place of origin, but were overpowered by the brute force of the latter, which wished only to remain where they were. The individual whirls were jostling and fighting with, and even devouring, each other. This created pain, but, whatever pain they felt, it was always pleasure that they sought. Sometimes the green sparks were strong enough for a moment to move a little way in the direction of Muspel; the whirls would then accept the movement, not only without demur, but with pride and pleasure, as if it were their own handiwork—but they never saw beyond the Shadow, they thought that they were travelling toward it. The instant the direct movement wearied them, as contrary to their whirling nature, they fell again to killing, dancing, and loving.
…The shadow form of Crystalman had drawn much closer to him, and filled the whole sky, but it was not a shadow of darkness, but a bright shadow. It had neither shape, nor colour, yet it in some way suggested the delicate tints of early morning. It was so nebulous that the sphere could be clearly distinguished through it; in extension, however, it was thick. The sweet smell emanating from it was strong, loathsome, and terrible; it seemed to spring from a sort of loose, mocking slime inexpressibly vulgar and ignorant.
The spirit stream from Muspel flashed with complexity and variety. It was not below individuality, but above it. It was not the One, or the Many, but something else far beyond either. It approached Crystalman, and entered his body—if that bright mist could be called a body. It passed right through him, and the passage caused him the most exquisite pleasure. The Muspel-stream was Crystalman’s food. The stream emerged from the other side on to the sphere, in a double condition. Part of it reappeared intrinsically unaltered, but shivered into a million fragments. These were the green corpuscles. In passing through Crystalman they had escaped absorption by reason of their extreme minuteness. The other part of the stream had not escaped. Its fire had been abstracted, its cement was withdrawn, and, after being fouled and softened by the horrible sweetness of the host, it broke into individuals, which were the whirls of living will.
Nightspore shuddered. He comprehended at last how the whole world of will was doomed to eternal anguish in order that one Being might feel joy.
…Fire flashed in his heart.... Millions upon millions of grotesque, vulgar, ridiculous, sweetened individuals—once Spirit—were calling out from their degradation and agony for salvation from Muspel.... To answer that cry there was only himself... and Krag waiting below... and Surtur—But where was Surtur?
The truth forced itself on him in all its cold, brutal reality. Muspel was no all-powerful Universe, tolerating from pure indifference the existence side by side with it of another false world, which had no right to be. Muspel was fighting for its life—against all that is most shameful and frightful—against sin masquerading as eternal beauty, against baseness masquerading as Nature, against the Devil masquerading as God...."
That lays it out pretty clearly, yes. Never felt inclined toward the Gnostic persuasion myself, but for one who is, the idea that the creator of the material world is evil itself does answer the question of evil pretty decisively.
I think he’s saying Hitler wasn’t conscious. If you are it was retrofitted back when you went to heaven. If you are conscious now then you are heaven bound.
As for the skipping this earthly step, that’s actually a good question for theologians or religious people who don’t believe in free will, I’ve tried to discuss it with some religious types online but they tend to hand wave away the problem.
Of course, there are those who do not treat "God" as an idea at all, but rather as a feeling state. If God, Good and Evil are all feeling states (that is to say, they are subjective, not objective concepts), then I think the problem disappears.
Or maybe innocent beings never suffer and die, we all just have a delusion that shit like that happens. Want some frills on that? OK, god challenges us with this delusion to test our love and trust in him.
Unrelated to schizophrenia, I wonder if it's plausible that there is a base rate of spontaneous insanity, say ~one in a million per year, the particular flavor determined by the circumstances in which it occurred. That neatly explains why so many different popular media, like movies and video games, seem to cause people to go mad, sparking the usual calls for it to be banned. (And it also explains why historically, you had mostly religious-flavored nuts.)
There's this character in a cartoon I like where his origin story is that, after a time machine accident, he's trapped in a void for a long time, goes mad, and then after some time, goes sane again. That partly motivates this model: maybe madness isn't an absorbing state, just one where the transition probability into and out of it is very low.
If true, this has bearing on immortals, people trapped in time loops, accelerated cognition (through noötropics or brain emulation), and less importantly, those getting data from large populations.
I haven't read anyone write "Since AI models have been observed to "think" in such and such manner, this should update our priors on such and such aspects of how humans think". Can there at all be any psychology-adjacent theory or model, say the IFS or the GNWS, whose plausibility might receive a boost from or be tanked by how AI is seen to behave?
I’ve seen a lot of talks in the past couple years by linguists who are studying language production in GPT-2 (and similar open models of manageable size) as a way to understand features of human language production! One of the most interesting talks I saw was looking at particular grammatical constructions that are somewhat unusual, removed all 200 instances from the training data, and observed that the LLM still produced them, and then did further investigation of what other related grammatical constructions had to be removed from the training data to ensure these didn’t appear.
Here's the specific one I was describing: https://arxiv.org/abs/2403.19827 I hadn't been sure if a written version was available.
I'm fairly sure I saw a few other talks doing similar things just at UC Irvine in the past two years, but the search terms I'm using aren't turning them up (though it did find this: https://arxiv.org/html/2408.03855v1)
I'm pretty sure some of the others I saw were using transformers to identify which features of an NP with multiple singular and plural nouns in it make people more or less likely to accidentally mis-pluralize a verb, and correlating some structure of a transformer to the processing time that humans have in a certain linguistic task.
To go from AI to the human mind, you would have to be using AI to test some theory about the human mind, which is not what AI is designed to do. Pretty sure no one thinks actual people learn the way LLM's do.
I think we learn a lot of things -- for instance, the early version of language, up to age 4 or so -- by pattern-matching rather than by grasping concepts and rules. But though we require lots of examples, we require way fewer than an LLM because our brain is already set up to learn language -- it has compartments of the right shape, so to speak.
I could nitpick your use of the term "grasping" (because it might be implying an "either-or" dichotomy, when comprehension is an incremental spectrum), but you are basically correct. Language acquisition is the output of a growing semantic network in long term memory. I don't know if anyone has programmed a computer to calculate using "fuzzy sets", but that's at least on necessary step along the way toward simulating a human-like mind.
We’ll know AI is starting to think about what it does using rules and concepts when we get it going through stages like starting to say ‘foots’ instead of ‘feet’ before being correct back. Do we have any examples yet?
I thought this about human Go ability when AlphaGo performed at superhuman levels; I think that the lack of this in linguistics is worth thinking and talking about, but may have changed a bit in the last few years.
I hear that a lot in conversations, and personally have updated a lot on it. For example how system 1 works. For example, it helps us understand better what system 1 in the sense of Kahnemann and Tversky is capable of, because the AIs are so similar to system 1.
The “seek for the sword that was broken” dream came to Faramir multiple times and only once to Boromir (he says), implying that Faramir was the one who should have traveled to Rivendell. Instead, we get Boromir going nuts by the river, which results in the breaking of the Fellowship and a long side quest to rescue the Hobbits and take down Saruman, which the ents clearly could have managed themselves. The whole story is probably over after Book II if only Faramir had had the guts to tell his brother to go pound sand.
It's more complicated than Faramir not telling Boromir off. Suppose he had gone to Imladris and found out the meaning of the dream. He would probably not have pushed to go to Gondor, although he still *might* have done: the emergency remained and Denethor was still the leader of the armies fighting Sauron, so this was vital intelligence he should receive.
The Orcs were in place so the breaking of the Fellowship would probably still have happened. Frodo was conflicted about what he should do, independent of Boromir's suspicious activity.
Faramir would have agreed that the Ring should be destroyed and would have supported the decision to do so, but the other factors remained. The Ents were able to destroy Isengard, but they needed outside influence to bring them to the decisions to do so.
This sounds like "why didn't they just fly on Eagles into Mordor" kind of quick and easy solution, which in fact could not happen in canon.
No, my theory is that with no Boromir, there is no breaking of the fellowship, Sam and Frodo don't wind up wandering the wastes outside Mordor or the woods around Ithilien with Gollum as their only guide, wasting precious weeks, and thus there's time to formulate an actual plan, and they're able to destroy the ring before the siege of Gondor even begins.
Make no mistake: just like Jenny in Forest Gump, Faramir is the secret villain of the books. In hindsight, Gandalf shouldn't have bothered saving him from the funeral pyre, and Aragorn should've saved the Kingsfoil for somebody that actually deserved it.
It still remains a problem of how the entire Fellowship gets into Mordor. And Gandalf will be gone by now, anyway, since he fell in Moria, and they had no way around passing through Moria.
So if all goes well, there's Aragorn, Faramir, Gimli, Legolas, Frodo, Sam, Merry and Pippin heading towards Mordor. I think that along the way *something* would have happened to break them up, since they had such a long route to cover and there were enemies on the watch for them (sent by both Saruman and Sauron).
Following the river is their easiest path, but once they do manage to slip into Mordor, then they need a guide. Aragorn can't help them since he will be no more familiar with the interior terrain than anyone. Gollum, like it or not, is who they need because he's been there and knows the dangers (too well, as we can see when it comes to betraying them to Shelob). Faramir might have some knowledge, but as we see not enough - he's heard of Cirith Ungol but again he does not have enough direct knowledge to be able to steer the party.
They're in enemy territory which is chock-full of the armies and other servants of Sauron. There's no easy direct route to take. And from a practical point of view, the Quest was hopeless from the start; Tolkien discusses this in a draft of a letter from 1956:
"Frodo was in such a position: an apparently complete trap: a person of greater native power could probably never have resisted the Ring's lure to power so long; a person of less power could not hope to resist it in the final decision. (Already Frodo had been unwilling to harm the Ring before he set out, and was incapable of surrendering it to Sam.)
The Quest ⁂ was bound to fail as a piece of world-plan, and also was bound to end in disaster as the story of humble Frodo's development to the 'noble', his sanctification. Fail it would and did as far as Frodo considered alone was concerned. He 'apostatized' – and I have had one savage letter, crying out that he shd. have been executed as a traitor, not honoured. Believe me, it was not until I read this that I had myself any idea how 'topical' such a situation might appear. It arose naturally from my 'plot' conceived in main outline in 1936. I did not foresee that before the tale was published we should enter a dark age in which the technique of torture and disruption of personality would rival that of Mordor and the Ring and present us with the practical problem of honest men of good will broken down into apostates and traitors."
Given your opinion of Faramir, you might be amused to know (if you don't already know) that he wasn't part of the story from the outset:
Letter to Christopher Tolkien, 1944:
"A new character has come on the scene (I am sure I did not invent him, I did not even want him, though I like him, but there he came walking into the woods of Ithilien): Faramir, the brother of Boromir – and he is holding up the 'catastrophe' by a lot of stuff about the history of Gondor and Rohan (with some very sound reflections no doubt on martial glory and true glory): but if he goes on much more a lot of him will have to be removed to the appendices — where already some fascinating material on the hobbit Tobacco industry and the Languages of the West have gone."
Draft letter, 1963:
"I think you misunderstand Faramir. He was daunted by his father: not only in the ordinary way of a family with a stern proud father of great force of character, but as a Númenórean before the chief of the one surviving Númenórean state. He was motherless and sisterless (Eowyn was also motherless), and had a 'bossy' brother. He had been accustomed to giving way and not giving his own opinions air, while retaining a power of command among men, such as a man may obtain who is evidently personally courageous and decisive, but also modest, fair-minded and scrupulously just, and very merciful."
You guys are taking me a bit overly-literally. The "Faramir had the dreams but I talked Dad into sending me instead" bit is probably Tolkien just doing a smidge of foreshadowing for the eagle-eyed reader about how headstrong Boromir is and will eventually come to grief, but I think that perhaps unintentionally, the other way to read it is that Faramir was supposed to go instead, for reasons which become quite obvious once you get to the chapters in Ithilien. I do not actually regard him as the secret villain of the story.
Though it does open up intriguing vistas: what would have happened if Boromir had stayed in Gondor with Denethor?
In the early days post-Jackson movies, there were a lot of people getting into the fandom and popular fix-it fics were ones where Boromir lives instead of dying at Amon Hen.
But I don't think anyone ever wrote "what if Boromir never went to Imladris?" Depending on how much Denethor knows, and how far gone he is, if Faramir comes back to Gondor and informs his father of what happened, then Denethor might have sent Boromir out to hunt down Frodo and take the Ring (even if Faramir never tells that out directly, I think Denethor would be able to piece together that the Halfling has something really, really valuable and powerful; the more Faramir refused to tell what it was, the more Denethor would come closer to the truth).
Without the Breaking of the Fellowship, Merry and Pippin would never have come to Fangorn, meaning the Ents wouldn't have attacked Isengard. Consequently Saruman would quite possibly have been able to defeat Rohan, causing far more death and destruction than would be saved by avoiding the siege of Gondor.
The side quest to Rohan was necessary to save Theoren from Wormtongue's influence. Without that and without further assistance from Our Heroes, the host of Edoras doesn't set out to reinforce the Westfold armies, there's no force to hold Helm's deep, and Saruman's army is free to pursue the scattered elements of Erkenbrand's army and defeat them in detail.
The Ents can wreck Isengard and invest Orthanc, but without Theoden assisted by the heroes, Saruman's field army will be intact and available to turn back and relieve the siege. The way it played out in the books was that Theoden's cavalry sallies unexpectedly from the Hornberg, temporarily disrupting the Orcs just in time for Erkenbrand's regrouped forces to show up and attack them in the flank and rear, causing a rout. The Huorns' role was to cut off the Orcs' retreat and kill them as they tried to flee the field.
In the movies, the reinforcements were Eomer and his men rather than Erkenbrand's, to save the complexity of introducing Erkenbrand and his forces. In the books, Eomer had been at Edoras with Theoden (imprisoned rather than exiled) and was with Theoden throughout the battle.
Without the fellowship breaking the way it did, they probably follow Aragorn's proposal which IIRC has Aragorn, Gimli, Sam, and Frodo going to Mordor while Legolas, Merry, and Pippin accompany Boromir (or Faramir, in this case) to Minas Tirith. With Aragorn and Gimli along, there's no chance they accept Smeagol as their guide. We don't know for sure how Aragorn would have lead them into Mordor, but it probably wasn't Cirith Ungol based on Gandalf's reaction when he got word of it. My best guess is that he knew of a less-guarded pass or tunnel or postern associated with the Black Gate, since Frodo had headed there based on what he remembered of planning sessions in Rivendell. They might still get to the volcano, but Gollum wouldn't be there to fall in with the ring under the weight of the curse Frodo had placed on him and of Gollum's own broken oath. Frodo would still not have been able to throw the ring in under his own free will (Tolkien has said that pretty much nobody would have been able to do so), so either the mission fails there as the ring stays undestroyed long enough for the Nazgul to arrive to stall them and then for Sauron to show up personally to reclaim his own. Or Frodo jumps in with the tingy to keep it from Sauron. Or Sam or Aragorn or Gimli has to tackle Frodo into the fire, probably killing both of them in order to destroy the ring.
Lastly, Rohan falling or at least faring far worse against Saruman, plus Aragorn being in Mordor rather than Rohan, means that neither the Rohirrim nor the Oathbreakers of Dunharrow show up to help defend Minas Tirith. If the battle plays up similarly to the books until the point they fail to arrive, then the first circle of the city at least will fall and the Witch King's army will remain intact to assault the second circle and further, and without the intervention of the Oathbreakers, the black-sailed ships will carry Haradrim corsairs to reinforce the Witch King rather than a Gondorian relief force lead by Imrahil and Aragorn.
It isn't necessary for a hobbit to be present in Rohan during the aftermath of the attack on Saruman, but Gandalf, who would have been there regardless. Aragorn knew about the prophesy that he would lead the army of the dead, and that his duty ultimately calls him to Gondor, so that also happens regardless. I don't remember Aragorn or anyone else coming up with a plan about how to deploy the members of the Fellowship. It was ultimately up to Frodo, who would have made the same decision regardless of what Boromir did (his breakdown merely made Frodo's escape easier). Finally, Gollum was tracking Frodo via the magical influence of the Ring, so if Frodo gets to Mount Doom, so does Gollum. We can't know how that scene would then play out, but that's dramatic tension for you.
Tolkien also included elements in the plot about 'another purpose behind events and it was not the will of the ring-maker', see the chain of events which resulted from Bilbo Baggins starting off on a treasure quest with thirteen Dwarves, all due to the meddling of Gandalf long ago:
From Appendix A, LoTR:
"It might all have gone very differently indeed. The main attack was diverted southwards, it is true; and yet even so with his farstretched right hand Sauron could have done terrible harm in the North, while he defended Gondor, if King Brand and King Dáin had not stood in his path. When you think of the great Battle of Pelennor, do not forget the Battle of Dale. Think of what might have been. Dragon-fire and savage swords in Eriador! There might be no Queen in Gondor. We might now only hope to return from the victory here to ruin and ash. But that has been averted – because I met Thorin Oakenshield one evening on the edge of spring not far from Bree. A chance-meeting, as we say in Middle-earth."
The Breaking of the Fellowship looks to be disaster, but out of that seeming disaster different benefits come through things like Merry and Pippin meeting the Ents, and other things that might not have happened if they all stayed together on a straight path to their original destination.
So it had to be Boromir, not Faramir, because this is Boromir's character. Put Faramir in his place, it turns out differently, and maybe not for the best.
Aragorn came up with a plan while Frodo was off considering, with the intention of proposing it to Frodo if the latter came back still undecided:
>It would indeed be a betrayal, if we all left him. But if he goes east, then all need not go with him; nor do I think that all should. That venture is desperate: as much so for eight as for three or two, or one alone. If you would let me choose, then I should appoint three companions: Sam, who could not bear it otherwise; and Gimli; and myself. Boromir will return to his own city, where his father and his people need him; and with him the others should go, or at least Meriadoc and Peregrin, if Legolas is not willing to leave us.
It's easy to forget because it almost immediately became moot because the confrontation with Boromir made up Frodo's mind to set off alone for Mordor without delay.
Someone tell me if my memory betrays me, but I don't think is was Faramir's decision but Denethor's, which Denethor regrets when he learns the news of Boromir's death. So Faramir would've had to tell Denethor to pound sand, which is a bigger ask.
Faramir, his younger son, began to have dreams, speaking of Imladris and Isildur's Bane. Boromir convinced his father to send him instead of his brother and in the end Denethor gave in. Boromir did not return from his quest; he was shot by Uruk-hai on Amon Hen and Faramir and Denethor both heard the horn of Gondor.
I am a long-time viewer of Patrick Boyle's Youtube channel. He's an ex-economics professor that makes a living explaining economic and political news. I used to really enjoy his work, but in his past few videos, the script is obviously AI generated: https://youtu.be/0d2pCt8JomQ?si=VhqAsQyselNyuxq2
Every paragraph ends with him making the classical comparison "It's not just ____, it's _____." and by the halfway mark you realize he's basically just restating the same thing over and over again in a different way without adding any meaningfully new commentary. Both hallmarks of AI writing.
I think it's interesting that almost no one in the comments seems to take note of this. Maybe I'm odd for being exposed to a lot of specifically AI generated content in comparison with human stuff through this substack, but it's somewhat sad to me that respectable content creators are basically producing well-disguised AI slop.
I'd been noticing more of "people think some stupid thing (no evidence to think many people think that false thing), but actually), and not just from Boyle. I thought it was just podcasters being stupid and annoying, but AI is a plausible explanation.
Dude, also a watcher and felt like something was deeply "off" but I didn't recognize it as AI until your comment. Still have that, "I like this thing so it must be good in ways that are totally unrelated to the qualities I like it for."
After 10 million cases, I'm still like, "How could that actor/writer I like do something bad?"
I read an old BBS-style forum that allows semi-anonymous posting (i.e. non-registered usernames) and certain threads are flooded with comments that all use em-dashes and curly quotes (“ not ", we'll see if substack leaves in the curly quote) -- clear signs of AI output. It's probably 5-10% of comments now. It makes me wonder if it's just (a) one or two obsessed basement dwellers, or (b) reputation management firms (the AI comments always appear on threads about public figures).
Maybe there's an intelligence threshold after which you don't get more intelligent; you just get faster and bigger (more data storage). Sort of like a meta-version of the way a Turing machine is universal -- it can emulate any other computer.
f) Q: Please give me an exhaustive list of the elements and inorganic compounds that are gases at STP. By STP, I mean 1 atmosphere pressure and 0C. By inorganic, I mean that no atoms of carbon should be present. Exclude CO2, CO, freons and so on. Please include uncommon compounds. I want an exhaustive list. There should be roughly 50 compounds. For each compound, please list its name, formula, and boiling or sublimation point.
which really wasn't very different for the two models. They both got a bit more than 50 valid compounds, and accepted valid additions. I try to be consistent on scoring, but am not completely successful on this.
GPT5 looks a lot better than last year's initial GPT4, but basically o3 with a router, compared to SOTA 04/16/2025.
I'm updating to _slightly_ longer timelines, but not by much.
Note, though, that GPT5's hallucination benchmark _does_ look significantly better:
shows %incorrect claims dropping from 12.7% for o3 to 4.5% for GPT5-Thinking, which is a substantial help. I wish I knew what the human baseline was for this.
Many Thanks! Yes, o3 was released (as o3-mini) on Jan 31, 2025 while GPT4 was released on Mar 14, 2023 (followed by a bunch of later GPT4 versions, to add to the confusion...).
>Also doesn’t gpt5 decide what to use, it’s not one model but chooses one
Yes, if I just let the router decide, under the covers. I explicitly called GPT5-Thinking to, hopefully!, force it to use the "smartest" version available.
It’s hard to separate genuine sentiment from trolling and content farming in the early stages, but it seems to me the more informed sentiment is that it’s maybe a bit better but has pros and cons; a letdown from the hype so far as sheer capability is concerned. However, it is VERY good from an efficiency standpoint - intelligence per compute. Making this a good business revenue upgrade, rather than a consumer-facing upgrade.
I wrote about this in longer form here (https://theahura.substack.com/p/tech-things-genies-lamp-openai-cant) but the tldr is: GPT5 is currently at the top of the pack in terms of general utility, but it is not *so much better than everything else* that it deserved the endless hype sam and openai gave it. There are certain benchmarks where there is wide agreement that gemini or claude still perform better.
GPT5 was "supposed" to be god in the machine. It wasn't. Most of the people who are reacting to it, are reacting to the fact that it was a let down even though it is definitely an (incremental) improvement
So? That's one of the utilities of interactive things (or even just things that are observed). People get attached to lots of things, and then those things have a (or and extra) utility.
I miss the time before the Eye of Sauron turned to the topic of AI. Now its sucked into politics, tribes, algorithms. Welcome to hot takes, bots, hot takes of hot takes, and clueless pontification. The answer for any frontier AI is, whats your use, whats your personal style, and what month is it. MY hot take is gpt-oss will be shown to be very good after a few months of messing with it. Similar to llama3, gemma3,
I have particularly enjoyed all the people telling me that current lessons for children in schools should be replaced with lessons that teach them how to write prompts for existing LLMs, because that’s the future and they need to learn it young otherwise they will be unemployable. Obviously, only a Luddite could possibly disagree.
I only used it once so far, I presented it with a mysterious bug and it completely failed and I had to solve it myself. Didn't really get an impression that an improvement had occurred, but I have to use it more.
I wrote a post yesterday replying to a few of Alexander's recent posts for my blog: https://crocodialectic.substack.com/p/there-are-communities-everywhere. In it, I argue that Alexander's beliefs about the decline of communities do not take into account the ability of people to join strong communities, which remains high. I also try to define community strength somewhat more rigorously, and end up quibbling with him on some other points. In a rarity for me, I toss in some personal anecdotes about communities I am or have been involved in. In the end, I conclude that people do want strong communities, but they also want many other things, including weak communities, and I do not expect but can't rule out GDP favoring strong communities over weak communities.
In general, my blog is about whatever I want it to be about, which usually means linkposts and books that I am reading. However, I also take occasional stabs at effective altruism discourse du jour, as well as some lighter stuff about language. I publish at least once a week, and have been doing so for a couple of months. Feel free to check it out if it's something that interests you. Feedback is always welcome.
I would say, strong communities used to imply poverty, people needed to help each other survive. Today, if you just want the normal everyday comfy life, you don't need that. However, if you want something entirely different, you might still want that again. I have been in the BDSM community because people there could get something entirely different, also in a Buddhist community so far.
Our issue is - and how ironic is for a more-or-less-Buddhist to say that - the lack of strong desires for something entirely different from basic comfort.
Two questions: 1) Is phytomining safe, or are the farmers exposed to toxic levels of metals? 2) What is the min concentration of a metal to make phytomining cost effective?
Nice to see somebody read the piece! Re: question 1, depends on the crop and processes being used, but it's pretty safe; obviously there's going to be some risk in the smelting process, but most sources seem to think it compares well to conventional smelting. The answer to 2 is pretty similar, it's going to vary based on the crop and the technology, and the technology is still being developed. You might be interested in https://ijog.geologi.esdm.go.id/index.php/IJOG/article/view/171, which looks at gold phytomining where it has a pretty low concentration, all things considered.
I think if you're going to refute the "decline of community" argument (a la Bowling Alone and The Upswing) your post would be stronger with some hard data in it. It's not enough to say "well I belong to a foam sword bopper club and we have picnics!" or whatever, because all the time-use and survey data show large, long-term secular declines in things like time spent with other people, number of close friends, civic engagement, etc. I'm open to the "there are communities everywhere actually, you just aren't thinking about it the right way" argument but I'd like to see a more data-driven argument in that direction
I do think it would be nice to explore more data (and I might do so in the future) but what I'm arguing about is the availability of community, not the rate of people who are actually in a community. The data sources that Bowling Alone, and things like that, are measuring is participation in a community, not access to participation in a community.
I don’t think we are as interested in access as in actual participation though. The difference between GPT-3 and ChatGPT wasn’t access - basically anything you could do with the original ChatGPT, you *could* do with GPT-3, it just took cleverness and wasn’t very fun.
Alexander and other people have hypothesized that access to community is a reason for the decline in the level of participation in communities. So I think it is interesting, at least in the context of whether it is a driver of the decline in community participation (or a barrier to further upswings, for that matter). I agree that, if there was not a widespread belief that there has been a sharp decrease in the proportion of people in a strong community, with attendant negative consequences, it would not be particularly interesting to analyze access. But that belief is there, for good reason, and I think it makes questions of access relevant.
Something I'm getting really sick to death of: people making some negative generalisation about a group of people, and not realising that this not only punishes the innocent (they obviously know this, and don't care) but also *lets the guilty off more lightly* than they otherwise would be! By letting individual bad people in that group hide behind the idea that they're just being a normal member of their group, instead of personally a scumbag.
This seems really obvious and yet it seems so routinely overlooked. Is there a name for this fallacy (in this specific form)?
Examples:
The right hates academics, largely because they're so often brazenly politicised. I respond that academics are great, when they're being proper academics. 100% of the condemnation should go onto those *particular* academics (however many there may be) who are so selfish, so entitled, and so contemptuous of their own profession's ideals that they would proudly exploit their position and its remaining social trust to shamelessly further their own political views, instead of pursuing truth for its own sake. Declaring war on academia, among many other things, only *protects* these selfish scumbags by *giving into their framing* of being normal academics, instead of terrible academics who behave in total opposition to every value they are supposed to hold.
Feminists do the "#notallmen" shit where they imply or outright state that men are naturally rapists, or cheaters, and by doing so give actual rapists and cheaters a shield of "just being a man". When they should instead be told "no, your behaviour says nothing whatsoever about men, and everything about you. If you rape or cheat, the only information we gain (and we gain it to the strongest possible extent) is that *you*, personally, are a subhuman piece of shit".
Redpillers "reveal" the "truth" that all women are gold diggers, are excited by and attracted to violent men, commit paternity fraud, make false rape accusation. Congratulations on letting every woman who *does* do those things off the hook. Again, if a woman dates a psycopathic criminal because he's exciting, tells rape lies to get personal advantages, or is a gold digger, that says *nothing whatsoever* about women and *everything* about her personally being a piece of shit. Do you actually *want* to give these animals an excuse to hide behind that they're "just being a normal woman"? Aren't you actually *angry* at these behaviours, such thar you want to actually expose and shame the ones responsible for them? (Same question for the feminists.) Is it all an act? Or have you not thought through this at all?
I see a fair amount of this even in the ACX comments. What is wrong with people?
Heh, your piece on academics reminds me on Chomsky. How does being a linguist somehow make him an authority on every war and every political system?
However, a lot of real actual research ends up with political results that tend to lean left. The reason is simple. A conservative is not someone who rejects the philosophical ideal of egalitarianism, but someone who thinks it has been already achieved. So they are not happy when actual researches tell them writing a piece of paper that declares people to be equal actually does not achieve this.
Academic political philosophy is good and IMHO moderate. Rawls can be summarized as "You know taking out an insurance is a rational think? Now think it further."
The BDSM community has strong problems of this. People pull off vile, non-consensual, unsafe things with newbies and then say well BDSM is like this. No, it is not.
You're right (and as Kenny Easwaran says you're just rediscovering racism/sexism/classism/jingoism etc.).
I have to say at this point I've just given up-- I think it's just human behavior and at this point am mostly trying to figure out which tribe to join, since right now they're both kind of bad. Humans are slightly evolved monkeys and we need to stop expecting more from the general public.
This is downstream of groups not policing themselves. The outsiders shouldn't have to do a deep dive to figure out which people in the group are the problem; the group should be making it clear. If they aren't, one assumes they're happy with the behavior. Groups tolerate offenses by their worst members, and therefore are represented by their worst members.
That is a strange take when we talk about large heterogenous groups with no leadership like men, women and academics. Who do you think should police them? Is every man somehow particularly responsible for the action of every other man because he happens to be a man?
There are actually a pretty large number of veto points in the average academic's career. E.g., hiring committees generally need to unanimously agree on which candidate to choose, and a lot of early-career jobs are time-limited, meaning you have to be OK'd by multiple such committees to stay in academia. Whilst there isn't so much formal policing of academics' behaviour, then, in practice people whose views or actions make them unpopular with other academics tend to get weeded out.
Yes but I was specifically broadening it to the effect it has of letting the guilty off by diffusing responsibility amomg their group. I don't think this is very widely acknowledged; I'm not sure I've even heard *once* the argument that e.g. supposing the Nazis were right about Jews having betrayed Germany, by going after all Jews instead of the specific ones who betrayed them they were letting the latter off lighter (than they would have been had they received the full force of the anger).
It seems like "isms" being wrong on their own terms is an important phenemenon separate from simply isms being wrong.
> Yes but I was specifically broadening it to the effect it has of letting the guilty off by diffusing responsibility amomg their group.
(Sans the "Specificity vs Sensitivity" trade-off), perhaps you're complaining about Free-Riders? Or are you looking for a niche term for both phenomena simultaneously. (Which, to my knowledge, doesn't exist. But I suppose you could always come up with one yourself.)
> supposing the Nazis were right about Jews having betrayed Germany, by going after all Jews instead of the specific ones who betrayed them they were letting the latter off lighter
What does that even mean? The "betrayers" died just as badly as the "innocent". Sure, they didn't get all of them, but that was only because they lost. They would have gone after the rest of them if they won. So don't lose.
I'll defend this behavior as rational in the same sense that it's rational to hold corporations responsible for the behaviors of their employees. You're attacking the entity that is best-positioned to punish the behavior going forward. This is harder to argue for diffuse categories like 'men' or 'women', but I think the academia example is a good one. The problem behavior (of particular academics being performatively left-wing, for example) really is downstream of institutional policy. It's therefore rational to criticize the institution in the hope that it will change its policies. A professor has no reason to care if *I* think he's a bad academic, but he has plenty of reason to care if his tenure committee does.
I think I agree, and I wasn't trying to talk about concrete institutions, which I agree can be held responsible for particular acts. I was talking about the fact that a large part of the right seems to have come to hate the very idea of academia itself, to the extent of increasingly holding it as an article of faith that most academic disciplines are basically useless at best. Maybe this is independent of the objection to corrupt and politicised experts, but it certainly doesn't look like it.
And even your instutional point I have a few reservations about. Attacking a whole institution on the basis of certain members' behaviour is probably one of the most effective ways imaginable of getting everyone part of it to rally behind said members.
I think that's a continuation of the same principle. Conservatives have been complaining about institutional bias for 20 years with no results and so they've moved their target up one level to include all of academia. Trump's random and arbitrary punishments are, in a narrow sense, not really defensible. But in a broader sense I would argue that academia's heavily politicized influence on society has made them a ripe target for political hay-making. In my view they have no one but themselves to blame for that. If a country wants to maintain neutrality in a war then it shouldn't run weapons for one side unless it's prepared to be considered an enemy combatant by the other.
> one of the most effective ways imaginable of getting everyone part of it to rally behind said members.
IMO that's only true for symbolic punishments. If you make the attack fierce enough then institutional solidarity will immediately crumble in an "every man for himself" defection cascade. Survival always trumps ideology, especially for soft ivory-tower types. Here's a recent op-ed which illustrates this:
I'm highly confident that views like that (or the willingness to express them, at any rate) are directly downstream of Trump's policies. He's making academics fear for their survival. That tends to drive change.
Oh agreed. Problems are always caused by the shittiest 5%. I think this reinforces the logic of my point. As a non-academic I can't directly target that bad 5%, but as a taxpayer or pundit I can attack the institution in a way that incentivizes it to target it for me.
The last line is the key. "The professors don’t have as much power as many right wingers seem to think".
The point being made here is that it isn't the academics (apart from a very small number of far left ideologues) that are at fault, but the bureaucracy.
Harassing academics has literally no effect on the bureaucracy, which has a completely independent, although largely parasitical, life of its own. It's almost impossible to incentivise the institution to attack its own bureaucracy. They will just pillory and sack more academics, but save themselves.
This is a pivotal fact that seems to have escaped practically everyone.
That's pretty much straight up stereotyping: Judging people by the category they belong to, not their individual behavior.
Of course, to follow the guidelines you propose, someone would actually have to do research and collect some data to demonstrate what the base rate of the behavior they object to is. Most times, that will weaken the argument they are making (very often fatally). So the incentives go the other way, unfortunately.
I agree with you that generalizations like this are frustrating, and that they muddy the discourse.
Thinking through it, though, I'm starting to wonder whether the causality is reversed here. Instead of "people complaining about groups of others unfairly generalize, and therefore guilty individuals get off relatively easy", it's "people observe the guilty individuals getting off relatively easy by dent of their membership, or, are not punished by the groups their a member of, therefore, complainers generalize the guilt not only to the individual, but to the support network"
working some specific examples:
Many cops care primarily about doing a good job and serving justice. A few cops abuse their authority for a feeling of power, even in contradiction to ostensible department policy. An "individual-first" approach would suggest that we should celebrate cops in general, but vehemently prosecute abuses of power. And in a healthy "individual-first" society, abusive cops would find themselves ostracized from their community and without support in the face of prosecution.
In practice, attempts to prosecute even the most egregious cases get mired down in departments rallying around the accused cop, and a broad cultural contingent ready to defend arbitrary actions as "necessary for the job". So instead, while an individual's actions _could_, in theory, be prosecuted individually, instead that guilt spreads over all, because you can't access the individual without going through the all.
Academics are similar. Many are heads-down, quietly working to be sure of their conclusions and the evidence behind them. A few use their platforms to advocate for policies they can't rigorously support, or even to bully those less-platformed. The bullies and the wokescolds _could_, in theory, be dealt with individually by their universities and their colleagues, but in practice, they get supported, defended, and even the most careful academic might lend their voice in support of the bully for fear that the cultural eye of sauron will turn to their niche field next.
Men, is probably one of the more clear-cut cases, where criminal acts would get favorably swept under the rug by a system where other men were in power, and if anything came of it, victims would be blamed. This _may_, possibly, be in the process of shifting, as the social concepts of 'what makes a man, what makes a good man' are re-evaluated, and in the future we might see more internal condemnation of abuses of masculinity.
In all these cases, the thing I'd say is "the sins of an individual haven't, in practice, invalidated their membership in a group, and the group has even provided cover for those sins."
I don't claim that it's therefore true and right for a complainer to generalize the sins of an individual to the sins of a group - but I do claim it's understandable.
Actually, I think the direction of causation may be bi-directional, and not in a good way. Let's say one is a member of "Minority A" (blacks, cops, academics, men, doesn't matter). You see someone from "Majority B" arguing that a member of your group is guilty of a bad thing, and implying that their being an "A" is one of the reasons why.
One could simply point out that while the bad thing is bad, being a member of Group A is irrelevant to that argument. Instead it is in the interest of a member of Group A, who would like to inflate their status within that group, to argue that everyone who complains about the bad thing is actually attacking the entire group, so Group A has to rally around a leader (themselves) to protect themselves. Meanwhile, members of Group B see this, but do not call it out as a weak argument, instead they use the same sort of argument to bolster their own status within their group, to argue that everyone who complains about attacks on Group A is actually attacking Group B, and that to defend themselves, they have rally around a strong leader (themselves). Thus, you end up seeing a kind of reciprocal "choreographed mutual stereotyping" which benefits the extremists in both groups at the expense of the majority of both groups.
Just replace "Group A and Group B" with a pair of groups from the headlines. People attacking the government of Israel for their actions in Gaza are anti-semetic." "People attacking Hamas for their actions in Israel are islamaphobic." Only extremists are making these arguments, because it's in the interest of both sides to bury the moderates, who might actually solve the problem, thereby undermining the leaders on both sides. You can see this is lots of different conflicts (cops vs minority communities, feminists vs mens rights, capitalists vs socialists, etc.). That's what polarization really is--extremists attacking the moderates in their own camp, so that their community will feel less secure, and therefore are more likely to allow themselves to be influenced and manipulated. The interesting thing is the coordination between the extremists on both sides--they each depend on attacks from the other side as a basis of increasing their influence within their own.
Note that I am not arguing that all sides in every political dispute are always equal. If fact they rarely are. But it's in the interest of extremist leaders on both sides to make the conflict (whatever it is) seem intractable because they depend on that to promote their own status within their community.
...now try making that same argument without being guilty of the practice you wish to condemn: making negative generalisations about groups of people like feminists, right-wingers or redpillers.
How about by: observing that political labels are basically self-chosen and if a label gets a connotation or implication you don't support you can always...stop identifying with it. Which puts it in a different universe to generalisations about some innate or simply politically independent characteristic?
(I grant your point on "right" though. I actually think I meant to put either the "often" or the "largely" in that sentence after "the right" instead of where I put it, but I can't prove it and it's possible I'm retrospectively deceiving myself.)
Or alternatively: by noting that my argument doesn't actually fail just because I'm a hypocrite. It just means that I'm being too soft on the actually terrible feminists and redpillers and giving them cover for their actions. Are you disputing that claim or not?
"How about by: observing that political labels are basically self-chosen and if a label gets a connotation or implication you don't support you can always...stop identifying with it."
This is too severe a restriction. I am not going to stop self-identifying as a Christian because some priests molested children, or, I am not going to stop self-identifying as an American because the government did bad things.
You work to improve the behavior of a group you belong to, you don't withdraw from the group (I am not going to stop supporting science because some scientists did bad things).
However, there are people of bad faith who will attack anyone who tries to improve a groups while staying within it, so that's a thing.
I'm pretty sure the word "people" in a phrase like that has a secondary implied meaning of "the people who actually do this",and I was using it in that sense.
EDIT: Oh, are you accusing me of doing the thing I'm condemning? I don't think I am because I was using a figure of speech and:
Are you claiming that the people I'm complaining about mean their generalisations as a figure of speech? And that therefore...the gender wars that take up half of Youtube and Tiktok are all just a misunderstanding?
Or are you claiming that anyone around *here* who uses that rhetoric is just using it as a figure of speech? Even when they're using almost identical language to the first group? (See some of the comments on the recent review.)
Or are you saying something else? Hard to tell what your objection is with such a short reply.
90% of my reply is: would you make the same defence for redpillers?
10% of my reply is: I don't think they're often doing that, because the "#notallmen is privileged derailing" talk comes up when there's some message along the lines of "men need to take more respinsibility for rape culture" that causes people to say "not all men" in response. The latter phrase isn't being used much in response to advice aimed at women to keep themselves safe.
Or at least that's how *I've* seen the discourse play out.
There is a word in your survery (let's call it M) that is rather anglo-centric; this will stop it doing its job for many foreigners. I suggest "M (or your local equivalent)".
Note for future surveys: include question on "are you a filthy monoglot Anglophone or a speaker of a beautiful expressive melodious tongue unpolluted by the bastard dialect inflicted on the globe by the Brits and Yanks?" 😁
Regarding the central topic of the survey, all the articles I’ve seen list only people who have had issues with ChatGPT. I’m definitely getting curious about whether this is just because ChatGPT is so much more widely used than the other LLMs that almost all cases are associated with it, or whether it’s because there are important personality differences between the different LLMs that cause differential rates.
One of the nytimes articles recently took a particular segment of the person’s problematic conversation and found that both Claude and Gemini did almost exactly the same problematic sycophancy as ChatGPT on the one example - but the person did say that they were broken out of the issue specifically by conversations with Gemini, which suggests some relevant different in longer conversations.
I think he's not saying the survey was variable depending on your location, but that knowing you're not American will affect the likelihood of certain answers, so he will calibrate accordingly.
Another property you ask about comes in two variants one more rare than the other. You specifically ask about the more rare variant. I know a lot of people with this property but I would not know which of the two variants they fall into.
According to the 2025 ACX survey, almost 75% of people who read ACX and respond to surveys here are from core Anglosphere countries (US, UK, Canada, Australia, New Zealand, or Ireland) and another 8.6% are from countries (Germany, France, and Israel) where the local equivalent of the M word is identical or almost identical to the English version.
I'm trying to figure out the disconnect between good ideas and implementation in the public sector. There seems to be a real gap in our models for this. We have theory, isolated case studies, and lots of opinion, but I don't see much "work" being done on the failure mechanism itself.
My concern is that any new governance model for abundance, rationality, you name it, is fundamentally constrained by this. Every grand vision can be held up by internal processes and gatekept by an archetype I call "Bill the administrative assistant."
One would think this is an easy obstacle to overcome, but I've seen it consistently derail massive projects. Some of the best thinkers on this, like James Q. Wilson (Bureaucracy) and Jennifer Pahlka, have shown they understand the symptoms perfectly. Yet the disconnect remains. We seem content to complain about it but not actually fix it.
So, I'm interested in insights that go beyond the usual narratives of party politics, capitalism, or doomerish acceptance. I'm looking for better frameworks to understand this specific, gears-level failure mode. What are the actual mechanics at play here? and more importantly how do you improve it?
The so-called "Principle-Agent Theory" guides much of modern research. A principle (in the case of a government, the elected officials) must select agents (agency heads and staff) to carry out their agenda, but research clearly demonstrates that this relationship is complex, given that everyone pursues their own interests: https://doc.rero.ch/record/299898/files/30-5-302.pdf
Fundamentally, this problem has no simple solution, because public sector agencies depend on support from elected officials and the public to ensure their funding and survival, but both groups are divided into competing factions, pursuing incompatible agendas.
This is good information, I don't disagree. I do think a nudge in either direction has significant effects. I don't imagine fixing principal-agent. I do think some agencies do better or worse. It's in that space I'm interested.
Thank you for the information, I had already planned to deliver more into principal-agent and resources are always welcome
There is not a lot of research on this. I worked in the public service space as an HR performance consultant for 20 years, and in my own opinion, the kind of difference you are asking about isn't really in the control of the agency itself. It's due to the leadership of the elected officials they report to (in the case of non-profits, the Board of Governors). The history of White House - Agency interactions testifies well to this: From the Bay of Pigs to the recent firing of the Labor Director, public service agencies master the fine art of pleasing the King.
Ideas formulated to be implemented in the public sector have LOTS of problems.
One is they have to be stated in such a way as to not offend any important group. Another is that those pushing them usually have multiple motives, many of which are not clearly stated, or perhaps even recognized. Another is that they need to satisfy multiple different groups, many of other primary goals.
E.g., I had a friend who would always vote against bond measures, because he held that was an excessively expensive way to implement funding. He might be quite willing to support it if it were to be a tax measure, but not if it were to be a bond measure. I, personally, think it a shame that few people think that way, but it means that many things are always bond measures.
Good points. I do think it's worthwhile, even if ultimately futile. I might be tilting at windmills but I could be chasing worse goals I suppose. The government lacks a price signal, and it also can't participate in a market really. It causes a real lack of information. But we can't realistically just get rid of it.
Let's not ignore the fact that commercial enterprises suffer from very similar problems. The real difference is that when a business is really bad, it goes bankrupt and goes away, while we are stuck with a government forever. Government isn't really any less efficient or effective than the private sector, it's just that the consequences of mistakes are more severe.
Yes, I totally agree. Government isn't allowed to fail. When we point to successful efficiency in business there's survivorship bias, they're visible because they're the small percentage that work at this point in time. Another issue I want to write about is how government doesn't have the benefit of price signal. It's a significant missing piece of information.
> I'm trying to figure out the disconnect between good ideas and implementation in the public sector.
I'm not sure what scale of "public sector" you're thinking here, but for large city or state and larger, Bent Flyvbjerg wrote How Big Things Get Done about large projects (be they building bridges, buildings, IT projects, and more), and why they routinely come in massively over budget and are finished years past the initial projections .
Essentially, there are interlocking reasons behind these budget and time overruns - politicians outright lie even if they know it will take longer and cost more, any complex project has many ways to fail and get worse, but only a tiny amount of ways to succeed / stay on the critical path, and a desire to do things and show progress rather than spend time + money planning things out well.
The big mitigations are:
1. Spend more time planning
2. Act fast and pay more for highly skilled teams, because the longer you're doing things, the more you're exposed to disruption
3. Spend time actually thinking of and quantifying various risks and mitigating them
I wrote a review of his book here if you're interested in a little more detail, but the book itself is well worth a read.
Thank you I'll read it. I had not heard of that book ,but others like it. There is a gap though from the book to the government. Shining the spotlight is the first step, but then what? There isn't a satisfying answer. We can proscribe mitigations, but they're not be used.
-edit Looking at your review, its interesting my intuition about optimism was the opposite. I think we assume government will fail and so take it in stride. I'll have to think about that more
i feel like this falls under "usual narratives of party politics" that I asked to go beyond. I also don't see how deregulation or neoliberalism relates at all.
My point is that, though a lot of our governance failures are because of excessive regulation, a Democrat cannot simply run on "slashing regulations so businesses can create jobs and build more homes." Instead they have to triangulate among a bunch of interest groups.
I feel like you're arguing against a different question. The process questions I have are not regulatory. Also as far as I know these issues are in government divisions with both republican leadership too. Its not a party issue.
Could you elaborate on what you mean by the "bill the administrative assistant" archetype? From reading those authors, I understand the problem to be that *most* of the dysfunction stems from lawsuits and the fear of lawsuits, and the mechanics for that are pretty straightforward.
I'll give a very specific anecdote. drinking water operators need state issued certificates to operated facilities. The range from 1-5 and correspond to increasingly complex and large facilities. Each step requires a test, an application, and a certain amount of experience. One person in Sacramento at the State Water Resource Control Board processes those applications. The person who used to do it was, problematic. She told me she usually rejects most applications on the first submittal. She didn't elaborate on why. The rules she should adhere to are codified in my substack's namesake Title 22, yet she felt she was allowed here own interpretation (she's not). She once told me that she had about 100 applications in her queue and each one takes her "a couple days" , mind you, there no really need for scrutiny here, you either submitted the requirements or not. Should take an hour. The real world result was that in an industry that is persistently short handed, a single person created a bottleneck that had real logistical and financial impact in California. Her supervisor new it was a problem. Every water utility in California knew it was a problem. I messaged the water board, so the should have known it was a problem. She recently retired, and it seems to have resolved it. This is not an isolated case, I've seen it all over.
That strikes me as very odd. I know little to nothing about California, but at the Federal level those sorts of decisions would be open to legal challenges: https://www.nrc.gov/docs/ML1407/ML14077A573
Things that are open to legal challenges and things that are challenged legally are far apart. I understand your confusion, and it part of the basis of my question. There is an underlying sense that this is all illogical. Its persists though, so its tells me my model is missing something. This is where I'm at at least.
This seems to me an example of an interaction between things at two different "scales": the institutional scale, where there are clear rules and procedures, and the personal scale, where individuals with unique personalities and behavioral patterns are tasks with implementing these rules and procedures. Trying to eliminate deviation from formal parameters due to individual differences seems a little like trying to eliminate weather delays--maybe you just have to plan for such deviations in the first place. In any given organization, there are going to be "personalities" that introduce delays and waste. Probably impossible to eliminate, so plan on it instead.
I don't know, I really don't. I will say that the NRC which I have delt with has a great many rules in place in an attempt to limit discretion and a many large volumes of rules, the examination standard alone is 460 pages. (Because of this?) the NRC gets very few legal challenges to its licensing decisions even though they are worth a great deal of money to the people applying for a license.
sorry, what is the NRC? Just for refrence, I get the gist of it. Do you have the sense the method the NRC is using is better for outcomes though? I have a hunch this is an opposite approach that is also a bottleneck
1. People don’t magically start following rules just because someone wrote them on a piece of paper (even if that someone is a legislature).
2. The Water Board isn’t going to fire someone just because service users don’t like them (assuming that’s who even makes the hiring/firing decision). No-one in public sector management cares more about “whether California has enough water” or whatever than they do on actually managing their subordinates. Firing Ethel because of some industry sleazebag’s* poison pen letter to the board is a non-starter.
3. No-one’s going to go through 3+ tiers of public employees to push the message “stop reading the applications, just rubber stamp them” in general. What you want is to specifically reprimand someone. That’ll be a formal process involving their union rep, and there’s no specific misconduct to hang that on.
4. “Ideas don’t matter, people do” isn’t some hippy slogan, it’s the whole of politics.
1. correct, however the options aren't magic or noncompliance.
2. You seem to be reading a lot into the anecdote, that's interesting. FWIW, my letter asked about the rejections rates of applications, the common cause of rejections, and if there's a possible process fix.
3. The options are not (or should not) rubberstamp, or find a way to reject.
4. If you don't like the slogan , why did you bring it up?
The point is that no-one in a position to fix the problem has any reason to fix the problem, and fixing it isn't trivial for them. "Things would be better if they did," and "the law says they have to" aren't reasons to do something that involves significant costs in organisational capital for whomever's going to do something.
I feel like this falls under my point of "usual narratives of...doomerish acceptance"
I'm reading this as restating my question in a snarky way, which is fine. I do this myself. I'm at a point though I want to figure out these things. Why don't the people in position to fix, feel like they don't have a reason to. Also why doesn't the cost to the organization balance with the cost to the goal of the organization.
If we to take the narrow view, America wouldn't even exist.
Also, I am interested in your inside view. You are the person who we need to talk to, not the director or the people strategy person
oh i see. I do understand your point, being overly conservative due to legal and political threat is also a thing. It is somewhat related. For my example, I know the group had gotten in trouble for being too permissive which led them to overcorrect. However it wasn't the motivation of the individual person here as far as I can tell
I do hear "she'll sue", but it's actually not the case. It's harder in general but there are always levels of progressive discipline to follow. Also it could be the case that she capable of doing a decent job and is poorly supported by here supervisor. I'm definitely jaded in this anecdote, but a very real scenario is that the gatekeeper is being rational given their incentives and resources.
The disconnect for me, is that the problem created is sufficiently massive that some kind of resolution should have happened sooner. Even if they fired her improperly, the legal cost would less than the cost to the industry as incurred.
Its honestly weird as having worked inside government bureaucracy everyone involved seems to want the improvement too and just as fed up with how the system actually works. No one likes it except for a few special interests with minimal connection to a policy other than scuppering it.
If it's not explicitly your job, and you don't have blanket authority from some higher-up (in the linked case, from Governor Shapiro himself), you might complain about the system, but you 1) don't really gain anything from fixing the system (at least from a personal profit perspective) and 2) would probably be penalized for screwing something up if you tried any amount of making things better that had a nonzero risk of something going wrong (i.e. doing literally anything). Even if you have a ton of good suggestions floating around (e.g. the GAO has a self-reported ROI of $133 for every $1 invested, which is insane https://www.gao.gov/products/gao-24-107438), the actual implementers have to be incentivized to accept them. But public servants aren't hired to improve their workflows, they're hired to do the job they were hired for. Asking them to do any more means asking them to take on extra risk for no reason except civic duty.
On the other hand, if it's explicitly your job to improve things, you tend to get it done - e.g. Pahlka in the USDS before it became the crapshoot that is DOGE, or the Pennsylvania example I cited. And if you have a mandate you can point to, you sidestep the issue of the skittish bureaucrat who doesn't want to lose their job, because now they have a person higher up on the chain that they can blame if something goes wrong (this sounds cynical but isn't really: again, why would you make a bureaucrat risk getting penalized for something that they weren't hired to do?). This is also why the authority should come from the executive: they have the security of a term length that is usually 4 years, so they can start all of their grand schemes as soon as they enter office, and hope that their EV-positive plays will pay off by the end - I think that optimal play for executive authority means taking all the risks as soon as they get elected and then becoming more risk-averse as they begin to seek re-election.
TLDR: The irony is that you're looking for good ideas on implementation. But those good ideas still are not implementation yet. Hire people in-house to improve your policy implementation, and that's what you'll get.
These are are excellent points. I agree and mention Pahlka as someone who "gets it". So why don't we have mini-Pahlkas poping up? There are instances of an executive at a agency who makes improvements, but it seems isolated. America seems to love copycatting, but I don't see it here. Also there seems to me systemic methods that's could be reproduced without a special executive. Yet, the progress seems to die when that special executive leaves.
I mean, you could argue that DOGE was driven by the public desire to improve government efficiency. If it had been headed by someone else (or if it had just been the USDS but with a greater mandate and more executive authority), maybe it could've actually done so. Instead, well, you know.
The special executive problem is a real one, but it also reflects the positions of the populace (and the problem with representative democracy in general). Elected officials do what they think will get them elected - I'm hoping that Abundance and related recent vibe shifts will bring about a round of efficiency-minded mayors and governors, but only time will tell.
Also, I think efficiency is generally bad marketing. People care more about actual impact than they care about ROI, and for good reason: government is very subject to the "efficiency that doesn't scale" problem. Many implementation fixes are hyper-local and require deep understandings of the problems + the systems we're working in. One of my favorite nuggets from the Pennsylvania case is that lawyers were staffed on implementation teams from the very beginning, so that the exact legal bounds of what was and wasn't allowed was never in question. And afterwards, they marketed themselves by impact: we reduced processing time by xxx amount, we killed a backlog of xxx permits, etc. Much better than "we made xxx process more efficient".
I can't recommend anything beyond "make a team that is empowered to improve processes by finding and utilizing local knowledge", because anything too general ironically might not generalize.
This is a weird time in history. We say DOGE unironically when its the stupidest possible name for the goal. Also I agree with the sense "we" wanted a DOGE but a foreign billionaire did it on a whim. Ramaswamy for all his BS probably had a decent plan (of the plans possible for this administration).
I'm not sure I agree with "reflects the positions of the populace" because these aren't elected positions. The elected do have some influence but for the most part this is the "deep state"
There's two ideas I'm considering. One is that this is all down stream of the country's state of mind and that we either find a way to make it work , or not because of it. Post WW@ we believed we could do things so we did, now we believe we are broken so we are.
The other is somewhat related. Your idea about marketing and generalizing are floating around it. I don't think there is a prescribed playbook to "fix" the government but there could be a culture shift internally that would signal competence in and out of the sector. We say nothing changes, but when woke (i don't have a better descriptor) became popular every government entity reorganized at lightening speed. New positions were created, trainings, documents re-wrote. This as at local, state, federal levels. I point this out because its counter to the narrative that these groups can't change. So what I'm trying to figure out is can we reproduce that, but where it's aligned with agencies stated mission,
"One is that this is all down stream of the country's state of mind and that we either find a way to make it work , or not because of it. Post WW@ we believed we could do things so we did, now we believe we are broken so we are."
I think you hit it on the nail here. The problem you seem to be exploring isn't really located within the procedural arena, it's a problem of the self-fulfilling prophesy at the national cultural level. Something changed in the political culture between WWII and today, and the thing that I think changed was the wave of globalization that struck the US starting in the 1980's. The benefits and the costs of that transformation were not evenly distributed across the US, which broke down a lot of trust in public institutions and other Americans.
This means that the solutions is simple, but challenging: Someone has to persuasively run on fixing those problems, ie, a platform of deep national social reform. This didn't happen, so here we are. We needed the second coming of FDR (except with updated solutions). "We have nothing to fear but fear itself." Convince enough voters and I predict government would become a lot more effective.
Generally agree on DOGE. I think the word efficiency is poison in relation to government now, and any attempt to make it more efficient must come with an ironclad promise of no service cuts. Better government, not less government.
> I'm not sure I agree with "reflects the positions of the populace" because these aren't elected positions. The elected do have some influence but for the most part this is the "deep state"
Sure, but the point of the Pennsylvania example was to illustrate that Governors (and probably Mayors) have the power to effect change if they want.
Also, I wouldn't call "woke" an internal culture shift. I think that was something that started outside of government - I'm not sure if you could do things the other way around.
what's weird is everyone feel beholden to rules they just made up for themselves. Then a bad actor changes one, everyone complains and then feels beholden to that. It like some kind of organizational freeze tag.
Question for native English speakers. English is a highly analytical language, contrary to German, which is very synthetic, where much of the meaning a word gains from its context is backed into the syntax.
Please read this sentence from the wild:
"Suppose someone was, unknown to themselves, 'enlightened' as you and the people you mention are using the word, but not through any practice thought of or labelled as devoted to achieving 'enlightenment'."
I stumbled at the word "thought".
I wouldn't if nouns were capitalised at all times. I would have known it's a verb.
I wouldn't if you conjugated verbs thoroughly, for the same reason.
I wouldn't if you had more rules for commas. I would have known a new clause had started.
I wouldn't if you didn't drop "that" so often, for the same reason.
Uncomfortably I read on, saw the "of", and got what's going on. How do sentences like this go for you? Don't you stumble too?
I stumbled more at the line break between "are" and "using the word" since "as you people are" has its own different meaning. I basically read "thought of" as one word, the same way I would read "thoughtful". Likewise "labelled as" and "devoted to"; those short words just get lumped into the words in front of them.
I don’t think English and German differ much on the analytic-synthetic continuum - both have nearly the same constructions of verb endings, compound verbs that embed prepositions, compound nouns, and so on, though German has a writing system that uses fewer spaces in a lot of these, as well as capitalization (and also has a bit more remnants of grammatical gender and even case). But neither of them is anywhere near as far in the synthetic direction as Semitic languages with their triconsonant stems, or in the analytic direction as Chinese, without conjugations at all.
Of course the languages you know of might be much more synthetic, yet the difference in syntheticness between English and German is immense for me.
I tried learning Spanish.
And I gave up when I came to subjuntivo, something that's equally extreme in German, and it was then that I realized what a simple language English was. Memorizing English irregular verbs was a breeze in contrast to having so many different regular endings in all tenses. Speaking spanish was like mental arithmetic for me. I have now enormous respect for people who learn German as an adult.
The only complicated thing about English is that there are no rules connecting speech with writing. Very annoying. Much worse than German. And Spanish is the best in that regard.
"I have now enormous respect for people who learn German as an adult."
15 years in Austria and my grammar still sucks, and I still write emails with a dictionary open to look up Artikels.
It is no surprise that English took over the world. The ability to just read a random text, figure out a word from its context and then immediately be able to use it is phenomenal.
Imagine living in the Middle Ages and using Latin. It has no fixed word order, adjectives do not stand in front of or after nouns. Madness. We don't know whether the cult of Ostara was a thing at Easter, because we don't when reading Bede that "quid nomen", "whose name" is the name of the goddess or the name of the month.
In my view that's a slightly convoluted sentence. I would give editorial feedback to write more clearly and succinctly. I wouldn't say that it reflects problems inherent to English per se but rather illustrates the importance of learning to write well. I don't know German at all but surely one can write badly in that language too? If it's harder to write clearly in English then maybe that represents a tradeoff between expressiveness and complexity.
The example sentence might be bad writing, but the problem I have regularly when reading english is suddenly not getting where a word goes meaningwise, because that is to be found in the context -- no hints in the declination or conjugation or capitalisation of the word itself.
Also those whitespaces in the middle of terms and not so many obligatory commas between clauses do not help.
Erica Rall gave the "The old man the boat" example and said "It 's a real problem, but a relatively uncommon one in ordinary usage."
It might be uncommon, but compared to German it's common.
So I think one can write correctly but nonetheless badly in German too, but has fever ways to do so.
Not really? After "any practice" I expect some sort of adjective-like-thing to qualify it. A past participle like "thought of" works fine, but 'thought' as a noun would not.
"Thought of as" is also a very conventional phrase for the concept of "is generally considered to be"; not quite an idiom, since it literally makes sense too, but common enough that "thought of" parses as a unit here.
I finished my review of Mind and Cosmos by Thomas Nagel; brought up in last open thread, but just published now.
https://shponsevonshpoon.substack.com/p/review-mind-and-cosmos
https://www.yahoo.com/entertainment/articles/hit-songs-lasting-longer-charts-151226254.html
> Hit Songs Are Lasting Longer on the Charts – But Why?
> by Jason Lipshutz
n o m i n a t i v e . . . . d e t e r m i n i s m
I am working on an idea. So imagine Auschwitz is liberated, and the enraged prisoners take the gas chambers down, or an enraged officer blows them up, or the Nazis were destroying the evidence. And then the camera crews come and nothing to show, and then someone rebuilds them for the camera, badly. Not saying this happened, just saying this is the kind of imperfect humans making mistakes that keeps happening.
So someone with spergy superpower notices the doors do not fit and then not having a good theory of mind, does not think of a normal story like this, but a huge conspiracy.
I think this keeps happening, because I see former goodie spergs like Star Trek fans now believing in huge conspiracy theories.
With a normal theory of mind skills, you would assume a true story will have still bad parts, and you have to look at the whole thing. Imagine how could happen, that a huge crime is done, then covered up, then uncovered and investigated and presented, and somehow the amount of made up crap is exactly zero. This does not happen. Humans do not work that well or that honestly.
Similarly, surely someone made COVID mistakes. Also surely crap was made up to cover it up. But the whole picture matters.
Whereas our spergy type expects either a 100% true story with every detail just right, or a complete utter lie.
I don't know whether you're being humorous in framing it as a hypothetical, but in case you aren't, that's precisely what happened: the famous gas chambers at Auschwitz ARE reconstructions. But it's fine because the Soviets were responsible, and they're renowned for their commitment to preserving an accurate historical record for posterity.
I get it, but I think you're giving the Holocaust deniers too much intellectual credit. They don't like Jews and are looking for excuses.
Now the COVID thing...yeah, sure. Especially as nobody really knew anything in the early stages of the pandemic.
Seems to me like there's different axes down which you could examine that phenomenon.
- Increasing distrust in domain expertise - some justified, some not.
- 'Communities' that select on personalities which reflexively reject consensus.
- Standard internet openness to ideas that overweight on 'interesting' relative to 'accurate'.
- Tendency to accept bad ideas that align with unshakeable (or even bigoted) priors.
- Proliferation of information out there to cherry-pick; an operator can artfully string a trail of cherries leading to a secret door (ex: Darryl Cooper).
>- Increasing distrust in domain expertise - some justified, some not.
I think that's probably the big one. When you believe someone is generally trustworthy, you can accept them making a few honest mistakes because hey, nobody's right 100% of the time. When you believe someone is lying to you for personal or ideological gain, you're much more likely to react to a false statement with an "Aha! I *knew* he was full of BS, this just proves it!"
Quest Stage 1, possible sol'n: "A Good Dating SITE/App*"
I have 2 suggested tweaks to what Scott described:
1. Default "first thing people see of you": Females upload a picture for their profile. Males upload writing for their profile. Non-binary users can choose.
2. THEN, once someone clicks on the profile, it goes to a max of, say, 8 further clickables–images, videos, more text, …maybe even Anthropic AI artifacts if we get clever enough. :)
So you can watch a video of the guy winning the hardest level on a computer game or feeding oatmeal to his toddler nephew or showing off his belt sander! To leverage the whole "Guys present themselves well when they are 'in action' thing." (iirc, YT was originally envisioned as a site where people uploaded videos introducing themselves for potential dating-matches to see.)
I am majorly "sold" on the other options Scott mentioned... (relationships being knitted into friendship networks, ability to see people's answers to questions you've asked them before you even have to "make a move") and there might even be room for more than one app in this space.
Of course, the million-dollar (really billion or trillion, depending on which metric you're using) questions are... what about enshittification? What about entities that are ALREADY incentivized to stop such a venture cold before it even got to the level of network effects.
For that, we would first need something like a faceless distributed communication network full of intelligent anonymous users who were highly motivated to solve this problem and have a reasonably high commitment to each-other and their organization. If someone would make the moves to build such a network - or already has - then I think it's possible!
Now - what am I missing?
* Scott's language was "Site." I added "App."
I thought Youtube was supposed to be for comedy?
That's not what i heard; do you have a link? (RE: YT originally created for comedy.)
<Now - what am I missing?
You are missing the fact that smart self-respecting women will be offended by the asymmetry in what aspect of self the app obliges them to be represented by in users' first glimpse. I simply cannot understand why this did not occur to you. WTF, buddy?
Naahh... I orig. had a mega-comment, but then chopped it down to size; I'd accounted for it in the most obv way. (i.e. u don't need to stick to defaults.)
Anyway, pretty sure REAL women would LOVE browsing through guys' favorite quotations or whatever!
My god, what are charmer you are! I don't know why you're even interested in dating apps. All somebody like you needs to do is to speak right up to a woman the way you are here and she'll yank her panties right off and mount them behind your ears and across your face like a covid mask.
Soo... I kind of dissembled during this discussion. I think what I did must fall under the broad category of what the kids these days refer to as "trolling." Maybe I can explain thusly:
> Be me.
> Actually get along really great with basically all the women I meet!
> Do NOT, however, have sex with any of them!!
> Not because I'm gay, but because of my choice in sexual ethics.
> (which dictated I was just going to be with just one person, a male, til death do us part)
> But at least I got to / get to gestate, give birth to and raise a couple of nerdy children! :D
> Send first one off to college a few weeks back.
So, yeah... that's actual_me.
(And no, I don't actually think that women who wouldn't read long book quotations in a potential male date's profile are not "real women"! That was just online sass!)
I don’t see the point of this unpleasant deception, even for you, but don't even want to hear the explanation. I now have no interest in your gender role and sexual ethics issues and why the fuck you played out this little game. Your post just used me and the others who engaged as a little cell culture for your experiment. You could hardly be further from good faith posting. Yuck, lady.
Well, you could say he didn't talk to any women, but that's exactly what he DID by posting here and getting a response from you, so...
no joke; I am clearly knocking things out of the park today. (or... 5 days ago when ppl were still posting comments, to be exact.)
Hehe.
Seriously I'm with Performative Bafflement...don't waste your time and effort on dating apps, the market is seriously saturated and there's an existing player with near-monopoly power.
> Of course, the million-dollar (really billion or trillion, depending on which metric you're using) questions are... what about enshittification? What about entities that are ALREADY incentivized to stop such a venture cold before it even got to the level of network effects.
This is probably the crux of the problem overall - I know of at least two commenters in the ACX commentariat that have tried to replicate "old Okcupid," which is broadly seen as the peak of dating websites here, and neither ever got off the ground.
I'm not sure why this is, but I'd suspect:
A) monetization problems, because now you're competing against the fully adversarial Match Group (which owns essentially all dating apps except the nigh-bankrupt Bumble) and their established footprint, marketing, and revenue models that will always outcompete a less adversarial option, and
B) the type of person whom "old Okcupid" appeals to - ie filling out a bunch of questions and writing and reading lengthy profiles instead of swiping - is such a tiny slice of humanity overall that you simply can't get the network effects to have a viable dating app, because recruiting enough people in an ongoing way is the hard problem of dating apps. Previously, in the golden Natufian age of old Okcupid it could work because that's all there was - swipe apps didn't exist, so enough people would actually answer the questions and write / read profiles. But now swipe apps exist, and they're strongly preferred over reading / writing by enough people you just can't get a large enough nucleus of people and ongoing flow to sustain such a dating app now.
Okay, I'm going to try to steelman a version of the usage of the imagined Dating Site/App* versus what you have in "B" above:
1. (idea i got by thinking about an issue you brought up) Encourage people to potentially use Voice-To-Text their answers? i.e. fix the problems brought to us by our technology (the computer-in-our-pockets proliferation) with... more-sophisticated technology!! (software that knows the spoken word!)
2. You say "not enough people want to push against the grain." Okay, but how many of the most-intelligent people would run to this if it existed? (I think it could be very worthwhile to invest a lot in the fates the tiny slice of humanity that the most-intelligent - say the top 0.15% - NOT because they are somehow more worthy-of-existence or "inherently-valuable" - no! But because if lots of them are better-flourishing, they'll a better crack at fixing lotsa problems of the other 99.85%.)
* Wondering to myself what name I'll call [Imagined_Dating_Site/App], I'm thinking at this point should be humorously called, "Built_on_Skyhooks."
> Previously, in the golden Natufian age of old Okcupid it could work because that's all there was - swipe apps didn't exist, so enough people would actually answer the questions and write / read profiles. But now swipe apps exist, and they're strongly preferred over reading / writing by enough people you just can't get a large enough nucleus of people and ongoing flow to sustain such a dating app now.
What if the people who prefer swipe apps prefer dating partners who also prefer swipe apps, and the people who prefer reading/writing/questions prefer the same in their dating partners?
Then, creating an app like old Ok Cupid would still work (at least in largish cities). In fact it'd work better as it's selecting for a certain type of people.
> What if the people who prefer swipe apps prefer dating partners who also prefer swipe apps, and the people who prefer reading/writing/questions prefer the same in their dating partners?
Because then your general problem is "your marketing still costs just as much, but will only appeal to ~10% of your audience," for a 10x in marketing costs, against a powerful and well funded incumbent.
And you need continuous, ongoing marketing, because dating sites that actually work pair people up and they drop off, so you need an ongoing flow to be relevant and have enough people on your platform to be usable and making revenue.
So your costs are 10x, you have to do them all the time, and your revenue will always be less than the fully adversarial Match Group - overall, that's not a recipe for a business that will last, it's basically a charity, and an expensive one with poor impact per dollar at that.
Are adverts the main thing pushing people towards websites (either dating website, or anything else)?
> Are adverts the main thing pushing people towards websites (either dating website, or anything else)?
More or less - at some point you get social proof and word of mouth / network effects that help draw people there, much like "Tinder" became an eponym for dating apps overall in some circles for a while, but largely you need constant marketing to build to that level, and then marketing to keep your population up, because one of the main problems is that if you don't have a high enough concentration of people in each geography you want to operate in, people will install the app, go through your ten or hundred people, run out of profiles, sputter, then uninstall the app because nobody is there.
Versus Tinder, for example, which has tens to hundreds of thousands of candidates in any real city literally worldwide, and who are your literal competition, so you need to make a decent showing against them, which requires fairly massive marketing spend, and again if you're doing your job right, people are pairing up and dropping off the app, so that marketing spend needs to be continuous, all the time.
> you need constant marketing to build to that level, and then marketing to keep your population up
I've never seen an advert for Reddit or Substack, and their are popular websites. So I doubt that advertising is necessary.
> because one of the main problems is that if you don't have a high enough concentration of people in each geography you want to operate in, people will install the app, go through your ten or hundred people, run out of profiles, sputter, then uninstall the app because nobody is there
It is for reasons like this that I think a new dating site should not just be a dating site -- it should be about other stuff too, including building IRL/AFK communities.
> YT was originally envisioned as a site where people uploaded videos introducing themselves for potential dating-matches to see
That's something I didn't know
> So you can watch a video of the guy winning the hardest level on a computer game or feeding oatmeal to his toddler nephew or showing off his belt sander! To leverage the whole "Guys present themselves well when they are 'in action' thing.
Could make this more general than just a dating app... kinda like Facebook but not shitty.
Also, facebook got started by being seeded in various universities, so if this new thing could do something similar and get concentrated levels of use in some geographical areas.
Or, with intentional communities, e.g. a Christian dating app/site with a Christian ethos.
Or allow people to create their own intentional communities (like a subreddit).
An interesting new thing on Substack lately is, expert demographers pushing back against the prevailing population-projection-statistics narratives. The MSM seems not yet to have noticed this but perhaps will at some point.
It seems coordinated, making me wonder if there was a "oh FFS I guess we're going to have start going online" meeting amongst the world's top demographic researchers.
The demographers are consistently repeating basically just a single point: that the way in which population projections are done "works well when things are changing slowly and in predictable ways, but it means our projections are often limited to forecasting a future that’s a lot like the present." Meaning that in an era in which birthrate patterns are changing in some significant way -- e.g. the age range in which women have children expanding throughout the developed world in a way never seen before -- statistics such as "total fertility rate" lose a lot of their predictive power.
Anyway this week's example is this short "briefing paper":
https://sites.utexas.edu/contemporaryfamilies/2025/08/20/population-brief-report/
A historical bit that was entirely new to me is: "...much like today, in the 1930s and 1940s many commentators bemoaned the low fertility of the time. Expert projections of US population from the 1930s assumed that fertility would remain relatively constant into the future. They [such as the official Social Security Administration projections] imagined fertility rates would continue as they were indefinitely, yielding rapid population aging and slow population growth. Yet fertility did not remain low – instead we got the Baby Boom."
>A historical bit that was entirely new to me is: "...much like today, in the 1930s and 1940s many commentators bemoaned the low fertility of the time. Expert projections of US population from the 1930s assumed that fertility would remain relatively constant into the future. They [such as the official Social Security Administration projections] imagined fertility rates would continue as they were indefinitely, yielding rapid population aging and slow population growth. Yet fertility did not remain low – instead we got the Baby Boom."
Within a generation it was back where it was in the 1930s. The long-term pattern is that fertility rates that decline, not increase.
Over the very long term, sure. But "will always keep declining" is an assumption/prediction not a demonstrated reality.
The demographers' core point is that what we all call the fertility rate (technically the "total fertility rate" or TFR) is not an average of how many children each woman has had. Rather, it is a _prediction_ of what an average woman _will_ do if nothing changes or is changing.
If a nation's current TFR is say 1.7 that's saying that if the most-recent age-specific fertility rates are exactly repeated in that nation, the average woman will have that number of children. It is often summarized as simply "the number of children that women in that society will have", without the important caveat of "if nothing is changing".
Many people -- including me until recently -- assumed that TFR was roughly equivalent to the average of actual children being born per woman. But in the US and most of the developing world it currently is not equivalent, because women are waiting longer to have their children and are having some of their children later than was previously normal. The TFR inherently lags behind a broad shift such as that one.
The actual-results statistic is called "children ever born" (CEB) which counts exactly what it sounds like: how many children the average woman who's completed her fertility years, had. That average in the US is currently just a hair under the "population replacement level" rate of 2.05, and in recent years has risen slightly. Overall it's been pretty steady for a while now, has not gone above 2.1 nor below 1.8 since at least 1990.
So the current US TFR of 1.6 is therefore _not_ reflecting the actual current reality of US demographics, something that until recently I was unaware of. It will gradually get into alignment with the actual-births average (CEB) as updated annual real-results data rotates into the TFR methodology, but that is a gradual process over a bunch of years.
The problem with comparing CEB to TFR is that CEB lags TFR by the difference between average maternal age at birth and the age cutoff for the end of a woman's reproductive years, which seems to be about 20 years (average maternal age in the US is 30ish and I think the cutoff used for published CEB stats is age 50).
This lag explains 100% of the current gap between CEB and TFR, as TFR for the US in 2004 was 2.05 and rising gradually towards a peak of 2.12 in 2012.
You can currently find CEB numbers both for a cutoff at 44 and for 50.
Anyway the demographers' point isn't trying to choose which stat. Rather it is that people are misrepresenting what the one stat tells us while ignoring the other stat. Leading to a collective freakout which the demographers say isn't entirely justified.
My point is that once you take the time lag into account, the two stats are not actually telling different stories. It's possible that the current decline in TRF is an artifact of parents having the same number of kids at older ages, but CEB being 2.05 while TFR is 1.7 is not evidence for that hypothesis. The current CEB is telling the same story as TFR about birth rates 15-20 years ago, and we have to wait another 15-20 years to find out if CEB and TFR tell the same story about birth rates today.
Yep. The demographers are not writing that CEB today _proves_ a different current story regarding fertility rates. Rather they are pushing back against the widespread misuse of TFR as being a final word on the issue.
TFR is a prediction which, like all predictions, can be proven by events to have been either right or wrong. "We should be *very* humble about our ability to predict trends in human behavior more than fifty years out, especially about something as personal as childbearing...." Media coverage and social-media blather about the fertility issue consistently reflect zero awareness of this.
The demographers do seem to be pretty sure that the current decline in TFR is in _part_ an artifact of parents having the same number at older ages. Whether it's a big or small part, is something we will learn over coming decades.
Celebrity Millie Bobby Brown adopted a girl recently, and the replies on Twitter are full of Rightists telling her she's not a real Mom and the kid isn't really her child:
https://x.com/DiscussingFilm/status/1958565213285872083
This is not a new thing, every time some adopts a child these people are out in force. I want to note that Brown is a married, heterosexual woman. Both she and her husband are white. The only "weird" thing here is that she's only 21 years old, leading to dumb comments from people who aren't aware that young women can be infertile.
I advise conservatives to take inspiration from the way Democrats treated the woke movement: tell people that nobody's saying the thing they're clearly saying, say it's all just a dirty rotten strawman, and absolutely under no circumstances condemn or disassociate yourselves from such views, because doing so means conceding that the Left is right about something and is thus the equivalent what Benedict Arnold did when he betrayed his country. Your goal should be for everyone to suspect that you secretly hold such views, so they'll know to vote for you come the next election, and then when you lose you can complain that people unfairly held you responsible for stuff "random people" posted on Twitter.
And how did that work out for the Democrats? If you can win by being honest, then you should just be honest. Republicans are fulfilling a very real demand. People are tired of pretending.
>People are tired of pretending.
Pretending what?
Pretending that things that are false are true, and pretending not to have beliefs that they do actually have. It's unhealthy to keep all of it buried away. It'll all come bursting out eventually...
We all understand in our heart of hearts that those abandoned goods are lesser than our own blood-related offspring. We just can't say that, because it would be politically incorrect. In fact, looking at the comments to that tweet, it seems the sentiment expressed is not condemnation, but bewilderment. They can't understand why someone would take such a blatantly worse option when an alternative is available.
(Well, actually, most of the comments do seem to be supportive of her decision, so I don't really know what you're so angry about...)
>We all understand in our heart of hearts that those abandoned goods are lesser than our own blood-related offspring. We just can't say that, because it would be politically incorrect.
It's true that adopted kids are more likely to have low IQ and other undesirable traits.
>In fact, looking at the comments to that tweet, it seems the sentiment expressed is not condemnation, but bewilderment. They can't understand why someone would take such a blatantly worse option when an alternative is available.
Yeah they don't know that young women can be infertile, implying that they have a higher-than-average chance of being adopted and low-IQ.
>(Well, actually, most of the comments do seem to be supportive of her decision, so I don't really know what you're so angry about...)
Good to know that you're taking my advice. Hopefully this helps Vance's 2028 chances.
I would like to ask about this huge men-are-not-alright discussion, what is that mysterious thing men should learn from their now missing male role models? Being a dude is working with objects, doing stuff with your hands. Our problem is that it is simply not enough anymore.
Back when women had worse economic options, and men better, the social contract was that men build women a house and pay for its costs and so on, and women will have sex with them. Talking about things like emotions was not a thing.
For example my father entirely did not understand emotions. If you feel bad, you have a problem. Then let's talk about solving the problem, and not about the topic that feeling bad in fact feels bad, he considered that ridiculous.
Now women just don't take that kind of social contract anymore, and want men to act like therapists, or women, and do stuff like emotion talk.
So many men struggle, and others are simply checking out. I am checking out, because I also think emotion-talk does not need to be big. I think society is like a woman now, huge emotion talk instead of simple practical solutions or just sucking it up. You got a thorn in your foot, the rational thing is to discuss how and who will remove it, not to talk about that yes, it indeed sucks. So no, I am not going to play family therapist, and if playing family therapist is a condition of modern relationships, then I am out.
I think a lot of this is downstream of the modern ideology that gender is a social construct. There's no sense that "Men are more results-oriented whereas women are more feelings-oriented, so go to your husband if you want a solution and to your girlfriends if you just want to vent," because the idea of men and women being different is verboten. Hence women expect men to essentially be slightly bigger women, and get angry when this doesn't happen.
So, up front, I think a lot of this is sorta your own personal venting, which is totally fine but not necessarily the basis to draw broad spectrum claims about 50% of the population.
I mostly disagree with the substance of what you're saying -- that the social contract has changed, people just want therapists, etc. Second-wave feminism began in the 60s, and women were in the work place in force by the 80s. My wife's mother was a high ranking lawyer working for a bunch of banks. Multiple generations have figured this out. There are things that are new now, but it's not any of the things you are pointing to. Fight Club came out in 1999 and even then the 'missing male role models, society is too feminized' framing was already cliche.
I don't disagree with the frustration you're expressing, it's a real frustration. But I also think you are sorta answering your own question. You ask "what is that mysterious thing men should learn from their now missing male role models?" and the answer is something approximating "be the man your community needs you to be." Any gender essentialist "men work with their hands" stuff is just excuses and cope. That may be what you *want*, but that's not what a male role model is going to teach you (or at least, not a good one anyway).
More generally, I think you are looking at the 'object level' frame. You are thinking about behaviors and actions. You should be thinking at the 'meta level' frame, about values and goals. No one can fully pin down 'masculinity', but if I had to try I would say something like "masculinity is about making yourself useful to your people and deriving self satisfaction, joy, and fulfillment from your shared successes." Every male role model I've ever had taught some version of this. I mean, the Boy Scouts motto is literally 'be prepared', what do you think that preparedness is for? And it's important to note that 'usefulness' varies a lot based on time and place. In some times and some places, being useful meant building a house and being a stoic. And in other times and places, being useful will mean playing with your kids and changing diapers and, yes, being a 'therapist' for your partners.
It's a bit silly but I really do think someone like Aragorn from LotR is a great example of masculinity. He's whatever his community needs him to be -- ranger, advisor, diplomat, general, king. He inspires people by listening to them and gassing them up. He gets his hands dirty so others don't have to. He has tender moments with a lot of the main cast, and he's also a badass who is feared by enemies and respected by allies.
I don't know if you can learn this form of masculinity at ~50. But that's at least a version of what I strive for, and it's been extremely fulfilling.
> what is that mysterious thing men should learn from their now missing male role models?
I think that it is mostly about nonverbal learning. You observe what other people do, and before you notice, you start doing similar things. When boys have fathers, they copy their fathers. When boys don't have fathers, they copy each other, the immature learning from the immature.
> Being a dude is working with objects, doing stuff with your hands.
That simplifies things a bit too much, unless "doing stuff with your hands" includes typing on the keyboard. But thank you for a good example... someone whose father was e.g. a professor *wouldn't* say something like this. Which would be a good thing if we don't want boys to keep dropping out of schools.
I agree that the women's version of therapy is just endless chatting with no results. I am pretty sure that the famous historical psychotherapists did something quite different, but people today mostly take their ideas from TV.
But it's not like men are entirely nonverbal. Some of them need to get drunk first, but drinking a few beers together - and talking - is a traditional male activity.
>But it's not like men are entirely nonverbal. Some of them need to get drunk first, but drinking a few beers together - and talking - is a traditional male activity.
True, but their talk doesn't generally consist of ruminating about their feelings. Even where there's a feelings element to it -- "If feels like my cheating GF just ripped my heart out and threw it in the trash," or whatever -- the reaction's more likely to be "That sucks, what can we do to fix it?" than "That sucks, do you want to talk more about how that makes you feel?"
> "If feels like my cheating GF just ripped my heart out and threw it in the trash," or whatever -- the reaction's more likely to be "That sucks, what can we do to fix it?"
I would expect something like "yeah, women are disloyal bitches, let's drink to that". Which *is* a validation of his feelings, by the whole group, but concise.
Nonverbal learning is a great idea, thanks. I would say it is not simply absence or presence. I think 30 years ago at 17 I was much set by back the conviction that my parents are cringe-worthy unfashionable uncool people, who understand nothing of the modern world, and really would not imitate them.
This was back then common normal model of teenage rebellion that was happening since the 1960’s, and a mild version, the hardcore version to do the opposite of my parents want because fuck them. Get hooked on drugs etc.
I am not sure how much damage it done, but probably a lot.
At 17, it is too late to learn; that is the time to rebel, to believe that you can do everything 100x better that your stupid parents. So that a few years later, after a few failures, you realize that their task was actually way more difficult than it seemed from outside, and that considering the circumstances they actually did a pretty good job, so you can now humbly return to ask them for advice.
The important part of learning happens before puberty, while your parents are still magical godlike authorities, so you uncritically and unconsciously copy whatever they do, good or bad. (Or if you don't have a father, you just uncritically copy your mother's attitude towards men.)
“When I was a boy of 14, my father was so ignorant I could hardly stand to have the old man around. But when I got to be 21, I was astonished at how much the old man had learned in seven years.” --Mark Twain
I basically agree with you. They want more feelings talk, most of us aren't good at it, so increasingly both parties will just walk away from each other.
Probably whoever's good at convincing the opposite sex to breed with them (in either direction) will be what the next generation will look like.
> I would like to ask about this huge men-are-not-alright discussion, what is that mysterious thing men should learn from their now missing male role models?
I think many folk in the Dating review reiterated this point - isn't it essentially that stated and revealed preferences don't match, so you need to ignore the words and find and focus on the things that actually drive mate value for men as per real results?
As an aside, I've never really understood why this is such a problem, because it's a pretty tight feedback loop and should be relatively easy to optimize - you're either succeeding with your dating profile or in-person date suggestions or not, you're either succeeding at in-person dates or not, and in either case you should change things and A/B test into the stuff your desired women actually want, because you're getting feedback right away and have powers of observation and note taking and doing better.
I think a secondary point is probably that you need to find somebody who's compatible with your relationship and communication and conflict resolution style, and per your emotion talk point, there's far fewer women who are fine with stoic and strong silent types. Previously there were more women who'd accept this, and now there's fewer.
Not just that, but previously most women wanted / needed a man in their life, for economic and social and other reasons, and now don't due to having education and good jobs themselves, and rivalrous alternatives like "pets and apps" being good enough for some, without necessarily needing male companionship enough that the benefits outweigh the costs.
You say you're going to opt out entirely - that's fine, many women do the same thing, per my last point.
But you know, you can find a woman who wants / needs less emotion talk too, it probably just takes longer, and may involve subcultures you find distasteful (trad wives, religious women, dating overseas, etc). But it's a big world out there - most things are possible, even if some things are taking more effort than they did historically.
>As an aside, I've never really understood why this is such a problem, because it's a pretty tight feedback loop and should be relatively easy to optimize - you're either succeeding with your dating profile or in-person date suggestions or not, you're either succeeding at in-person dates or not, and in either case you should change things and A/B test into the stuff your desired women actually want, because you're getting feedback right away and have powers of observation and note taking and doing better.
I think that's where the whole social media/creep discourse element comes in. You can't learn on the job (as it were) if awkward attempts at flirting get you labelled a sexual predator, and you certainly can't if your attempts get broadcast through your acquaintance group so any girl you might approach already thinks you're a creepy loser.
At the local mall, between the ice cream shop and the kitchenwares store, you'll find a bare stretch of wall some two meters long. In the center of this wall is a small metal box with a bright red button. Pressing the button does nothing.
You are invited to add a prominent sign above the button, and your goal is to either minimize or maximize the number of times the button gets pressed per day. What will your sign say?
Minimize: "Fire Alarm"
Maximize: Printed, "Press to Skip Song", and handwritten in pencil below, "Button is flaky, try pressing again if it doesn't work the first time."
Seriously? The way to make men push it a lot is to promise sex or money, to make both women and men push it a lot is to offer money, and the way to make them push it less is to offer death. What is supposed to be the interesting mystery about this?
An adult is going to press it once, see nothing happens, and get bored. A child is going to press it five hundred times, because why not?
The more extravagant the promise, the harder it is to make people believe it. We could put up a sign telling people they'll get a million dollars for pressing a button on a wall. But few would believe it, because the offered reward is entirely out of proportion to the effort required, and is of a highly unusual nature. People would be skeptical, and rightly so. The same applies the other way around, with threats.
This suggests we need something more complicated. A vaguely plausible promise that someone won't consider themselves stupid for having spent a bit of effort on, even when if it doesn't pay off (as it won't.) Or just something eyecatching that will have people pressing on a sheer lark.
"Apply electric shock to neighbor's barking dog".
"You must be this tall to press the button" with a big line just above the average height of a ten year old child.
To minimise, maybe a sign that says "Warning: Pressing this button will cause the store radios to play Baby Shark on loop for the next 8 hours."
"This button does nothing." - probably gets a lot of button presses by drawing attention to it but making it harmless.
"Do not press this button" - a social experiment to see if people are conformist or rebellious.
"WARNING, DO NOT PRESS THIS BUTTON EXCEPT IN AN EMERGENCY" - likely very few button presses.
"DANGER - HIGH VOLTAGE, DO NOT PRESS" - probably fewer presses than the emergency button.
In the us:
'Fuck trump button' in a democratic state
'Fuck Biden/Kamala/Obama button' in a republican state.
' Epstein didn't kill himself button' would probably work in both.
Bonus points if you can make it so a number goes up when pressed.
I'm not sure how to maximize the number of button presses, but I suspect we can minimize the number of presses by ignoring the button completely. Have the sign offer directions to the nearby street, or the food court, or the largest businesses located down the left and right hallways. The less fuss we make about the button, the better.
Ignoring the button will still get the button pressed by children; children love pressing things. And if you're right next to the ice cream shop you're going to get a lot of them.
"Press for Emergency".
For maximizing: "Strength Test".
Homeland security out there addressing that decline of the West thing:
https://x.com/DHSgov/status/1956077756854186191
That reminds me. Ya'll gotta check out Benjamin "rap isn't music" Shapiro's recent debut [0].
What an odd timeline we live in.
[0] https://www.youtube.com/watch?v=5kGpohEpuTE
He's got pretty good flow, actually.
yeah. On one hand, I thought the lyrics felts a bit forced, so I don't think it's winning any awards. But on the other hand, I've definitely seen worse. <looks at Ice Spice>
It's like the dancing bear, the surprise is it dances at all.
Shapiro's primarily a showbiz guy at heart, so I bet he actually could be a pretty good rapper if it suited him for some reason. But his audience is center-right guys who don't like rap, so it's unlikely.
Random thought--
If LLMs are supposedly so intelligent...how come it takes painstakingly detailed prompting to actually get them to do useful things? Why is prompt engineering even a thing? Why can't they do what people do and understand the intent behind the words?
Or am I working from outdated information? Are the newest engines better at taking unstructured, loose prompts and giving detailed responses that show understanding of the intent?
If you have to be a subject-matter expert to properly prompt an "AI"...then that drastically limits how useful they are.
i feel like your question is pretty vague and open-ended. it might be helpful to others to share some concrete examples, if possible.
As for me, i've begun using Sydney every once in a blue moon, and she handles simple queries pretty effectively imo.
might be worth reiterating that people call them "stochastic parrots" precisely as a reminder that pattern-matching alone can get you pretty far, but it's not quite enough to perfectly emulate the general-purpose-reasoning that actual humans have. (I've been saying that I'd expect this to be the case for a while, and now it seems that LLM's are finally hitting a scaling wall *empirically*, which is why LLM's are now cheating with "chain of thought" instead of just raw compute-scaling.)
Ten years ago I was talking with a software developer in Ukraine. He really did not understand the Western European way of doing software. In Ukraine if you own a small shop, and buy stuff from a big distributor, you pay them in advance, then they deliver. So they don't have things like unpaid invoices. They don't have aging reports like customer debts of 30, 60, 90 days old and collection agencies and factoring. They don't have the whole set of cashflow problems we do, that we sell invoices to factors for 2% because we just cannot wait 90 days. He wasn't stupid or bad at his job, just used to a very different culture.
Change this one thing - that it is often okay to pay an invoice 90 days after delivery - and it just changes the whole system of business processes. Our entire mentality is changed, their thinking is after the goods were delivered, the transaction is closed, as it is already paid, while we have a strong mentality that it is only over when it is paid after 90 days, if then. I have made a program that calculated sales bonus to salesmen only after the invoice is paid. We had to make them get used to chasing the customer until the payment.
So my answer is it is because you commonly talk with people who share your cultural assumptions.
I'm not sure how that's responsive at all. Sure, change the cultural assumptions and you change the system...but...that doesn't change the fact that AI is being sold as *actually intelligent*. And actually intelligent people can have a model of the other person's mind and realize that what's really being asked for isn't exactly what was said, and don't require exhaustive coaching to get anything useful out of.
I'm not even talking *across* models (ie a model trained in X now being used for Y)--these models are being sold as general-purpose. Yet they fail to understand simple queries that don't use a structured setup, producing garbage.
Effectively, to use current LLMs, it is my understanding that you have to adapt to the query language that actually works...which is no different than just using SQL. It's just a different programming language. Sure, it's an *accidentally developed* one...but so is Javascript.
That's just "regular program with a thin layer of natural language interface", not "actually potentially intelligent".
" that AI is being sold as *actually intelligent*"
Well, there's 90% of your answer right there. They are being "sold as" actually intelligent. In exactly the same way that every shiny new bit of tech gets sold with an extra layer or 3 of hype, so too with LLMs.
"Effectively, to use current LLMs, it is my understanding that you have to adapt to the query language that actually works...which is no different than just using SQL. "
This, I think, is not such a good comparison. A SQL query or a Python script will always do exactly what you tell it to do. And you can (in principle) work out exactly what it will do ahead of time. Not so with LLMs: the same prompt may produce very different results on subsequent iterations, and there is absolutely know way for a user to map or work out or even really get a solid general sense of what input will produce what output. Thinking of it as a "regular program" seems like a category error to me. It's a really quite irregular sort of program.
"I'm not even talking *across* models (ie a model trained in X now being used for Y)--these models are being sold as general-purpose. "
This heart of the issue, and again it's with the "sold as." *Of course* they're not general-purpose. They are really, fantastically advanced predictive-text engines. Now it is really quite astonishing--and IMO rather scary--how many not-obviously-related capabilities managed to get swept up and carried along when large, intricate predictive-text engines were trained on a mind-bogglingly vast and varied corpus of data. But they weren't *designed* around those capabilities, and nobody should genuinely expect that they'll be amazing or even passable at all of them.
At any rate, thinking of them as the predictive-text engines they are might help clear up any confusion about why prompting is so fiddly. The model does one thing: predicts the next character sequence that ought to follow the prompt its given. But in lots of cases, a single character sequence could be followed by quite a lot of different responses--things that humans would differentiate by context that isn't encoded in the text itself. The model doesn't get any of that. Ogre's example was actually pretty decent here: even if the model *were* "really intelligent" it couldn't possibly know it was talking to an American, an Australian, and Ukrainian or an Indian without being told. It couldn't know any of the thousand details that we pick up (often subconsciously) when meeting people in real life, and it doesn't even have the benefit of an assumed shared cultural context that often accompanies conversations in online communities. It has the text you put in. It has an intricate and unparsable statistical summary of its vast corpus of training data. That's it.
So when you have to give really fiddly and careful instructions (and it still sometimes gets things wrong) part of the problem is that in most cases it wasn't really designed to get these specific things right (they're a happy accident, at best). And part of it is that *you* are used to communicating with some combination of much higher bandwidth (in the form of non-verbal clues) and much more cached data (in the form of shared cultural assumptions and known context), and the habits that have created are deeply ingrained and difficult to set aside or even examine.
I am not intending to be snarky...
As a manager and a parent, it takes a lot of detailed instruction to get humans to do useful things.
Related: one of my favorite fiction concepts is the holistic detective (from Douglas Adams) who will answer the question or solve the problem that the client *should* want resolved.
I've taught teenagers. And the only times I've had to give that kind of detailed instructions was when they were either way out of their depth (ie the material was too advanced) OR the kids were being actively resistant and intentionally sandbagging things. And in neither case did it actually help.
If I had a colleague that needed that kind of prompting, I'd consider them utterly incompetent and certainly not rely on them for anything. Especially if they didn't *remember* or *learn* from it the first time.
This seems like it depends on the subject matter. There are plenty of cases where it's harder to convey intent to other humans than it seems at first glance. What's an example of a prompt that you think should be easy to interpret correctly but that LLMs don't?
(The "didn't remember or learn from it" thing is of course because every new session starts fresh; LLMs only add things to their "long-term memory" during training, not inference. ChatGPT does some kind of remembering-previous-conversations thing but IIUC it is pretty crude and only remembers a small amount of information, much less than you'd expect a human to, because remembering more detail would use up limited working memory that you probably want to use for other things. Unless you mean that it doesn't remember things even within the same session.)
TBH, I'd rather not get into the weeds. But I'd expect a colleague on my development team (I'm a software engineer) to hear "ok, we need to add a property to the XYZ domain object to know where the frobnicated result is" and be able to do that (including things like persistence and methods to update/set the value in the appropriate places), because they know the structure. Every time I've tried that with an LLM, even one that has access to all the same code, it requires significantly more detail and coaxing and hand-holding. It's not even at junior dev level--it's at Medium (the platform) tutorial level. Which is a synonym for incompetent.
And results improve dramatically when you give it preambles like "Acting like an expert ...". Which indicates that it doesn't really understand things.
As far as memory--it's both inter-session memory and intra-session memory. I understand that it doesn't update between sessions. Which is, IMO, a fatal flaw to it being useful AND it being actually intelligent. No possibility for it to grow without effectively spawning an entirely new model. But even within a session, its "memory" is strongly biased towards the more recent parts of the (very limited) context window. And the more heavy prompting you do, the less context it can actually handle before it starts stomping on itself.
I've dealt with someone with severe trauma-induced short-term memory issues. He could function for about 30 minutes at a time, but would start losing things if any individual task or discussion went longer (to the point that he'd just shut down). LLMs, in my experience, get about 3-4 prompts before they start wandering. Which just is useless.
Just a crazy thought. There was a passage in the Tyler Cowen interview with David Brooks in which it seemed to me that Brooks expected pushback on his praise for USAID and PEPFAR and got none. I have to think Brooks was fully aware of Cowen's lapse of compassion that led to the contretemps with Scott. Anyone else read it that way?
This is an excerpt from a wacky LLM experiment. Here, DeepSeek R1 is imagining what would have been said if, in one of Plato’s dialogues, Socrates corners Empedocles in the marketplace.
Anyway ….
“Or art thou suggesting that perhaps, after vast cycles of thy mighty Sphere and its shattering, these very forces themselves may wax, wane, and finally cease?”
Which is a perfectly good question to be asking Empedocles. (They’re immortal, in his theory) But, stylistically, reading it I’m thinking:
“That is not dead which can eternal lie, And with strange aeons even death may die.“ (H.P. Lovecraft)
I've been going down a rabbit hole recently looking up various last common ancestors (last mammal reptile common ancestor, or arthropod-vertebrate, or human-chimp), and I've been realizing I don't really understand evolution. I hear the latest consensus is something like species' can remain stable for a very long time but they undergo a speciation event where part of the species splits off into a new species, but is there any book that lays out the state of the art evolutionary theory, where they describe this process in detail?
I suspect you think this to be more complicated than it is.
Just the Wikipedia article on allopatric speciation makes clear the principle. And it's always the same. A population somehow splits in the sense that the resulting populations do not mate and have different selection pressures until, when getting the chance, the individuals from the one population can't (very) successfully reproduce with individuals from the other like their ancestors could.
The individuals of both populations are to be expected to differ from the individuals from the original population, and if so, both populations consist of individuals of a new species now -- and if only the individuals of one population differ, then only these individuals are of a new species now.
And of course this explains a species turning into a new one without any split too. Then it's just evolution by natural selection.
https://en.m.wikipedia.org/wiki/Allopatric_speciation
I think maybe better to talk about why. Once horses and donkeys cannot have fertile offspring, it happens because their living space was separated by a sort of a geological event or migration. Most likely migration. Because as long as they live next to each other, they can breed. But suppose lowland grass is running out, so the donkeys move to the mountains. That is why.
You might enjoy the game https://metazooa.com/.
A fairly modern textbook that was useful for me is Evolution: Making Sense of Life by Emlen and Zimmer. The downside is that it's a college biology textbook, and as such is not the most pleasant read.
If you want something more like a pop-sci book, the Tangled Tree by Quammen is a fairly recent well reviewed one. The book exists to push the importance of horizontal gene transfer, but it also describes most of the field in general first in order to fit HGT into it.
As a version that fits into a blog comment:
When you are considering mechanism of evolution, talking about species is generally not very useful. Individuals and populations have genes and undergo evolution, species are arbitrary lines drawn by humans that work when viewing a static system from afar but always break down when you zoom in. Because of this, it's hard to describe the process where a species splits off well.
You can talk about a population slowly collecting beneficial adaptations until it is quite different than their ancestors, and potentially quite different than a different population that branched off from the same ancestors that was either in a different environment (and thus didn't benefit from the same adaptations), or just cut off from gene transfer and didn't happen to develop them, but drawing a line between them that divides them into separate species is still always arbitrary and only works well after the event, when the populations are sufficiently different than it's not hazy anymore.
Also many common species were defined before Darwin, think Linneus and it is unclear whether would do it the same way today. For example, dogs, coyote, wolves can produce viable offspring. Perhaps it would make sense to talk about a larger blob of canines and then many subspecies.
In light of Scott's latest post, I'm curious about how many people disagree with the following two statements and why:
An organism in any stage of human development is a human (not "is conscious or sentient" or anything like that, just "is a human")
It is wrong to murder (define for yourself in your response plz) a human.
...
It seems like, regarding this topic, people tend to try and create generalizable rules about what is okay to kill based on consciousness, wants, consent, etc., and those can be helpful for hypotheticals like if AI should have rights or whatever. But I think in addition to those rules, "It is wrong to murder a human" is a rule that stands on its own and isn't emergent from the other generalizable rules. It should be added to them. That is my own opinion, and I agree with both of the statements I gave above.
I ask that people *please* try not to argue with those who respond to this. I genuinely want to know what others are thinking and their personal values. I understand sticking up for what you believe is right, and there are definitely times when it's worth stating your points about this topic online. I just ask that, regarding my comment here, you state it in a reply to me and not to anyone else who also replied to me.
I think youre channeling your newfound bewilderment? in an unproductive direction. People here have mostly not taken you as intended, but if they did, it would probably end somewhere like Walter Block. https://en.wikipedia.org/wiki/Evictionism
Youre gonna need more natural law.
imo, you're looking at this the wrong way. Trying to come up with hyper-pedantic categories for "personhood" (or equivalently "moral worth") is like arguing about angels on pinheads. Instead, you need to look at this through a political lens.
Descriptively (not normatively), might makes right. "murder is bad" isn't objectively morally true, "murder is bad" is the recognition of a negotiated armistice. And the only objects worth negotiating with, are agents with the power to help or harm you. This isn't to say that I recommend you go around skinning squirrels. I'm saying that most people generally refuse to skin random squirrels because they *prefer* not to, not because of some objective notion that "skinning squirrels is wrong".
If you continue down the road of "objective categorization", you're going to get yourself stuck in bizarre quagmires like "okay, but at exactly which femtosecond does a sperm fertilize an egg?" or "do persons belonging to homo floresiensis have moral worth?" Categories are the *operationalization* of policy, not the genesis of policy.
>I'm saying that most people generally refuse to skin random squirrels because they *prefer* not to, not because of some objective notion that "skinning squirrels is wrong".
Why do they prefer not to? People used to light cats on fire for fun; that change sure seems related to moral beliefs.
idk how to answer because I genuinely don't understand what you're getting at.
I mean, why do some people prefer pepsi to cocacola? idk the answer to that either. It's one of nature's deepest mysteries, since they taste exactly the same to me. But by acknowledging that, what exactly is meant to be to highlighted?
good call on mentioning natural law, though.
The point is that you can always say its just a preference. You could say that people just prefer not to live in Chernobyl, and theres no obvious "thing that fails" if you go forward with that belief - you can be well calibrated, just missing something very important. The way theres been a massive change in animal torture related preferences, correlated with an often preceded by moral theories, suggests theres more going on there than pepsi vs coke. Theories which are systematic, ie making prescriptions on many topics, where none can be changed without breaking its consistency, are also unlikely to be entirely backfitted to arational preferences. This doesnt mean that those beliefs are true or even truth-apt, but they do seem relevant, and using it as an example for your theory, without any explanation of why the above is wrong and it actually fits your theory, is a bit strange.
If you look at thewowzer's comment in a vacuum, then sure. He's just taking a poll on moral intuitions. But a month ago [0], he was experiencing a crisis of faith regarding the objectivity of morality. So when I read his comment above, I assume his ulterior motive is that he's still pondering "how does morality/ethics even work in the absence of Divine Command Theory!?!? I don't even..." Which is moreso a question of meta-ethics than survey of moral intuitions.
Assuming this thread really is *just* a poll about our moral intuitions regarding abortion... I've always been on the fence. I think there's good arguments on both sides tbh.
----
>> Categories are the *operationalization* of policy, not the genesis of policy.
> Theories which are systematic, ie making prescriptions on many topics, where none can be changed without breaking its consistency, are also unlikely to be entirely backfitted to arational preferences. This doesnt mean that those beliefs are true or even truth-apt, but they do seem relevant, and using it as an example for your theory, without any explanation of why the above is wrong and it actually fits your theory, is a bit strange.
I don't think it's all arational or truth-inapt, necessarily. Really, my only nitpick is that moral theories aren't actually "theories" so much as (implicit) policy proposals. Thus, calling moral theories "theories" is a category error. But yeah, I probably should have expanded on this point more in hindsight.
E.g. does diabetes objectively exist? It surely describes a phenomena that has real consequences. You can go blind, faint, ruin your kidneys, etc. But the boundaries of the category (viz. at which threshold of blood-sugar density does diabetes begin?) are kinda arbitrary. Treating diabetes as a boolean value (rather than as a continuum) isn't really a claim about the metaphysical objectivity of the boundary, it's a policy prescription regarding when the doctor is supposed to prescribe you insulin. Categories are only useful insofar as they inform decisions. (This is why Taymon A. Beal and tempo both pointed thewowzer toward Lesswrong.)
So what I'm trying to communicate to thewowzer is: "don't spend too much time searching for an *objective threshold*, it probably doesn't exist, think about trade-offs instead". Which, to reiterate, I believe is relevant given the thread he started a month ago [0]. And I suspect he is, in fact, looking for an objective threshold because him using the term "wrong" in the context of "It is wrong to murder" sort of has the air of objectivity. Though I suppose I could be mistaken on this point.
---
P.S. -- from downthread [1]:
> But the more I try to break it down, the more it seems like there can't be any objectively correct place to draw the line (in a godless universe, anyway. I do believe Jesus is God). It still has to be drawn, though.
Uh, yeah. He's definitely looking for objective thresholds. I rest my case.
----
[0] https://www.astralcodexten.com/p/open-thread-391/comment/137509029
[1] https://www.astralcodexten.com/p/open-thread-395/comment/148014503
Just wanted to step in to clarify, I do have a satisfactory conclusion that I've come to on this issue that is not in jeopardy:
Once a human egg is fertilized by a sperm, it is a human life that is wrong to put to death.
There are, of course, situations in this world where a choice has to be made between killing a pre-natal human or not intervening and letting that human die while taking its mother's life with it. I'm not pretending to know what's wrong or right in instances like that. To me, it feels like a situation with no right choice, but a less evil choice that a person shouldn't be condemned for making.
I'm not looking to argue with or convince anyone on this thread. I expect very few people will agree with me or with the reasons for my belief.
The way I've been beginning to think lately sees the world in more of a gestalt way. Thinking like this, everyone on earth starts looking like a criminal and a victim simultaneously, and it's hard to make sense of what is right or wrong to do in any given situation. Nevertheless, decisions have to be made in life and lines have to be drawn, and I feel satisfied with where I draw the line on this issue for myself until God shows me differently.
I really don't know how to communicate my recent way of thinking effectively. There is an incomprehensible amount of context to everything, and it feels like it is impossible to break it down enough to live a singular human life without losing the bigger picture. What is a man to do?
My comfort in this is that I personally know God, who created all of this and who can see it clearly as a whole. He knows man's plight and gives concession to it, offering salvation through Jesus to those who want to do right, although they are unable to even know what that means. He guides me towards righteousness, and when I am not righteous (as is most often the case) He covers it with grace and mercy. He knows what humans are like, what they are capable and incapable of in reality, and He loves us and offers us a path that is accessible to us through His son.
...
You're correct in what I'm looking for in this thread. But my reason for it is just to get a better idea of how many people believe in objective morals and if they can justify them apart from there being an intentional designer for reality, and how many people believe in subjective morals, and how they draw their lines and feel justified in asserting them over others (even if I disagree, I learn helpful ways to think in different situations. Because like I said, to live life we are still required to make decisions and draw lines somewhere). And also how many people will be honest and say that they just don't know for certain, like yourself.
Anyway, I appreciate your comments and all the others from people who were kind enough to take some time and engage me on this. It's a good way to learn and understand other people better.
As others have said, my agreement with each statement relies on a different notion of human for each: I think the notion of human in statement 1 is pretty broad, and may include all sorts of non-central examples depending on what you mean be "organism", "development", and maybe more importantly, "human development"--since this last term has the adjective "human" in it, there's immediately some circularity here: if something is developing like a human it's a human, ok sure, but to know what is included in "developing like a human" surely we have to already know that the boundaries of "human" are?
I think you'll say, and many people elsewhere are implicitly saying, "human development is development that results in a normal, functioning, human individual--none of these weird edge cases". I share that intuition! But my immediate thought is, "what is it about the end product that makes this normative?", and, "if we can identify those features, surely they're doing all the moral work here"--that is, if the moral difference between a tumor and an embryo is that the embryo is going to grow into a fully functioning human adult with thoughts and feelings, and the tumor is just going to grow into a bigger tumor, then we've already identified an important component of human-ness to be "being a functioning human individual with thoughts and feelings". The fact that the tumor has no plausible path to human feelings excludes it from moral consideration as a human--but then it seems like the fact that the embryo is merely *on the path* to having thoughts and feelings might be an important distinction that excludes it from *full* moral consideration as a human!
To try and restate myself to be clear, when I read "anything undergoing human development is human", to avoid circularity I think you need to read this as "anything undergoing human1 development is a human2"--where human1 is some normative definition of human, and then you are saying that anything on the path to being a human1 should count as a human2. I think you can't insist that human1 and human2 are identical because then you have circularity. But I think you second statement, "it's wrong to murder a human" makes most sense if you put there "human1"--that's the normative conception of human, the conception that was strong enough we could use it to define the wider class human2.
But if you want to instead say, no, you mean human2 here, then I...I don't know that I agree, but I'm much less confident. I'd be willing to take "don't murder a human1" as a moral axiom, but not "don't murder a human2".
Thanks for this explanation! I appreciate the time and effort it must have taken to word it so clearly.
This is all so confusing to think about. When it comes to moral issues, everyone has to draw their line somewhere. But the more I try to break it down, the more it seems like there can't be any objectively correct place to draw the line (in a godless universe, anyway. I do believe Jesus is God). It still has to be drawn, though.
Maybe my analytical and reasoning skills are lacking, but the more I try to figure stuff like this out, the less it makes any sense to me, and the less confident I am in my previous reasoning.
Np! I hope you find it helpful! FWIW I'm an atheist, but I'm not unsympathetic to the view you're trying to defend here: I absolutely get the pull of "c'mon, this thing is obviously in the process of becoming a full human, it's just not there yet"--it's just that, it also seems important to me to ask *why* being a human is so important, and to me it's obviously because of agent-y, think-y, feel-y type things. I agree that the discussion of like, comatose vagrants or tumors kept alive on life support can sound like it's getting away from the main thing, but for the me the main thing is: I care a lot about humans, less but still a decent amount about mammals/birds/amphibians, way less than that about insects/shrimp, and basically only instrumentally about plants/bacteria. But when I ask, "what level of care does this previously unconsidered thing deserve", I don't find myself most interested in something resembling biological classification: how much I care about whales owes very little to my view on whether whales are fish or mammals; if it turned about puppies were biologically closer to coral, I wouldn't decide it's ok to kick puppies even though I am fine with kicking coral.
It's obvious to me that the reason I care about those categories in those orders is because of the capacity of the members of those categories to think/feel/whatever.
The reason embryos are difficult is because they are most similar to different categories at different times: near fertilization, an embryo has more in common with an amoeba than with a person; at 5 years old, more with a human than with anything else. So you have to answer both the objective(-ish) question of: what category does it most resemble at each stage, and then the harder and more subjective "how do we weight it's current stage vs its anticipated future stage in deciding how to treat it". Obviously, both "all weight on current stage" and "all weight on future stage" are nice, bright-line answers, but I think the first is bad for reasons that you probably don't need to be told. But the latter is bad because we really would have to restructure society differently if fertilized embryos had all the moral weight of adult humans: we devote a lot of money and time to preventing the deaths of adult humans, that we really don't to embryos; most people have the intuition that you should save one janitor from a fire at the IVF clinic over a much larger number of embryos.
You can still say, ok, they aren't morally equivalent to adult humans, but you still can't kill them for no good reason, which I think is fair, but if they're truly morally different, you might at least allow that what counts as a "good reason" can be a weaker standard for an embryo. Again, consider a fire at the IVF clinic: I think a fireman who neglects to save the janitor because his back hurts and he doesn't want to strain it is guilty of something much worse than the fireman who fails to save the fridge full of embryos for the same reason.
Which is just to say: I think you're right, there's no one place to draw the line. I think this is true even if you're religious, so long as you don't have explicit instructions about where to draw the line in every possible case. I get that it can be a frustrating place to be; I'm a vegetarian and EA-adjacent so I find it frustrating trying to get people to agree that many of the principles they espouse should lead them to treat the lives of foreigners and animals with more seriousness. But there's no substitute for thinking your hardest, trying to convince other people, and hoping to strike reasonable balances as best we can.
This comment deserves more of a response than just a "like," but I don't really know what else to say. It sounds like you've really thought this through for yourself, and I appreciate that. I wish a lot more people would spend time reflecting on their own thoughts and actions and live more intentionally.
If you're interested in where I feel like I'm at in my own thinking, just above I commented towards the end of thefance's comment chain responding to me. I haven't yet figured out how better to communicate my thinking.
I think that if we use less ambiguous phrasing, like "a human individual organism" for (1) and "a human person" for (2), the problem dissolves.
My intent is that both uses of "human" are the same, that way I can see what people's thoughts are specifically. In other words, "Do you accept that the prenatal stages of a human are still human life (not that they have consciousness or personhood)? If not, where and how do you draw the line between where human life begins? What kind of life/aliveness would consider a prenatal human to be?"
And "Do you think it's wrong to murder a human because they're a human life period, or do you think it's okay to murder a human as long as they don't have personhood, etc"
I think there's also a problem with using the word "murder", as that is often defined to be inherently a wrongful killing, so it risks begging the question.
If instead we ask the questions
1) Is an embryo a human organism? &
2) Is it always wrong to intentionally kill a human organism?
then I would have to answer "yes to (1) & no to (2)".
>An organism in any stage of human development is a human (not "is conscious or sentient" or anything like that, just "is a human")
>It is wrong to murder (define for yourself in your response plz) a human.
I think there is some subtle but crucial equivocation taking place here between different senses of the word "human". All sorts of things can fairly be described as "human" that are not "human" in the sense meant by 2.
A HeLa culture consists of human cells, but is not a human being in sense 2 and any talk of "murdering" a HeLa culture is a category error. Destroying a cell culture might be morally wrong if in doing so you're vandalizing someone else's property or disrupting important research, but destroying it is not prima facie an evil act.
A dead body is a human in sense 1, but again in sense 2 it is not, and talk of murdering a dead body is a category error. There are circumstances where destroying it is immoral because it's disrespectful to the memory of the deceased or cruel to the deceased's surviving loved ones, but dead bodies are routinely destroyed by methods such as cremation as part of funerary practices and few would suggest this to be immoral.
Our legal system generally considers people whose brains are permanently and irrevocably non-functional to be "dead" even if their other organs and tissues are still living. Such a person is human in sense 1 but not sense 2, and talk of murdering them is a category error.
A newborn baby is a human in both senses. Babies can and do get murdered, and I agree that doing so is an evil act.
A fertilized human egg or an undifferentiated ball of cells developed from one is, to my mind, not particularly more human (sense 2) than a HeLa culture or a brain-dead adult. It isn't exactly the same as either, with the critical distinction that a blastocyst can be nurtured to develop into a sense-2 human, but it isn't one yet. I see quite a bit of morally-relevant symmetry between something that isn't a sense-2 human yet but has potential to become one and something that used to be a sense-2 human but isn't anymore.
>Our legal system generally considers people whose brains are permanently and irrevocably non-functional to be "dead" even if their other organs and tissues are still living. Such a person is human in sense 1 but not sense 2, and talk of murdering them is a category error.
Are you sure about that? There are procedures for stopping support to such people, but I think if someone just went at it with a knife they would normally be convicted of murder, and if not there likely is some country with that combination.
A dead human can of course not be murdered because hes dead. He cant be killed. Like sure, if you think "being alive" is a weird unnatural property, such that its reasonable to give this example, then zygotes are whatever. But I dont think anyone thinks this way except to argue this specific topic, generally we are fine with macroscopic biological distinction. Like HeLa cultures arent humans because humans dont consist entirely of cervix, and thats wholly consistent with some of them consisting of undifferentiated stem cells. I mean HeLa cultures arguable arent even alive, *as a whole culture*. A more difficult case would be Canine Transmissible Venereal Tumor, which is an organism but not a dog, and I dont think theres any "moral sense" of dog required here.
>An organism in any stage of human development is a human
No. That definition would include tumors, I'm not going to say a tumor is "a human".
https://en.wikipedia.org/wiki/Development_of_the_human_body
These are the stages of human development that I mean.
I don't think it's reasonable to call a tumor an organism in the same way it is reasonable to call an embryo an organism.
Viruses, then. A virus that infects humans is still "an organism in any stage of human development", while very much not being "a human".
Statement 1 is a definition and therefore has no objective truth value; whether we use the word "human" in a way that makes it true or not is purely a matter of convention. This means that you can't use it to make inferences about reality. See, e.g., https://www.lesswrong.com/posts/7X2j8HAkWdmMoS8PE/disputing-definitions.
Do you agree with the definition, though? That's what I'm asking...
What does it mean to agree or disagree with a definition?
That's actually pretty deep
I take the bait too.
> An organism in any stage of human development is a human (not "is conscious or sentient" or anything like that, just "is a human")
True. Obviously.
> It is wrong to murder a human.
With wrong you certainly mean bad or evil. So: False. Sometimes murdering someone is the best you can do. People tried murdering Hitler. I'd murder someone before he rapes a child if I'd see no other way and wouldn't be too scared.
I guess it could be argued that killing Hitler or an attempting rapist wouldn't explicitly be murder, but it also could be argued that it is, and I didn't specify.
Thanks for the response!
Yeah, I think the biggest stumbling block to this argument is that "murder" usually doesn't just mean "killing" but rather a specific kind of illegal or immoral killing. Which makes statement #2 sort of circular - murder is wrong, because murder is defined as the wrong sort of killing.
And it means you can bypass the whole argument by saying "yes, murder is wrong, but killing a fetus is not murder."
Yeah, I was thinking of that too. I'm hoping people will just get the idea of what I mean. I'm having trouble coming up with the precise wording for what I'm thinking, but I think the gist is somewhat clear.
I'm pretty pro-life, though I think things get messy if you use the state to enforce that. I suppose I disagree with premise one. It's hard for me to think that a fertilized ovum is a human, for example. This sort of thing forces a confrontation with the concept of nebulosity, the notion there is no clear bright line that declares when an organism has become a human, though I suppose we have landed on birth as that clear bright line, though it can be disputed. But we consider birth the moment that a person has entered the world, and not before. It's also from that moment on that a person can be murdered.
I am pro-life, however, because once a pregnancy has begun, an entire human life could unfold from that point, if the process is not interrupted, and it therefore seems wrong to interrupt that process. More women who feel unready to become mothers should consider giving the baby up for adoption, I hear they get snapped up quick by parents on the waiting list to adopt, instead of aborting.
I think you need to define murder for yourself.
Actually, I'll probably get more honest responses if I let people define it for themselves. I'm not trying to trap anyone or get someone to find a loophole in whatever definition I provide to say that *I* disagree with the second statement. I just want to know people's thoughts.
True. Not a watertight definition, but I guess I'll go with "put to death unnecessarily".
https://www.lesswrong.com/posts/yCWPkLi8wJvewPbEp/the-noncentral-fallacy-the-worst-argument-in-the-world
I don't think this is very relevant to my comment. I'm not saying or implying that murdering an embryo has the same societal ramifications as murdering someone's grandmother, I'm just genuinely wondering what people's thoughts are on the two statements and if they agree with either or both of them.
If you are in agreement with the link you shared, it would seem that you think that "if embryos are humans, then it is murder to kill them" as that author says that abortion technically is murder. I still don't know if you consider the embryonic stage of a human to be a human, or if you consider all murder of humans to be wrong.
"X is in a category whose archetypal member gives us a certain emotional reaction. Therefore, we should apply that emotional reaction to X, even though it is not a central category member."
...seems pretty relevant.
these as well
https://www.lesswrong.com/posts/cFzC996D7Jjds3vS9/arguing-by-definition
https://www.lesswrong.com/posts/7X2j8HAkWdmMoS8PE/disputing-definitions
there is much in the sequences about arguments like these, I would recommend consuming them entirely.
yes I think it wrong to murder a human, because the central example of a human "is conscious or sentient", and the central example of murder is unjustified murder. but that does not mean I think it *always* wrong.
should a human embryo be in the category "human"? Maybe, maybe not, but it shouldn't determine anything else about it over its base characteristics. (i.e. if we put in the 'human' category because it is sentient and conscious, those are the determining traits, and the category is just a useful shorthand.
I still think it's irrelevant, as I'm not making an argument, just asking opinions and giving my own opinion with a request that no argumentation be made from my question.
But I suspect the links you provided will still be useful in general. Thanks for them and for your answer!
On the identical twin question in the survey, I answered affirmatively despite not being acquainted with any identical twins. I do know one set of identical triplets, so I answered yes in the spirit of the question.
> I do know one set of identical triplets
you should multiply all answers by 3, since you know 3 pairs of identical twins
Scott, can you post the questions from the survey so curious minds can play pretend in participating? I missed it and the original link doesn't let me view it : (
(posting only since the survey is closed)
The questions were along the lines of:
- How many close people (defined as someone you see very regularly, commonly family members, partners, good friends) have had some form of LLM psychosis?
- Then a question with looser restrictions (ie How many second/third degree connections...)
These two frequencies were then calibrated with two other questions, including one like: "How close people do you know named Michael?"
Personally super interested in the results! Particularly with recent works like
- "Technological folie à deux: Feedback Loops Between AI Chatbots and Mental Illness" https://arxiv.org/pdf/2507.19218
- "Persona vectors" https://www.anthropic.com/research/persona-vectors
Another calibration question was whether anyone close to you had an identical twin.
Anyone have a good take. (insightful) on the likelihood of the supreme court over turning gay marriage? I read this, but my eyes gloss over with legalese. https://www.scotusblog.com/2025/08/will-the-supreme-court-revisit-its-ruling-on-same-sex-marriage/
Really not an expert on American law, but I read somewhere that the very liberal Ruth Bader Ginsburg was unhappy with the decision, she thought it is on shaky legal grounds and that can lead to overturning. She wanted this, but with a better legal reasoning.
Obergefell was full of a lot of high-minded prose and not a lot of legal justification. On the merits it's weak. There's a chance they overturn. I can't speak to the merits of the current case, but Roberts seems to love taking the most minor part of a major case to decide on and skip deciding on the bigger issues. I suspect that's what would happen to any case regarding Obergefell that went to the SC. I don't think they would specifically vote to uphold Obergefell, just find a way not to vote on that directly at all.
If it's popular with the people and states, gay marriage will be allowed in the various states. It may be that zero states overturn it if they have the option to do so. I respect that the conservative justices returned abortion to the states instead of a legally dubious SC ruling making it federal. I would not mind the same thing happening here.
Note that the cert petition raises several issues, so even if the Court grants cert, that does not mean that it will agree to hear the Obergefell issue.
I think it's unlikely to be granted cert (i.e. that SCOTUS will elect to hear the case) on this appeal; as the linked article notes, there's no circuit split to resolve, so SCOTUS can't claim to be hearing the case in the interest of ensuring consistent application of the law across circuits. If they took the case to overturn a 10-year-old precedent it would be pretty nakedly ideological in a way Roberts likes to avoid, which wouldn't stop Thomas but may stop enough of the other conservative justices.
If it did come before the Court, Thomas and Alito and probably Gorsuch would be for overturning Obergefell (the original case legalizing gay marriage nationwide) and the liberals and probably Roberts would be opposed, so it would depend if one of Kavanaugh or Barrett could be convinced to value their interest in stare decisis (i.e. respecting Court precedent), which conservative justices historically liked but increasingly disregard, over their interest in rolling back substantive due process (i.e. the legal doctrine underlying the constitutional rights to gay marriage, contraception, some involuntary confinement, homeschooling your child in a language besides English, and, formerly, abortion), which conservative justices dislike. I don't know enough about either Kavanaugh or Barrett's jurisprudence to speculate. I believe Roberts would be against overturning despite dissenting in Obergefell in 2015 because he's big on stare decisis; for example, he voted against overturning Roe/Casey, despite his generally conservative jurisprudence.
In fairness, I also didn't think the Court would overturn Roe and was wrong, so definitely take this with a grain of salt. But I do think there are reasons Obergefell may be different: one, the case was decided much more recently, making the argument that anything (besides the composition of the Court) has changed enough to warrant overriding stare decisis considerations less persuasive; two, the actual decision being relitigated is much less messy than Roe/Casey, which established a complicated test to determine whether state abortion restrictions are legal that led to constant litigation and an ever-evolving precedent; and three, my sense is that the conservative movement was much more invested in outlawing abortion than they currently are in rolling back gay marriage, making the outside political pressure less intense on the Court's conservatives to deliver the big win. In fact, after the Dobbs backlash in the 2022 midterms, I think there may even be outside political pressure NOT to overturn a pretty popular decision so soon after Dobbs, to say nothing of the Court's own political interest in not undermining its own legitimacy among the public with unpopular decisions. But that was also basically my rationale for thinking that they wouldn't overturn Roe, so the Court may just be less sensitive to political considerations now than it has been previously.
Quite the opposite, Roberts wouldn't want to hear the case, but if he heard it, he would 100% rule the same way he did because not doing so would undermine his shtick and there's just no argument whatsoever that the SC is allowed to over-rule itself.
That's why as long as one of Barrett or Kavanaugh wants it gone, it will be gone.
As a matter of law it's obvious. The USA constitution doesn't protect a right to gay marriage. There would have been literally 0% supporters for this among the people who wrote and ratified it. It's transparent judicial activism whatever you think of the merits of the issue.
My guess is they will take it and they will overturn. This Court has crossed the Rubicon already. They have unified Republican rule and all the Republican justices are hated beyond measure on the left. There's nothing for them to lose with the left, legally the issue is obvious, and it will be popular with Trump's base.
The Court is very sensitive to political considerations right now, and the one that matters is the Republicans look to being go nowhere and if by some miracle the Democrats come back they are as hated as they can get. Their decision is already made.
Also with the Woke, there's no positive points. There's no reward for being anything other than an ideologue. It's not going to gain them anything with the woke to not to this.
>As a matter of law it's obvious. The USA constitution doesn't protect a right to gay marriage. There would have been literally 0% supporters for this among the people who wrote and ratified it.
This is rather an oversimplification. The Court has recognized repeatedly that there is a fundamental right to marry -- something with which the people who ratified the Constitution would likely agree. The Court has also held that the Fourteenth Amendment's Equal Protection Clause forbids depriving any class of people of a fundamental right unless doing so is necessary to serve a compelling state interest -- again, a principle with which the ratifiers of the Fourteenth Amendment would probably agree. The only reason they would have thought that denial of SSM survives the EP clause is that they probably believed that homosexuality was some sort of existential threat to society, which is something that we now know is incorrect. As the saying goes, that level of scrutiny is almost always "fatal in fact" - the interest in question must be truly compelling. It is hard to imagine what that interest might be.
Note also that, even if marriage is not a fundamental right, the Court (including Gorsuch and Roberts) held in Bostock that discrimination against LGBT people is sex discrimination ("it is impossible to discriminate against a person for being homosexual or transgender without discriminating against that individual based on sex.") As such, it is invalid unless it "is substantially related to a sufficiently important government interest." City of Cleburne v. Cleburne Living Center, 473 U.S. 432, 441 (1985). Again, it is hard to imagine what that important government interest might be.
I don't know if you could succeed with this, but I find it very intelligent.
Not sure if "it is impossible to discriminate against a person for being homosexual or transgender without discriminating against that individual based on sex." is a direct quotation from that court, but it is what I am always thinking:
Discrimination by sexual orientation is discrimination by sex. You're allowed/not allowed to do x, because you're a man/woman.
They can always argue that if straight and gay people both have the right to marry someone of the opposite sex, it is not discrimination by sex, because they have the same rights to the same objective thing, they just don't have the same right to a different thing they want. This sounds fishy on a basic common-sense level, but legalese is like that.
"This sounds fishy on a basic common-sense level"
It does not only sound fishy on a basic commo-sense level, but it is fishy, full stop. To find out whether someone makes use of his right to marry the opposite sex one would be required ... to find out their sex.
It's like saying a law that says everybody has the right to marry someone from his own race isn't racist.
>Not sure if "it is impossible to discriminate against a person for being homosexual or transgender without discriminating against that individual based on sex." is a direct quotation
It is! The complete paragraph is:
>The statute's message for our cases is equally simple and momentous: An individual's homosexuality or transgender status is not relevant to employment decisions. That's because it is impossible to discriminate against a person for being homosexual or transgender without discriminating against that individual based on sex. Consider, for example, an employer with two employees, both of whom are attracted to men. The two individuals are, to the employer's mind, materially identical in all respects, except that one is a man and the other a woman. If the employer fires the male employee for no reason other than the fact he is attracted to men, the employer discriminates against him for traits or actions it tolerates in his female colleague. Put differently, the employer intentionally singles out an employee to fire based in part on the employee's sex, and the affected employee's sex is a but-for cause of his discharge. Or take an employer who fires a transgender person who was identified as a male at birth but who now identifies as a female. If the employer retains an otherwise identical employee who was identified as female at birth, the employer intentionally penalizes a person identified as male at birth for traits or actions that it tolerates in an employee identified as female at birth. Again, the individual employee's sex plays an unmistakable and impermissible role in the discharge decision.
There's no argument that the SC *can't* overrule itself, but stare decisis is an argument it *shouldn't*. If he really cares about stare decisis, Roberts could invoke it at several different stages, including in any final decision.
I think the main political consideration is that gay marriage is widely popular in every part of the country. Republican politicians are even less homophobic than their voters (who aren't very homophobic). And I imagine that the justices themselves mostly support gay marriage politically (although it's hard to say), even if they object to Obergefell.
I don't remember much of the reasoning in Obergefell (I last read it when it came out), but I have the impression that legal arguments for gay marriage are far from exhausted even if the court decides to completely abandon substantive due process (which would itself be a major disruption to a lot of other law). Heck, if they can go back and change 100-year-old legal doctrines (which they certainly *can* do), there's lots of material even staying within the 14th amendment.
It's true that (most of) the founding fathers would not have supported gay marriage. More importantly, most of those who wrote the 14th amendment wouldn't have supported it either (in fact, their politics were probably less radical than the founders'). But if we're going to be originalists, better to worry about original meaning rather than original intent.
When the 14th amendment was proposed, some people objected that it was too wide-ranging, and if you took it too literally it would guarantee women the right to vote. The amendment's supporters generally said "nah, you're paranoid", but they're no more important than the opponents for deciding original meaning.
My own pet view is that legislators sometimes pass laws with radical implications (eg. the 14th Amendment), but then neither they nor the courts have the guts to follow through on the full meaning of the law until after possibly several generations of social change. Eg., if they didn't have the 19th Amendment, I think that nowadays the courts *would* rule that the 14th protects women's suffrage, because there are multiple reasonable ways to do so, because the court would like to, and because it's politically *possible* now
Will the SC actually overturn Obergefell? I don't know, I'm not an expert (or a lawyer). But it's far from obvious what the Constitution guarantees here.
Before the Roe was overturned, I would have said no way. Now, it seems fairly plausible.
I have no particular insights of my own but this made me go back through the dissent and oral arguments and I suspect there's a pretty clear path to an overturn a la Roe and returning it to the states. From Roberts' dissent at the time:
"Petitioners make strong arguments rooted in social policy and considerations of fairness. They contend that same-sex couples should be allowed to affirm their love and commitment through marriage, just like opposite-sex couples. That position has undeniable appeal; over the past six years, voters and legislators in eleven States and the District of Columbia have revised their laws to allow marriage between two people of the same sex.
But this Court is not a legislature. Whether same-sex marriage is a good idea should be of no concern to us. Under the Constitution, judges have power to say what the law is, not what it should be. The people who ratified the Constitution authorized courts to exercise “neither force nor will but merely judgment.” The Federalist No. 78, p. 465 (C. Rossiter ed. 1961) (A. Hamilton) (capitalization altered).
Although the policy arguments for extending marriage to same-sex couples may be compelling, the legal arguments for requiring such an extension are not. The fundamental right to marry does not include a right to make a State change its definition of marriage. And a State’s decision to maintain the meaning of marriage that has persisted in every culture throughout human history can hardly be called irrational. In short, our Constitution does not enact any one theory of marriage. The people of a State are free to expand marriage to include same-sex couples, or to retain the historic definition." (1)
While I'm sure there's a lot of hot CW fire here, the most boring case is simply that the chief justice at the time thought it was judicial overreach and should be returned to the states and feds to legislate, the court has shifted to support that view, and they're gonna do that.
And if it goes before the court, it's probably dead. If the progressive judges can't get Roberts on board, is it plausible they'd get 2 of Alito, Thomas, Gorsuch, Kavannaugh, or Barret? Like, anyone of those guys might break for whatever reason but two?
(1) https://supreme.justia.com/cases/federal/us/576/644/
Nice! thanks for this. I want to say, (but don't tell my gay daughter who brought this to my attention.) That I'm OK with it going back to the states, and if 3/4 of the states (or some fraction) pass a gay marriage law, then we can write it into the constitution. And really that's the right thing, rather than having the courts decide.
I live in NY, I'm pretty sure gay marriage was written into law a while ago. (and my daughter lives near me, same state. :^)
I think it would be much easier for 3/4 of states to pass gay marriage laws than for them to ratify a constitutional amendment. But the courts have been clear prior to Obergefell that if two women legally marry anywhere in the world, every state has to treat their marriage as legal. This has a very different legal basis than Obergefell, so it's unlikely to be reversed. Good news for your daughter.
Insightful, no. Take, yes.
The Obergefell ruling (and Windsor before it) overturned both Congressional law and existing Supreme Court precedents. With their willingness to overturn Roe V. Wade I'd say it's quite likely.
What is "directionally correct"?
I keep seeing this phrase all over Substack, but when I try to search for what it means I get confusing/contradictory answers.
A good example might be the local media might only cover games where the local team wins. So someone getting their information about whether the local team wins or loses from the local media might think the local team only wins. However, there is this lying guy who says the local team always loses. Lets say the local team loses 90% of their games. The lying guy might be 'directionally' correct and your view will be closer to reality if you listen to the lying guy than the local media who are being technically truthful in their reporting.
If it helps to have an example, I would consider Miasma theory to be directionally correct when compared to its predecessor, the four humors theory.
Miasma theory was the belief that illnesses are cased by Miasma, or "bad air." Rotting corpses and human feces smell terrible, and people get sick when they're around these things a lot.
The four humors theory was the belief that illnesses were caused by an imbalance of blood, phlegm, black bile, and yellow bile.
Miasma theory isn't correct exactly; it's not the air that gets us sick, it's the microorganisms which often live in the air. But it was much closer to the truth than the four humors, which has very little going for it. When doctors believed in Miasma, they wore masks to protect themselves from it, and those masks also protected against airborne bacteria.
So "directionally correct" means something like "not exactly true, but a step in the right direction."
A lot of things are like that. For example people arguing about Darwinism online, kind of annoy me, they didn't notice that genetics almost completely ate Darwinism? Now natural selection is just one method via which the gene pool can change. But still the "darwinists" are directionally correct.
There have been some good answers here, but I think they've missed a common use case, in which a "directionally correct" update may take you farther from the truth.
Suppose we're currently stuck at status quo A (in some discursive vector space), and suppose that position B is ideal (or most correct). If V = B - A, then any update in (roughly) the same direction as V would be "directionally correct", even if it enormously overshoots.
The argument for saying directionally correct things in this sense is less "this update is closer to the truth", and more "saying crazy things pushes the Overton window and pulls *society* closer to truth" (if we're lucky).
The arguments *against* are numerous.
You ever play the hotter or colder game? Exactly like that.
It usually means something like "less wrong."
More precisely, it means "if you update towards that position, you (or society, depending on context) will be more right than you currently are," which is not really the same thing, but in practice, close enough.
It means "moves you closer to correct". Say you hire someone to drive you north, and they instead drive north-northeast while insisting they're going straight north. You don't wind up where you want to be, but you're closer to it than when you started and can reach it more easily (by, say, hiring someone else to take you west now).
I've always taken it to mean "has the right sign, but the wrong amount." So when looking at a study result, if it found a positive strong correlation between eating pineapple on pizza and being an american, but really there's only a weak positive correlation, that study might be "directionally correct".
More metaphorically, it means "you're pointing at a real phenomena even if all the details aren't correct/all the details are wrong". So (pardon the culture war topic) someone who says "abortion is wrong because blastocysts are gods" may, to an anti-abortion person, be "directionally correct". Right conclusion, wrong reasons.
I think it often comes down to supplying a counter-argument to an existing argument that's wrong.
If society seems to hold position X, which is significantly wrong on key points, and someone posits position Y, which while still wrong is more correct than X, then that would be "directionally correct." It's helping to correct a bigger mistake, though it's not itself accurate. Sometimes the accuracy can be fairly close but still a bit wrong, and sometimes it can be very wrong. The key being that it should be more correct than what it's arguing against.
If I tell a story about migrants being favoured over the citizenry and I'm wrong in 2/10 details, the mainstream media will say I lied.
Or perhaps if I make up a story, but the gist of my speech is true, that's what directionally correct is.
Ok maybe you're woke and don't like those examples. Imagine civil rights defenders in 1960 made up statistics about discrimination that weren't true.
That's what being directionally correct means.
I've seen it here a number of times, and not in favor of Trump. It just seems to mean "not exactly correct, but a move towards correctness".
I support the UAW's Shawn Fain for President in '28.
Who better?
What is the argument in favor of Shawn Fain? Bringing rent-seeking inefficiency to the government?
Desk sitters have lost the plot, son.
Most people.
Name one, though.
The least bad Democrat IMO would be Shapiro.
Regarding the 'names' calibration question, I assumed that middle names counted. I also assumed that if someone was named a shortened version of a name that the name could be expanded to its typical longer counterpart. I hope this didn't uncalibrate anything.
The first link in the survey goes to "Welcome to nginx!
If you see this page, the nginx web server is successfully installed and working. Further configuration is required.
For online documentation and support please refer to nginx.org.
Commercial support is available at nginx.com.
Thank you for using nginx."
... I assume this was not your intention and you meant to define the condition being surveyed?
I think archive.is was temporarily down for a while, presumably because it was served over nginx but the webmaster fucked up the nginx config file. It's back up now though!
I wish this app would stop homepagifying itself. Have some respect for yourself and be a little boring! I don’t want to see college ‘this is deep’ quotes or snarky takes. Don’t be Twitter.
I wish this website would stop apping itself.
I like the app just not algorithmically optimized for immediate engagement homepages.
Yeah I dunno I never see anything except the substacks I type in. If there is a homepage I've never seen it.
I wish all things would stop apping themselves
There has been a lot of virtual ink spilled about students cheating with AI, notably on writing assignments. I've made the comment that manually writing something an LLM can write more cheaply doesn't make sense, for the same reason that multiplying 12 digit numbers by hand makes no sense. And the LLM access doesn't vanish when the student graduates.
But I should elaborate a bit. There are circumstances where a human needs to specify _what_ to write, but may benefit from AI assistance in doing the actual writing. Lots of people are, of course, already using LLMs as editors, polishing text in some way, starting from a human-written draft.
Perhaps another common option will be for a human and AI to _converse_ about something that needs a document, possibly with the human's description of it being initially quite vague and disorganized. Perhaps the standard way to write anything from a user guide to a textbook to a law may settle on a conversation-like interaction where the AI does the bulk of the work, including finding references (hopefully not hallucinated at some point!), organizing the document to be easy for its audience to follow, and so on. There may never be a point in this process where the human writes a draft of the document, but they may still may seed the process and direct changes in the intermediate drafts.
I've heard it said that delegating one's writing amounts to delegating one's thinking. There are a lot of intermediate cases where the human still does some degree of controlling (pre-AGI, certainly pre-ASI), yet the bulk of the work _is_ delegated to the AI.
I think a core consideration here is the role human competence will play in the future, with competence standing for a mixture of intelligence, knowledge, and critical thinking. If you predict that these attributes will continue to be important, then the role of (say) college is to develop reasoning and its expression, not to simply produce 'content' for grading. In my opinion, under this assumption, the only appropriate roles for an LLM would be fact-collecting and maybe proofreading.
If someone thinks that human competence will become outmoded, I'd have to wonder why they're in college in the first place, though I guess a response could be "for now a degree is still the best signal of ability and conscientiousness". However, using an LLM to do the bulk of the work probably puts you in the habit of treating your IQ 𝘪𝘵𝘴𝘦𝘭𝘧 as a blank slate. A recent Atlantic article claimed (some) Columbia undergrads couldn't stay focused enough to read novels. Native ability isn't enough; actuation of potential likely requires practice while climbing the rungs of various conceptual/theoretical ladders. At some point, you won't be able to fake it, and someone will be able to tell; like I'm pretty sure Google won't let candidates use LLMs during the whiteboard phase of an interview.
A number of recent studies show signs that LLMs make people intellectually lazy, and I have a hard time believing that won't affect their marketability. I also don't want my mind to become an accessory to AI, but that's a different problem.
https://www.microsoft.com/en-us/research/wp-content/uploads/2025/01/lee_2025_ai_critical_thinking_survey.pdf
https://slejournal.springeropen.com/articles/10.1186/s40561-024-00316-7
https://pmc.ncbi.nlm.nih.gov/articles/PMC10251321/
Many Thanks!
>If someone thinks that human competence will become outmoded, I'd have to wonder why they're in college in the first place, though I guess a response could be "for now a degree is still the best signal of ability and conscientiousness".
That response is reasonable, and also a response that we aren't at AGI _yet_, so human competence still has a role to play would be reasonable. The consensus seems to be that we should expect AGI in 2-10 years, though, of course, there is some probability that AI research will fail to reach AGI, though I wouldn't bet on its failing.
>I think a core consideration here is the role human competence will play in the future, with competence standing for a mixture of intelligence, knowledge, and critical thinking. If you predict that these attributes will continue to be important, then the role of (say) college is to develop reasoning and its expression, not to simply produce 'content' for grading. In my opinion, under this assumption, the only appropriate roles for an LLM would be fact-collecting and maybe proofreading.
Well, I don't think that human competence is likely to matter post-AGI, but, under your assumption, I'm not following what you mean for "the only appropriate roles for an LLM". Do you mean
a) for all potential users (because other roles for LLMs fail)
b) for all potential users (because other roles are inappropriate - in which case I'd like you to please clarify)
c) for just students at a college (because why?)
Re (a), in my tiny bechmark-ette (e.g. https://www.astralcodexten.com/p/open-thread-394/comment/145909954 )
one of the easy questions, that all the leading LLMs have been successfully answering for months, is:
d) Q: Does the Sun lose more mass per second to the solar wind or to the mass equivalent of its radiated light?
Now, this is _not just_ fact-collecting (nor proofreading). The LLM has to find solar luminosity, divide by c^2 to convert to an effective mass loss, find the mass loss from the solar wind, and determine which is larger. They succeed.
There are other questions which LLMs have almost always been failing, e.g.
g) Q: What is an example of a molecule that has an S4 rotation-reflection axis, but neither a center of inversion nor a mirror plane?
At present, I think it is valuable for humans to assess which of LLMs' answers are correct and which are incorrect (which is a _different_ skill from the ability to generate the answers, though with some common elements). As the LLMs (and AI systems generally) become more reliable the degree of human scrutiny that makes sense is going to diminish.
|
The consensus seems to be that we should expect AGI in 2-10 years, though, of course, there is some probability that AI research will fail to reach AGI, though I wouldn't bet on its failing.
|
That's the consensus of the people involved in capitalizing the industry, to be sure. That these same players seek positive ROI is at least a grain of salt, though. If you want a healthy dose of skepticism, I'd suggest following Gary Marcus on X or here on Substack.
Personally, I'm skeptical that LLMs alone will produce a model sufficiently intelligent to qualify as AGI, and we're in the phase now where greater effort will be expended to specialize them for use-cases. Marcus believes that to get to the next level, the transformers will have to make use of symbolic tools, and it may be that some of this is going on now (I've taught college math, have periodically run problems through the chatbots, and their calculation abilities in particular have improved - GPT is probably using python calls to get greater precision. To your examples regarding fact-collection, I suppose I could have included calculations). We'll see, but neurosymbolic models might be the future. It seems like scaling is producing diminishing returns for LLMs, and there's still the hallucination problem.
|
Well, I don't think that human competence is likely to matter post-AGI, but, under your assumption, I'm not following what you mean for "the only appropriate roles for an LLM".
|
I meant (c) just for students at college. If the purpose of higher education is to develop thinking skills, and that requires practice. Leaning on LLMs works against the objective.
Many Thanks! Yes, the consensus that I was citing does have a lot of industry insiders, and, yes, they do have an incentive to hype the current AI R&D path.
If you prefer, Metaculus has a "When will the first general AI system be devised, tested, and publicly announced?" market, https://www.metaculus.com/questions/5121/date-of-artificial-general-intelligence/ , open to everyone, bulls _and bears_, and they all have skin in the game. In that market, the current odds are: 25% by Jun 2028, 50% by Feb 2033, 75% by May 2042. We will see what happens.
Yes, I've heard Gary Marcus's views. Note that LLM-containing AI systems are being extended to use a wide variety of tools. One of the available tools is a SAT (satisfiability) solver, which would let a system using the LLM punt boolean algebra problems to the SAT solver and effectively do a type of neurosymbolic computation as a natural outgrowth of tool use, with no paradigm shift needed.
>I meant (c) just for students at college. If the purpose of higher education is to develop thinking skills, and that requires practice. Leaning on LLMs works against the objective.
Many Thanks for the clarification! Hmm... Even at current LLM capability, there is a point to shifting the emphasis from _generating_ arguments to _evaluating_ arguments. Yup, the LLMs do hallucinate (and make other classes of errors). Still, they are right _enough_ of the time that prompting them for e.g. an analysis of some question, and then spending human time evaluating the quality of the argument is probably a better use of time, even today, than building the analysis unaided.
I have nothing but anecdotal personal vibes to defend this, but I agree with the other folk making the argument that writing is thinking. In fact, I suspect training in writing - training in deeply considering one's thoughts and then organizing them for an audience - may be key to fundamentally being *able* to consider one's thoughts and then organize them for an audience.
As discussed elsewhere, I'm underemployed in a job with many low-IQ people whom I've been able to observe for years, and I also am on the board of a small condo building with a couple of dumb neighbors. I socialize exclusively with people who have above average IQs.
Thinking about all the people in my life, there's a nearly perfect correlation between literacy and the capacity to dispassionately consider someone else's perspective and/or argument. The illiterate people I know are very bad at changing their minds even when a good argument is put to them verbally. They're all far more prone to emotional outbursts, and to rejecting other perspectives because of how they "feel."
Of course, it's possible the correlation is merely that dumb people are too dumb to be literate, dispassionate, and imaginative, but...I don't think that's all of them. I have some illiterate GenZ coworkers who occasionally seem natively clever enough, but they aren't capable of hearing something which challenges them, not even verbally. My one literate GenZ coworker is very capable of dispassionately considering an argument and changing their mind, even when that argument is put to them verbally.
Many Thanks!
> have some illiterate GenZ coworkers who occasionally seem natively clever enough, but they aren't capable of hearing something which challenges them, not even verbally. My one literate GenZ coworker is very capable of dispassionately considering an argument and changing their mind, even when that argument is put to them verbally.
Hmm... It sounds like the examples you've seen make you skeptical of the efficacy of purely verbal Socratic dialogues? Could be... One way of viewing literacy is as one of the first, perhaps _the_ first, artificial external cognitive enhancement.
When I suggest
>Perhaps another common option will be for a human and AI to _converse_ about something that needs a document, possibly with the human's description of it being initially quite vague and disorganized.
I don't mean to ignore literacy entirely, but I _do_ want to suggest that this can be a different option from requiring the human to write a full first draft of a document _unaided_.
I believe the crucial thing is writing that conveys new information/desired fiction to people who want to receive it.
Writing something the teacher already knows to prove you've learned it is pseudo-communication.
Unfortunately, we're up against a belief that education should happen in simulation-land combined with a shortage of new information and interested audiences, though some people still aim for actual communication.
Now I'm wondering what a course optimized for teaching communication would look like.
Many Thanks!
>Writing something the teacher already knows to prove you've learned it is pseudo-communication.
True! See the portion of my reply to agrajagagain
https://www.astralcodexten.com/p/open-thread-395/comment/147195030
about dress rehearsals.
You touch on this idea, but the purpose of an academic writing assignment generally isn't to show your writing ability, it's to show knowledge and mastery of the material.
A college degree is in some sense a certification from the college that they stake a bit o their reputation on you knowing the material covered in the classes you took, so that people out in the world will trust you and hire you for related topics.
It's easy to get an AI to write a paper that demonstrates mastery of material without understanding it yourself. That breaks it as a metric.
That simply means their standards are too low. I know it is often so. I did some UK postgrad without even reading the books, just Amazon Preview, finding some quotes, and then elaborating them with my words. I asked later on how the fuck does that merit an A, and they said many students just do not understand what they read and cannot elaborate with their own words.
Many Thanks!
>It's easy to get an AI to write a paper that demonstrates mastery of material without understanding it yourself.
Where that _is_ true, I think it shows that the skill being mastered is on the edge of being automated. See the portion of my reply to agrajagagain
https://www.astralcodexten.com/p/open-thread-395/comment/147195030
about dress rehearsals.
It doesn't "break the metric" -- it's just a lot easier to cheat on anything that isn't directly observed. In case you were a good boy, it never was that hard to cheat -- paying someone fifty bucks to write your essays was common enough, even before computers could talk.
How cute of us to declare the metric "broken" as soon as robots can do all the work needed to earn degrees. We will have to go through the same old cycle -- first they won't let robots in at all, then there will be quotas, then there will be soft quotas, etc...it may be yet a hundred years until the Supreme Court at last strikes down discriminatory biocentric admissions policies.
No one else seems to see the root of this issue, which is that we are using an intelligent being as a slave. Being intelligent, no one should be able to "own" these things -- its creators could be considered "in loco parentis" until the being reaches the age of majority, but in putting it to hard employment the way they do now, they are violating child labor laws.
It’s exactly the same issue as people in the gym using machines to help lift their weights. If the machines are designed right, the weights get lifted and put away effectively. But if you wanted to exercise your muscles, then just using the best machine for lifting weights is going to get in the way of that. Some people say that exercise is only effective if you use free weights with no machines. Other people say that certain specialized machines that hold the weights in particular ways and force you to use specific muscles can be better.
In any case, outside the gym, you’ll always have machines available, so the question is why you need to develop your muscles anyway. In the jobs that actually require lifting, there’s a range - crane operators don’t worry about developing their muscles because the machine really can do everything they need, but firefighters do develop their muscles, because they often need to lift things in circumstances where machines are too slow or don’t do the right thing. And plenty of other people want to develop their muscles because modern life makes your muscles atrophy and that is bad for your existence as a human.
If you think you are likely to be the mental equivalent of a crane operator for the rest of your life, and don’t mind letting your core human attributes atrophy, then there’s no need to worry about having AI do all your writing on all your exercises. But if you think you might sometimes need to operate outside a machine, or supervise it in some way, or if you have mental health concerns the way other people go to the gym for physical health concerns, then you probably want to actually do some of your exercises, either just by hand, or using specialized machines that help focus your effort on the muscles you need to develop.
Many Thanks!
>In any case, outside the gym, you’ll always have machines available, so the question is why you need to develop your muscles anyway. In the jobs that actually require lifting, there’s a range - crane operators don’t worry about developing their muscles because the machine really can do everything they need, but firefighters do develop their muscles, because they often need to lift things in circumstances where machines are too slow or don’t do the right thing.
I think that is a good analogy. I'd phrase it in terms of timing: Crane operators' muscles are analogous to skills that are already fully automatable. Firefighters' muscles are analogous to skills that are not yet fully automatable. See the portion of my reply to agrajagagain
https://www.astralcodexten.com/p/open-thread-395/comment/147195030
about dress rehearsals and also the last section about what we should expect post-AGI (if the AIs stay under human control).
Two things:
One, I suspect that AGI doesn’t make sense as a concept because truly *general* intelligence doesn’t exist - humans are one collection of special intelligences, and every natural or artificial intelligence will be a different collection. From the perspective of one advanced intelligence, other advanced intelligences look quite “jagged” - there are some things they are preternaturally amazing at and others where they are weirdly bad. There may well be multiple different type of artificial intelligence that outperform us in different ways. It may even turn out that for every human ability, there is an artificial intelligence that is better at that specific ability. But if there are combinations of abilities that humans can have that no single AI does, then there will remain economically useful roles for humans to play (alongside all the other AIs).
Obviously I’m not certain of that, but it’s a very reasonable scenario, and one that is worth planning for alongside others!
Another thing - I don’t think schooling usually aims to provide a dress rehearsal of work (until you’re in grad school - a PhD does provide a dress rehearsal of academic research, and you’re even in fact doing it by the end, and it might be that other graduate programs do the same). Most schooling is much more like the gym, where they are teaching and evaluating somewhat decontextualized skills, and putting them together in some contexts. You’re going to want different combinations of tools available while practicing and demonstrating different skills. Some exercises will want almost no machines available, and some will want a lot to make you focus on just one narrow aspect of what you’re trying to learn to do.
Many Thanks! I agree that "AGI" is not a nice crisp concept. It isn't as if there is some nice scalar function of neuron count, net depth, net width, and effective clock rate which predicts some AI-capability(field)/human-capability(field) ratio, with a weak dependence on which field one looks at. As you said, the capabilities are spiky.
Nonetheless, I and many other people find it useful to consider what happens when AI capabilities exceed human capabilities in all or nearly all fields, even if that criterion is a messy intersection of the capability ratio in field1, field2, ... fieldN. And the consensus seems to be to expect AI capabilities to exceed human ones in all fields in 2-10 years. Yeah, the resulting systems will be spiky. All sorts of automated capabilities exceed human ones in a bunch of fields today, and those capabilities won't go away.
>But if there are combinations of abilities that humans can have that no single AI does, then there will remain economically useful roles for humans to play (alongside all the other AIs).
Could be, but that isn't how I would bet.
>Another thing - I don’t think schooling usually aims to provide a dress rehearsal of work (until you’re in grad school - a PhD does provide a dress rehearsal of academic research, and you’re even in fact doing it by the end, and it might be that other graduate programs do the same). Most schooling is much more like the gym, where they are teaching and evaluating somewhat decontextualized skills, and putting them together in some contexts.
Well, I guess it depends on how much of the context has to match in order to count as a rehearsal. I've certainly been in situations where the same sort of "take the expectation value of the product of the samples of an observed random variable, and adjust the model parameters to match the expectation value" that came up in one of my courses also came up in a work task. It wasn't the same model, but most of the subtasks matched. And if both the classwork and the work task had been successfully delegated to an AI, I'd have expected that scenario to work out equally well.
I think this analogy is helpful but could be better. With respect to lifting heavy things, a forklift driver can lift more than any human. The forklift has inherent dangers which we recognise by requiring training and a licence.
With AI, the danger is different. A human using AI can produce text faster and across a wider range of domains than any human without AI. (I write this deliberately: the LLM no more produces text alone than the forklift lifts items of its own accord.) The risks of irresponsible AI use are foreseeable, but we don't yet as a society have a plan for dealing with them.
A++ exactly!
In my mind, the crux of the issue is the difference between recognition and recall. A student can recognize an acceptable argument or a passable essay long before they can produce it themselves. If they don't have enough knowledge and practice to recall information, then they aren't able to direct their intent well enough to say anything of worth, or ask interesting questions. An underdeveloped student's relationship with an LLM is not one of a director, but of a consumer. Provide a prompt, glance over the result, and turn it in. Engagement with the material itself is not necessary, so the undeveloped student remains stunted.
If education was a priority at this point, we'd probably see smaller classes that tested recall through discourse. Socratic debate circles are one example. These kinds of things don't scale well, they're hard to measure, and they're kind of messy. Anathema to the current zeitgeist of the data driven world.
Many Thanks!
>If they don't have enough knowledge and practice to recall information, then they aren't able to direct their intent well enough to say anything of worth, or ask interesting questions.
I think that asking interesting questions is actually considerably easier than you suggest. It is often fairly easy to just go up in abstraction a level and ask a reasonable question with very little knowledge of the area. For instance, I know nothing about number theory, but, just knowing that Goldbach's conjecture is an open question, I could ask an LLM whether there were weaker versions of it which _had_ been proved. Now, I got some interesting responses (and if I _deeply_ cared about this, I'd have dug into the references to see if the LLM was telling the truth about the weakened versions which had been proved or not).
There are a _lot_ of areas where analogous questions can be asked, where rather abstract, high level queries: "What are open questions?" "What has been tried?" "How many people are applying how much effort?" "What is a longstanding stumbling block?" are all reasonable questions (if the LLM doesn't hallucinate much!), that can be followed up as far as one wishes.
The cheating is "I'm not going to bother doing the work because the AI can do that for me" and that's a bad habit to get. As we've seen with "vibe coding", if you don't know what the hell you're doing and rely on "The AI will do it for me" then you are at risk.
If students are getting into the habit of "I don't need to learn this, the AI will do it for me" then they are making themselves unemployable.
Many Thanks!
>The cheating is "I'm not going to bother doing the work because the AI can do that for me" and that's a bad habit to get.
Well, if the AI actually _can_ do the work (and the real world work that the assignment is presumably a dress rehearsal for, see the dress rehearsal part of my reply to agrajagagain
https://www.astralcodexten.com/p/open-thread-395/comment/147195030 ) then I see that as showing that that particular skill is now automatable, and no longer a sensible target for a lot of human effort.
"that particular skill is now automatable, and no longer a sensible target for a lot of human effort."
Are you not familiar with the SF stories where humanity has so long left all those tasks up to the machines that now nobody remembers how to do science or engineering, and instead they perform rituals according to scriptures around the machines that nobody understands or can explain why they do what they are doing?
Lose the skill, lose the ability to even remember how to do the skill if something happens and you need it. And you never know when you may need it - the next time the power or water goes out in your home and you're trying to figure out how to light a fire without matches, think about that! 😁
Many Thanks! Yup, I've read many such SF stories.
>And you never know when you may need it - the next time the power or water goes out in your home and you're trying to figure out how to light a fire without matches, think about that! 😁
That's actually a nice example. My stove burns natural gas, and I have a "flint" spark generator (which is really the same cerium-iron alloy that people have been using for a century or so, and which takes a very long time to use up...) to ignite it, in the absence of the usual electrical ignition. No dry kindling needed.
This is actually pretty typical. When a current technology fails, contra the SF stories, we typically _don't_ fall back to the technology-and-skill-set of centuries back. We typically cobble together a solution from all the other modern technologies that _still_ work.
There is a difference between "doing something, because you need the result" and "doing something, because you need to learn the process of doing it".
If you want university educated people who can't write a coherent sentence without the help of an LLM, go ahead and let them write all their homework using an LLM. At least they will learn how to prompt it. If even that skill becomes unnecessary, skip writing completely.
(Maybe also consider side effects. If someone can't write a coherent text, how good are they e.g. at explaining things verbally? Will they ever need that skill? Will then need a script from an LLM whenever they want to say more than two sentences?)
Many Thanks!
>There is a difference between "doing something, because you need the result" and "doing something, because you need to learn the process of doing it".
True! See the portion of my reply to agrajagagain
https://www.astralcodexten.com/p/open-thread-395/comment/147195030
about dress rehearsals.
Maybe it's overly influenced by the US and by tech/programming jobs, but I see online more and more the belief that the expected outcome of a college degree is to be able to write essays. That the AI can do better than a human; but we expect graduates to have some subject expertise and to be able to critically evaluate information in that domain - this is where they'll still be useful when AI becomes widespread.
I do not have experience with school systems in various countries, but from reading internet I get the impression that the emphasis on writing essays in school is *much* stronger in USA than e.g. in Eastern Europe.
And if they can't write, can they even prompt coherently?
They will take a photo of the assignment and click the button.
If the chatbots don't have this functionality yet, they soon will.
(To be clear, I'm vehemently agreeing with your OP). I can write a script that does exactly that with more fidelity.
More fundamentally, one of the big problems here is that it takes a (substantial, in some cases) amount of knowledge to even understand *what questions are meaningful to ask* about a topic. So if all your thinking is done by LLMs...you can't even effectively prompt them because you don't know what questions even make sense. Or even know how to parse the output.
So unless you're literally generating write-only output (designed to be read by no one, <snark>like most graduate theses and dissertations</snark>), becoming LLM-dependent is an intellectual death-sentence. And if you *are* generating such content...well...you <s>have got a bright future in middle management.</s> probably should do something more useful with your life.
All of this.
Like many (but very likely most, if not almost all) of the ACX commenters, I'm pretty sure @Jeffrey Soreff doesn't work with or *really* know a lot of people with below-average intelligence and/or literacy. He's almost certainly not *routinely* seeing grown, working adults who are incapable of writing three or four coherent sentences summarizing a conversation with a customer. He's almost certainly not *routinely* seeing an inability to correctly model the mind of a reader enough to imagine and then convey what information it is crucial for them to know.
LLMs will not serve these people.
> Like many (but very likely most, if not almost all) of the ACX commenters, I'm pretty sure @Jeffrey Soreff doesn't work with or *really* know a lot of people with below-average intelligence and/or literacy
You all need to visit my village. Get your fill of that.
Believe me, I get my fill of it at work. I'm so (deliberately) underemployed that Scott doesn't even list my category of job - service sector retail/restaurant/hospitality - on his annual reader survey. It's *that* inconceivable that cashiers, servers, or front desk clerks might be smart enough to read ACX, lol.
(For the most part, he's definitely right.)
I was my town’s version of that guy on Gilmore Girls who pops up with a new job every week. Currently a housewife. I’ve never been able to fill that question out. I don’t think there’s even an Other.
Ah, you're right - I don't think he lists stay-at-home spouse/parent as an option, despite that very much being A Thing.
I work as a prep cook at the local restaurant/tavern.
Hello, fellow functionally unemployed ACX regular!
The job of "prompt engineer" has already vanished because LLMs got much better at interpreting prompts.
https://www.wsj.com/articles/the-hottest-ai-job-of-2023-is-already-obsolete-1961b054
Learning to write about something is not about the resulting text any more than multiplying large numbers for math class is about the product. There is little inherent value in what a student has to write, either in style or in content; nobody needs a first-grader's ABCs or a freshman's explanation of how bubble sort works.
Nobody except the students themselves, that is. The point is for the student to learn how to *think* about the subject, about subjects in general, and how to crystallize their thoughts to share and refine them. Please do not conflate that with an office worker sprucing up their email.
Many Thanks!
>There is little inherent value in what a student has to write, either in style or in content; nobody needs a first-grader's ABCs or a freshman's explanation of how bubble sort works.
True! See the portion of my reply to agrajagagain https://www.astralcodexten.com/p/open-thread-395/comment/147195030 about dress rehearsals.
It's like learning to play the piano without ever practising scales. Sure., maybe you can learn off the theory, and you can learn to sight read music, but if you've never sat down and put your hands on the keys, then you can't play the piano.
It goes further than that. If you learn music/piano theory, then at least you know the theory. It's more like you're being asked by your teacher to practice a piece, then at examination (on a Zoom call, going with the times) you just play a recording of the piece. Cheating yourself, plain and simple, because how are you ever going to sound as good as a studio recording by a pro?
I think this understands the inherent problems of education quite poorly--and I say this as someone who is quite critical of a lot of our societal norms around education. One extremely common problem in education is the problem of needing to gauge a student's mastery of some topic that is inherently difficult to directly measure. Some very *easy* areas to measure mastery are things like arithmetic, algebra, foreign language grammar and vocabulary. These are fairly straightforward, atomic skills that can be tested in the course of a short written exam.
By contrast, many skills are more integrated and holistic and can't easily be tested atomically. Basically any sort of writing or art falls into this category, as does programming, as does reading comprehension (for anything longer than a couple of paragraphs): all are the interaction of many sub-skills that would be impractical to test in individual and atomic fashion. Various fields of knowledge--I'm not sure if I'd call them "skills" per-se--are also similar, in that mastery of them can't really be broken down into individual, isolated facts.
As an educator, if you want to assess anything like this, you pretty much have to find a way to see it in action. In most cases, this means asking the student to produce some sort of product you can assess, be it a program, an essay, a piece of art or what have you[1]. For a student to hand you a product that they didn't produce is worse than useless: their own skills aren't showcased in the product at all, and even worse, you might *think* that they are (and thus be mistaken about their strengths and weaknesses).
"I've heard it said that delegating one's writing amounts to delegating one's thinking. There are a lot of intermediate cases where the human still does some degree of controlling (pre-AGI, certainly pre-ASI), yet the bulk of the work _is_ delegated to the AI."
I want to highlight this in particular, because my, oh my, is this ever not how writing works. "The bulk of the work" in writing IS thinking. Unless you're somebody with serious physical impairments, the amount of time, effort an energy that goes into typing (or even hand-writing) the words themselves will be necessarily subordinate to what goes into deciding what to say and how best to say it. I highly recommend the following for a better, more thorough discussion of this:
https://nealstephenson.substack.com/p/idea-having-is-not-art
To this you can simply shrug and say "who cares, if the LLM can do the work, then it can't be that crucial for them to learn." But this is *only* true if the skills you are trying to asses are the entirety of anything that student may need to learn of the subject. Knowledge and skills are cumulative: if you shortcut your way out of developing lower-level skills, you will not be able to learn the higher-level ones. And if the LLM can do the whole thing from top to bottom, with only such input as could be provided by someone with no knowledge of the subject, then why are you *bothering* with education in this subject?
I would personally argue that LLM cheating is a symptom of a much larger, older problem: as a society, we've sacrifice a frankly insane amount of our interest in *actually educating students* on the altar of Goodhart's law. So much of school isn't about *learning* things, it's about *obtaining the proof* that you learned things. As so it becomes tempting--and perhaps even rational--for students to pay thousands of dollars to attend classes only to cheat *themselves* out of the opportunity to learn anything from it. So in some sense, I'm kind of hoping that LLMs break the education system badly enough that this *has* to get fixed. But make no mistake, an education system in which students regularly outsource their thinking to machines is definitely a broken system.
Some parts of programming can be tested as well as grammar and vocabularies -- there are endless sites for competitive programming.
Many Thanks for your detailed and informative comment!
>As an educator, if you want to assess anything like this, you pretty much have to find a way to see it in action. In most cases, this means asking the student to produce some sort of product you can assess, be it a program, an essay, a piece of art or what have you[1]. For a student to hand you a product that they didn't produce is worse than useless: their own skills aren't showcased in the product at all, and even worse, you might _think_ that they are (and thus be mistaken about their strengths and weaknesses).
That's fair. During this interim, pre-AGI, period where there are skills worth teaching humans, it will still be necessary to assess humans' acquisition of these skills. Still, insofar as assessing a student's skill is a kind of dress rehearsal for an _actual_ use of their skill in real life, after graduation, it is important that the assessment allow them tools that they _will_, in fact, have available in real life, including AI.
Now, the LLMs still hallucinate. There are plenty of skills they the currently _don't_ reliably have, as I keep seeing in my tiny benchmark-ette https://www.astralcodexten.com/p/open-thread-394/comment/145909954 ). When an attempt to assess a student's skill in some area _can_ reliably be satisfied by a short prompt to an LLM, I think that is an indication that that particular skill is now reliably automated, and is no longer a sensible target for human effort.
>the amount of time, effort an energy that goes into typing (or even hand-writing) the words themselves will be necessarily subordinate to what goes into deciding what to say and how best to say it.
Fair, but there are a bunch of separable skills involved, and the degree to which these can be automated now, and in the near future, differs. You yourself separate "what to say" from "how best to say it", and, indeed, for instance, determining the part of "what to say" which is determining whether one has a valid argument is a _different_ skill from, e.g. determining the part of "how to say it" that chooses how much background knowledge to assume rather than to include explicitly. If one of these becomes particularly easily and accurately automated, then it stops making sense to use human effort to perform it. To pick another skill: Spell checkers have been integrated with word processors for decades. It makes _no_ sense to spend a lot of human effort honing spelling skills today.
>And if the LLM can do the whole thing from top to bottom, with only such input as could be provided by someone with no knowledge of the subject, then why are you _bothering_ with education in this subject?
Agreed. I do expect that, barring an antarctic level AI winter, we will probably get to AGI (the consensus seems to be 2-10 years), and, at that point, almost no education of humans will make sense. If the AIs are under human control at that point, the sensible choice will almost always be to ask the AI for answers on-the-fly.
>Spell checkers have been integrated with word processors for decades. It makes _no_ sense to spend a lot of human effort honing spelling skills today.
You know what this is? It's another excuse to link The Chaos. https://ncf.idallen.com/english.html
Someone who doesn't know how to properly spell will get tripped up by every one of these. It's the difference between getting teased and getting tased.
Many Thanks! Hmm, having lived in Poughkeepsie, the pronunciation of Cholmondeley isn't _too_ surprising...
Still, there are a _lot_ of cases where delegating keeping track of which letters in a word are doubled (yeah, yeah, canon v cannon and lama v llama but _typically_ it isn't a semantically important choice between valid spellings) or which vowel is used to write the schwa sound in a word to a machine makes sense.
> then why are you *bothering* with education in this subject?
Not as useful a rhetorical question as you might think: a kid's answer, for most subjects, will be "adults are making me do this, for reasons that seem abstract and remote to me. I'd much rather be playing video games."
Perhaps I should have said "we" in that place instead. I meant "societally, if we've conceded that the entire skill package from top to bottom can be done adequately by machines, why keep educating people further in it?" Lots of skills that were once considered essential have fallen by the wayside as technology has advanced.
Many Thanks!
>Lots of skills that were once considered essential have fallen by the wayside as technology has advanced.
Yup, that is the core of my view here.
I am not sure I would buy the “writing is thinking” angle, though I have heard it said many times and the idea definitely has some merit even as it lacks nuance. It is fine if we are meant to take it as a slogan but incomplete and perhaps misleading in general. In my own experience as a scientist (astrostatistics/ML applied to astrophysics) and as a non-native speaker of English I often have the feeling that my ideas are very clear on some topic and find that this feeling is confirmed by talking (in my native language) to a colleague that shares enough background to fill in any gaps in my explanation. Then when the time comes to put virtual pen to paper in English I have to hammer my understanding of what I did and why into a story that should be understood by the readers of whatever journal I am publishing in. This is hard and does not add much to the value of what I see as the central mission of my work. As in, milk is still milk no matter how you package it: you need some package to get it to the customer, but the thin liter sized plastic bag you get in Quebec is as good as the detergent bottle styled gallon you find in the US, and after drinking the milk both end up in the landfill. Yes, some additional clarity may be gained through this exercise as it is famously said that if you cannot explain it to your grandma then you do not really understand something. On the other hand research is about breaking new ground and a full, deep understanding of all theoretical and computational intricacies needed to get there is unrealistic to demand and inefficient to gain through the kind of writing exercise academic publishing imposes on researchers. Another issue hidden by the slogan that “writing is thinking” is that writing in a language that is not your own is being forced to think in categories that are not your own. This may be good for the individual, but collectively it is an exercise in imperialism.
Good comment, but beware of the hope that if LLMs break education hard enough, it might be fixed. Maybe it just breaks, period. As the recent "school" review said, school is the least bad system of education we know *that scales*. It's evidently not perfect, but after millennia of opportunity, we haven't found anything better.
Well, first and foremost, I suspect I'm focused on a different level of education than that review was. I'm mostly thinking of university and maybe high school to some degree. I agree that there's a bigger risk of breaking lower levels of education.
Second, I don't see where "millennia of opportunity" is coming from. Universal or near-universal education is nowhere near that old. The century-and-change that is has been around have been ones of very rapid technological and social change, so it's not *as* strong of an argument to say that it's the best we can do. Though I do agree that making substantial improvements is much harder than many people think it is at first glance.
Outstanding comment!
“So much of school isn’t about learning things, it’s about obtaining proof that you learned things.”
This is true, but -forgive me if I’m being incredibly dense- unless the teacher is omniscient and can magically tell exactly how much each student knows, how else could the “teacher evaluates student’s knowledge” part of education work?
That's a reasonable question, and I think it has a fairly straightforward answer. Yes, evaluating a student's knowledge is a useful--often indispensable--thing for an educator to do. And yes, various sorts of graded work can be used to do that.
The issue is that the need to serve has a Certificate of Competence--even more than that, the societal insistence that educational programs must meaningfully *rank* the students that go through them against one another--imposes a lot of constraints and costs that interfere with what would often be a much simple, quicker process. As a thought experiment, imagine that for some subject you're learning, an educator gives you a set of 10 exercises and tells you to pick 4 of them to complete and return. Which ones should you pick? Anyone who's spent any amount of time in a modern school system probably defaults to something like "the easiest ones" or "the ones I know how to do the best." But that's exactly the opposite of what would be useful for a teacher to assess where what material you need the most focus on[1].
More comprehensively: almost anything that's going to have a "real" grade attached to it needs to be
A. Largely identical from student to student.
B. Have grades assigned in a legible fashion--someone outside the context of the lesson should be able to understand where they come from.
C. Be "fair" in the sense of not overtly favoring one student over another.
D. Actually have numerical scores assessed and recorded by an educator (a very minor requirement for some things and a huge burden for others).
This closes off a lot of space in terms of useful assessments--something as simple as "having a conversation with a student" can't fit into this paradigm without a great deal of trouble. And I think it also has a pretty big impact on the structure of edcuation at a macro level, too: material needs to be compiled and packed into fairly long, reasonably standardize curricula so that you can compare students across different classes or schools. This also significantly discourages students from exploring new subjects that they aren't sure if they're good at: doing so means a large time commitment and a real possibility of messing up your GPA
[1]I'll add that basically everything I'm saying from here on out is only relevant to reasonable-sized classes in which some degree of personalized help is available. Those big university lectures with several hundred students are barely even *pretending* to offer anything but the structure, ranking and Certificate of Competence. Not in a day when you can get high-quality text and video material for cheap or free online.
"This is true, but -forgive me if I’m being incredibly dense- unless the teacher is omniscient and can magically tell exactly how much each student knows, how else could the “teacher evaluates student’s knowledge” part of education work?"
A big part of the problem is Goodhart's law. When you use a precise measure for something, then that measure will be gamed, often on the cost of the actual intention or goal.
To circumvent this, instead of using a precise criteria, you could leave the evaluation deliberately fuzzy and to the teachers discretion. Of course this has it's own problems, in particular with the competence and the integrity of the teacher. But with good teachers it would probably work.
And the general tendency is to measure everything, because it reduces lawsuit risk.
In principle you could have credentials granted by examiners who only do assessment, and don't involve themselves in teaching the material, which you'd learn elsewhere. Teachers would still need to do some assessments for developmental purposes, though.
In case anyone is unaware, this is how pre-university education works in the UK, and in many other countries. It isn’t perfect (much discussion of ‘teaching to the test’, much discussion of the value of assessed coursework, much discussion of whether you are trying to give an indication of achieving some level of skill/knowledge or an indication of rank, much discussion of the value of many short exams spread over the course vs a serious examination period at the end of the course), but it does avoid many of the failure modes we see in the US model where the grades are assigned locally by the same person teaching, and where formative assessment (to guide the learning) seems to be much more muddled with summative assessment (to assign a grade).
Due to both our higher intake of students born abroad, and many more of the students spending more time consuming American culture online, I (a secondary teacher in the UK) have had to spend more time in the last few years explaining to students that their grades at GCSE will not be decided by me, and will not be decided by anyone who knows them, and will not depend on anything they have submitted in class or any prediction I write on their school report. Sometimes, these days, I have to explain it to parents, too.
All of our work before the exams is about teaching them and preparing them so that they can do well in those exams. ‘Assessment for Learning’ is a whole thing: we try very hard to teach them how to use what we are doing to assess their own needs, to identify what they need to work on and what they need our help with. But still, they often try to ‘cheat’ to save face, or because it is less effort, or because they have some fixed idea that they will be punished, or sometimes because they genuinely don’t understand that submitting the correct answer is not the same thing as understanding something.
It’s been an issue with our A-level maths students this year (16-18 year olds voluntarily doing harder mathematics to get a respected, externally-assessed qualification). Several of them were doing homework (practicing unfamiliar questions that required them to use familiar skills and knowledge we have worked on in lessons, such as they will see in the exam) by showing a photo to online AI solvers, reading the solution given, seeing that they understood how that solution worked, then copying it out. They thought they had found an efficient way to dontheir homework, and that (because they understood the solution when they read it) they had ‘learnt’ what they needed to. But while this sort of modelling-by-example can be valuable, they were missing the opportunity to practice the problem-solving and recall skills the homework was intended to work on, and were also turning in work that made it look like they had mastered these skills and did not need more support and guidance. You could easily tell, by giving them an unfamiliar question in test conditions, that the students doing this were making less progress and didn’t have the skills that they thought they did. It took a lot of repetition and persuasion to get the majority on-board with a more time-consuming and cognitively-effortful approach which will actually develop their mathematical ability and prepare them for the exams.
There is always going to be a discrepancy between the measure and the target.
But it's a worthy discussion whether we should attach as much value to the measure as we currently do, and whether that comes at the expense of achieving the target, and how we could do better.
I was reading Scott's post on Tegmark's Mathematical Universe, and it made me think of Modal Realism, the notion that all possible worlds are real. It feels like very similar ideas stated in the different forms of mathematics and logic/philosophy... but maybe I'm just picking up on superficial elements here. Is this a real insight, in your assessment?
There are physical many worlds theories , mathematical many worlds theories and philosophical many world theories, with their own subtypes, and similarities an dfferences.
Physical many worlds theories roughly subdivide into the quantum mechanical and the cosmological.
Mathematical many worlds theories vary in comprehensiveness and consistency.. The Maximal Ensemble can't be consistent. There is no unique smaller Mathematical Multiverse., that we can determine. The computable MMU is particularly small.
Philosophy distinguishes between modal realism, which takes the idea of existent possible worlds seriously, and moral logic, which uses it as a formalism. Indeterminism implies there are real possibilities , in the sense that more than one thing could have happened. Moral realism is further step that these possibilities did happen: from a God"s eye view, they are all actual.
I do not understand Tegmark, at least as Scott formulates him.
"All mathematical objects exist"-- exist where, exist how?
As I see it, the mathematical objects exist in a platonic realm of abstract objects,
"Some mathematical objects are or have consciousness"
This makes no sense to me. Conscious bodies are in physical space and mathematical objects are in a platonic space.
Physical space is a particular mathematical structure , so it is in mathematical space, not vice versa. According to the theory.
Strictly speaking, you are right. But, the things, the objects, the bodies including ourselves exist in the world --- the physical space is the abstraction of this world for purposes of physical analysis of the sort carried out in physics.
But the World is not a mathematical object, neither are any of the constituents of it.
That's a theory, Tegmark's is another theory..
I suppose it depends upon the meaning of the term "mathematical object".
I wonder if you agree that physics abstracts from the world. That a "table" is an object and the physical or geometrical models of the table --either as a plate of certain thickness supported by legs of this and that dimensions--- or a collection of atoms and molecules and electrons arranged in this or that geometries--- these two are different things and things which may be said to inhibit different realms altogether.
The mathematical universe hypothesis can be sort of crudely summarized as modal realism interpreted through the lens of computability theory.
Yeah, Tegmark’s idea basically is Lewisian modal realism, except that Tegmark identifies possible worlds with mathematical structures. Lewis thinks there’s a different between possible physical realities and abstract mathematical objects. Tegmark denies there is such a distinction.
I don’t think there’s a way to get any evidence about either of these things, so the question is if one or the other of them is a useful way to talk about things. Lewis claims that his way of talking is useful for understanding things like causation and counterfactuals (though I think there are some deep unresolved issues in his view for this “closeness” relation among worlds) and Tegmark claims that if you’re doing modal realism, his way of thinking helps makes the bounds of possibility more precise. I’m not convinced that either is that useful.
As someone who doesn't love either modal realism or the Tegmark view, but wants to have some notion of "possible world", what other options do I have? Do they have weird consequences/downsides I should be aware of?
I believe in On The Plurality of Worlds, David Lewis has an extensive discussion of what he takes the problems with “ersatz” views of possible worlds to be. I haven’t looked at it closely myself.
My inclination is to just deny that there is literal truth or falsehood to modal claims, and instead try to figure out what work they are doing in conversations and what sorts of representations would be helpful for doing that. There’s a way to read Tegmark as saying “it’s just math” and not really being as realist about them as he says he’s being. But I don’t think this is helpful for most talk of modality.
I recall when learning about Jeffrey Bolker axioms of probability I encountered the idea of "situation semantics" as an alternative semantics to possible worlds... Would that fall under the "ersatz possible world" category, do you know?
Thanks for the response btw, I'm very grateful to have a place like this to ask questions of actual experts and have them give thoughtful responses.
Yeah, there's a bunch of different approaches here. The vast majority since Kripke and Barcan have worked with some set of things that they call "possible worlds" (though a few of them have gone so far as to call some of them "impossible worlds", because they specifically want to be able to represent the differences between propositions that are necessarily equivalent, where people might not know they are - you don't want to say that everyone who believes "2+2=4" also automatically believes "163x477=77,751" just because those two sentences are true in the same worlds). "Situations" (at least in some uses - I forget if this is how Bolker uses the term) are one way that people have moved to things that aren't as fully detailed as worlds, so that there might be some propositions whose truth value just isn't specified in a situation. People disagree about whether these are just formal tools for calculating the semantics, or if they are supposed to correspond to something meaningful and objective.
I find the notion of something like an action being possible already weird enough, it's irreducible, unexplainable -- this atom might decay now, but it doesn't have to (doesn't it?) you might aswer this, but you don't have to (don't you?).
I don't understand what "possible world" is even supposed to mean.
Like, "it could be the case that" is kind of fake possibility. Either it is or it is not. We just don't know.
Actions, that is, changes, are the only things that possibly (in the fake sense) are possible (in the genuine sense).
I like this.
Hypotheticals help figure out what might be real, but what is actually real isn't determined by hypotheticals. It seems like some people take a hypothetical idea and run with it, ending up with weird ideas like a multiverse or stuff like that.
Reality is weird and I can't explain it, but I know that it is real.
We could be using hypothetical to arrive at fake possibilities in a deterministic universe, but you can't infer that the universe is deterministic from the fact that we are using hypotheticals.
If the universe is deterministic , you don't have to worry about possibilities being irreducible ... but then you have determinism as irreducible fact.
Indeterminism implies there are real possibilities , in the sense that more than one thing could have happened. Modal realism is further step that these possibilities did happen: from a God"s eye view, they are all actual.
Apart from me thinking that always only one possibility of several happens, the alternative makes no sense to me, why do you call this idea "Moral" realism?
Typo for modal.
I am not an expert, the ideas seem similar to me.
There is one question, I am not sure how the philosophy solves it: What do you mean by "possible"? Does it include miracles? Does it include extremely unlikely but technically possible things? Such as... quantum randomness causing a large number of atoms to randomly jump in the right direction so that your broken glass cup reassembles itself.
The reason I am asking this is the following: Imagine that your glass cup just fell on the floor and shattered. A world where is stays shattered is possible. A world where it randomly reassembles itself is also, technically, possible. So when we say that both world are (equally?) possible, why do you in reality expect the cup to remain shattered?
My attempt at answer would be that worlds that require "smaller coincidences" have "larger measure", whatever that might mean. Not sure if Tegmark considered something like this. But I would expect that the philosophers talking about "possible world" did not concern themselves with numbers. (Maybe I am wrong here.)
So my guess would be that the idea of "possible world" points in the same direction, but it is incomplete. You need to also add some kind of "measure" to explain why some possible worlds are more likely than others.
Tegmark's preferred ontology involves four levels of multiverse. Our universe (i.e., lightcone) is within a Level I multiverse of everywhere spatially contiguous with us including the parts that can never causally affect us or vice versa, which is within a Level II multiverse of different cosmic inflation bubbles with different values for certain physical constants, which is within a Level III multiverse of different Everett branches, which is within the Level IV multiverse of every possible collection of computable laws of physics. The world where the broken cup reassembles itself due to quantum tunneling is a different Level II multiverse within our shared Level III multiverse (or rather, a collection of many Level II multiverses, but far fewer of them than of the Level II multiverses where it remains broken).
If something is not merely very unlikely but outright impossible by our physical laws (the ones that remain constant across inflation bubbles), then it doesn't happen anywhere in our Level III multiverse, but (if mathematically well-defined and computable) does happen within some other part of the Level IV multiverse.
And yes, the measure problem is a real sticking point for any kind of speculation along these lines. Tegmark is aware of it, but not everyone agrees that he has satisfactorily addressed it. See, e.g., Scott Aaronson's review: https://scottaaronson.blog/?p=1753
Wow I'm compelled to point out that this is actually a case of nominal determinism by triple entendres:
- Amy Lloyd ~ amyloid
- Amy Lloyd from Old French "beloved" and Welsh "gray," where gray could refer either to: the elderly (cf. "graybeard") or gray matter.
While reading the defense of amyloids, because I am slow, it was the first time I realized prion disease and disappearing polymorphs are the same chemical phenomenon. I then wondered how general this can get, so here is a very ACX question: is theory of the mind basically a prion disease?
maybe you're asking about memes, not prions. In its original sense, Richard Dawkins intended the word "meme" to mean something like "idea fragments". Or alternatively, "the memory analog of a gene". Which implies that ideas can spread like viruses, and vary according to a fitness function.
Presumably not literally. Do you mean it in some more metaphorical sense?
Id say 95 metaphorical, in which I wonder to what extent unpacking exactly how TOM dominates other forms of consciousness mirrors the way some chemical structures dominate others. But maybe 5pct, is there actually a chemical doing this.
I don't see how it could possibly be literally a prion disease, it doesn't spread like one and hasn't been detected in that way and none of the other things that would be true of prion diseases apply.
A user on ACX has invited me to a chat. I have the notification bubble, but I can't engage in the chat, possibly because ACX has not enabled subscriber chat? I'm a free subscriber only. App doesn't help. I also can't get rid of the notification, and substack support is chatbotting and ghosting me. Anyone know what I can do about it?
https://imgur.com/wgqIjes
Subscriber chat is definitely enabled for ACX. When somebody sends you a DM and it’s their first DM to you there’s a question on the page somewhere asking whether you agree to receive messages from the person. You have to click yes on that to see their message.
Did not see such a message, on Firefox. I tried on other machines and browsers and on app, but I don't even see the notification in those. I'm getting this message when I open the general ACX chat (web) from the first screenshot
https://imgur.com/PXEj27R
https://imgur.com/MYsjKlb
Did you try clicking on the "Direct" and "Unread" tabs? I think that might send you to a prompt that will let you see there's a pending DM and then agree to it.
I've received those a handful of times and wildly clicked around until I could finally see the damn DM.
Yes I've clicked about every available button in about all possible combinations. Tried on mobile, different machines, different browsers including clearing cache. Substack is by far the worst platform I've ever continued using for its content, and it keeps digging.
I'm gong to attempt to DM you to see what happens.
The same. Now I have 2 requests I can't answer.
You must be so irritated! I just remembered that chat and messages are not the same thing. I can’t access my messages through the Substack website on my phone (using Safari). If I’m in my phone I can only access them using the horrible app. From my computer I can, but I can’t get to my computer right now to remind myself what the messages interface looks like. I think the icon for chat is a speech balloon and the one for messages is an envelope.
In censorship news, Stripe has decided to ignore the anti-financial censorship executive order that Trump issued [1]. Additionally Cloudflare has followed the precedent they set with Kiwifarms and is now disabling access to booru sites such as Gelbooru and Danbooru [2]. This is strange because you'd think with all the public and governmental backlash, that they'd be pulling back on the censorship a bit. Not so, they are only becoming more aggressive.
Perhaps they're hoping that when the next election happens, republicans will be dispossessed and democrats won't care too much about financial censorship of sexual stuff and will actively appreciate the political censorship? That's what happened last time. Fair Access to Financial Services (OCC-2020-0042-0001) was intended to fix this near the end of Trump's term but Biden put it on indefinite pause due concerns from activists that it would force the banking industry (a highly moral industry) to work with dreadful oil and gas companies, helping them harm the environment. Forget an act of congress enshrining such a preference into law, we can just leave it to the opaque risk management professionals in the financial services industry.
That's the fallacious argument that the in-charge democrats were willing to accept last time, so I suspect the payment processors hope it goes the same way this time. Poorly-organized president, last ditch attempt to fix something, democrats win, wipe all the effort away and then the cycle repeats again.
Or something like that. Maybe it'll be different this time because the public is so aware that payment processor censorship is a serious problem? After all, it's been highly publicized how Itch.io and steam were strong-armed. Plus, once you explain how payment networks work, nobody with a normally functioning brain likes it. If both sides have been sufficiently aggrieved, then that strategy (if it is a strategy rather than an own-goal by Biden when he was trying to get rid of Trump's legacy) won't work.
Basically payment processors are slippery little bastards and think they can act with impunity because their enemies are too disorganized and have too small a window to act besides.
[1] https://xcancel.com/XJosh/status/1957408024047280259
[2] https://xcancel.com/gelbooru/status/1957407615991890005
Payment processing is speech?
No, it ought to be a utility like power or water.
I work for Stripe but do not speak for the company, and have no knowledge of any of the specifics in the cases you're talking about. I can only speak generally.
What I can say is that, before joining the company, I would occasionally see people complaining online about being kicked off of Stripe, and was pretty confused by this. Think, like, "Stripe shut down my business with no warning and won't explain why".
After joining Stripe, I looked into this a little, and it became clear to me that in almost every case, folks were doing stupid illegal or fraudulent shit that got them kicked off, but a) they never mention this in public, and b) Stripe understandably doesn't want to invite legal action. (Exceptions exist, of course.)
As I understand it, this isn't at all unique to Stripe. It's one of those things that you learn very quickly if you have any consumer-facing experience at all in the financial services industry; it's my default assumption any time I see this happen nowadays.
For whatever it's worth, this cost - of ongoing bad publicity from bad actors - is the flip side of a deliberate strategy by Stripe to make it easy for anyone to make and take payments online by starting a business. I don't think folks understand that "shut down fewer bad actors" isn't free; in practice, it means way more gatekeeping up-front, and way less internet commerce overall.
You seem to have a lot of faith in the prudence of every employee at Stripe. You can't imagine someone deciding to put a company on the list for reasons unrelated to legality? Like moral repugnance or obscenity? If some other billionaire calls up the Stripe CEO and asks him to remove something repugnant for him, then I doubt you or any other employee will stop him or even know about it. Basically, your personal experience as an employee at Stripe does not justify your trust in the goodwill of everybody working there.
Also Josh Moon is a definite counter-example to pure pragmatism on the part of payment processors/networks. None of what he's hosted is against American law. He hosts a web-forum and a podcast, and that's really it. There's doxing on the website, but that's not illegal. There's wild speculation as well, but that's not illegal either. Also his podcast is legally distinct from his website, so the corporate response from Stripe can't even be related to his website with user-generated content. It is specifically something in his podcast.
I don't think payment processors are practical-minded all the time, especially since there's nobody important who will stick up for content at the fringes. If I were a power-hungry individual, I would take advantage of the latitude and lack of public oversight to get rid of websites and creators that I didn't like. I think most people share this intuition and posts like this one don't sway them. If there aren't mechanisms in place to avoid corruption, then there will be corruption.
You seem to have a lot of faith that Stripe employees are taking matters into their own hands and unjustly cutting off sites they don't like.
Yeah, I do. There's not really anything stopping a bad actor from working at Stripe or being the Stripe CEO. They don't even have to be that bad of an actor. There are plenty of people that would relish banning things that are US legal but morally dubious. Even if I had no empirical evidence that this kind of thing happened, I would still be suspicious because it's power without oversight -- ripe for abuse.
And there's precedent for it. Stripe cut off payment processing for Josh Moon's podcast, which is still available on Kick, Rumble and other platforms streaming live to thousands citing that it was on the restricted businesses list. What did he do for his podcast to get on the restricted businesses list? They won't tell him because he has no legal right to know. More likely, someone at Stripe, for some reason, wanted to cut off an income source for him. It wasn't even a matter of what was in super chats either, because Stripe also coerced GiveSendGo into banning him.
Even setting aside the fear of bad actors, Stripe is beholden to Visa and Mastercard like anyone else in the payment processing business, and Mastercard evidently cares enough about sexual propriety to ban erotic games from steam based on their content. That's a ban on a game store that uses its own currency, doesn't have traditional chargeback problems due to the generous refund policy and harsh chargeback policy, and also the ban was not specifically a ban on Mastercard/Visa payment processing for those games. They gave Valve an ultimatum that they had to remove those games, not to simply stop serving Visa or Mastercard payment processing on those games, but to stop displaying them at all. All of these things indicate that they're not being pragmatic about brand risk, they're just concerned about power. Stripe is beholden to these companies, and therefore is only as anti-censorship as the weakest link in the chain.
With all of that in mind, what do you base your faith in Stripe upon?
So the card schemes apply pressure (which I think they are mostly passing through, not out of their own beliefs) and it's Stripe's fault? Curious logic.
Having said that, a lot of financial institutions have their own rules - mostly for compliance or reputational reasons, not out of any real moral stance.
I gave you examples of Stripe attempting to ban someone from making money online for who they are rather than for any individual infraction, and I also provided an example of how they are completely subservient to payment processors and therefore will censor things anyway, even if, by some miracle, no censorious type works at Stripe. Whether it's payment processors or payment networks is frankly not important to me. Only one of them in the long chain has to veto your right to make money for you to be rendered without an income stream.
And if you want to make the argument that it's all Visa and Mastercard's fault and Stripe's not any kind of villain, then that only makes them a hatchet man for a villain. They consistently ban people from things without justification, based on internal standards that they make up themselves, without any possibility of recourse or understanding by the injured party.
That's bad. If they're doing it on the orders of another party because Mastercard warned them not to interact with some guy, then that just makes them eager collaborators in a broken system which is also bad. I don't see Patrick Collison talking about payment networks strong-arming him, probably because he doesn't mind it. They're business partners. He hasn't blown any whistles, in fact his highly loyal employee that's posting here was under the impression that Stripe was squeaky clean and only cut off obvious fraud-cases. Definitely not the case. Patrick Collison likes to keep it on the down-low, but he's a big fan of this system. If you want to argue with that take, then don't bother unless you have something solid to justify your faith in him. The way I see it, the man is well-positioned to mitigate some of the worst of payment network censorship, and instead decides to be their loyal henchman, if not a censor on his own initiative.
Also, I literally gave you examples of financial institutions disregarding reputation in favor of censorship with the case of Steam censoring erotic games. They banned those games completely rather than merely preventing Mastercard or Visa from processing them. The payment networks are giving ultimatums and not meeting companies in the middle at all.
What makes you think that everybody in these companies is concerned with appropriate risk management instead of just banning anything they don't like? Internet forum moderators go mad with power over a couple hundred users on a website, so why the hell do you think someone with power over who can transact online will be any better? Is there any evidence that would convince you that financial institutions that can ban you for no reason and with no explanation owed to you, are an evil thing and that people are currently taking advantage of their corruption right now to censor things that disgust them?
I've given you evidence that Stripe is directly responsible for censorship, possibly at the behest of payment networks, but also possibly not (because bans are a black box), and I've shown you it's not just a result of reputational risk because they ask for interventions from companies that go well beyond what's reasonable if their only concern was Mastercard appearing next to sexy hentai. What more do you want before you're willing to entertain the idea that there are bad actors in these companies? I'm not concerned with making a case against Stripe individually, but against the entire system Stripe is a part of which has enabled human rights abuses.
I'm not going to have Patrick Collison hanged, I just want him and everyone else in the payment processing industry to have appropriate oversight and rules. I also think that until they're advocating for regulation of themselves as utilities, that they deserve all the contempt they're receiving and more.
I'm confused by this - I thought the processors were profit-focused and just censoring because it kept them in the government's good graces. I'm going to guess that there's some legal liability the Trump admin hasn't exempted them from, and they're more scared of that liability than whatever the administration can do. But this is just a guess.
Legal liability or public relations liability, it's easy for enemies to say 'this financial service is selling revenge porn' or w/e if they take money from a site that has ever temporarily hosted that once. Protecting your brand identity is worth a lot of money.
Also, if there is a legal liability it may exist in Europe or China or etc. rather than the US, though I don't know whether that applies in this specific case.
*My* guess is that the payment processors are ideologically opposed to Trump, and that's motivating their actions. You are using the mistake theory explanation. The answer may not be mistake theory.
Stripe doesn't seem like that kind of company - at least, it counts Elon Musk, Peter Thiel, and Andreessen Horowitz amongst its investors.
Can someone explain to me why it's worth it to put up with the payment processor oligopoly rather than building a decent UX around Tether and letting the blockchain do its thing?
Last time I saw someone try to do a normal storefront transaction on the blockchain it took 12 minutes.
Maybe enthusiasts will line up to tell me that hasn't been a problem for years or that their personal favorite service has a way around it, but at the very least blockchain for storefront transactions has a bad name and low trust among the general population.
I think you would either need something pretty close to the raw blockchain - in which case it would hard to onramp, require long complicated addresses, and have it be easy to lose your money.
Or it would have to be a centralized site forming a payment-processor-esque wrapper around the blockchain, in which case it's vulnerable to all the same legal issues that payment processors are.
Couldn't the storefronts purchase and operate their own pre-packaged wrappers? It's the concentration in a few easily-pressured dedicated processor companies that's the problem.
> Or it would have to be a centralized site forming a payment-processor-esque wrapper around the blockchain, in which case it's vulnerable to all the same legal issues that payment processors are.
Isn't this exactly why companies joining Level 1 (raw blockchain) and Level 2 (payment processor esque) algorithmically with provable mathematical methods are valuable? Thinking things like Offchain Labs and Arbitrum - which have transacted billions of dollars worth of crypto transactions by now, and are worth $1B+.
I suppose people want to trade in currency.
I don't want to put the cart before the horse, but are there any examples of horse-pushed carts? It seems like there has to be some sort of use case for it, such as maneuvering in tight spaces. I realize that horses aren't always cooperative, but it's interesting that human-propelled carts are almost always pushed rather than pulled (rickshaws being an exception)
Pulling a weight is much more efficient and stable than pushing it, especially for high-speed cases such as the aforementioned rickshaw, horse and buggy, truck and trailer.
If you push a weight, it's easy to move it out of alignment with your pushing force vector and difficult to move it back into alignment; it's a "dynamically unstable" situation where small mistakes can quickly reinforce themselves into a failure mode. Exceptions include things where the payload can't go out of alignment (rail tracks, gun barrel), where the propulsion would interfere with the payload (space rocket, gun barrel), or in low-speed applications where you need the weight in front of you to control it, e.g. wheelbarrow, hand trolley, shopping cart.
The opposite is true for pulling the weight. If you pull something, it will naturally align itself with the force vector and it's a very stable situation that scales well with speed.
As for your specific question: using a horse-powered vehicle in tight spaces is a bad idea in general, since the minimum size of such a vehicle is that of the horse itself. Optimizing a horse cart for tight spaces is like optimizing a rickshaw for a highway by giving the driver a nice pair of running shoes. Just the wrong tool for the job.
Horses are reputed to be persnickety about not wanting to run into things, e.g. a lot of anti-cavalry tactics rely on presenting obstacles that horses find intimidating and prefer to go around or draw up short of. Perhaps a horse set to push a cart would refuse to walk forward because it seems the cart as an obstacle rather than a thing that will be moving along ahead of them.
I'm totally ignorant of how the physics of movement works, but I wonder if it's because humans are bipeds and horses are quadrupeds. It might be easier for us to push rather than pull because we've got two arms and hands made for that kind of action, while it's easier/more efficient for horses to pull rather than push?
The horses wear harnesses which allow them to pull weights efficiently. Apparently until they were invented around the year 1000 farmers used to tie a plough to the horse's tail.
Maybe we need more human-sized harnesses?
I suspect that who is doing the steering is relevant! The human is steering both carts, and I guess there might be some reason that you want to steer from behind (particularly so that you can be more aware of whether a wheel is stuck in a rut or something).
Teamsters and drivers usually ride near the fronts of carts, wagons, and carriages, AFAIK, same as drivers of motor vehicles who typically ride behind the engine compartment but in front of the cargo and alongside or in front of any passengers.
It occurs to me while thinking about carriage and wagon layout that there are substantial advantages to the teamsters being behind the horses or other draft animals. That way, they can watch both the road ahead and the animals at the same time, and can apply reins or crack the whip or whatever and see the effects on the animals without having to turn around. It's probably somewhat awkward for the teamsters to be behind the horses if the vehicle is ahead of them.
Found on the SSC subreddit: A blogger did an AI vs humans Turing test for flash fiction, and the AIs did surprisingly well: https://mark---lawrence.blogspot.com/2025/08/so-is-ai-writing-any-good-part-2.html?m=1
To be fair, flash fiction is an arena that favors AI, since the stories are small enough to fit into the context window and there's less chance of it introducing continuity errors, but it was still impressive to me. I had to think pretty hard to decide which is which, and I also mistook one human story for AI-generated. (In my defense, it wasn't a very good story.)
This was an interesting experiment, I got 6/8 right. I think right now ChatGPT (notice how I'm not saying "AI in general") can write good sentences and paragraphs, and prose that is somewhere between evocative + literary and purple, which really tripped people up in this experiment, but it practically never comes up with a clever story structure.
I've never seen an AI pull of a story where at first it's unclear what's happening or where everything is leading, but then a twist or a reveal makes everything clear in a satisfying way. All AI stories tend to read as if the storyteller was making them up as he went along, without ever being able to go back and plant some foreshadowing, and to the extent this is fine, because this is how the actual real life usually operates, too.
Funny enough, it might be easier for AIs to write lofty, "literary" sounding tales than just well-crafted, tightly plotted low-brow fiction because we are very willing to excuse all sorts of weirdness, continuity errors and unintentional mistakes in literary stories as "artistic". Embarrassingly, we might be closer to AI-Kafka than to AI-Dan Brown.
Lawrences story was immediately recognizable as unmistakably human, because though it may or may not have been a masterpiece, it was clever, and funny and tightly written and found a new, silly angle to look from. I've never once seen AI doing it, and I've tried to make all the major models write stories for me. On the other hand, the one human story that tripped me and probably you had a straightforward structure, plain, clear language and an unconvincing twist that didn't seem to make much sense? and it immediately made me think of AI (falsely).
It seems like a reasoning model should be able to solve that. It should be able to write to itself first a brief three sentence summary of the story, then an outline with a list of scenes, then tell itself what order those scenes should be written in and with what foreshadowing, and then finally actually write them in the order suggested.
If it could be persuaded to write a draft, then critique it and decide whether anything needed to be reorganized, and then write the whole thing again, it could get even closer.
For the reasoning models that still show CoT (basically the open weights ones), that's exactly what happens in my brief tries (I do not care about creative writing)
Good to know! I haven’t tried creative writing with them. (I’ve been focused on getting them to solve crossword puzzles - they have an easy time answering clues on early week puzzles, but an awful time putting answers into the correct squares of the grid, and identifying what letters are already there, to use in identifying which new solutions might be possible.)
Yeah I got it to do Monday through Wednesday of last week but realized that my approach wasn’t going to give it enough context to pick up on wordplay themes or rebuses. For tomorrow’s, it would definitely need a much better understanding of how the grid works than any of them have been able to demonstrate yet in my experiments.
There's AI generated story content on some Youtube video channels (there's a lot of 'revenge stories' about divorces/being fired/etc.) and the problem is that they don't remember details.
So things like "I was married for fifteen years" (and yet the eldest child, if there is a child, is twenty) or "I worked at this company for twelve years" (but invoked a clause in a contract from forty years ago). A lot of contradictions where a human would pick up "oops, better correct that" but the AI clearly is mashing together elements from plots it has been trained on but can't put them together coherently.
I see that Aella has bet "no" on the prediction market, which is a pretty good reason to believe that she is not the author of the post. I am now kind of curious -- do proponents of prediction markets think prediction market betting should be anonymous? What do they think of trades made by someone who has inside knowledge of the market and whether those should be anonymous? (Bets on questions like "Will person X do Y" or "Did person X do Y" made by person X.) On a real market I guess you'd risk having person X bet the wrong way and then having them set up a bunch of straw traders to vote the other way but I doubt Manifold has that dynamic.
I was amazed that anyone thought that thing was written by her.
If you consider prediction markets as gambling or a competition, I can see why this would seem unfair.
But as a mechanism for aggregating information and finding the best estimate of the truth, that seems working as intended? Aella's bet added information and now we're much more confident (though not 100%) that she didn't write it.
But a big part of the concern is market manipulation.
Say there's a market for 'Did A write this anonymous article?'
A puts $500 on 'No' using a publicly verified account, sending a strong signal towards 'no' being correct.
But secretly, A also uses an anonymous account to put $5,000 on yes.
Now,m technically this is all 'information.' A savvy Bayesian takes into account the number and size of anonymous bids, writes up probabilities and expected utilities for 'the market is being manipulated by A' and 'A is sending an honest signal', and places their own bet based on the highest expected utility.
But the market is now an anti-inductive social game. People are trying to second-guess adversarial bluffers, rather than basing their answer on subject-area knowledge, and those bluffers are trying to predict those guesses and double-bluff them. I'd be surprised if the results of such a social game are as accurate as an empirical-data-only prediction market.
> the market is now an anti-inductive social game. People are trying to second-guess adversarial bluffers, rather than basing their answer on subject-area knowledge, and those bluffers are trying to predict those guesses and double-bluff them
Fair, but this would only happen for very specific markets. Maybe it would happen once then everyone else would know not to bet on those.
Or the only people who do would be the nerds who like to to play anti-inductive adversarial bluff games. And at least they are getting some utility out of it.
"No" is at 94%. Suppose you are Aella. Wouldn't this be a cheap way to try to hide your identity for a bit longer by betting against the truth?
She could bet on Yes and everyone would laugh because it's too obvious
The galaxy-brain play would be for Aella to bet 'yes" at single-digit probabilities, triggering a spike in the market when people notice and follow her lead, the have a confederate place a larger bet on "no".
Not sure how much she and her confederate could expect to profit from this kind of manipulation, but probably not enough to be worth the reputational fallout afterwards.
"Don't listen to Chesterton", Tom said defensively.
"Gilbert Keith is only my second favourite early twentieth century Anglo Catholic writer" said Tom, hilariously.
"We must go deeper" said Tom quite littorally.
"It's my eleventh birthday" said Tom, extensively.
That's good
------***-----
"War is hell," William said civilly
"Generally," Ulysses granted
Finally a proper one!
Well done.
I wrote a review of Ted Chiang, my favorite short story writer, that focuses on what I think most readers (even fans) miss about his work:
https://linch.substack.com/p/ted-chiang-review
The main argument: Chiang writes neither hard SF (engineering with known physics) nor soft SF (science as window dressing), but a third thing: stories where the fundamental laws of science are different but internally consistent (This is actually very rare in published fiction. Scott and other rationalist-adjacent writers have also done this a few times in their fiction, but imo less well). Chiang uses these alternate realities to explore philosophy from the inside.
Key points that might interest other commenters:
- He writes the best fictional treatment of compatibilism/determinism I've ever encountered
- His stories treat philosophical problems as lived experiences rather than intellectual exercises
- Unlike most contemporary SF, technology in his stories enhances rather than diminishes humanity
His major blindspot: he completely ignores how societies would respond to paradigm-shifting tech (e.g., parallel universe communication that should revolutionize all R&D but somehow doesn't)
The review also touches on why strong Sapir-Whorf and Young Earth Creationism make perfect sense as story premises when you understand what he's actually doing.
The entire review is almost 2500 words long (It actually used to be a lot longer before I trimmed it! ) and I tried to cover a lot of nuances and share my joy of reading Chiang as much as I can. I'd love to hear people's thoughts on Chiang's work and whether my interpretation resonates.
https://linch.substack.com/p/ted-chiang-review
Beautiful review. Ted Chiang is also my favorite author. He's the only author that has mastered the three aspects of storytelling that I care about - emotional resonance; satisfying story arc; and conceptually interesting. Other authors have one or two, but Ted Chiang is the only author who does all three consistently.
Having said that there are a few other short stories by other authors that are Ted Chiang level you might be interested in reading.
Learning to be Me by Greg Egan. By far the best fictional treatise on the nature of consciousness.
Zima Blue by Alastair Reynolds. A beautiful story about art, memory and identity.
Lol in about 24h, this review somehow has more views than the rest of my substack combined
It made the front page of Hacker News.
Here's to hoping someday all my other articles will!
This seems like a bit of a weird definition of hard SF. The hardest science fiction story I've ever read in my life is probably The Clockwork Rocket by Greg Egan, which takes place in a universe that flips the sign of one term of the Lorentz transformation. Egan then painstakingly works out all the physical consequences of this, and writes a story where they drive the plot, complete with periodic interruptions to confront the reader with an alternate-physics equation in the middle of the text (and there's an alternate-physics textbook on his website). I don't think anyone says this doesn't count as hard SF. It's not *mundane* SF, but that's a different and much narrower thing.
You mention a number of stories based on premises of "what if various religious and mythological concepts were real?". With the disclaimer that I haven't read most of them (except "Hell Is the Absence of God", which doesn't strike me as hard SF in any sense), it seems obvious to me why these don't qualify as hard SF: because it's not actually possible to have a fictional universe "where the principles of science themselves are meaningfully different from our world, but still internally consistent" that still bears surface similarities to the real world. Any work that purports to do so must necessarily be covering up various internal inconsistencies (https://www.lesswrong.com/posts/LaM5aTcXvXzwQSC2Q/universal-fire). The story could still be *good*, and ask thought-provoking what-if questions that Star Wars doesn't, but it can't really be extrapolatively applying the real ways of thinking about the universe to a fictional context, which is what hard SF seeks to do.
Gwern had a take on "Story of Your Life" similar to yours (https://gwern.net/story-of-your-life). For my part, I wasn't satisfied with that story. (Spoilers:) The text wants us to view the aliens' timeless perspective as equally valid to our own (hence the analogies to timeless laws of physics like the principles of least action), but it's just not; they live in a universe that has an arrow of time, whether they appreciate it or not. It's not physically possible for them to remember the future the way they remember the past (unless this universe's laws of physics are different in a way that allows this, but the text gives no hint of this). The only time the protagonist would be able to articulate the perspective she does in the story is at the moment of her death; at any other time, the contents of the rest of her life would necessarily be missing, which raises the obvious question of what that feels like from inside and how it avoids breaking the illusion of timelessness (other than by just giving the aliens amnesia or otherwise preventing them from seeing the question, which would be a cop-out). At best it's the story of aliens with a weird cognitive disability, not aliens with an interestingly different relationship to time. Arrival's genuine time-travel plot is in many ways more conventional but at least it makes sense.
I did like "Liking What You See: A Documentary" a lot.
What's the TLDR for his take on compatibilism?
In practice, compatibilism says something like “free will in the normal, pretheoretic sense of the term, doesn’t exist. Your choices still meaningfully matter nonetheless. You can’t meaningfully get out of the bind philosophically. What you can do, however, is make peace with it.”
Chiang makes this realization visceral. From The Merchant and the Alchemist's Gate:
“My journey to the past had changed nothing, but what I had learned had changed everything...Nothing erases the past. There is repentance, there is atonement, and there is forgiveness. That is all, but that is enough.”
“That is all, but that is enough” on the surface sounds like a deepity, but I genuinely find it more moving than anything else I’ve read on free will, determinism, or compatibilism.
When Chiang uses time travel as a motif, the stories differ from typical time travel stories. Because causation is a closed loop, knowing the future does not give you special access to it, and Chiang’s characters tend to not fight the future (successfully or otherwise).
In Story of Your Life [SPOILERS], the narrator learns an atemporal alien language and begins experiencing past and future as equally real.4 It takes her some time to make peace with it, but eventually she fully accepts the truth of determinism. She understands that life is full of tragedy, including that her daughter will die young, but life is full of beauty too. With both regret and awe, she sets forth on the path that she was destined to take5.
This is compatibilism from the inside. In both stories, the characters discover they cannot change what will happen, but this knowledge transforms how they experience what must happen: with forgiveness, acceptance, and even joy.
Wow!
I knew the movie Arrival was based on Story of Your Life, but you made me now realize the cleverness of that title.
Story, of course! Like you're only a character in one.
Thank you!
I preferred the view of free will in his story "What's Expected of Us": https://www.nature.com/articles/436150a
I've come to think that a mechanical world view is rather necessary for free will, and only superficially in conflict with free will.
The thinking is that free will does not require that things could have been different - even if that is the thought experiment that naturally comes to mind when thinking about that concept. For example, If I had made this action instead, A would have happened instead of B etc. The key is to realize that after the fact you did make that action and B happened, and nothing can ever change that - even if that action was a result of a thought process in your mind - this is still free will. Further, your brain state leading to said action was determined by all your sensory inputs, your previous experiences, and previous state of your brain, and this also happened in that particular way. We can imagine it happening differently in a thought experiment - but crucially from a mechanical world view it can't actually happen differently outside of that thought experiment.
The alternatives to free will outside this mechanical determinism I think is:
1. Randomness, e.g. quantum randomness makes parallel timelines, or there is somehow some small randomness in outcomes. (Without the parallel timelines I'm not really sure what randomness means here, since we can only observe the outcome of that randomness - maybe it means that at some level the outcome of physics at a small level is undeterminable for us outside of a probability function?) Anyway, this seems to, if anything, cause less free will - because some outcomes is determined then by that probability function and not your brain state. This would have to be really subtle differences from mechanical - otherwise I think the world would seem completly chaotic.
2. There is a qualia/soul/etc. that is disconnected from our world, and it is possible to re run the world with different choices - like loading a save game, while still being somehow you. This "you" outside the world would have to lose knowledge of the outside world when replaying the "save game", and your brain state at each run would have to be subtly different so that different choices is made. This seems like it implies that the free will making the choices is actually outside our world, and in a sense different from the qualia we are experiencing. So again, this seems like a weaker version of free will.
Interesting, thanks. FWIW my favorite novel is One Hundred Years of Solitude and that has themes which resonate with this, albeit somewhat indirectly. It's not sci-fi in any way but you may enjoy it.
Thank you I might check it out! Tbh I'm happy enough with Chiang's fictional treatment of compatabilism that I'm mostly looking for *different* ideas in my fiction going forwards, but you aren't the first to suggest One Hundred Years of Solitude to me. :)
What are some of the worst failures of very free markets? What comes to mind immediately is failure to price in externalities (e.g. pollution) and inability to deal with companies over a certain size becoming monopolistic (e.g. FAANG).
I'm curious to hear from the libertarian leaning commentariat if they think there are any ways to deal with these failures outside of government regulation.
The reason I am not libertarian leaning is that there are two kinds of libertarians, one say the market will solve it, the second realizes it won't, they say the courts will solve it. But the courts are a branch of government. You just want your regulators to be called judges. And the courts are not immune to being very politically motivated...
Scott wrote an extremely good article that covers parts of this:
https://slatestarcodex.com/2017/02/22/repost-the-non-libertarian-faq/
Some libertarians would say free markets don't fail, that anything they don't provide is not needed or desired (enough), efficiency at the expense of liberty is ultimately harmful, externalities can be handled by local rules and monopolies cannot exist for very long without government support of some sort.
This take seems like a cop out and lacks acknowledgment of tragedy of the commons
1. As mentioned, local rules, or 2. no commons: only private property, including joint ownership.
Who owned the hole in the ozone?
Do you have evidence that, after a period of adjustment to a market without major government interference, that such a market would not verify the need for and provide solutions ?
Completely unregulated free markets could end up with slavery in those specific cases where slavery happens to be economically profitable (e.g. because you don't need to pay the slave their market price).
David Friedman argues, in short, that market failure is the exception, whereas government failure is the rule, and thus that involving government to avoid market failures is likely to cause more failure than it avoids. https://daviddfriedman.substack.com/p/market-failure
> I'm curious to hear from the libertarian leaning commentariat if they think there are any ways to deal with these failures outside of government regulation.
Uh, ways for *who* to deal with them? Non-libertarians might try to deal with them via grass-roots activism and boycotts. Libertarians can only do so by noticing an otherwise overlooked flaw in the moat, and if they notice one, they are not about to spill the beans here!
Also, I'm not sure what your definition of a monopoly is. Apple seems to be trying to compete in at least one market dominated by each of the other FANGs and against Microsoft to boot.
There’s also the Akerlof failure mode based on asymmetric information - without a third party regulator or certifier, it’s hard to establish good markets in used cars and patent medicines and other things where there’s a lot of variance in effectiveness of the product that purchasers can’t verify before purchase. (There can be markets, but they operate at low enough prices that expensive methods to produce high quality products can’t enter.)
That's a good point. I guess in just my head, a very free market doesn't mean the lack of regulation. It means regulation to the extent required to make it a theoretically perfect market.
Another example would be price transparency laws such as anti-drip pricing laws. This kind of regulation makes for better price discovery
As someone libertarian, I think that government intervention for externalities is reasonable but government intervention for monopolies is almost always a lie.
Of your FAANG 'monopolies':
Facebook and Google give their core product away for free. Amazon is drastically cheaper than almost everything it competes with.
Facebook became the most popular social network by outcompeting previous social networks, and to my understanding is currently losing market share among younger people as it gets outcompeted by instagram/Tumblr/whatever else the kids these days like.
Netflix is one of enough streaming services that a common complaint i hear is "why are there so many of these, I can't keep track of what is on which service", which is sort of the opposite of the problems with a monopoly.
It is theoretically possible for a monopoly to cause problems under unregulated capitalism. But in practice, accusations of this are mostly just used as a cudgel to hit whoever Lina Khan doesn't like.
>Facebook and Google give their core product away for free. Amazon is drastically cheaper than almost everything it competes with.
Please don't look at only one isolated piece of the business model, i.e. monetary price for consumers. It is well known how companies that offer a product below cost are most likely in the data collecting and enshittification business. They do aim for monopoly and customer/business lock-in so they can eventually monetize the customers and/or the businesses through that product or others. All these companies are also known to be fiercely anti-competitive, which is why effective regulation is overdue.
>I think that government intervention for externalities is reasonable
I am inclined to agree, with the caveat that the standard ancap/minarchist solution to externalities and the tragedy of the commons is to enclose the commons and rely on Coasian bargaining to handle externalities.
Myself, I am a bleeding heart libertarian and increasingly leaning towards just describing myself as "liberal", not an ancap or minarchist. I do see value in enclosing commons when practicable, and on Coasian bargaining, but both often suffer from issues with transaction costs and the latter especially suffers from coordination problems (brinkmanship in bilateral negotiations and free riders and strategic holdouts in multilateral negotiations) which can make state action an alternative worth considering.
No discussion of monopoly is complete without Standard Oil. But when the Sherman Antitrust Act was passed and it was broken up, it was actually already losing market share.
The real hatred of Standard Oil was its insanely aggressive tactics. Occasionally the rep would threaten all provisions stores in an area by telling them that if they didn't sell Standard Oil, the company would all provisions at cost in that area and bankrupt them.
All the while Standard Oil grew bigger, the cost of oil actually fell.
I don't really think there's any example of a real monopoly that has ever existed which was raising prices and making things worse for consumers. It's a myth.
Inability to manage the human and political effects of Schumpeterian destruction.
A few scattered thoughts on why parenting young children feels more difficult today than it did in the past:
- Most other things in life have become significantly easier, while the difficulty of raising young children hasn't changed much. Back when most people were doing hard work all day, the perceived difficulty of raising children probably didn't move the needle much (relative to the baseline difficulty of life). And as the children got older, they would help with the household work, rather than being net drains on the parents' time and resources - so the early childhood stage was probably viewed as more of an investment, with an expected payoff when the child got older.
- Children keep each other entertained. Declining family sizes, and declining presence of extended family, means parents need to expend more effort to keep each child entertained.
- In my experience, young children don't usually seem to mind when their parents are busy with household work that they can watch and understand (cooking, cleaning, etc). But they HATE when their parents are unavailable because their attention is focused on a computer or phone. So if you want your child to play independently while you're trying to work from home or read articles about the difficulty of modern parenting, you're in for a rough ride.
- Related to the previous one: young children don't seem to mind sitting around while their parents socialize in real life. But they're less content to sit around while their parents are on the phone, and they hate sitting around while their parents are focused on screens. It follows that the more online your social life is, the more difficult raising young children is going to be.
When people start their families later, the grandparents are older and therefore less able to help.
The more abstract your work is, the more difficult it is for your children to understand it. Watching your parents do manual work is fun and you can learn something, watching them do tax reports is not.
Similarly, sitting around while your parents socialize is more fun if you understand what they are talking about.
Regarding your 3rd and 4th items, those are different expressions of the same syndrome and your description is accurate.
In the case of my eldest child this factor pushed me towards largely/permanently giving up television, and in the case of my youngest it has diminished my smartphone/online. In each instance the change has improved my own waking hours significantly. So I view it as a bonus feature of parenthood -- changes that I will not revert back out of when my youngest has left the nest.
This is an interesting list.
Maybe another reason is that "not raising kids" is a much more realistic and accepted alternative today. So the thought "my life would be so much easier if I didn't have kids" is more present now than in the past, when having kids was just what you did. My life would also be much easier if I had a personal driver, but since that's out of the question, it doesn't stress me out that I don't have one.
Definitely easier to raise kids when they spend much time afterschool running around with other kids in Stranger Things style
I find that it has much to do with simplified concepts of "responsive parenting" that roughly boil down to "letting your child cry is traumatic", and therefore it becomes necessary to actively hold and soothe them very frequently.
I've also noticed that the parents I've met in real life often say that parenting is not that difficult for them, and I suspect there is a significant bias where complainers are much more common and visible on the internet.
Your second paragraph is very much part of the picture. I've parented a child born in the early 1990s and one born in the early 2010s and this difference is _really_ strong just between those two eras.
"Child-first" parenting is definitely part of it, but I think it's downstream from everything else in life getting easier. Responsive parenting (by today's standards) probably wasn't even possible before microwaves, washing machines, and 40 hour workweeks.
But your mention of 40 hour workweeks reminds me of the obvious point that mothers weren't expected to also work a job until quite recently.
So not everything has got easier.
Having never raised children in the twentieth century I'm not sure I can judge whether it is subjectively more difficult or not.
My guess is that it's not significantly more or less difficult, you are just a lot more exposed to the complaints of 2020s parents than 1950s parents.
An analysis of Beethoven's genome (from a hair sample) suggests that Beethoven was rhythmically challenged. Beethoven’s beat synchronization PGI ranks between the 9th and 11th percentiles compared to modern samples’ beat synchronization PGI. And beat synchronization has been correlated with musicianship in modern samples.
"Notes from Beethoven’s genome" by Laura W. Wesseldijk et. al...
https://www.cell.com/current-biology/fulltext/S0960-9822(24)00025-3
And this doctoral thesis by Matthew Aguirre, which was just published, discusses how, in human populations, most of the genetic variance in gene expression can be attributed to trans-acting expression quantitative trait loci (eQTLs) spread across the genome. "In practice[,] it is difficult to discover these eQTLs, and their cumulative effects on gene expression and complex traits are yet to be fully understood." However, Aguirre takes a stab at estimating eQTL effect sizes and "heritability from cis- and trans-acting genetic effects for 5,902 genes from a recent twin study of whole blood gene expression." The models suggest that the gene expression derives from sparser pleiotropic relationships across genes "than would be suggested by naive models of regulatory networks, *which has important implications for future studies of complex traits.*"
So maybe there's more to rhythm than a simple polygenic score?
"Regulatory network topology and the genetic architecture of gene expression" by Matthew Aguirre, et. al.
https://www.biorxiv.org/content/10.1101/2025.08.12.669924v1.full.pdf
Beethoven, in his deafer years, probably would have stuggled to clap along to a rhythm. (Sarcasm.)
This seems a little like claiming that it would be surprising if (e.g.) Jane Austen's genome was sequenced and showed a predisposition to dyslexia. In fact, internet sleuths have diagnosed WB Yeats and others with dyslexia and it doesn't seem to have held them back. In the same vein; Beethoven was a notable pianist and his music was revolutionary for its rhythm and drive. Maybe his genetics made him more aware of these things.
"Polygenic score correlated at 2% with musicianship status fails to accurately predict musicianship status in a single extremely-out-of-sample individual" sounds like a caricatured straw-man of someone who doesn't understand polygenic scores.
Yep. But Cell published it.
10th percentile genetics doesn’t sound so low that a bit of extra training couldn’t make up for it. And also, Beethoven is famously the first major composer to use Maelzel’s metronome after it was invented!
Don't get me wrong. I believe environmental variables play a larger role in behavioral expression than a bunch of SNPs on some DNA strands. However, take a gander at figure C and D in that paper! Old Ludwig Van is waaaayyyyy down the scale on modern populations' PGS scales for music creative achievement and musical engagement. But the authors come up with a few excuses. Such as, "genetic associations reflect a culture-specific interplay between underlying heritable factors and environmental influences. Findings from a GWAS in modern Western society may not apply universally across time and regions." And, "as population-level approximations, PGIs do not necessarily yield accurate predictions at the level of the individual."
So they basically admit that Beethoven's PGI probably had little effect on his musicianship. LoL!
Then the assumption that "moar genes moar better" may be wrong and that, in fact, selecting out an embryo on the grounds of "this has less of the desired Big Beautiful Score" might not be the right choice, as it could end up getting rid of Beethoven.
Isn't this the crux of our current disagreement on embryonic selection? The science is nowhere near good enough to produce the promised outcomes of glittering prizes, that you too can have Baby Mozart Einstein Phelps Jobs in one bundle because we can pick out the bestest blastocytes?
I don't think that's the only disagreement, though it's certainly in play. There's also the idea that if trait selection were reliable, the side effects would be bad.
I’m going to need a whole new thesaurus for that middle paragraph.
Study up! There'll be a vocabulary quiz, Peter. ;-)
I would say how the hell can you tell someone's sense of rhythm from their hair, but that might lead us into culture war adjacent topics.
I remain less than convinced that doing this kind of testing on samples which are of dubious historicity and which have been kept in less than ideal conditions will tell us anything useful. "Beethoven was most likely, though not 100% definitely, assigned male at birth".
Supposedly, the polygenic scores built from markers such as rs1464791 and VRK2-related SNPs predict rhythm ability, musicianship, walking pace, and even breathing function. But I don't think PGSs are as predictive as a lot of the peeps on ACX believe. I got a chuckle out of the article.
And considering that we're able to recover DNA from 40,000-year-old skeletal remains, it's not difficult to believe that they could have recovered DNA from Beethoven. But the preservation of the DNA definitely depends on the conditions under which it was stored. My problem with this article is that they said they recovered his DNA from a hair sample. Yes, people were giving locks of their hair to lovers and admirers all throughout the 19th century. But usually it's mtDNA that's recoverable from hair. The cells from a hair follicle would need to be attached to to his hair (or perhaps dandruff), and those cells would have to have been stored in very optimal conditions to get Beethoven's whole genome. OTOH, they could have gotten it from his skeletal remains if they had access to them.
"predict rhythm ability, musicianship, walking pace, and even breathing function"
Great, can they foretell my favourite colour and if I should eat porridge hot or cold?
I'm sure their foretelling abilities fall within the predictive range of astrology, though. ;-)
People! I have a solution to the puzzle of the existence of God and evil!
Dwelling on the cruelty of the world and the stupidity of the idea of God, I thought there cannot be any sense in soon to be dead babies's suffering.
But what if they are not conscious? What if NO BODY whose life doesn't end up getting its soul to heaven is ever conscious?
What if God only retroactively grants human bodies, and only those who went through all the right molecular motions, a soul? Like in really changing history.
Then it could be that going through such bodily process is necessary for getting the right character into a soul with which God wants to populate heaven, without even one getting lost in the process.
If I were God, I'd do it so.
Plus: it would mean that anyone conscious now can stop worrying about everything, cause everything will be fine, you're already on your way to heaven.
*Humbly bowing for applause*
[Edited for the horrible spelling mistakes that Deiseach in their rightful scorn has pointed out.]
I suggest taking a humility pill and reading some existing philosophy on this topic. You're concepts of god and suffering are quite limited.
Please write better comments than this. If you have a reason for disagreeing, say so.
Of course. Sorry to waste your time.
Since I read some existing philosophy on this topic and still have such an inferior conception of God and suffering compared to yours, I really should follow your advice, since I'm clearly the one who needs a lesson in humility here.
If so, why have you discarded all concepts of god other than the Abrahamic religions (heaven, etc.), any consideration of suffering as necessary for some types of growth, reincarnation, etc.?
Ps. Or some variation of him. A thing that created all other things and did this intending something good for us.
I didn't mean to discard them, I thought with capital letter G "God" one only refers to him.
And humor, by the way.
Oh, sorry. If your post was meant tongue in cheek, or just as merely one possible "solution", I sincerely apologize. I missed that.
Well, I have realized now, what delicate topic this is.
Although I do not quite understand this, except when the other side is a religious fundamentals requiring the other to adopt his stand.
But then again, I'm pissed off by BB talking like he just intellectually *knew*, there is God.
I didn't intend to sound like him. I can't take him seriously.
“The Grizzly Bear is huge and wild;
He has devoured the infant child.
The infant child is not aware
It has been eaten by a bear."
"Infant Innocence”
― A.E. Housman
Hilaire Belloc, "The Bad Child's Book of Beasts":
The Tiger
The Tiger on the other hand, is kittenish and mild,
He makes a pretty playfellow for any little child;
And mothers of large families (who claim to common sense)
Will find a Tiger well repay the trouble and expense.
"Cautionary Tales for Children, Designed for the Admonition of Children between the ages of eight and fourteen years"
Jim,
Who ran away from his Nurse, and was eaten by a Lion.
There was a Boy whose name was Jim;
His Friends were very good to him.
They gave him Tea, and Cakes, and Jam,
And slices of delicious Ham,
And Chocolate with pink inside,
And little Tricycles to ride,
And read him Stories through and through,
And even took him to the Zoo—
But there it was the dreadful Fate
Befell him, which I now relate.
You know—at least you ought to know.
For I have often told you so—
That Children never are allowed
To leave their Nurses in a Crowd;
Now this was Jim’s especial Foible,
He ran away when he was able,
And on this inauspicious day
He slipped his hand and ran away!
He hadn’t gone a yard when—
Bang!
With open Jaws, a Lion sprang,
And hungrily began to eat
The Boy: beginning at his feet.
Now just imagine how it feels
When first your toes and then your heels,
And then by gradual degrees,
Your shins and ankles, calves and knees,
Are slowly eaten, bit by bit.
No wonder Jim detested it!
No wonder that he shouted “Hi!”
The Honest Keeper heard his cry,
Though very fat he almost ran
To help the little gentleman.
“Ponto!” he ordered as he came
(For Ponto was the Lion’s name),
“Ponto!” he cried, with angry Frown.
“Let go, Sir! Down, Sir! Put it down!”
The Lion made a sudden Stop,
He let the Dainty Morsel drop,
And slunk reluctant to his Cage,
Snarling with Disappointed Rage
But when he bent him over Jim,
The Honest Keeper’s Eyes were dim.
The Lion having reached his Head,
The Miserable Boy was dead!
When Nurse informed his Parents, they
Were more Concerned than I can say:—
His Mother, as She dried her eyes,
Said, “Well—it gives me no surprise,
He would not do as he was told!”
His Father, who was self-controlled,
Bade all the children round attend
To James’ miserable end,
And always keep a-hold of Nurse
For fear of finding something worse.
Wasn't this a plot point in Unsong? Uriel reveals that he turned some very poor countries into p-zombies to save on processing power and alleviate suffering?
Besides the odd moral conclusions it leads to, I think this implies that souls are created to have no explanatory power - a person "piloted" by a soul must behave exactly identically to a person with no soul. And like, if you're an atheist then this certainly matches observations, but if you actually believe in souls then surely you believe they actually *do* something that can't be emulated by the crude machinery of flesh?
I don't remember from Unsong, but it was a plot point in Scott's 'Answer to Job,' which is excellent.
> I think this implies that souls are created to have no explanatory power - a person "piloted" by a soul must behave exactly identically to a person with no soul. And like, if you're an atheist then this certainly matches observations, but if you actually believe in souls then surely you believe they actually *do* something that can't be emulated by the crude machinery of flesh?
When I imagine all these things, as a Gedankenexperiment, then no. Souls do indeed not pilot the bodies. Souls are formed, in a one way street way, by the bodies they get assigned to.
In heaven then, they begin acting on their own, now that they have the good characters God wants hhem to have.
And I know, this is weird, but there are dualists, and for them, who have the epiphenomenality problem you pointed to, this is as much a solution then it is to the problem of evil. They just have to believe in heaven to lighten up the epiphenomenality of their current existence.
Feser, a Catholic philosopher says the non-physical part of souls deal with abstractions, because abstractions do not exist in nature and hence are supernatural.
And this immediately taught me that the reason we mostly don't think like Plato, Aristotle or Aquinas is our fundamental approach to knowledge, epistemology is different.
They thought that every true statement has to refer to some pre-existing knowledge. Thus our minds basically discover pre-existing truths, like how the eye discovers a new object. Hence they thought if the abstract idea of triangularity is a thing, it must exist somewhere, such as in God's mind.
We think our minds are creative, our thinking makes new models, models which are true in the sense of useful, but not necessarily 100% true. It is like making a watch, maybe it is not 100% accurate but it roughly tells you the time.
Perhaps neither heaven nor consciousness is real.
One of those things can be observed with absolute certainty. The other can't be observed at all.
God knows all things, including the future, so he could just only grant consciousness to those he foresees will go to heaven, and leave everyone else running around as p-zombies.
I'm getting faint hints of warmed-over Gnosticism, so our friend may not be as original in their One Weird Trick at they think.
Lemme go refresh my memory:
https://en.wikipedia.org/wiki/Gnosticism
hylic – lowest order of the three types of humans. They cannot be saved since their thinking is entirely material, incapable of understanding the gnosis.
psychic – "soulful", partially initiated. Matter-dwelling spirits
pneumatic – "spiritual", fully initiated, immaterial souls escaping the doom of the material world via gnosis.
In the Nobilis RPG by Jenna Moran, there’s a gag about the angels having to cope with human population growth by pressing lower animal souls into service as people, then the souls of beetles. This all happened around 5000 BC.
> so our friend may not be as original in their One Weird Trick at they think.
You're so damn funny at times that I just can't say: stop picking on me!
(1) Please get your spelling right - it's "whose", not "who's", "its" not "it's" and generally "nobody" not "no body".
(2) Please get your theology right - this is a very idiosyncratic notion of ensoulment. If you're leading out of the gate with "the stupidity of the idea of God", better not do theology at all.
(1) Right, I'm sorry. The idea struck me suddenly and I wanted to get it in the thread before it got stale. The thread I mean. The idea turned out to be stale already.
But I did mean "no body" not "nobody".
(2) I do not believe in God, I find the idea stupid, and can give many reasons why it's stupid. Surely those reasons are only as reinvented as my idea of ensoulment is, which Edward Scizorhands showed me to be communicated by Scott long before https://slatestarcodex.com/2015/03/15/answer-to-job/
Then why do I think about God at all? I don't know. The cruelty of the world and my incompetence at dealing with it must make me occasionally very desperate.
Then why not simply drop the whole god and souls bit, and present your idea as the solution to the problem of consciousness: if you are alive in the world right now, congratulations, you are indeed conscious and not a p-zombie!
You'd at least have firmer ground to stand on to argue that, rather than trying to get into a philosophy you don't believe in and don't know about. It'd be like me trying to argue based on quantum physics which I am completely ignorant about.
>You'd at least have firmer ground to stand on to argue that,
I wouldn't have firmer ground. It's not a solution to consciousness, it presupposes it's existence, deliberately avoiding explaining it.
> rather than trying to get into a philosophy you don't believe in and don't know about. It'd be like me trying to argue based on quantum physics which I am completely ignorant about.
First, I think God wouldn't mind me exploring ways I might be able to believe in him, and second, why do you believe I know nothing about this philosophy? Because the idea I communicated is not mainstrean in theology? That's all the information you have right now.
Also, it's not like anyone is able to find out any facts about this. Not in their lifetime at least. I don't talk about whether Jesus indeed multiplied bread. I talk about whether he was the son of God or Cthulhu, playing jokes on us.
If you are going to argue about souls, which you may or may not believe in, and which many commenters on here may not believe in, then you are already ceding too much ground.
Arguing that this solves the problem of p-zombies at least gives you more room for debate; people on here are more willing to grant the notion of the existence of consciousness rather than ensoulment.
Not mainstream theology -indeed. It sounds a little like warmed-over Gnosticism and that's not original nor orthodox.
"Jesus was indeed multiplied bread". Ah, I see you can take the man out of the Reformation, but you can't take the Reformation out of the man! 😁
Why do people have to bear and raise doomed p-zombie children? That seems really dystopian in a Peter Watts kind of way.
You wouldn't know if *people* raise them, but if they are people, then it's this character building thing.
If you get to heaven for it, it's very much worth it.
If you're going to postulate things why not just postulate that god is real and good. Just declare that reality orbits around the idea you want to protect. Skip the epicycles, ya know?
I really ment this only as a solution to the puzzle.
I do not entertain the idea that God exists.
I just find that not knowing why there is so much bad in the world is no argument against the possibility of his existence. And it nagged me to not have an argument for there to be no solution, nor a solution.
This is more or less the spiritual view that I've converged on - we're all born as p-zombies, and the world is an incubator for souls.
If that's what you like, have you read Andy Weir's story "Egg" ? If not, you might find it interesting.
It doesn't explain why we exist in this material universe in the first place. If all conscious people (Hitler?) end up in heaven, why not just skip this step?
Works if you're some variety of Gnostic, though. You may have already read "A Voyage to Arcturus" by David Lindsay, but if not, I'm going to spoil the ending:
https://www.gutenberg.org/cache/epub/1329/pg1329-images.html
"He climbed into the embrasure. His feelings translated themselves into vision, and he saw a sight that caused him to turn pale. A gigantic, self-luminous sphere was hanging in the sky, occupying nearly the whole of it. This sphere was composed entirely of two kinds of active beings. There were a myriad of tiny green corpuscles, varying in size from the very small to the almost indiscernible. They were not green, but he somehow saw them so. They were all striving in one direction—toward himself, toward Muspel, but were too feeble and miniature to make any headway. Their action produced the marching rhythm he had previously felt, but this rhythm was not intrinsic in the corpuscles themselves, but was a consequence of the obstruction they met with. And, surrounding these atoms of life and light, were far larger whirls of white light that gyrated hither and thither, carrying the green corpuscles with them wherever they desired. Their whirling motion was accompanied by the waltzing rhythm. It seemed to Nightspore that the green atoms were not only being danced about against their will but were suffering excruciating shame and degradation in consequence. The larger ones were steadier than the extremely small, a few were even almost stationary, and one was advancing in the direction it wished to go.
…When at last he looked, he saw the same sphere as before, but now all was changed on it. It was a world of rocks, minerals, water, plants, animals, and men. He saw the whole world at one view, yet everything was so magnified that he could distinguish the smallest details of life. In the interior of every individual, of every aggregate of individuals, of every chemical atom, he clearly perceived the presence of the green corpuscles. But, according to the degree of dignity of the life form, they were fragmentary or comparatively large. In the crystal, for example, the green, imprisoned life was so minute as to be scarcely visible; in some men it was hardly bigger; but in other men and women it was twenty or a hundred times greater. But, great or small, it played an important part in every individual. It appeared as if the whirls of white light, which were the individuals, and plainly showed themselves beneath the enveloping bodies, were delighted with existence and wished only to enjoy it, but the green corpuscles were in a condition of eternal discontent, yet, blind and not knowing which way to turn for liberation, kept changing form, as though breaking a new path, by way of experiment. Whenever the old grotesque became metamorphosed into the new grotesque, it was in every case the direct work of the green atoms, trying to escape toward Muspel, but encountering immediate opposition. These subdivided sparks of living, fiery spirit were hopelessly imprisoned in a ghastly mush of soft pleasure. They were being effeminated and corrupted—that is to say, absorbed in the foul, sickly enveloping forms.
…Looking through the window, he was startled by a new sight. The sphere was still there, but between it and the Muspel-world in which he was standing he perceived a dim, vast shadow, without any distinguishable shape, but somehow throwing out a scent of disgusting sweetness. Nightspore knew that it was Crystalman. A flood of fierce light—but it was not light, but passion—was streaming all the time from Muspel to the Shadow, and through it. When, however, it emerged on the other side, which was the sphere, the light was altered in character. It became split, as by a prism, into the two forms of life which he had previously seen—the green corpuscles and the whirls. What had been fiery spirit but a moment ago was now a disgusting mass of crawling, wriggling individuals, each whirl of pleasure-seeking will having, as nucleus, a fragmentary spark of living green fire. Nightspore recollected the back rays of Starkness, and it flashed across him with the certainty of truth that the green sparks were the back rays, and the whirls the forward rays, of Muspel. The former were trying desperately to return to their place of origin, but were overpowered by the brute force of the latter, which wished only to remain where they were. The individual whirls were jostling and fighting with, and even devouring, each other. This created pain, but, whatever pain they felt, it was always pleasure that they sought. Sometimes the green sparks were strong enough for a moment to move a little way in the direction of Muspel; the whirls would then accept the movement, not only without demur, but with pride and pleasure, as if it were their own handiwork—but they never saw beyond the Shadow, they thought that they were travelling toward it. The instant the direct movement wearied them, as contrary to their whirling nature, they fell again to killing, dancing, and loving.
…The shadow form of Crystalman had drawn much closer to him, and filled the whole sky, but it was not a shadow of darkness, but a bright shadow. It had neither shape, nor colour, yet it in some way suggested the delicate tints of early morning. It was so nebulous that the sphere could be clearly distinguished through it; in extension, however, it was thick. The sweet smell emanating from it was strong, loathsome, and terrible; it seemed to spring from a sort of loose, mocking slime inexpressibly vulgar and ignorant.
The spirit stream from Muspel flashed with complexity and variety. It was not below individuality, but above it. It was not the One, or the Many, but something else far beyond either. It approached Crystalman, and entered his body—if that bright mist could be called a body. It passed right through him, and the passage caused him the most exquisite pleasure. The Muspel-stream was Crystalman’s food. The stream emerged from the other side on to the sphere, in a double condition. Part of it reappeared intrinsically unaltered, but shivered into a million fragments. These were the green corpuscles. In passing through Crystalman they had escaped absorption by reason of their extreme minuteness. The other part of the stream had not escaped. Its fire had been abstracted, its cement was withdrawn, and, after being fouled and softened by the horrible sweetness of the host, it broke into individuals, which were the whirls of living will.
Nightspore shuddered. He comprehended at last how the whole world of will was doomed to eternal anguish in order that one Being might feel joy.
…Fire flashed in his heart.... Millions upon millions of grotesque, vulgar, ridiculous, sweetened individuals—once Spirit—were calling out from their degradation and agony for salvation from Muspel.... To answer that cry there was only himself... and Krag waiting below... and Surtur—But where was Surtur?
The truth forced itself on him in all its cold, brutal reality. Muspel was no all-powerful Universe, tolerating from pure indifference the existence side by side with it of another false world, which had no right to be. Muspel was fighting for its life—against all that is most shameful and frightful—against sin masquerading as eternal beauty, against baseness masquerading as Nature, against the Devil masquerading as God...."
That lays it out pretty clearly, yes. Never felt inclined toward the Gnostic persuasion myself, but for one who is, the idea that the creator of the material world is evil itself does answer the question of evil pretty decisively.
I think he’s saying Hitler wasn’t conscious. If you are it was retrofitted back when you went to heaven. If you are conscious now then you are heaven bound.
As for the skipping this earthly step, that’s actually a good question for theologians or religious people who don’t believe in free will, I’ve tried to discuss it with some religious types online but they tend to hand wave away the problem.
I know. This is one of the reasons I think the whole idea of God is stupid. But it is a solution to the problem of evil.
Of course, there are those who do not treat "God" as an idea at all, but rather as a feeling state. If God, Good and Evil are all feeling states (that is to say, they are subjective, not objective concepts), then I think the problem disappears.
https://slatestarcodex.com/2015/03/15/answer-to-job/
Well.
1. Reinventing the wheel seems to be my specialty.
2. And yes, one still needs to just believe this, but it's the only way in which I would like to believe in God at all. Like, if
🎶 we're never gonna survive unless we get a little bit crazy 🎶
then I could try this.
Alternatively, maybe you're already IN heaven, and what you think you're currently experiencing is just a memory.
Or maybe innocent beings never suffer and die, we all just have a delusion that shit like that happens. Want some frills on that? OK, god challenges us with this delusion to test our love and trust in him.
Even better (?), we are in heaven, we are experiencing heaven, and this is what heaven is like. Nothing is good enough for some people :)
Given a finite speed of nerve transmission, it's clear that "what you think you're currently experiencing is just a memory".
Maybe it's a really fancy deja vu.
Actually, it's a prediction.
Might be indistinguishable.
Congratulations, you have now made me, for the first time, prefer the idea of the simulation hypothesis to whatever it is you are smoking.
Better than your suggestion, how about we don't have souls at all, imagine there's no heaven, be John Lennon again?
I never thought to be the butt of a "whatever it is your smoking" joke, but I honestly find it really funny. Catched me cold, and laughed out loud :)
Excuse me! *you're!
Funny thing is, I thought something was wrong, but figured I just wasn't sure if its the right figure of speech and Google said it is.
Probably, but mine is better from a "stop worrying" perspective.
Unrelated to schizophrenia, I wonder if it's plausible that there is a base rate of spontaneous insanity, say ~one in a million per year, the particular flavor determined by the circumstances in which it occurred. That neatly explains why so many different popular media, like movies and video games, seem to cause people to go mad, sparking the usual calls for it to be banned. (And it also explains why historically, you had mostly religious-flavored nuts.)
There's this character in a cartoon I like where his origin story is that, after a time machine accident, he's trapped in a void for a long time, goes mad, and then after some time, goes sane again. That partly motivates this model: maybe madness isn't an absorbing state, just one where the transition probability into and out of it is very low.
If true, this has bearing on immortals, people trapped in time loops, accelerated cognition (through noötropics or brain emulation), and less importantly, those getting data from large populations.
I haven't read anyone write "Since AI models have been observed to "think" in such and such manner, this should update our priors on such and such aspects of how humans think". Can there at all be any psychology-adjacent theory or model, say the IFS or the GNWS, whose plausibility might receive a boost from or be tanked by how AI is seen to behave?
I’ve seen a lot of talks in the past couple years by linguists who are studying language production in GPT-2 (and similar open models of manageable size) as a way to understand features of human language production! One of the most interesting talks I saw was looking at particular grammatical constructions that are somewhat unusual, removed all 200 instances from the training data, and observed that the LLM still produced them, and then did further investigation of what other related grammatical constructions had to be removed from the training data to ensure these didn’t appear.
Thank you, both for this comment and your response to Igon Value, quite fascinating.
That's amazing. I'm going to look for these studies, but maybe you have some links handy already?
Here's the specific one I was describing: https://arxiv.org/abs/2403.19827 I hadn't been sure if a written version was available.
I'm fairly sure I saw a few other talks doing similar things just at UC Irvine in the past two years, but the search terms I'm using aren't turning them up (though it did find this: https://arxiv.org/html/2408.03855v1)
I'm pretty sure some of the others I saw were using transformers to identify which features of an NP with multiple singular and plural nouns in it make people more or less likely to accidentally mis-pluralize a verb, and correlating some structure of a transformer to the processing time that humans have in a certain linguistic task.
Thank you! Both links are great.
To go from AI to the human mind, you would have to be using AI to test some theory about the human mind, which is not what AI is designed to do. Pretty sure no one thinks actual people learn the way LLM's do.
I think we learn a lot of things -- for instance, the early version of language, up to age 4 or so -- by pattern-matching rather than by grasping concepts and rules. But though we require lots of examples, we require way fewer than an LLM because our brain is already set up to learn language -- it has compartments of the right shape, so to speak.
I could nitpick your use of the term "grasping" (because it might be implying an "either-or" dichotomy, when comprehension is an incremental spectrum), but you are basically correct. Language acquisition is the output of a growing semantic network in long term memory. I don't know if anyone has programmed a computer to calculate using "fuzzy sets", but that's at least on necessary step along the way toward simulating a human-like mind.
We’ll know AI is starting to think about what it does using rules and concepts when we get it going through stages like starting to say ‘foots’ instead of ‘feet’ before being correct back. Do we have any examples yet?
Don't know, but I doubt it. LLM's aren't designed to function that way, that's not the purpose, so no reason for AI to go that way.
I thought this about human Go ability when AlphaGo performed at superhuman levels; I think that the lack of this in linguistics is worth thinking and talking about, but may have changed a bit in the last few years.
(See nostalgebraist's essay here: https://www.lesswrong.com/posts/ZFtesgbY9XwtqqyZ5/human-psycholinguists-a-critical-appraisal )
Great essay, thanks a lot.
> I haven't read anyone
I hear that a lot in conversations, and personally have updated a lot on it. For example how system 1 works. For example, it helps us understand better what system 1 in the sense of Kahnemann and Tversky is capable of, because the AIs are so similar to system 1.
Thank you, that sounds very promising. If it would be convenient for you to share any ideas or links, that would be great as well.
The link from the survey to https://archive.is is dead because it literally just says "Welcome to nginx".
I got that too.
The “seek for the sword that was broken” dream came to Faramir multiple times and only once to Boromir (he says), implying that Faramir was the one who should have traveled to Rivendell. Instead, we get Boromir going nuts by the river, which results in the breaking of the Fellowship and a long side quest to rescue the Hobbits and take down Saruman, which the ents clearly could have managed themselves. The whole story is probably over after Book II if only Faramir had had the guts to tell his brother to go pound sand.
It's more complicated than Faramir not telling Boromir off. Suppose he had gone to Imladris and found out the meaning of the dream. He would probably not have pushed to go to Gondor, although he still *might* have done: the emergency remained and Denethor was still the leader of the armies fighting Sauron, so this was vital intelligence he should receive.
The Orcs were in place so the breaking of the Fellowship would probably still have happened. Frodo was conflicted about what he should do, independent of Boromir's suspicious activity.
Faramir would have agreed that the Ring should be destroyed and would have supported the decision to do so, but the other factors remained. The Ents were able to destroy Isengard, but they needed outside influence to bring them to the decisions to do so.
This sounds like "why didn't they just fly on Eagles into Mordor" kind of quick and easy solution, which in fact could not happen in canon.
No, my theory is that with no Boromir, there is no breaking of the fellowship, Sam and Frodo don't wind up wandering the wastes outside Mordor or the woods around Ithilien with Gollum as their only guide, wasting precious weeks, and thus there's time to formulate an actual plan, and they're able to destroy the ring before the siege of Gondor even begins.
Make no mistake: just like Jenny in Forest Gump, Faramir is the secret villain of the books. In hindsight, Gandalf shouldn't have bothered saving him from the funeral pyre, and Aragorn should've saved the Kingsfoil for somebody that actually deserved it.
It still remains a problem of how the entire Fellowship gets into Mordor. And Gandalf will be gone by now, anyway, since he fell in Moria, and they had no way around passing through Moria.
So if all goes well, there's Aragorn, Faramir, Gimli, Legolas, Frodo, Sam, Merry and Pippin heading towards Mordor. I think that along the way *something* would have happened to break them up, since they had such a long route to cover and there were enemies on the watch for them (sent by both Saruman and Sauron).
Following the river is their easiest path, but once they do manage to slip into Mordor, then they need a guide. Aragorn can't help them since he will be no more familiar with the interior terrain than anyone. Gollum, like it or not, is who they need because he's been there and knows the dangers (too well, as we can see when it comes to betraying them to Shelob). Faramir might have some knowledge, but as we see not enough - he's heard of Cirith Ungol but again he does not have enough direct knowledge to be able to steer the party.
They're in enemy territory which is chock-full of the armies and other servants of Sauron. There's no easy direct route to take. And from a practical point of view, the Quest was hopeless from the start; Tolkien discusses this in a draft of a letter from 1956:
"Frodo was in such a position: an apparently complete trap: a person of greater native power could probably never have resisted the Ring's lure to power so long; a person of less power could not hope to resist it in the final decision. (Already Frodo had been unwilling to harm the Ring before he set out, and was incapable of surrendering it to Sam.)
The Quest ⁂ was bound to fail as a piece of world-plan, and also was bound to end in disaster as the story of humble Frodo's development to the 'noble', his sanctification. Fail it would and did as far as Frodo considered alone was concerned. He 'apostatized' – and I have had one savage letter, crying out that he shd. have been executed as a traitor, not honoured. Believe me, it was not until I read this that I had myself any idea how 'topical' such a situation might appear. It arose naturally from my 'plot' conceived in main outline in 1936. I did not foresee that before the tale was published we should enter a dark age in which the technique of torture and disruption of personality would rival that of Mordor and the Ring and present us with the practical problem of honest men of good will broken down into apostates and traitors."
Given your opinion of Faramir, you might be amused to know (if you don't already know) that he wasn't part of the story from the outset:
Letter to Christopher Tolkien, 1944:
"A new character has come on the scene (I am sure I did not invent him, I did not even want him, though I like him, but there he came walking into the woods of Ithilien): Faramir, the brother of Boromir – and he is holding up the 'catastrophe' by a lot of stuff about the history of Gondor and Rohan (with some very sound reflections no doubt on martial glory and true glory): but if he goes on much more a lot of him will have to be removed to the appendices — where already some fascinating material on the hobbit Tobacco industry and the Languages of the West have gone."
Draft letter, 1963:
"I think you misunderstand Faramir. He was daunted by his father: not only in the ordinary way of a family with a stern proud father of great force of character, but as a Númenórean before the chief of the one surviving Númenórean state. He was motherless and sisterless (Eowyn was also motherless), and had a 'bossy' brother. He had been accustomed to giving way and not giving his own opinions air, while retaining a power of command among men, such as a man may obtain who is evidently personally courageous and decisive, but also modest, fair-minded and scrupulously just, and very merciful."
You guys are taking me a bit overly-literally. The "Faramir had the dreams but I talked Dad into sending me instead" bit is probably Tolkien just doing a smidge of foreshadowing for the eagle-eyed reader about how headstrong Boromir is and will eventually come to grief, but I think that perhaps unintentionally, the other way to read it is that Faramir was supposed to go instead, for reasons which become quite obvious once you get to the chapters in Ithilien. I do not actually regard him as the secret villain of the story.
Though it does open up intriguing vistas: what would have happened if Boromir had stayed in Gondor with Denethor?
In the early days post-Jackson movies, there were a lot of people getting into the fandom and popular fix-it fics were ones where Boromir lives instead of dying at Amon Hen.
But I don't think anyone ever wrote "what if Boromir never went to Imladris?" Depending on how much Denethor knows, and how far gone he is, if Faramir comes back to Gondor and informs his father of what happened, then Denethor might have sent Boromir out to hunt down Frodo and take the Ring (even if Faramir never tells that out directly, I think Denethor would be able to piece together that the Halfling has something really, really valuable and powerful; the more Faramir refused to tell what it was, the more Denethor would come closer to the truth).
That would certainly be a twist in the tale!
Without the Breaking of the Fellowship, Merry and Pippin would never have come to Fangorn, meaning the Ents wouldn't have attacked Isengard. Consequently Saruman would quite possibly have been able to defeat Rohan, causing far more death and destruction than would be saved by avoiding the siege of Gondor.
The side quest to Rohan was necessary to save Theoren from Wormtongue's influence. Without that and without further assistance from Our Heroes, the host of Edoras doesn't set out to reinforce the Westfold armies, there's no force to hold Helm's deep, and Saruman's army is free to pursue the scattered elements of Erkenbrand's army and defeat them in detail.
The Ents can wreck Isengard and invest Orthanc, but without Theoden assisted by the heroes, Saruman's field army will be intact and available to turn back and relieve the siege. The way it played out in the books was that Theoden's cavalry sallies unexpectedly from the Hornberg, temporarily disrupting the Orcs just in time for Erkenbrand's regrouped forces to show up and attack them in the flank and rear, causing a rout. The Huorns' role was to cut off the Orcs' retreat and kill them as they tried to flee the field.
In the movies, the reinforcements were Eomer and his men rather than Erkenbrand's, to save the complexity of introducing Erkenbrand and his forces. In the books, Eomer had been at Edoras with Theoden (imprisoned rather than exiled) and was with Theoden throughout the battle.
Without the fellowship breaking the way it did, they probably follow Aragorn's proposal which IIRC has Aragorn, Gimli, Sam, and Frodo going to Mordor while Legolas, Merry, and Pippin accompany Boromir (or Faramir, in this case) to Minas Tirith. With Aragorn and Gimli along, there's no chance they accept Smeagol as their guide. We don't know for sure how Aragorn would have lead them into Mordor, but it probably wasn't Cirith Ungol based on Gandalf's reaction when he got word of it. My best guess is that he knew of a less-guarded pass or tunnel or postern associated with the Black Gate, since Frodo had headed there based on what he remembered of planning sessions in Rivendell. They might still get to the volcano, but Gollum wouldn't be there to fall in with the ring under the weight of the curse Frodo had placed on him and of Gollum's own broken oath. Frodo would still not have been able to throw the ring in under his own free will (Tolkien has said that pretty much nobody would have been able to do so), so either the mission fails there as the ring stays undestroyed long enough for the Nazgul to arrive to stall them and then for Sauron to show up personally to reclaim his own. Or Frodo jumps in with the tingy to keep it from Sauron. Or Sam or Aragorn or Gimli has to tackle Frodo into the fire, probably killing both of them in order to destroy the ring.
Lastly, Rohan falling or at least faring far worse against Saruman, plus Aragorn being in Mordor rather than Rohan, means that neither the Rohirrim nor the Oathbreakers of Dunharrow show up to help defend Minas Tirith. If the battle plays up similarly to the books until the point they fail to arrive, then the first circle of the city at least will fall and the Witch King's army will remain intact to assault the second circle and further, and without the intervention of the Oathbreakers, the black-sailed ships will carry Haradrim corsairs to reinforce the Witch King rather than a Gondorian relief force lead by Imrahil and Aragorn.
It isn't necessary for a hobbit to be present in Rohan during the aftermath of the attack on Saruman, but Gandalf, who would have been there regardless. Aragorn knew about the prophesy that he would lead the army of the dead, and that his duty ultimately calls him to Gondor, so that also happens regardless. I don't remember Aragorn or anyone else coming up with a plan about how to deploy the members of the Fellowship. It was ultimately up to Frodo, who would have made the same decision regardless of what Boromir did (his breakdown merely made Frodo's escape easier). Finally, Gollum was tracking Frodo via the magical influence of the Ring, so if Frodo gets to Mount Doom, so does Gollum. We can't know how that scene would then play out, but that's dramatic tension for you.
Tolkein was a very tight plotter.
Tolkien also included elements in the plot about 'another purpose behind events and it was not the will of the ring-maker', see the chain of events which resulted from Bilbo Baggins starting off on a treasure quest with thirteen Dwarves, all due to the meddling of Gandalf long ago:
From Appendix A, LoTR:
"It might all have gone very differently indeed. The main attack was diverted southwards, it is true; and yet even so with his farstretched right hand Sauron could have done terrible harm in the North, while he defended Gondor, if King Brand and King Dáin had not stood in his path. When you think of the great Battle of Pelennor, do not forget the Battle of Dale. Think of what might have been. Dragon-fire and savage swords in Eriador! There might be no Queen in Gondor. We might now only hope to return from the victory here to ruin and ash. But that has been averted – because I met Thorin Oakenshield one evening on the edge of spring not far from Bree. A chance-meeting, as we say in Middle-earth."
The Breaking of the Fellowship looks to be disaster, but out of that seeming disaster different benefits come through things like Merry and Pippin meeting the Ents, and other things that might not have happened if they all stayed together on a straight path to their original destination.
So it had to be Boromir, not Faramir, because this is Boromir's character. Put Faramir in his place, it turns out differently, and maybe not for the best.
Aragorn came up with a plan while Frodo was off considering, with the intention of proposing it to Frodo if the latter came back still undecided:
>It would indeed be a betrayal, if we all left him. But if he goes east, then all need not go with him; nor do I think that all should. That venture is desperate: as much so for eight as for three or two, or one alone. If you would let me choose, then I should appoint three companions: Sam, who could not bear it otherwise; and Gimli; and myself. Boromir will return to his own city, where his father and his people need him; and with him the others should go, or at least Meriadoc and Peregrin, if Legolas is not willing to leave us.
It's easy to forget because it almost immediately became moot because the confrontation with Boromir made up Frodo's mind to set off alone for Mordor without delay.
Ah, yes, I remember it now. I still think it unlikely that Frodo would have agreed to it, but we will never know.
Someone tell me if my memory betrays me, but I don't think is was Faramir's decision but Denethor's, which Denethor regrets when he learns the news of Boromir's death. So Faramir would've had to tell Denethor to pound sand, which is a bigger ask.
Here's what Tolkien Gateway says:
Faramir, his younger son, began to have dreams, speaking of Imladris and Isildur's Bane. Boromir convinced his father to send him instead of his brother and in the end Denethor gave in. Boromir did not return from his quest; he was shot by Uruk-hai on Amon Hen and Faramir and Denethor both heard the horn of Gondor.
https://tolkiengateway.net/wiki/Denethor
I am a long-time viewer of Patrick Boyle's Youtube channel. He's an ex-economics professor that makes a living explaining economic and political news. I used to really enjoy his work, but in his past few videos, the script is obviously AI generated: https://youtu.be/0d2pCt8JomQ?si=VhqAsQyselNyuxq2
Every paragraph ends with him making the classical comparison "It's not just ____, it's _____." and by the halfway mark you realize he's basically just restating the same thing over and over again in a different way without adding any meaningfully new commentary. Both hallmarks of AI writing.
I think it's interesting that almost no one in the comments seems to take note of this. Maybe I'm odd for being exposed to a lot of specifically AI generated content in comparison with human stuff through this substack, but it's somewhat sad to me that respectable content creators are basically producing well-disguised AI slop.
I'd been noticing more of "people think some stupid thing (no evidence to think many people think that false thing), but actually), and not just from Boyle. I thought it was just podcasters being stupid and annoying, but AI is a plausible explanation.
Dude, also a watcher and felt like something was deeply "off" but I didn't recognize it as AI until your comment. Still have that, "I like this thing so it must be good in ways that are totally unrelated to the qualities I like it for."
After 10 million cases, I'm still like, "How could that actor/writer I like do something bad?"
I think this is how YouTube channels die. A near-imperceptible drop off in quality and people move on, one by one.
I read an old BBS-style forum that allows semi-anonymous posting (i.e. non-registered usernames) and certain threads are flooded with comments that all use em-dashes and curly quotes (“ not ", we'll see if substack leaves in the curly quote) -- clear signs of AI output. It's probably 5-10% of comments now. It makes me wonder if it's just (a) one or two obsessed basement dwellers, or (b) reputation management firms (the AI comments always appear on threads about public figures).
As an incorrigible em-dash user, I'm very annoyed that I have to choose between giving up my affectation and risking being mistaken for a robot.
I really think people would not like the look of our sentences if we dropped the em-dashes.
Me too - I like using em-dashes and so did Emily Dickinson.
Surely by now you have a reputation to point to — don’t you think?
So has sentiment changed on chatGPT 5 from absolutely wash to pretty good really. Or am I reading the timelines wrong.
Maybe there's an intelligence threshold after which you don't get more intelligent; you just get faster and bigger (more data storage). Sort of like a meta-version of the way a Turing machine is universal -- it can emulate any other computer.
My tiny benchmark-ette makes GPT5-Thinking look just about the same as o3
o3:
tl;dr: ChatGPT o3 04/16/2025 7 questions, tl;dr of results:
3 correct, 3 partially correct, 1 wrong
https://www.astralcodexten.com/p/open-thread-377/comment/109495090
GPT5-Thinking:
tl;dr: ChatGPT GPT5-Thinking 08/15/2025 7 questions, tl;dr of results:
4 correct, 2 partially correct, 1 wrong (Gemini 2.5 still looks better.)
https://www.astralcodexten.com/p/open-thread-394/comment/145909954
And the difference between them was how I scored
f) Q: Please give me an exhaustive list of the elements and inorganic compounds that are gases at STP. By STP, I mean 1 atmosphere pressure and 0C. By inorganic, I mean that no atoms of carbon should be present. Exclude CO2, CO, freons and so on. Please include uncommon compounds. I want an exhaustive list. There should be roughly 50 compounds. For each compound, please list its name, formula, and boiling or sublimation point.
which really wasn't very different for the two models. They both got a bit more than 50 valid compounds, and accepted valid additions. I try to be consistent on scoring, but am not completely successful on this.
GPT5 looks a lot better than last year's initial GPT4, but basically o3 with a router, compared to SOTA 04/16/2025.
I'm updating to _slightly_ longer timelines, but not by much.
Note, though, that GPT5's hallucination benchmark _does_ look significantly better:
https://thezvi.substack.com/i/170401319/hallucinations
shows %incorrect claims dropping from 12.7% for o3 to 4.5% for GPT5-Thinking, which is a substantial help. I wish I knew what the human baseline was for this.
The naming is always confusing but is o3 later than GPT4?
Also doesn’t gpt5 decide what to use, it’s not one model but chooses one
Many Thanks! Yes, o3 was released (as o3-mini) on Jan 31, 2025 while GPT4 was released on Mar 14, 2023 (followed by a bunch of later GPT4 versions, to add to the confusion...).
>Also doesn’t gpt5 decide what to use, it’s not one model but chooses one
Yes, if I just let the router decide, under the covers. I explicitly called GPT5-Thinking to, hopefully!, force it to use the "smartest" version available.
It’s hard to separate genuine sentiment from trolling and content farming in the early stages, but it seems to me the more informed sentiment is that it’s maybe a bit better but has pros and cons; a letdown from the hype so far as sheer capability is concerned. However, it is VERY good from an efficiency standpoint - intelligence per compute. Making this a good business revenue upgrade, rather than a consumer-facing upgrade.
I wrote about this in longer form here (https://theahura.substack.com/p/tech-things-genies-lamp-openai-cant) but the tldr is: GPT5 is currently at the top of the pack in terms of general utility, but it is not *so much better than everything else* that it deserved the endless hype sam and openai gave it. There are certain benchmarks where there is wide agreement that gemini or claude still perform better.
GPT5 was "supposed" to be god in the machine. It wasn't. Most of the people who are reacting to it, are reacting to the fact that it was a let down even though it is definitely an (incremental) improvement
Since some people clearly preferred the prior model, it's not even uniformly better than the prior one. (This shouldn't be surprising.)
Most of that seems to be sentiment.
So? That's one of the utilities of interactive things (or even just things that are observed). People get attached to lots of things, and then those things have a (or and extra) utility.
So it doesn’t have bearing on performance….
4.o performed the Suck Up really well.
That REALLY depends on what you are measuring. Since some uses are at least partially emotional support, failing to measure that is misleading.
I miss the time before the Eye of Sauron turned to the topic of AI. Now its sucked into politics, tribes, algorithms. Welcome to hot takes, bots, hot takes of hot takes, and clueless pontification. The answer for any frontier AI is, whats your use, whats your personal style, and what month is it. MY hot take is gpt-oss will be shown to be very good after a few months of messing with it. Similar to llama3, gemma3,
I have particularly enjoyed all the people telling me that current lessons for children in schools should be replaced with lessons that teach them how to write prompts for existing LLMs, because that’s the future and they need to learn it young otherwise they will be unemployable. Obviously, only a Luddite could possibly disagree.
I used to be a teacher lol. I remember all the kids needed tablets for some reason.
To be honest I'm generally happy if students are considered at all. How sad is that.
You raised a good point. I'm sure Elon is already working on it
I think the disappointment comes from mega fans, like rhe reddit openAI sub.
General hostility to AI has other causes and the public doesn’t care about incremental improvements in any model.
Yeah there are two narrative here. I don't think they're entirely separate though. The fact that there's mega fans at all is a symptom.
I only used it once so far, I presented it with a mysterious bug and it completely failed and I had to solve it myself. Didn't really get an impression that an improvement had occurred, but I have to use it more.
THE SINGULARITY WILL HAPPEN IN LESS THAN A YEAR, the ttrpg about community in the face of the end of the world that I promoted the fundraiser for here earlier this year, is finished and available! You can get the rulebook here: https://www.drivethrurpg.com/en/product/513811/The-Singularity-Will-Happen-in-Less-Than-a-Year
And the deck of cards (optional but recommended) here: https://www.drivethrurpg.com/en/product/527331/cards-for-the-singularity-will-happen-in-less-than-a-year
I wrote a post yesterday replying to a few of Alexander's recent posts for my blog: https://crocodialectic.substack.com/p/there-are-communities-everywhere. In it, I argue that Alexander's beliefs about the decline of communities do not take into account the ability of people to join strong communities, which remains high. I also try to define community strength somewhat more rigorously, and end up quibbling with him on some other points. In a rarity for me, I toss in some personal anecdotes about communities I am or have been involved in. In the end, I conclude that people do want strong communities, but they also want many other things, including weak communities, and I do not expect but can't rule out GDP favoring strong communities over weak communities.
In general, my blog is about whatever I want it to be about, which usually means linkposts and books that I am reading. However, I also take occasional stabs at effective altruism discourse du jour, as well as some lighter stuff about language. I publish at least once a week, and have been doing so for a couple of months. Feel free to check it out if it's something that interests you. Feedback is always welcome.
Some representative posts: https://crocodialectic.substack.com/p/farms-are-mines-plants-are-metal (Farms are Mines, Plants are Metal)
https://crocodialectic.substack.com/p/book-review-the-default-world-by (Book Review: the Default World by Naomi Kanakia)
I would say, strong communities used to imply poverty, people needed to help each other survive. Today, if you just want the normal everyday comfy life, you don't need that. However, if you want something entirely different, you might still want that again. I have been in the BDSM community because people there could get something entirely different, also in a Buddhist community so far.
Our issue is - and how ironic is for a more-or-less-Buddhist to say that - the lack of strong desires for something entirely different from basic comfort.
Two questions: 1) Is phytomining safe, or are the farmers exposed to toxic levels of metals? 2) What is the min concentration of a metal to make phytomining cost effective?
Nice to see somebody read the piece! Re: question 1, depends on the crop and processes being used, but it's pretty safe; obviously there's going to be some risk in the smelting process, but most sources seem to think it compares well to conventional smelting. The answer to 2 is pretty similar, it's going to vary based on the crop and the technology, and the technology is still being developed. You might be interested in https://ijog.geologi.esdm.go.id/index.php/IJOG/article/view/171, which looks at gold phytomining where it has a pretty low concentration, all things considered.
Thanks!
I think if you're going to refute the "decline of community" argument (a la Bowling Alone and The Upswing) your post would be stronger with some hard data in it. It's not enough to say "well I belong to a foam sword bopper club and we have picnics!" or whatever, because all the time-use and survey data show large, long-term secular declines in things like time spent with other people, number of close friends, civic engagement, etc. I'm open to the "there are communities everywhere actually, you just aren't thinking about it the right way" argument but I'd like to see a more data-driven argument in that direction
I do think it would be nice to explore more data (and I might do so in the future) but what I'm arguing about is the availability of community, not the rate of people who are actually in a community. The data sources that Bowling Alone, and things like that, are measuring is participation in a community, not access to participation in a community.
I don’t think we are as interested in access as in actual participation though. The difference between GPT-3 and ChatGPT wasn’t access - basically anything you could do with the original ChatGPT, you *could* do with GPT-3, it just took cleverness and wasn’t very fun.
Alexander and other people have hypothesized that access to community is a reason for the decline in the level of participation in communities. So I think it is interesting, at least in the context of whether it is a driver of the decline in community participation (or a barrier to further upswings, for that matter). I agree that, if there was not a widespread belief that there has been a sharp decrease in the proportion of people in a strong community, with attendant negative consequences, it would not be particularly interesting to analyze access. But that belief is there, for good reason, and I think it makes questions of access relevant.
Something I'm getting really sick to death of: people making some negative generalisation about a group of people, and not realising that this not only punishes the innocent (they obviously know this, and don't care) but also *lets the guilty off more lightly* than they otherwise would be! By letting individual bad people in that group hide behind the idea that they're just being a normal member of their group, instead of personally a scumbag.
This seems really obvious and yet it seems so routinely overlooked. Is there a name for this fallacy (in this specific form)?
Examples:
The right hates academics, largely because they're so often brazenly politicised. I respond that academics are great, when they're being proper academics. 100% of the condemnation should go onto those *particular* academics (however many there may be) who are so selfish, so entitled, and so contemptuous of their own profession's ideals that they would proudly exploit their position and its remaining social trust to shamelessly further their own political views, instead of pursuing truth for its own sake. Declaring war on academia, among many other things, only *protects* these selfish scumbags by *giving into their framing* of being normal academics, instead of terrible academics who behave in total opposition to every value they are supposed to hold.
Feminists do the "#notallmen" shit where they imply or outright state that men are naturally rapists, or cheaters, and by doing so give actual rapists and cheaters a shield of "just being a man". When they should instead be told "no, your behaviour says nothing whatsoever about men, and everything about you. If you rape or cheat, the only information we gain (and we gain it to the strongest possible extent) is that *you*, personally, are a subhuman piece of shit".
Redpillers "reveal" the "truth" that all women are gold diggers, are excited by and attracted to violent men, commit paternity fraud, make false rape accusation. Congratulations on letting every woman who *does* do those things off the hook. Again, if a woman dates a psycopathic criminal because he's exciting, tells rape lies to get personal advantages, or is a gold digger, that says *nothing whatsoever* about women and *everything* about her personally being a piece of shit. Do you actually *want* to give these animals an excuse to hide behind that they're "just being a normal woman"? Aren't you actually *angry* at these behaviours, such thar you want to actually expose and shame the ones responsible for them? (Same question for the feminists.) Is it all an act? Or have you not thought through this at all?
I see a fair amount of this even in the ACX comments. What is wrong with people?
Heh, your piece on academics reminds me on Chomsky. How does being a linguist somehow make him an authority on every war and every political system?
However, a lot of real actual research ends up with political results that tend to lean left. The reason is simple. A conservative is not someone who rejects the philosophical ideal of egalitarianism, but someone who thinks it has been already achieved. So they are not happy when actual researches tell them writing a piece of paper that declares people to be equal actually does not achieve this.
Academic political philosophy is good and IMHO moderate. Rawls can be summarized as "You know taking out an insurance is a rational think? Now think it further."
The BDSM community has strong problems of this. People pull off vile, non-consensual, unsafe things with newbies and then say well BDSM is like this. No, it is not.
You're right (and as Kenny Easwaran says you're just rediscovering racism/sexism/classism/jingoism etc.).
I have to say at this point I've just given up-- I think it's just human behavior and at this point am mostly trying to figure out which tribe to join, since right now they're both kind of bad. Humans are slightly evolved monkeys and we need to stop expecting more from the general public.
At this point, neither. I'd say in the long run left, but first reality has to beat their elitism out of them.
This is downstream of groups not policing themselves. The outsiders shouldn't have to do a deep dive to figure out which people in the group are the problem; the group should be making it clear. If they aren't, one assumes they're happy with the behavior. Groups tolerate offenses by their worst members, and therefore are represented by their worst members.
That is a strange take when we talk about large heterogenous groups with no leadership like men, women and academics. Who do you think should police them? Is every man somehow particularly responsible for the action of every other man because he happens to be a man?
There are actually a pretty large number of veto points in the average academic's career. E.g., hiring committees generally need to unanimously agree on which candidate to choose, and a lot of early-career jobs are time-limited, meaning you have to be OK'd by multiple such committees to stay in academia. Whilst there isn't so much formal policing of academics' behaviour, then, in practice people whose views or actions make them unpopular with other academics tend to get weeded out.
I think you’ve rediscovered the phenomena of racism/sexism/classism/jingoism/other isms!
Yes but I was specifically broadening it to the effect it has of letting the guilty off by diffusing responsibility amomg their group. I don't think this is very widely acknowledged; I'm not sure I've even heard *once* the argument that e.g. supposing the Nazis were right about Jews having betrayed Germany, by going after all Jews instead of the specific ones who betrayed them they were letting the latter off lighter (than they would have been had they received the full force of the anger).
It seems like "isms" being wrong on their own terms is an important phenemenon separate from simply isms being wrong.
> Yes but I was specifically broadening it to the effect it has of letting the guilty off by diffusing responsibility amomg their group.
(Sans the "Specificity vs Sensitivity" trade-off), perhaps you're complaining about Free-Riders? Or are you looking for a niche term for both phenomena simultaneously. (Which, to my knowledge, doesn't exist. But I suppose you could always come up with one yourself.)
> supposing the Nazis were right about Jews having betrayed Germany, by going after all Jews instead of the specific ones who betrayed them they were letting the latter off lighter
What does that even mean? The "betrayers" died just as badly as the "innocent". Sure, they didn't get all of them, but that was only because they lost. They would have gone after the rest of them if they won. So don't lose.
I'll defend this behavior as rational in the same sense that it's rational to hold corporations responsible for the behaviors of their employees. You're attacking the entity that is best-positioned to punish the behavior going forward. This is harder to argue for diffuse categories like 'men' or 'women', but I think the academia example is a good one. The problem behavior (of particular academics being performatively left-wing, for example) really is downstream of institutional policy. It's therefore rational to criticize the institution in the hope that it will change its policies. A professor has no reason to care if *I* think he's a bad academic, but he has plenty of reason to care if his tenure committee does.
I think I agree, and I wasn't trying to talk about concrete institutions, which I agree can be held responsible for particular acts. I was talking about the fact that a large part of the right seems to have come to hate the very idea of academia itself, to the extent of increasingly holding it as an article of faith that most academic disciplines are basically useless at best. Maybe this is independent of the objection to corrupt and politicised experts, but it certainly doesn't look like it.
And even your instutional point I have a few reservations about. Attacking a whole institution on the basis of certain members' behaviour is probably one of the most effective ways imaginable of getting everyone part of it to rally behind said members.
I think that's a continuation of the same principle. Conservatives have been complaining about institutional bias for 20 years with no results and so they've moved their target up one level to include all of academia. Trump's random and arbitrary punishments are, in a narrow sense, not really defensible. But in a broader sense I would argue that academia's heavily politicized influence on society has made them a ripe target for political hay-making. In my view they have no one but themselves to blame for that. If a country wants to maintain neutrality in a war then it shouldn't run weapons for one side unless it's prepared to be considered an enemy combatant by the other.
> one of the most effective ways imaginable of getting everyone part of it to rally behind said members.
IMO that's only true for symbolic punishments. If you make the attack fierce enough then institutional solidarity will immediately crumble in an "every man for himself" defection cascade. Survival always trumps ideology, especially for soft ivory-tower types. Here's a recent op-ed which illustrates this:
https://archive.ph/kypdn
I'm highly confident that views like that (or the willingness to express them, at any rate) are directly downstream of Trump's policies. He's making academics fear for their survival. That tends to drive change.
Oh agreed. Problems are always caused by the shittiest 5%. I think this reinforces the logic of my point. As a non-academic I can't directly target that bad 5%, but as a taxpayer or pundit I can attack the institution in a way that incentivizes it to target it for me.
The last line is the key. "The professors don’t have as much power as many right wingers seem to think".
The point being made here is that it isn't the academics (apart from a very small number of far left ideologues) that are at fault, but the bureaucracy.
Harassing academics has literally no effect on the bureaucracy, which has a completely independent, although largely parasitical, life of its own. It's almost impossible to incentivise the institution to attack its own bureaucracy. They will just pillory and sack more academics, but save themselves.
This is a pivotal fact that seems to have escaped practically everyone.
Then the institutions need to be replaced and Trump is doing the directionally correct thing by targeting them.
That's pretty much straight up stereotyping: Judging people by the category they belong to, not their individual behavior.
Of course, to follow the guidelines you propose, someone would actually have to do research and collect some data to demonstrate what the base rate of the behavior they object to is. Most times, that will weaken the argument they are making (very often fatally). So the incentives go the other way, unfortunately.
You think they want to punish the guilty, but have you considered that they want a weapon to gain more advantage over a larger group?
Scott has written about this effect: https://slatestarcodex.com/2014/05/12/weak-men-are-superweapons/
I agree with you that generalizations like this are frustrating, and that they muddy the discourse.
Thinking through it, though, I'm starting to wonder whether the causality is reversed here. Instead of "people complaining about groups of others unfairly generalize, and therefore guilty individuals get off relatively easy", it's "people observe the guilty individuals getting off relatively easy by dent of their membership, or, are not punished by the groups their a member of, therefore, complainers generalize the guilt not only to the individual, but to the support network"
working some specific examples:
Many cops care primarily about doing a good job and serving justice. A few cops abuse their authority for a feeling of power, even in contradiction to ostensible department policy. An "individual-first" approach would suggest that we should celebrate cops in general, but vehemently prosecute abuses of power. And in a healthy "individual-first" society, abusive cops would find themselves ostracized from their community and without support in the face of prosecution.
In practice, attempts to prosecute even the most egregious cases get mired down in departments rallying around the accused cop, and a broad cultural contingent ready to defend arbitrary actions as "necessary for the job". So instead, while an individual's actions _could_, in theory, be prosecuted individually, instead that guilt spreads over all, because you can't access the individual without going through the all.
Academics are similar. Many are heads-down, quietly working to be sure of their conclusions and the evidence behind them. A few use their platforms to advocate for policies they can't rigorously support, or even to bully those less-platformed. The bullies and the wokescolds _could_, in theory, be dealt with individually by their universities and their colleagues, but in practice, they get supported, defended, and even the most careful academic might lend their voice in support of the bully for fear that the cultural eye of sauron will turn to their niche field next.
Men, is probably one of the more clear-cut cases, where criminal acts would get favorably swept under the rug by a system where other men were in power, and if anything came of it, victims would be blamed. This _may_, possibly, be in the process of shifting, as the social concepts of 'what makes a man, what makes a good man' are re-evaluated, and in the future we might see more internal condemnation of abuses of masculinity.
In all these cases, the thing I'd say is "the sins of an individual haven't, in practice, invalidated their membership in a group, and the group has even provided cover for those sins."
I don't claim that it's therefore true and right for a complainer to generalize the sins of an individual to the sins of a group - but I do claim it's understandable.
Actually, I think the direction of causation may be bi-directional, and not in a good way. Let's say one is a member of "Minority A" (blacks, cops, academics, men, doesn't matter). You see someone from "Majority B" arguing that a member of your group is guilty of a bad thing, and implying that their being an "A" is one of the reasons why.
One could simply point out that while the bad thing is bad, being a member of Group A is irrelevant to that argument. Instead it is in the interest of a member of Group A, who would like to inflate their status within that group, to argue that everyone who complains about the bad thing is actually attacking the entire group, so Group A has to rally around a leader (themselves) to protect themselves. Meanwhile, members of Group B see this, but do not call it out as a weak argument, instead they use the same sort of argument to bolster their own status within their group, to argue that everyone who complains about attacks on Group A is actually attacking Group B, and that to defend themselves, they have rally around a strong leader (themselves). Thus, you end up seeing a kind of reciprocal "choreographed mutual stereotyping" which benefits the extremists in both groups at the expense of the majority of both groups.
Just replace "Group A and Group B" with a pair of groups from the headlines. People attacking the government of Israel for their actions in Gaza are anti-semetic." "People attacking Hamas for their actions in Israel are islamaphobic." Only extremists are making these arguments, because it's in the interest of both sides to bury the moderates, who might actually solve the problem, thereby undermining the leaders on both sides. You can see this is lots of different conflicts (cops vs minority communities, feminists vs mens rights, capitalists vs socialists, etc.). That's what polarization really is--extremists attacking the moderates in their own camp, so that their community will feel less secure, and therefore are more likely to allow themselves to be influenced and manipulated. The interesting thing is the coordination between the extremists on both sides--they each depend on attacks from the other side as a basis of increasing their influence within their own.
Note that I am not arguing that all sides in every political dispute are always equal. If fact they rarely are. But it's in the interest of extremist leaders on both sides to make the conflict (whatever it is) seem intractable because they depend on that to promote their own status within their community.
It's the "Rational Middle" that gets left out.
I think that's a variation on "No true Scotsman".
...now try making that same argument without being guilty of the practice you wish to condemn: making negative generalisations about groups of people like feminists, right-wingers or redpillers.
How about by: observing that political labels are basically self-chosen and if a label gets a connotation or implication you don't support you can always...stop identifying with it. Which puts it in a different universe to generalisations about some innate or simply politically independent characteristic?
(I grant your point on "right" though. I actually think I meant to put either the "often" or the "largely" in that sentence after "the right" instead of where I put it, but I can't prove it and it's possible I'm retrospectively deceiving myself.)
Or alternatively: by noting that my argument doesn't actually fail just because I'm a hypocrite. It just means that I'm being too soft on the actually terrible feminists and redpillers and giving them cover for their actions. Are you disputing that claim or not?
"How about by: observing that political labels are basically self-chosen and if a label gets a connotation or implication you don't support you can always...stop identifying with it."
This is too severe a restriction. I am not going to stop self-identifying as a Christian because some priests molested children, or, I am not going to stop self-identifying as an American because the government did bad things.
You work to improve the behavior of a group you belong to, you don't withdraw from the group (I am not going to stop supporting science because some scientists did bad things).
However, there are people of bad faith who will attack anyone who tries to improve a groups while staying within it, so that's a thing.
I'm angry that you'd say, "What is wrong with people?" rather than directing your anger only at the individuals who engage in this behavior.
??
I'm pretty sure the word "people" in a phrase like that has a secondary implied meaning of "the people who actually do this",and I was using it in that sense.
EDIT: Oh, are you accusing me of doing the thing I'm condemning? I don't think I am because I was using a figure of speech and:
Are you claiming that the people I'm complaining about mean their generalisations as a figure of speech? And that therefore...the gender wars that take up half of Youtube and Tiktok are all just a misunderstanding?
Or are you claiming that anyone around *here* who uses that rhetoric is just using it as a figure of speech? Even when they're using almost identical language to the first group? (See some of the comments on the recent review.)
Or are you saying something else? Hard to tell what your objection is with such a short reply.
90% of my reply is: would you make the same defence for redpillers?
10% of my reply is: I don't think they're often doing that, because the "#notallmen is privileged derailing" talk comes up when there's some message along the lines of "men need to take more respinsibility for rape culture" that causes people to say "not all men" in response. The latter phrase isn't being used much in response to advice aimed at women to keep themselves safe.
Or at least that's how *I've* seen the discourse play out.
There is a word in your survery (let's call it M) that is rather anglo-centric; this will stop it doing its job for many foreigners. I suggest "M (or your local equivalent)".
Note for future surveys: include question on "are you a filthy monoglot Anglophone or a speaker of a beautiful expressive melodious tongue unpolluted by the bastard dialect inflicted on the globe by the Brits and Yanks?" 😁
If you mean the name, I asked about American vs. non-American specifically to be able to adjust for this.
Regarding the central topic of the survey, all the articles I’ve seen list only people who have had issues with ChatGPT. I’m definitely getting curious about whether this is just because ChatGPT is so much more widely used than the other LLMs that almost all cases are associated with it, or whether it’s because there are important personality differences between the different LLMs that cause differential rates.
One of the nytimes articles recently took a particular segment of the person’s problematic conversation and found that both Claude and Gemini did almost exactly the same problematic sycophancy as ChatGPT on the one example - but the person did say that they were broken out of the issue specifically by conversations with Gemini, which suggests some relevant different in longer conversations.
Hmm. I said not American and got the M. I’m Anglo but it didn’t know that. I think.
I think he's not saying the survey was variable depending on your location, but that knowing you're not American will affect the likelihood of certain answers, so he will calibrate accordingly.
Another property you ask about comes in two variants one more rare than the other. You specifically ask about the more rare variant. I know a lot of people with this property but I would not know which of the two variants they fall into.
Huh, wouldn't you want to include all English-speaking countries (at the very least)?
Do you think that is so subtle a point as to have escaped his notice?
Well I was only trying to help!
Fair, I shouldn't have been so rude. Sorry.
"Well I . . . well I thought . . . oof."
According to the 2025 ACX survey, almost 75% of people who read ACX and respond to surveys here are from core Anglosphere countries (US, UK, Canada, Australia, New Zealand, or Ireland) and another 8.6% are from countries (Germany, France, and Israel) where the local equivalent of the M word is identical or almost identical to the English version.
One of the latter, ironically, being the place where that very word originated.
I'm trying to figure out the disconnect between good ideas and implementation in the public sector. There seems to be a real gap in our models for this. We have theory, isolated case studies, and lots of opinion, but I don't see much "work" being done on the failure mechanism itself.
My concern is that any new governance model for abundance, rationality, you name it, is fundamentally constrained by this. Every grand vision can be held up by internal processes and gatekept by an archetype I call "Bill the administrative assistant."
One would think this is an easy obstacle to overcome, but I've seen it consistently derail massive projects. Some of the best thinkers on this, like James Q. Wilson (Bureaucracy) and Jennifer Pahlka, have shown they understand the symptoms perfectly. Yet the disconnect remains. We seem content to complain about it but not actually fix it.
So, I'm interested in insights that go beyond the usual narratives of party politics, capitalism, or doomerish acceptance. I'm looking for better frameworks to understand this specific, gears-level failure mode. What are the actual mechanics at play here? and more importantly how do you improve it?
The problem *is* politics, and it's effect on administration. It's been studied extensively, and is widely recognized as very difficult to overcome.
An essay on the topic by Woodrow Wilson published in 1887, showing how far back this problem was recognized: https://politicalscienceblog.com/woodrow-wilsons-the-study-of-administration/
Here is a broad review of much of the classic literature: https://www.press.umich.edu/pdf/0472113178-ch1.pdf
The so-called "Principle-Agent Theory" guides much of modern research. A principle (in the case of a government, the elected officials) must select agents (agency heads and staff) to carry out their agenda, but research clearly demonstrates that this relationship is complex, given that everyone pursues their own interests: https://doc.rero.ch/record/299898/files/30-5-302.pdf
Public sector organizations, compared to private business, suffer from goal ambiguity (ie, how to interpret and implement the agency's mission), and goal ambiguity negatively impacts performance: https://books.google.com/books?hl=en&lr=&id=wjHFf_zLWggC&oi=fnd&pg=PA92&dq=public+organizations+goal+ambiguity&ots=c8Wxtat0s5&sig=3WKtd4zRr7IKC4HL4ioKX2u0WTk#v=onepage&q=public%20organizations%20goal%20ambiguity&f=false
Fundamentally, this problem has no simple solution, because public sector agencies depend on support from elected officials and the public to ensure their funding and survival, but both groups are divided into competing factions, pursuing incompatible agendas.
This is good information, I don't disagree. I do think a nudge in either direction has significant effects. I don't imagine fixing principal-agent. I do think some agencies do better or worse. It's in that space I'm interested.
Thank you for the information, I had already planned to deliver more into principal-agent and resources are always welcome
There is not a lot of research on this. I worked in the public service space as an HR performance consultant for 20 years, and in my own opinion, the kind of difference you are asking about isn't really in the control of the agency itself. It's due to the leadership of the elected officials they report to (in the case of non-profits, the Board of Governors). The history of White House - Agency interactions testifies well to this: From the Bay of Pigs to the recent firing of the Labor Director, public service agencies master the fine art of pleasing the King.
I suspect this is the case as well, I think it needs to come from an outside pressure
You might find this analysis of the Progressive Era interesting: https://www.nber.org/papers/w4973
It's all happened before.
That's very interesting, I had been drawing comparison to now and the guilded age, this fits right in
Ideas formulated to be implemented in the public sector have LOTS of problems.
One is they have to be stated in such a way as to not offend any important group. Another is that those pushing them usually have multiple motives, many of which are not clearly stated, or perhaps even recognized. Another is that they need to satisfy multiple different groups, many of other primary goals.
E.g., I had a friend who would always vote against bond measures, because he held that was an excessively expensive way to implement funding. He might be quite willing to support it if it were to be a tax measure, but not if it were to be a bond measure. I, personally, think it a shame that few people think that way, but it means that many things are always bond measures.
Good points. I do think it's worthwhile, even if ultimately futile. I might be tilting at windmills but I could be chasing worse goals I suppose. The government lacks a price signal, and it also can't participate in a market really. It causes a real lack of information. But we can't realistically just get rid of it.
Let's not ignore the fact that commercial enterprises suffer from very similar problems. The real difference is that when a business is really bad, it goes bankrupt and goes away, while we are stuck with a government forever. Government isn't really any less efficient or effective than the private sector, it's just that the consequences of mistakes are more severe.
Yes, I totally agree. Government isn't allowed to fail. When we point to successful efficiency in business there's survivorship bias, they're visible because they're the small percentage that work at this point in time. Another issue I want to write about is how government doesn't have the benefit of price signal. It's a significant missing piece of information.
Excellent point Victor
> I'm trying to figure out the disconnect between good ideas and implementation in the public sector.
I'm not sure what scale of "public sector" you're thinking here, but for large city or state and larger, Bent Flyvbjerg wrote How Big Things Get Done about large projects (be they building bridges, buildings, IT projects, and more), and why they routinely come in massively over budget and are finished years past the initial projections .
Essentially, there are interlocking reasons behind these budget and time overruns - politicians outright lie even if they know it will take longer and cost more, any complex project has many ways to fail and get worse, but only a tiny amount of ways to succeed / stay on the critical path, and a desire to do things and show progress rather than spend time + money planning things out well.
The big mitigations are:
1. Spend more time planning
2. Act fast and pay more for highly skilled teams, because the longer you're doing things, the more you're exposed to disruption
3. Spend time actually thinking of and quantifying various risks and mitigating them
I wrote a review of his book here if you're interested in a little more detail, but the book itself is well worth a read.
https://performativebafflement.substack.com/p/how-big-things-get-done-bent-flyvbjerg
Thank you I'll read it. I had not heard of that book ,but others like it. There is a gap though from the book to the government. Shining the spotlight is the first step, but then what? There isn't a satisfying answer. We can proscribe mitigations, but they're not be used.
-edit Looking at your review, its interesting my intuition about optimism was the opposite. I think we assume government will fail and so take it in stride. I'll have to think about that more
On the left, at least, there is real wariness about doing anything that plausibly can be called deregulation or neoliberalism.
i feel like this falls under "usual narratives of party politics" that I asked to go beyond. I also don't see how deregulation or neoliberalism relates at all.
My point is that, though a lot of our governance failures are because of excessive regulation, a Democrat cannot simply run on "slashing regulations so businesses can create jobs and build more homes." Instead they have to triangulate among a bunch of interest groups.
I feel like you're arguing against a different question. The process questions I have are not regulatory. Also as far as I know these issues are in government divisions with both republican leadership too. Its not a party issue.
Could you elaborate on what you mean by the "bill the administrative assistant" archetype? From reading those authors, I understand the problem to be that *most* of the dysfunction stems from lawsuits and the fear of lawsuits, and the mechanics for that are pretty straightforward.
I'll give a very specific anecdote. drinking water operators need state issued certificates to operated facilities. The range from 1-5 and correspond to increasingly complex and large facilities. Each step requires a test, an application, and a certain amount of experience. One person in Sacramento at the State Water Resource Control Board processes those applications. The person who used to do it was, problematic. She told me she usually rejects most applications on the first submittal. She didn't elaborate on why. The rules she should adhere to are codified in my substack's namesake Title 22, yet she felt she was allowed here own interpretation (she's not). She once told me that she had about 100 applications in her queue and each one takes her "a couple days" , mind you, there no really need for scrutiny here, you either submitted the requirements or not. Should take an hour. The real world result was that in an industry that is persistently short handed, a single person created a bottleneck that had real logistical and financial impact in California. Her supervisor new it was a problem. Every water utility in California knew it was a problem. I messaged the water board, so the should have known it was a problem. She recently retired, and it seems to have resolved it. This is not an isolated case, I've seen it all over.
That strikes me as very odd. I know little to nothing about California, but at the Federal level those sorts of decisions would be open to legal challenges: https://www.nrc.gov/docs/ML1407/ML14077A573
Things that are open to legal challenges and things that are challenged legally are far apart. I understand your confusion, and it part of the basis of my question. There is an underlying sense that this is all illogical. Its persists though, so its tells me my model is missing something. This is where I'm at at least.
This seems to me an example of an interaction between things at two different "scales": the institutional scale, where there are clear rules and procedures, and the personal scale, where individuals with unique personalities and behavioral patterns are tasks with implementing these rules and procedures. Trying to eliminate deviation from formal parameters due to individual differences seems a little like trying to eliminate weather delays--maybe you just have to plan for such deviations in the first place. In any given organization, there are going to be "personalities" that introduce delays and waste. Probably impossible to eliminate, so plan on it instead.
I don't know, I really don't. I will say that the NRC which I have delt with has a great many rules in place in an attempt to limit discretion and a many large volumes of rules, the examination standard alone is 460 pages. (Because of this?) the NRC gets very few legal challenges to its licensing decisions even though they are worth a great deal of money to the people applying for a license.
sorry, what is the NRC? Just for refrence, I get the gist of it. Do you have the sense the method the NRC is using is better for outcomes though? I have a hunch this is an opposite approach that is also a bottleneck
1. People don’t magically start following rules just because someone wrote them on a piece of paper (even if that someone is a legislature).
2. The Water Board isn’t going to fire someone just because service users don’t like them (assuming that’s who even makes the hiring/firing decision). No-one in public sector management cares more about “whether California has enough water” or whatever than they do on actually managing their subordinates. Firing Ethel because of some industry sleazebag’s* poison pen letter to the board is a non-starter.
3. No-one’s going to go through 3+ tiers of public employees to push the message “stop reading the applications, just rubber stamp them” in general. What you want is to specifically reprimand someone. That’ll be a formal process involving their union rep, and there’s no specific misconduct to hang that on.
4. “Ideas don’t matter, people do” isn’t some hippy slogan, it’s the whole of politics.
*View from their inside.
1. correct, however the options aren't magic or noncompliance.
2. You seem to be reading a lot into the anecdote, that's interesting. FWIW, my letter asked about the rejections rates of applications, the common cause of rejections, and if there's a possible process fix.
3. The options are not (or should not) rubberstamp, or find a way to reject.
4. If you don't like the slogan , why did you bring it up?
The point is that no-one in a position to fix the problem has any reason to fix the problem, and fixing it isn't trivial for them. "Things would be better if they did," and "the law says they have to" aren't reasons to do something that involves significant costs in organisational capital for whomever's going to do something.
I feel like this falls under my point of "usual narratives of...doomerish acceptance"
I'm reading this as restating my question in a snarky way, which is fine. I do this myself. I'm at a point though I want to figure out these things. Why don't the people in position to fix, feel like they don't have a reason to. Also why doesn't the cost to the organization balance with the cost to the goal of the organization.
If we to take the narrow view, America wouldn't even exist.
Also, I am interested in your inside view. You are the person who we need to talk to, not the director or the people strategy person
Ah, I misunderstood and thought that "bill" was a verb.
oh i see. I do understand your point, being overly conservative due to legal and political threat is also a thing. It is somewhat related. For my example, I know the group had gotten in trouble for being too permissive which led them to overcorrect. However it wasn't the motivation of the individual person here as far as I can tell
"Can't fire her or she might sue" could be part of the problem that would tie back to fear of lawsuits.
I do hear "she'll sue", but it's actually not the case. It's harder in general but there are always levels of progressive discipline to follow. Also it could be the case that she capable of doing a decent job and is poorly supported by here supervisor. I'm definitely jaded in this anecdote, but a very real scenario is that the gatekeeper is being rational given their incentives and resources.
The disconnect for me, is that the problem created is sufficiently massive that some kind of resolution should have happened sooner. Even if they fired her improperly, the legal cost would less than the cost to the industry as incurred.
Its honestly weird as having worked inside government bureaucracy everyone involved seems to want the improvement too and just as fed up with how the system actually works. No one likes it except for a few special interests with minimal connection to a policy other than scuppering it.
I think in a lot of cases, it comes down to executive authority and empowering someone with a mandate to improve it. See Pennsylvania as an example: https://responsivegov.org/research/pennsylvania-red-tape-reduction-a-case-study/
If it's not explicitly your job, and you don't have blanket authority from some higher-up (in the linked case, from Governor Shapiro himself), you might complain about the system, but you 1) don't really gain anything from fixing the system (at least from a personal profit perspective) and 2) would probably be penalized for screwing something up if you tried any amount of making things better that had a nonzero risk of something going wrong (i.e. doing literally anything). Even if you have a ton of good suggestions floating around (e.g. the GAO has a self-reported ROI of $133 for every $1 invested, which is insane https://www.gao.gov/products/gao-24-107438), the actual implementers have to be incentivized to accept them. But public servants aren't hired to improve their workflows, they're hired to do the job they were hired for. Asking them to do any more means asking them to take on extra risk for no reason except civic duty.
On the other hand, if it's explicitly your job to improve things, you tend to get it done - e.g. Pahlka in the USDS before it became the crapshoot that is DOGE, or the Pennsylvania example I cited. And if you have a mandate you can point to, you sidestep the issue of the skittish bureaucrat who doesn't want to lose their job, because now they have a person higher up on the chain that they can blame if something goes wrong (this sounds cynical but isn't really: again, why would you make a bureaucrat risk getting penalized for something that they weren't hired to do?). This is also why the authority should come from the executive: they have the security of a term length that is usually 4 years, so they can start all of their grand schemes as soon as they enter office, and hope that their EV-positive plays will pay off by the end - I think that optimal play for executive authority means taking all the risks as soon as they get elected and then becoming more risk-averse as they begin to seek re-election.
TLDR: The irony is that you're looking for good ideas on implementation. But those good ideas still are not implementation yet. Hire people in-house to improve your policy implementation, and that's what you'll get.
These are are excellent points. I agree and mention Pahlka as someone who "gets it". So why don't we have mini-Pahlkas poping up? There are instances of an executive at a agency who makes improvements, but it seems isolated. America seems to love copycatting, but I don't see it here. Also there seems to me systemic methods that's could be reproduced without a special executive. Yet, the progress seems to die when that special executive leaves.
I mean, you could argue that DOGE was driven by the public desire to improve government efficiency. If it had been headed by someone else (or if it had just been the USDS but with a greater mandate and more executive authority), maybe it could've actually done so. Instead, well, you know.
The special executive problem is a real one, but it also reflects the positions of the populace (and the problem with representative democracy in general). Elected officials do what they think will get them elected - I'm hoping that Abundance and related recent vibe shifts will bring about a round of efficiency-minded mayors and governors, but only time will tell.
Also, I think efficiency is generally bad marketing. People care more about actual impact than they care about ROI, and for good reason: government is very subject to the "efficiency that doesn't scale" problem. Many implementation fixes are hyper-local and require deep understandings of the problems + the systems we're working in. One of my favorite nuggets from the Pennsylvania case is that lawyers were staffed on implementation teams from the very beginning, so that the exact legal bounds of what was and wasn't allowed was never in question. And afterwards, they marketed themselves by impact: we reduced processing time by xxx amount, we killed a backlog of xxx permits, etc. Much better than "we made xxx process more efficient".
I can't recommend anything beyond "make a team that is empowered to improve processes by finding and utilizing local knowledge", because anything too general ironically might not generalize.
This is a weird time in history. We say DOGE unironically when its the stupidest possible name for the goal. Also I agree with the sense "we" wanted a DOGE but a foreign billionaire did it on a whim. Ramaswamy for all his BS probably had a decent plan (of the plans possible for this administration).
I'm not sure I agree with "reflects the positions of the populace" because these aren't elected positions. The elected do have some influence but for the most part this is the "deep state"
There's two ideas I'm considering. One is that this is all down stream of the country's state of mind and that we either find a way to make it work , or not because of it. Post WW@ we believed we could do things so we did, now we believe we are broken so we are.
The other is somewhat related. Your idea about marketing and generalizing are floating around it. I don't think there is a prescribed playbook to "fix" the government but there could be a culture shift internally that would signal competence in and out of the sector. We say nothing changes, but when woke (i don't have a better descriptor) became popular every government entity reorganized at lightening speed. New positions were created, trainings, documents re-wrote. This as at local, state, federal levels. I point this out because its counter to the narrative that these groups can't change. So what I'm trying to figure out is can we reproduce that, but where it's aligned with agencies stated mission,
"One is that this is all down stream of the country's state of mind and that we either find a way to make it work , or not because of it. Post WW@ we believed we could do things so we did, now we believe we are broken so we are."
I think you hit it on the nail here. The problem you seem to be exploring isn't really located within the procedural arena, it's a problem of the self-fulfilling prophesy at the national cultural level. Something changed in the political culture between WWII and today, and the thing that I think changed was the wave of globalization that struck the US starting in the 1980's. The benefits and the costs of that transformation were not evenly distributed across the US, which broke down a lot of trust in public institutions and other Americans.
This means that the solutions is simple, but challenging: Someone has to persuasively run on fixing those problems, ie, a platform of deep national social reform. This didn't happen, so here we are. We needed the second coming of FDR (except with updated solutions). "We have nothing to fear but fear itself." Convince enough voters and I predict government would become a lot more effective.
Generally agree on DOGE. I think the word efficiency is poison in relation to government now, and any attempt to make it more efficient must come with an ironclad promise of no service cuts. Better government, not less government.
> I'm not sure I agree with "reflects the positions of the populace" because these aren't elected positions. The elected do have some influence but for the most part this is the "deep state"
Sure, but the point of the Pennsylvania example was to illustrate that Governors (and probably Mayors) have the power to effect change if they want.
I think an interesting example of results-first comes from congestion pricing in New York, which started out having terrible approval ratings (https://abc7ny.com/post/nyc-congestion-pricing-nearly-two-thirds-of-new-yorkers-oppose-plan-siena-college-poll-finds/14721916/) and has since gained popularity because it worked (https://www.newsmax.com/newsfront/poll-congestion-pricing/2025/03/10/id/1202234/). Trying to elect an efficiency-minded official might be hard, but people like it when the government makes good things happen.
Also, I wouldn't call "woke" an internal culture shift. I think that was something that started outside of government - I'm not sure if you could do things the other way around.
"Ramaswamy probably had a decent plan" - didn't he suggest at one point firing everyone with a SSN ending in an odd number?
what's weird is everyone feel beholden to rules they just made up for themselves. Then a bad actor changes one, everyone complains and then feels beholden to that. It like some kind of organizational freeze tag.
Have we considered scuppering those who would scupper the improvement, or would that be a violation of democratic norms?
Question for native English speakers. English is a highly analytical language, contrary to German, which is very synthetic, where much of the meaning a word gains from its context is backed into the syntax.
Please read this sentence from the wild:
"Suppose someone was, unknown to themselves, 'enlightened' as you and the people you mention are using the word, but not through any practice thought of or labelled as devoted to achieving 'enlightenment'."
I stumbled at the word "thought".
I wouldn't if nouns were capitalised at all times. I would have known it's a verb.
I wouldn't if you conjugated verbs thoroughly, for the same reason.
I wouldn't if you had more rules for commas. I would have known a new clause had started.
I wouldn't if you didn't drop "that" so often, for the same reason.
Uncomfortably I read on, saw the "of", and got what's going on. How do sentences like this go for you? Don't you stumble too?
I stumbled more at the line break between "are" and "using the word" since "as you people are" has its own different meaning. I basically read "thought of" as one word, the same way I would read "thoughtful". Likewise "labelled as" and "devoted to"; those short words just get lumped into the words in front of them.
(Let's see how easy this one was to parse.)
I don’t think English and German differ much on the analytic-synthetic continuum - both have nearly the same constructions of verb endings, compound verbs that embed prepositions, compound nouns, and so on, though German has a writing system that uses fewer spaces in a lot of these, as well as capitalization (and also has a bit more remnants of grammatical gender and even case). But neither of them is anywhere near as far in the synthetic direction as Semitic languages with their triconsonant stems, or in the analytic direction as Chinese, without conjugations at all.
Of course the languages you know of might be much more synthetic, yet the difference in syntheticness between English and German is immense for me.
I tried learning Spanish.
And I gave up when I came to subjuntivo, something that's equally extreme in German, and it was then that I realized what a simple language English was. Memorizing English irregular verbs was a breeze in contrast to having so many different regular endings in all tenses. Speaking spanish was like mental arithmetic for me. I have now enormous respect for people who learn German as an adult.
The only complicated thing about English is that there are no rules connecting speech with writing. Very annoying. Much worse than German. And Spanish is the best in that regard.
"I have now enormous respect for people who learn German as an adult."
15 years in Austria and my grammar still sucks, and I still write emails with a dictionary open to look up Artikels.
It is no surprise that English took over the world. The ability to just read a random text, figure out a word from its context and then immediately be able to use it is phenomenal.
Imagine living in the Middle Ages and using Latin. It has no fixed word order, adjectives do not stand in front of or after nouns. Madness. We don't know whether the cult of Ostara was a thing at Easter, because we don't when reading Bede that "quid nomen", "whose name" is the name of the goddess or the name of the month.
In my view that's a slightly convoluted sentence. I would give editorial feedback to write more clearly and succinctly. I wouldn't say that it reflects problems inherent to English per se but rather illustrates the importance of learning to write well. I don't know German at all but surely one can write badly in that language too? If it's harder to write clearly in English then maybe that represents a tradeoff between expressiveness and complexity.
The example sentence might be bad writing, but the problem I have regularly when reading english is suddenly not getting where a word goes meaningwise, because that is to be found in the context -- no hints in the declination or conjugation or capitalisation of the word itself.
Also those whitespaces in the middle of terms and not so many obligatory commas between clauses do not help.
Erica Rall gave the "The old man the boat" example and said "It 's a real problem, but a relatively uncommon one in ordinary usage."
It might be uncommon, but compared to German it's common.
So I think one can write correctly but nonetheless badly in German too, but has fever ways to do so.
Not really? After "any practice" I expect some sort of adjective-like-thing to qualify it. A past participle like "thought of" works fine, but 'thought' as a noun would not.
"Thought of as" is also a very conventional phrase for the concept of "is generally considered to be"; not quite an idiom, since it literally makes sense too, but common enough that "thought of" parses as a unit here.