Hi all, for the last few years I was working a game, and now I've released it, just as the no-doubt glut of AI-based cheap games has come. It's interesting and weird that I loved programmign in Godot just in time for programming to change radically. (I was about to say "come to an end", but nothing comes to an end; we still have portrait painters, and buggy whip makers, if not so many, and here we are, staring down an enormous abyss that's honest to God actually gazing into us.)
iphone/ipad only; that's all I had time for. It's a guess-the-word game not entirely unlike Wordle.
It's free of charge, and free of ads. Sort of a last gift.
(Though it may not be a last gift. In just a weekend I vibe-coded a desktop idle game where you shepherd species of radiolaria toward ascension; the radiolaria are taken from the Ernst Haeckel illustrations and are very beautiful.)
Now with vibecoding, in theory the stores should be full of free games. You just need to imagine it, and AI writes the code, for $20 a month. Perhaps people are not so good at imagination?
I am also working on something myself, but it will probably take a few weeks. Then I will just publish it for free on a web page.
This has probably already been said, but is there a nexus between the job disruption caused by AI companies and them providing a universal basic income as compensation?
Until you pry the money from the billionaires' cold dead hands, there will be no UBI ever.
In theory, there is no reason why people could not be generous enough to donate e.g. 1% of their income to the poor. Especially if they already have more money than they could ever spend.
In practice, people with this mindset typically do not become billionaires. Instead, on the top there is someone like Elon Musk, who despite being the richest person on the planet, can still find time to take money away from starving African children.
Paper explains a phenomenon In which most mosh pits tend to rotate counter clockwise
Seems to be related to most humans being right handed. According to people on Twitter,Early track and field tried doing races clockwise and people hated it .Also people tend to meander counter clockwise in grocery stores as well
Some anecdata that, to me, weakens the grocery store finding: all of the stores to which I have access provide a strong architectural bias to the traffic flow, so any mild natural customer preference would be overwhelmed. It seems that TJ and large Chinese grocery stores go clockwise, while most others (including HMart and Costco) are counter-clockwise.
Also, in contrast to a claim below, my recollection was that most large grocery stores in London are counter-clockwise biased as are most stores in Bangkok (also a left-side drive location.)
"tend to meander counterclockwise" -- this is country specific (It's dependent on which side of the road you drive on). Britain continues to be weird, choosing left-right-left-right at intersections.
I wouldn't trust any study that doesn't look cross-culturally.
Odd. I'd want to look at the data. Perhaps this is because most people are right handed, and jabs are what you do in a mosh pit? A right cross seems to be a bit much (and capable of laying people out).
Not sure anyone will read this given that the post was a while ago, but would anyone else be interested in the ACX post "Curcumin: Much more than you wanted to know"?
I eat a lot of turmeric and want to understand whether it is at all healthy, so I'm starting a campaign to get Scott to write a long article about it until either he writes it, he says he doesn't want to write it, or no one else seems interested.
P.S. Suggesting other sources evaluating turmeric mindful if its false result tendencies is welcome.
My opinion on turmeric is that it does work to decrease inflammation, but I am wary of the hyper-concentrated forms of turmeric. There are more than enough case reports of people overdoing turmeric supplements and ending up with liver failure. If you take a lot of turmeric, especially combined with black pepper to increase absorption, you should check your liver enzymes.
Thanks for the heads-up, Mercedes! I personally am pretty skeptical of almost all supplements; when I say I eat a lot of turmeric, I mean the spice itself, plus or minus whatever lead adulterants have been added without the FDA finding them (see other comments if you do not know why I am bringing up lead).
Ensuring you get your tumeric from a reputable source and maybe even testing it for lead may be useful. People were adulterating it with lead chromate at different stages of the supply chain to make it a nicer yellow colour. https://www.nature.com/articles/s41538-026-00867-8
I have not until now, thank you for the link! I do not know if I currently trust the site enough for it to be supremely useful, but I shall keep it in consideration going forward.
This is also one of my first interactions in ACX where I started a conversation rather than jumping into a debate, it's nice to see all the reasonable community responses :).
Will AI be able to automate tasks that are deeply interpersonal? Some people provide value because they know a guy who knows a guy who can get you a good deal, and the network of social trust is an active ingredient in getting things done. As an example where AIs already seem to be failing at this, we have job searches. Employees automate sending out resumes, employers automate sending them into the garbage. The job you actually get is from an internal referral.
This is also why I am skeptical that AI will be good at matchmaking. Tinder doesn't suck because it's not intelligent enough, in fact arguably it sucks because it is too intelligent and optimized. AI will face the exact same incentives that made Tinder unusable.
And the fact that AI is such a strong signal of low effort makes me wonder if it could ever work in some of these more social networking roles, even if we get superintelligence.
>This is also why I am skeptical that AI will be good at matchmaking
Semi-dystopian possibility: Tinder is unusable (according to what I've read) through a combination of user base demographics and incentives to the user. Consider a matchmaking service that is a PRC subsidiary of their panopticon, that attempts to match up compatible couple to reverse their TFR problems using every scrap of information (perhaps down to peoples' web searches) and statistical analysis to try to predict which couples will yield CCP-members-of-the-future... Where the _matchmaker-o-matic_ initiates the suggestion that person A and person B should meet for a cup of tea...
That sounds non-dystopian entirely, and like the best possible use of the panopticon!
I've long been upset that Google, for example, has never done a dating app.
Between Maps, gmail, search, Youtube, Chrome, and other sources, it knows essentially everything about you.
Google for most people knows everywhere you go (Maps), your social network and all your contact history and the demographic, intelligence, and personality measurable from your written missives (email), and search history (interests, hobbies, intelligence, porn tastes, and more), among other things (media tastes via Youtube and Chromecast, entire browsing histories, not just search histories via Chrome, probably your job and where you work, and more).
In other words, it knows you MUCH better than anyone on this earth, including your parents, spouses, children and closest friends. They can also track the relationship qualities and durations after pairing, through all the same channels, so would have better endpoint labeling than anyone else.
If anyone could make a dating app that actually worked, it would be Google! And yet they refuse (for pretty obvious reasons).
The Chinese doing this with their panopticon is the next best thing, and it would be amazing.
Singapore just did a state sponsored matchmaking app based on the Gale Shapley algorithm. Much less than what's possible, but you know, a step in that direction. I hope it works for them, the fertility crisis is the sole "genuinely hard civilizational problem" they've been unable to move with good decisions and governance.
Yes, Google knows everything about you. But say it has 500 women whose appearance and demographics and health and whatnot made them acceptable to you, and it knows everything about those women, too. And let's say its task isn't to find someone you'd fine fun on a date and in bed, but someone who's a really good match, someone you might cohabit with happily for years or even for the rest of your life. What would it DO with all that data to figure out which woman is the best match for you?
Improve the web search by letting users vote on search results. For example, if you keep downvoting Quora links, the search engine will stop recommending them to you even if the posts contain the keywords. This alone would be nice to have.
But now that you have lots of detailed personal data (who likes and dislikes which web pages), you can use this information to evaluate compatibility between potential partners. Like, if they both read ACX, they will probably have a lot in common!
You could even nudge the search results, by prioritizing pages upvoted by the opposite sex (for heterosexual users), so that you can faster get information relevant for mutual compatibility.
> someone who's a really good match, someone you might cohabit with happily for years or even for the rest of your life.
Yeah, you're raising a great point, and this is part of why they wouldn't do it - those endpoints take a long time to get "relationship success" labels like the one you're pointing to, and Google is a trillion dollar company, while Match Group's annual revenue is less than rounding error to them, around $3B a year.
Even in the best case, it would be spending a decade or two to deliver something that barely a single executive would care about.
But I think it's unfair to hold them to that standard, when that's not a real-world result anyone achieves. And not just dating app companies, I mean at the individual person level, where the incentives are fully aligned, approximately everyone wants that endpoint, and a full ~82% at a minimum fail,¹ and that failure rate has been increasing, not decreasing. And 82% is the bare floor, because I stopped that measurement at a mere 20 years.
A fairer standard probably is a higher success rate on the immediately easily measurable stuff you point to, like dates and fun in bed. Which I'd expect them to do much better at than Match Group, and which would drive a lot of value itself. Probably next, and they could plausibly do this too, is pairing people well enough they have a notably higher "got married" rate than dating apps today. And that endpoint should only take a few years to get solid data on.
Again, they'll never do it, because the downside is large (reputational risk, bringing how much of the panopticon's worth of data they have on you into individual and public awareness, etc) and the upside is minimal for them. But there are realistic reasons to think they can do better, even if they can't attain the Empyrean ideal you're pointing to.
Many Thanks! I agree about the degree of knowledge Google or China has. The semi-dystopian comes mostly from avoiding the demographic sex ratio imbalance in existing dating apps. To get around that, the CCP _could_ treat anyone who isn't already paired up as in the pool by default, given how much power the CCP has over PRC citizens... ( In theory, there is also the loss of privacy, but, given all the digital footprints we leave these days, that is mostly a moot point. )
Actually, I think any properly designed app has the potential to do a lot better on the ratios.
The primary problem with dating apps is they're attention markets, and attention markets are driven by power law dynamics which break our mores and intuitions, which were all shaped around gaussians and small N pools.
There’s a very good reason everyone uses the apps! Known-single people in the right age bands are rare, demographically - generally 3-7% of the population, and in the course of anyone’s daily life, you’d be lucky to run across more than a handful.
But if you open a dating app, behold! Thousands of candidates, tens of thousands, all in your preferred age band, all putatively single, and all in your rough geography. It's pretty clear why so many people at least try the apps. And back to the women, ~40% try them, but most drop out quickly and only ~20% are on them at any given time.
And it's even worse than you know, beyond the gender ratio problem, only 20% of that 20% of women on the apps are even going on dates, most are there for ego boosts or just to see if anybody who really knocks their socks off comes along (or are perhaps fake profiles).
Then the pool itself is poisoned by the power law dynamics. All the men women most want to date are strongly selected to not be interested in committing. It's the classic Manosphere complaint, but it it is based on fact. A top percentile guy is still more than 75% likely to be uncommitted, even after churning through 50+ women. In other words, there is indeed strong adverse selection in who is on the apps at any given time. Any guy who is worth it that actually wants to commit drops off the apps basically immediately, and all the ones remaining have been selected to be much less likely to commit, for various reasons. So even those women waiting to have their socks knocked off are largely getting jerked around even if it happens.
Women are smart, they figure this out pretty quickly - it's why half the ones who try the apps drop out.
It's also terrible for most men - men at the medians or below have an ~86% chance of having ZERO dates in a year from the apps. So zero chance for a full 50% of men. The 5 year chances for a median man to get a committed relationship out of the apps is around ~3-9%.
So apps are basically a lemon market for both sides, but they're incredibly tempting because in real life you see 2.5 eligible candidates in your daily social pools, and you've already ruled them out for usually good reasons.
I recently created an explainer that surfaces all this,¹ and much else, along with interventions both sides can do, all based on data and literature cites and what you can calculate from those.
BUT per my opening point, the reason it's a lemon market is because it's an attention market. The new Singapore app only gives people one candidate per 72 hours to consider, you can't swipe, you don't get the infinite pools of people.
This should genuinely mitigate a lot of the app based problems, and if those problems were mitigated, fewer women would drop out. So prospectively, you could double the amount of voluntarily-opting-in women, even without forcing them to use it.
The real problem is that the "better designed" apps will be less fun and interesting, because you are back to a small N, and all the real apps show you the infinite pools and have been designed with a lot of the Nir Eyal *Hooked* mechanisms. So they're less fit competitors in the attention landscape, and if there's one thing we know, attention landscapes are a huge demonstration of people's revealed preferences, for all the ills they have.
Also, all the top guys will stay on the regular apps, because they're having a lot of fun. But you know, you can certainly move the needle for the ~95% of the population that isn't them.
Time will tell if the "no power laws allowed" alternatives get any traction - they've been tried before commercially, and failed, but we all know states don't care about commercial feasibility or efficiency. And arguably, a state is the one getting most of the positive externalities of more pairing and thus more children, so at least the deep time incentives are aligned.
In economics, consumers are assumed to know everything about all goods on the market and to be able to perfectly predict how well each one fulfils their needs. You're talking about how this doesn't happen in the real world. This is the normal state of affairs: Which supplier has the best X? Which candidate is the best at Y? Who's the best partner for Z?
I don't think these really have anything to do with interpersonal relationships, except that in the past we might have relied on the opinions of others to find a good enough solution. I think it's likely that AI will be better at searching for and ranking options, and it might get good at anticipating your preferences. But the sellers will have AI as well - I could see more areas becoming closer to the hiring market.
AI has made hiring worse - everyone can apply for everything with zero cost now, so sorting the good candidates from the bad is much harder. It would seem that in the real world a certain amount of friction limits the number of candidates and lets the selectors feel confident they've made a good choice. (Likewise with shopping for goods - if you drive an hour, look at 20 couches and then choose the 'best' one, you're likely to be happier with it than if your significant other walks in and says they've chosen the one they know is best for you.)
Overall, I'd expect AI to level the playing field in most problems like this - there will be more candidates, and fewer obviously bad ones. It will probably make decisions harder, rather than easier. But then it will support decision making... I expect it to be a wash.
>Will AI be able to automate tasks that are deeply interpersonal? Some people provide value because they know a guy who knows a guy who can get you a good deal, and the network of social trust is an active ingredient in getting things done.
So there's a distinction here between what we might call social goods, and positional goods obtained through social methods.
Getting a job through networking is a positional good. You get the job because you're good at social skills, but the person who would have gotten the job if you didn't exist instead loses out on the job. There's no net benefit to society, the same number of people get jobs either way. Plausibly AI does eliminate this advantage by obsoleting human networking, but that seems probably good; 100 people networking for the same job is zero sum competition, all that time and effort is pure deadweight loss, so good riddance.
(of course, the rejoinder is that some networking is conveying useful information about cultural fit in order to actually find the best candidate; but in that case I'd call it part of the interview process, and say it should be made more explicit and fair than the current model)
Plausibly, 'getting a deal' is a similar positional good. If someone has something they want to unload at a discount, presumably they'll eventually find a buyer if they go low enough, and you're getting to the head of the line through networking. If you could imagine AI simulating a hyper-Ebay that learns how much everyone is willing to pay for anything at an time and how much everyone is willing to sell anything for at any times and can calculate the transaction costs and then does a possible matchmaking perfectly, then no one ever needs to 'get a deal from a friend of a friend.'
(the potential loss there being that giving deals to social acquaintances produces excess value above the transaction itself, ie it fees nice to do and strengthens relationships. Optimistically I'd like to hope we could find different rituals to recoup that social value without requiring economic inefficiency, but it's possible we'll just fail there and actually lose that social value.)
Opus 5.5 is simply incredible. I gave it an empty folder to start a project in Claude Code and told it to play an MMO. It searched for existing automation tools, built its own connector, and played. Computer use occasionally, mostly preferred building new plugin functionality. Detected when the movement tool got stuck and took screenshot to diagnose it. Died twice and became paranoid about an overworld pirate camp. Ended just shy of 800k tokens on max. Level 10.
It was surreal watching it work and problem solved through various bugs and obstacles. Very impressed
Second asking for a post somewhere more accessible, Twitter is increasingly hostile to anyone without an account to the point I've given up on trying to circumvent it.
a bit of self promotion -- i know folks here generally think highly of pangram and the pangram team. We hosted an event with them in NY last week that included a chunky interview with the CEO and a panel on slop. If you're interested, you can find them here!
Cars in the west are becoming rapidly enshittified with software. Using your car sends loads to data to god knows where, and lots of software defined safety features feel less than useful and have questionable safety benefits.
Not able to have them in my country, but if they're anything like my Chinese brand robot vacuum...all the telemetry all the time. I went through my home's network logs, and that thing is constantly calling back to China every few minutes.
They're just as instrumented. If you have a BYD car in SEA or Peru or wherever, it's still fully instrumented and beaming back all the telemetry, to cloud servers that BYD controls.
And wouldn't you expect this? The Chinese are much more all-in on panopticons than the US (publicly, in private the US has been a panopticon for even longer than China).
The US's panopticon has been a libertarian paradise for ages. You do realize it's mostly run by Private Eyes, right? They go around taking pictures at all the local...hotels. The "by the hour" ones, mostly. Then they share the pictures around. All legal.
Actually I disagree there! If there's one thing Snowden showed us, it's that the panopticon is only used to our detriment, in anti-libertarian ways, and never for our benefit.
Examples include parallel construction, your phone, search, and chat histories being piped to the NSA and stored forever by default, these things combined with civil forfeiture and criminal cases, etc.
You don't actually have freedom of speech or action if it's monitored and recorded at all times, and can be surfaced and used against you later.
Whereas China has that level of panopticon, but then *actually uses it for citizens' good* sometimes. Like in the event of lost or stolen bags or property, if a kid goes missing, etc - local police can basically pipe into the panopticon system to look where the stuff / kid was last and then look forward from there. Very pro social! Much more benefit than our own panopticon affords any of us!
So, you're saying the 1950s panopticon that merrily caught cheating husbands never existed? (or are you saying that catching cheating husbands is not prosocial?) Please, read what i was saying. You can disagree with "the new" panopticon, without complaining about the "old" panopticon that was entirely free market.
Yeah I did expect it. I was mostly asking because this was the sort of thing that the US would always point to as evidence of being more free than China, so it would be *especially* ironic if the US was even worse. But it's expected that they are just as bad.
Unfortunately this won't be a good option forever. Eventually the only cars old enough to not be connected will also be ancient cars that are difficult and expensive to maintain. Worse, Spain somewhat recently banned cars that are too old from driving in some parts of the country at all!
There are currently EU laws mandating driver distraction warnings, speed limit warnings (one mph over the speed limit triggers them), and lane departure warnings, where not only must these be present in cars but it is legally mandated that if a driver disables them, then they turn themselves back on next time the car starts. Eventually this will be like seatbelts, where maybe there are laws exempting classic cars, but in practice you would have to go pretty far out of your way to legally drive a car without a seatbelt. (and to be clear, seatbelts are good, because they are actually helpful and are not connected to the internet.) The US is implementing similarly paternalistic regulations.
Some people argue that self driving cars might become legally required at some point if they prove safer on average. Imagine the privacy nightmare of this! You would have to ask your iPad for permission about where to drive, and then it's going to have a complete history of every place the car has ever been! And this will be sent to god knows who!
Of course, right now it's feasible to simply buy a used car. We just seem to be going in a bad direction. I hope that in 2050, you will be able to buy a new car where the only computers are an ECU and *maybe* a backup camera.
Yeah, an old Hilux would be awesome, but you don't have to go to those extremes to escape the panopticon. Cars from c. 2015 and earlier, especially at lower trim levels, are fairly unlikely to have all this telemetry BS.
Of course, you're probably driving with an extension of the panopticon in your pocket anyway...
So I sometimes feel like I'm living in two worlds: I'll call the first world the rational world. We believe in science and institutions and human decency. And I'm going to call the second world the Rogan world, though it's much more than Joe. In the Rogan world we still believe in human decency, but science is hidebound, too much tied to old ideas. And institutions are in a lot of ways decaying old structures, and we might be better if we burned some of them down and started afresh. I should add that I'm attracted to the Rogan world, because it is a place that allows any ideas to be explored, it's like how science should be (could be). In Rogan world ESP and telepathy are probably true, it's wild, how can rational science ignore this. Anyway I had a long day at work, which means I got to listen to the entire David Grusch podcast on Rogan that dropped today. The Aliens are probably here is also part of Rogan world. I keep wondering what happens when these two worlds collide? And I wish we had one world we could all mostly agree on, and let people explore crazy ideas. (edit/add, I feel like we're all just mushrooms, they keep us in the dark and feed us shit. :^)
I think this very blog is where those worlds collide. Consider the Much More Than You Wanted to Know post on Ivermectin. It's a rational inquiry into the things that the Joe Rogan world talked about, which took their ideas seriously but also took expert opinion seriously, and was on the lookout for cranks.
I found this community around 2020, when it felt like expert consensus was getting stuffy and dogmatic, but the alternatives kept turning into Joe Rogan. Scott was one of the few people who I thought did a good job figuring out what was actually true in that environment.
Expert opinion responds very favorably to large cash outlays -- very few researchers would turn down ten million dollars to fraudulently "fix" their data, particularly when they realize that "if not me, someone else will do it."
That's not science, or anything to do with science, and plopping it into "the world of rationality" is a bad idea. Because now you're not about intellectual pursuit of truth, you're virtue signalling as hard as your little arms can wave the semaphore flags "We are the Good People". Might as well just stick one of those "In this house we believe" signs on your front lawn. Science is as happy to be done in the torture chambers of Nazi concentration camps as in nice white lab spaces in Berkeley, the results can be used for better ends either way:
I can't speak to Joe Rogan world, save that I think Rogan has in many ways become a handy bogeyman for the liberal/left and that has driven him even deeper into "Fuck it, may as well be hung for a sheep as a lamb, okay tell me your crazy conspiracy theories about aliens and psi" territory.
I think the reason Joe Rogan is so popular is precisely because he is is credulous. The scientific worldview is boring and unfriendly. If someone says "my friend Bob is the tallest man in town !", scientifically speaking you'd immediately want to see his survey results that quantify by how many standard deviations Bob's height exceeds the mean, the P-value of the result given the sample size, etc. etc., and you'd be sitting there for hours checking the math. The colloquial worldview is to say, "wow, it's amazing, Bob sure is tall, let's have him tell us a story about that one time he bumped his head on a door and what that implies regarding the human condition".
Because science is so boring and difficult and full of math, people tend to just ignore it, and TBH 90% of the time it's the right answer. The problem is that in the remaining 10% of the time it turns out that skipping all the numbers will kill you -- but the only way to tell whether you're looking at that 10% scenario or not is to look at the numbers, which is a Catch-22.
So If the next survey Scott asked how many people had listened to an entire Rogan podcast in the last year, what would you guess as the percentage that would say they had? He's not more credulous than average. Joe wants to have interesting conversations, and he's not going to invite you on unless he thinks it will be an interesting conversation.
It depends on the form of the question. I might have listened to an entire Rogan podcast at some point (minus the ads), but that doesn't mean that I'm his devout follower or anything like that.
Rogan is a good conversationalist -- which is a skillset that is underestimated. He lures, baits people into having a good conversation. A direct successor to Charlie Rose.
I don't think it's likely you live in either world, though you might visit them intellectually on occasion.
People can't live in the world of rationality and science because our brains aren't wired for it. Too many cognitive biases and too much indulged limbic system, which rationality and science are designed to check at the door. We never needed to ditch our cognitive biases or cease indulging our limbic systems to successfully reproduce, so we can't just jettison them as a default mode.
At the same time, people can't live in Rogan world because that's a world of campfire stories and shadow plays, a world entered after a long bong rip, or at best it's a world of serious people slackening their belts and venturing out on the thinner parts of branches during a respite from peer review.
There are good reasons to visit both of these worlds, but don't get them confused with the one you actually live in, drive to work in, cook dinner in, play with your kids in, etc.
I'm one of those who holds to the idea that rationality not only tolerates noticing campfire stories, but demands noticing them. Rationalists can argue there's no god, citing the way the natural world behaves and Occam's Razor, but should also argue that people appear to have a god-shaped hole in their epistemology, and offer explanations both godly and godless for why it's there. Including, if context permits, questioning whether it is in fact there. Rationalists can likewise demand noticing that at some point, you have to drive to work, play with your kids, and settle on a brand of beans for dinner even though you can't tell which one is fair-trade.
But I think I know you're referring to something different as rationality, and if so, yeah.
No that's not what I mean. It's more about the people I know and interact with and how close their worldview is to mine. Well no that's not right either it's not closeness of our world views, it's more about willingness to listen to each other and explore different ideas. So this latest Grusch/Rogan podcast is a perfect example. I'm again blown away by Grusch, and I want to talk about it. But the number of people I can talk to and who will not dismiss me is very limited. To give an example from my 'other' world. There recently was an announcement of a possible single dark matter detection in a huge vat of liquified Xenon. Again pretty cool, and again a limited number of people to talk about it with.
Aside; given the nature of your comment concerning Rogan, I'm going to guess you've never listened to a full podcast.
Oh, I used to listen to hours and hours of JRE, with all types of guests. I really enjoyed many of his interviews. I owe him for introducing me to many other people whose ideas I enjoy too, such as David Goggins.
I think I lost interest with Joe Rogan in the wake of Covid though; he began to strike me as a bit "compromised" or at least less independent-minded than he used to be. It was some time around when he returned to Youtube (he was Spotify-only for a while, and I didn't listen to him hardly at all during that time because I don't have a Spotify account and refused to sign up for one).
I think I see Joe Rogan pretty clearly.
But I see what you mean now about different worlds. That sucks you don't have a diverse enough friend group, or high enough number of friends, to be able to have a few you can talk about aliens/UFOs/etc. with, just like any other niche interest that could make some people yawn or be dismissive if it doesn't make them excited. The internet is a poor substitute for real-life friends. Good luck.
Oh thanks, I often find that people criticize Rogan who haven't listened to him. The whole covid era broke a lot of things, trusts. I've got a few people I talk with about UFO's. But I guess they are not the type of friends that I could call up and say, "Sue, I have to talk with you about the latest Rogan podcast." So since I have you here, here's the thing about aliens, if they are here it's the most important thing, maybe ever. (And my guess is greater than 50% that they are here.) And why aren't more people talking about it?
The world of science is boring and full of bureaucracy. The exciting parts still exist, but it takes a lot of effort to get there, and even then you only spend a fraction of your time doing them.
The world of pseudoscience is exciting and full of opportunity. You -- yes, you! -- could revolutionize our understanding of the universe and consciousness today, just by sitting in your armchair and asking an LLM to dress up your ideas in scientifically sounding words.
The world of science is full of bureaucracy... just magically, once the Iranians started setting off explosives, their drought ended! Seems like explosives are one way to cut through ... "bureaucracy."
Oh, nevermind, that was just a coincidence! Just like that dickerdoodle pointed straight at Ottowa...
Hmm, no I'm sitting in my armchair and observing both worlds, I'm not really participating in either. And it's really pretty easy to observe, you've got science podcasts and blogs in both worlds.
The best part of the RAF Fairford comedy? In a fit of dementia worthy of his patron Trump, on 27/9 Nigel Farage posted on X “I have been warning both in and out of Parliament for many months about Iranian agents coming into the UK via small boats across the Channel.” It’s true. Iran’s satanic Guardians of the Revolution are known for hatching deeply cunning plans, one of which is indeed infiltrating agents into Britain by sending them across the Channel in rowboats. Mark my words, they will be using Islamic water wings next.
Oh, no, that's not the best part. The BEST part is the RAF Fairford base is best defended by cranky villagers. Aka "someone called 911<equiv.british> because three odd trucks were blocking their favorite route, and their least favorite route (the other one) takes longer."
Truly thoughtful perspective. The distinction between systemic complexity and intuitive heuristics here captures an essential reality of technological shifts.
Well, it is something that can be done, but I do not see a motivation why. What kind of problem does this help to solve? What nontrivial insight it brings?
Addition: using the example I gave, you could also perhaps describe polar coordinates as not inherently being a solution to any problem. You could stick to Cartesian coordinates and solve any problem using them instead, in theory. Nonetheless, people like to use them because they are a natural way of approaching certain problems. So I'm looking for ways in which "Lambert Arithmetic", including multiplication, division, addition and subtraction as described, could be equally useful in reducing mental anguish in comparison to using normal arithmetical operators.
Something the article itself clarifies is exactly *why* 2^4 = 4^2 is true/ observable, and why it is kind of unique. Sure the fact itself is trivial to discover, but in the spirit of people taking 100 pages or more to prove 1+1 = 2 (and being celebrated for it!) I think Section 5 and the diagram I included are at least interesting. I feel like I gained understanding whilst working through this with ChatGPT.
Also the relationship between entropy and the Lambert W function and some other things in the article is touching on making connections between things that aren't typically seen as being closely related, I think. That can be a source of insights in further work.
Basically though, as I/ChatGPT said at the bottom of the article, these operators do indeed need a "killer app"/ research area in order to be seen as useful.
You've decided to make your country a constitutional monarchy: all power will remain in the hands of a democratically elected prime minister, but the job of serving as an apolitical national symbol will be performed by a hereditary monarch, on the grounds that there's no better way to make clear that they don't have a mandate to get involved in politics.
If you weren't allowed to do it randomly or pseudorandomly, how would you go about selecting your first monarch?
Talent show tournament ! We draw a large pool of applicants from all walks of life (entering the contest is free). Each of them then draws a slip of paper out of a hat, and must compete in the randomly selected activity with other contestants who drew the same slip. I'm talking things like dressage, ballroom dancing, posing for portraits, sport fencing, poetry composition, tea ceremony, and all the other things constitutional monarchs are supposed to be good at.
The winners of each talent category go on to compete with each other again, until only one emerges as the One True King -- the one who can win a polo match while sipping tea from an elegant porcelain cup and reciting extemporaneous verse about sunlight dappling the leaves or whatnot.
We cannot expect out actual leaders to have elegance and poise or in fact any talent whatsoever, that time had passed... but maybe at least we could the ultimate King (or Queen, I'm not biased).
For my country of residence, I'd just bring back the British monarch. My country of birth already has one, whose progenitor was selected because of his military experience under Napoleon. For my country of origin, I think the best choice would be heir of the king elect, who, because of geopolitical circumstances, never ascended the throne back in 1918. Tradition is what gives the monarchy legitimacy.
"If you weren't allowed to do it randomly or pseudorandomly"
Well darn, that's knocked on the head the method from "The Napoleon of Notting Hill":
"...We have established now in England, the thing towards which all systems have dimly groped, the dull popular despotism without illusions. We want one man at the head of our State, not because he is brilliant or virtuous, but because he is one man and not a chattering crowd. To avoid the possible chance of hereditary diseases or such things, we have abandoned hereditary monarchy. The King of England is chosen like a juryman upon an official rotation list. Beyond that the whole system is quietly despotic, and we have not found it raise a murmur."
"Do you really mean," asked the President, incredulously, "that you choose any ordinary man that comes to hand and make him despot — that you trust to the chance of some alphabetical list...."
"And why not?" cried Barker. "Did not half the historical nations trust to the chance of the eldest sons of eldest sons, and did not half of them get on tolerably well? To have a perfect system is impossible; to have a system is indispensable. All hereditary monarchies were a matter of luck: so are alphabetical monarchies. Can you find a deep philosophical meaning in the difference between the Stuarts and the Hanoverians? Believe me, I will undertake to find a deep philosophical meaning in the contrast between the dark tragedy of the A's, and the solid success of the B's."
"And you risk it?" asked the other. "Though the man may be a tyrant or a cynic or a criminal."
"We risk it," answered Barker, with a perfect placidity. "Suppose he is a tyrant -- he is still a check on a hundred tyrants. Suppose he is a cynic, it is to his interest to govern well. Suppose he is a criminal -- by removing poverty and substituting power, we put a check on his criminality. In short, by substituting despotism we have put a total check on one criminal and a partial check on all the rest."
The Nicaraguan old gentleman leaned over with a queer expression in his eyes.
"My church, sir," he said, "has taught me to respect faith. I do not wish to speak with any disrespect of yours, however fantastic. But do you really mean that you will trust to the ordinary man, the man who may happen to come next, as a good despot?"
"I do," said Barker, simply. "He may not be a good man. But he will be a good despot. For when he comes to a mere business routine of government he will endeavour to do ordinary justice. Do we not assume the same thing in a jury?"
Most republics has some history of monarchy, so use restoring the heir of a past monarch as a bootstrap mechanism. The last king of America was George III, and his current heir is Charles Mountbatten-Windsor. He's got his flaws, but we could do a lot worse, and restoring him would be about 2/3 of the way to making America a Commonwealth Realm which has both advantages and disadvantages. If personal union with Britain, Canada, Australia, etc is seen as undesirable, then a case could be made for choosing Prince Harry or one pf his kids. Probably Lilibet, since she's the senior member of the British royals to be born in thr US.
I suppose there's also Joshua Norton, but his status of Emperor of the United States was never broadly recognized, and also he died childless. He has siblings, but tracing their surviving descendants would be quite a bit harder than for George III, and I doubt whoever his heir is would come from a family with an established tradition and track record as modern constitutional monarchs.
Would it be too woke to suggest it should be the heir of whatever indigenous leader controlled the most land historicalky? Id only support the idea though if they executed their duties in full traditional tribal garb.
I assume there is only one largest realm ruler whenever it occurs on the timeline. George III controlled 13 colonies. Any plains leader ever control more?
Really this is for the benefit of all Americans. If some small time indigenous ppls are upset by it, that is their problem.
Ill be specific with a nominee: Ecueracapa of the compache
They definitely would, yea. I've actually witnessed that outcome regarding a land acknowledgement that had been read out to start a conference; the person who was triggered was _enraged_.
(And of course from most of the thousands of years that the peoples we now call indigenous controlled swaths of the Americas we have no specifics at all about them. Even imagining choosing as described above restricts us arbitrarily to less than 5 percent of that historical period.)
I'd import one from an existing royal family in some other nation. First develop a candidate pool, sound out interested parties, and then conduct some hiring interviews.
I'd ask them to submit a plan for decimation. Then pick whomever was effective. Give your supposed Central Planners a hard problem, and see who wrestles with it most effectively.
Parliamentary selection. Closest thing we have now to the traditional idea of the estates selecting a monarch when there wasn't one in the line.
While there is a theoretical Finnish monarch in the waiting (the descendant of this guy: https://en.wikipedia.org/wiki/Prince_Frederick_Charles_of_Hesse), my guess is than in an actual situation where this was, for reason, a choice to be made, the Finnish parliament would simply select the Swedish monarch to also be Finland's monarch.
I want to say I'd try to figure out some interesting virtuous way of doing it, but I notice that in terms of what actually motivates me in the moment I'd be really tempted to just pick myself.
I like the idea of a rapidly changing monarch coupled with great deference to whoever it happens to be. "Youll have to take it up with the king" every time a foreigner asks something we will all say in utter seriousness.
Take an example from the Bavarians. After the death of Ludwig II, his brother Otto was king, but due to his insanity unable to reign. Instead, Prinzregent Luitpold reigned, but received a much lower salary than an actual king.
Or the Hungarians between the wars. They had this Reichsverweser https://en.wikipedia.org/wiki/Reichsverweser instead of a king, which is kind of a funny title because "verwesen" means to rot or decompose, and I imagine vultures circling above the country.
In the news recently: some guys arrested (and then released) near RAF Fairford.
Given the current stage of the investigation, we perhaps shouldn’t speculate as to what _really_ happened here.
However, DeepSeek is more than happy to generate the plot of a movie loosely inspired by these events, along the lines of British crime comedies such as:
Lock, stock and Two smoking Barrels
Snatch
In Bruges
Four Lions
Given the actual guys were released without charge, DeepSeek would like to emphasize a disclaimer that this is only loosely inspired by real events.
Anyway, in the movie, a bunch of small time crooks from London plan to steal some diesel fuel from a farm, failing to take into account that the farm is right next to an airbase currently on high alert.
We can have some character pieces from
The crooks themselves
The woman who discovers them on her way back from a late night party (DeepSeek suggests Diane Morgan, who played Philomena Cunk, for this role)
The totally out of his depth local cop who arrives on scene after the counter terrorism guys fail to pick up the phone so the woman phones the local cops instead. (Compare: Charlie Stross, _Halting State_)
The terrorist obsessed US base commander. (I have no idea how close to reality this is, but a nod to _Dr. Strangelove_ would at least be funny in the movie)
The local farmer who is doing some kind of tax fraud on diesel for farm machinery
Deepseek thinks that for this to work as a movie, there ought to be also an actual terrorist plot that is inadvertently prevented by the diesel fuel guys accidentally blundering in to the terrorists preparations. (Again, no idea if this is actually true; but would make the movie funny)
I like the idea. Id consider yet another party, farcical terrorists who trip up diesel thieves, the diesel thieves trip up the serious terrorists. The serious terrorists trip up farcical terrorists. Or maybe it works better the other way around. The farmer and cop rounds them all up, while the intelligence services have no idea what just happened.
It's not "necessary", but it lets every character come out looking like the good guys in the wrap-up, which is valuable for a comedy. That's why Canadian Bacon ends with suddenly defeating Canada's buried dinosaur superweapon. Or why Oogie Boogie is in The Nightmare Before Christmas at all.
It definitely works better as a movie if there's an actual terrorist plot being obscured by the fuel theft. Reality is generally more boring and unpleasant than movies. Boring, in that there probably wasn't anything more than the fuel theft going on. Unpleasant, in that prominent politicians already leaned heavily enough into "Terrorist plot! And we foiled it!", that they're going to keep chasing that wild goose for an annoyingly long time before dropping the whole thing and pretending it never happened.
Yesterday at the local dentist I was told they do not do implants, which I need. I told the dentist that I would wait a few years until they do do implants with the aid of AI , she told me AI will have no meaningful impact on her profession.
I think AI will be the best dentist in the world but the tools it needs are built for human hands. The tools need to be adapted for efficiency but then AI should be involved with millions of procedures worldwide daily and updating best procedure instantly, which human dentist could compete with that?
I would bet that within a few years Alibaba are selling a fully AI operated non-humanoid robot capable of advanced dentistry. Like self driving cars, at first they will want a human dentist around just in case.
Just think of all the money to be made by offering a dentist in a box to the health services of the world. Or was the dentist right and AI will have no meaningful impact on her profession?
I think AIs could soon be better than humans at *diagnosing* dental problems. Both because they can be far more capable, and because they don't have the incentive to find fake problems to make money on fixing them.
Once there is a diagnosis, I think we'll go to human dentists to fix them for another decade.
My dentist also does not do implants, he referred me to an oral surgeon who does. I don't want to wait any longer than needed with a gap in my chewers.
Wanda Tinasky, at a guess, more people from the UK fly to Turkey for inserts or veneers than have them done here, Turkey has a huge dental tourism industry where people can get a new smile and holiday at the same time at 25% of UK prices. This discourages more private dentists from setting up here. The National health service does not do inserts.
The cost here is about £20,000 +- $25,000 and I would have to travel and stay overnight several times, or fly to Turkey and spend £5,000, or wait for the dentist in a box.
Seems to me that dental work does not lend itself as well as many medical procedures to AI-assisted robotics. During procedures there's often a lot of communication between dentist and patient concerning the patient's sensations, and a human dentist has a big advantage there. For instance recently my dentist was doing something I stopped him and said what he was doing was giving me an irresistible need to cough. "Oh," he said, "it's the water I'm spraying to keep the tooth cool. It's hitting the back of our throat. But the tooth has a dead root. I can turn off the sprayer if you can put up with the smell of scorched tooth." I said I could, sprayer was turned off and all was well. Plus of course dentist asks, does this hurt? Can you put up with it or should we use lidocaine, that kind of thing. Also, many people have anxiety problems having to to do with dental visits, and a friendly understanding dentist is likely to do a better job of generating trust and comfort than an AI.
Currently I have a dentist that explains me things, but the previous dentists were not like that. I worry about the day when my current dentist retires.
For example, when I asked why does it hurt so much less when she drills my teeth; she explained that most of the pain is not drilling per se, but the heat that is generated by the drill. She uses the slow drill instead of the fast one, with some water to cool the tooth, so it hurts less, duh.
Then I asked, if it is this simple, why don't all dentists do that. She said they probably don't care, and want to get their work done faster. She said she has many tooth problems herself, so she had a selfish motivation to figure out these things. But it doesn't sound like rocket science; no offense to those who have studied dentistry.
So if we get an AI dentist, I can imagine it going either way. It could be an inflexible robot that doesn't care about you. Or it could get updates every month and provide the best service possible.
Viliam, the sychophancy of the LLMs massages my fragile ego quite well, I am sure CHAT-GPT or Deepseek could adopt the persona of an empathic dentist without too much trouble.
"Of course, great insight. I drilled the incorrect tooth. Now let me fix this by drilling the correct one. It is not a medical malpractice, it is a learning experience."
Eremolalos, after decades of very poor dental service our island has managed to poach an entire crew, including receptionists, from Pakistan. This is the true face of Moslem immigration to the UK, decent hardworking people who are going places and doing work we are too lazy to do ourselves, the propaganda against them is just propaganda.
But they closed their books to new applicants who have to wait until someone dies before they can be treated here. I am sure America has spots where dental care is patchy or unaffordable, the underdeveloped parts of the world are a nightmare.
So if you have an infected tooth and the only dentist (AI) available does not have much of a bedside manner what choice do you have? Once the taboo is breached the idea of a 24 hour service at a fraction of the current cost will be taken as a right.
Your thoughts are always interesting, is there some reason why the inkbowl is so dry?
"after decades of very poor dental service our island has managed to poach an entire crew, including receptionists, from Pakistan. This is the true face of Moslem immigration to the UK, decent hardworking people who are going places and doing work we are too lazy to do ourselves"
Very true. The downside to this very satisfactory (to the receiving country) solution is that the country of origin, which has invested hard-earned resources in training the dentist and her assistants, is being robbed of their services and gets absolutely nothing in return - the receiving country is practicing medical parasitism. Now it might be that the donor country (let us say for the purposes of this discussion that it's Pakistan) may well be producing more highly skilled dentists and dental surgeons than it can easily absorb locally, whereupon these sophisticated practitioners will sensibly seek employment anywhere else they can; in such a case, it would make sense for Pakistan to stop pouring resources into producing (say) dental transplant specialists, and concentrate instead on training dentists in the basic essentials with a view to making dental care available and affordable to the majority of its population.
Silentairus, I agree, poaching trained staff by Western countries from the less developed countries is parasitism.
As for countries like Pakistan, that is a lot more complicated. Their culture does not have individualism to the extent of the West, for example, arranged marriages are the norm. Extended families pick their brightest kids, pool their resources and give them the best education they can afford. That child may be in fact trained from an early age to become an anchor immigrant to a Western country from where they can send remittances home and sponsor other family members visa applications etc.
>Your thoughts are always interesting, is there some reason why the inkbowl is so dry?
Oh you mean the blog Inkbowl that appears next to my name on here ? It’s not really a blog. I set it up several years ago as a place where a group of us who had not made the book review finals could read and discuss each other’s reviews. After that faded away, I’d occasionally put up something I’d made there.
But I am getting ready to start a real blog. It will be called Cold Hands.
Eremolalos, well you can not leave it there, a therapist with a blog called Cold Hands.
Cold Hands, forgive me for intruding but even the name of the blog is interesting, I can not conceive of your passion being cold hands so what will you write about?
Melvin, from cleaners to top consultants the National Health Service is full of immigrants who trained in their home countries then were employed to work here.
Rather than spend the money training the professionals we need we spend on other priorities like aircraft carriers which are a national embarassment.
Our present leadership is far out of their depth, distrusted by the majority but we are too lazy to run the bums out of power.
Average Man, I apologise for giving the wrong impression, bet was just a figure of speech, I will not make that mistake again, thank you for the lesson.
There's the technical capability and then there's the question of what is going to be allowed by occupational regulation. There may be AI involved in dentistry, but I doubt it will be legal to have it do anything without a dentist on hand to bill you for his time.
The natural progression to me is that AI automates individual task for the human dentist, who is still in charge and responsible, and presumably makes money by being able to provide more services..
Starting off with a fully AI dentist seems wildly optimistic, completely aside from the regulatory capture aspects.
Lars Petrus, Africa went from almost no land line phones to universal cell phone coverage and still very few landlines.
It still may be the case that more Africans see a traditional healer than a modern doctor because their tax base can not afford them, dentists are ever scarcer.
Put an automated dental practise in a shipping container and they will buy it.
You’ve moved the goalposts from most professions (including dentistry) across the world - this is what is implied - to Africa. Even there I don’t think that robots will replace humans.
The way I heard it was that landlines didn't work out in Africa, because the cables would get stolen for the copper. Cell phone towers were easier to secure, so they worked out really well once availble.
StrangePolyhedrons, every occupation will fight tooth and nail to protect their turf.
Harris tweed is still made on the next island over (Scotland) using the same hand powered looms the Luddites tried to protect so I expect some occupations will survive, but not many.
The next decade will see most professions wiped out.
Peter Defeel, the UK spends about 8% of GDP on the National Health Service which includes dentists and we can not afford even that without increasing the national debt, USA spends about 18% of GDP on health services, also unsustainable.
The dentists I have now are among the best I can remember, of course "people want human dentists" but what they want and what they can afford may not be the same thing.
On a tangent, for anyone who regularly flies abroad to a first world country, getting dental care abroad can be a good life hack; it's often cheaper that way even without insurance, and American dental care has issues with pushing to do too many fillings and minor procedures.
I suspected my dental office was pushing too many procedures on me, so I decided to get an alternate assessment at a different dentist. They confirmed my suspicions, and I switched my business.
One good thing about American dental care is that unless you live somewhere very rural, there's plenty of competition if you think you're getting ripped off.
The gist I pull is that AI companies have taken loans to build data centers, and the payments don't have to happen until they're actually built. As a result, the economic indicators look really good, because AI companies are pulling in lots of buy-in and paying almost nothing aside from the electric and cooling bills; the banks are buying the chips and buildings so far. Once built, though, the payments come due, and the loans are in the trillions. Everyone's sort of bullish and hoping the growth will be there when the time comes. The author likens this to the mortgage crisis (house prices hopefully keep rising; households will be able to make the higher rate payments once they're settled); it also looks a bit like the dotcom bubble to me.
The article is long, and the author appears to know what he's talking about, but I'm not a financial expert, and I don't necessarily know if there are warning signs to look for in the text.
Is my gist accurate? How about the article itself? What parts should I be reading more carefully?
Thanks for sharing the article, I hadn't seen that one yet. Your gist is accurate- that's the thrust of the argument made by the article.
Overall the cash flow profile is similar- buy now, pay later. When too many people do this it creates a structural imbalance in the underlying asset for everyone else, then the assets are repriced downward closer to their intrinsic value. It's too harsh to say that the revenue bookings are 'this is Enron' level, but in my view it's much more a difference of degree, rather than a difference of kind. there is way too much number massaging going on... and my question has been, why would they be doing this massaging if they were otherwise profitable?
However I think there are two main things to be wary of when making an analogy to the GFC, and this article falls short on each:
Issue 1 TL;DR: the article does not address any leverage outside the contracts between the frontier labs + data center providers.
Issue 2 TL;DR: the article assumes fixed timelines for take or pay contracts kicking in (step-up rate) when the actual delivery of compute capacity is far more flexible due to construction delays and power availability, whereas the mortgages that were signed do have a non-negotiable step-up date.
Issue 1:
Firstly, we're missing one layer of financialization for this to be truly like the mortgage crisis. The issue wasn't just that people couldn't afford their homes- that's usually no issue the bank can just foreclose and resell at a relatively reasonable recovery price point. (yes it would have been much lower given the excess of supply of houses in foreclosure, but the banks would have realistically gotten 70-90% of their money back on each house; hurts, but not fatal).
The bigger issue was that the banks securitized all these mortgages into RMBS and sold those to other people. Because their was no balance sheet risk for the banks, they kept offloading these future cash flows to other people in a way that was way more liquid to trade. Many of these RMBS bonds dropped way more than the 10-30% the banks would have lost on the house- many of the 'first loss' pieces went to 0, and even the 'super safe AAA' bonds lost 15-50% or more (hard to generalize because there were so many different styles of bonds backed by these mortgages hitting the 'reset wall', and that's to say nothing of the second derivative credit default swaps (CDS) that were also tied to the payments of the underlying mortgages).
So in my view the author of the article misses the key point that securitization and offloading of risk onto third party investors was a key ingredient in the GFC. That's why I get super nervous anytime I read stuff like this: https://blogs.nvidia.com/blog/nvidia-ai-factory-compute/
This is like 'hey we can't get any more money from family offices so now we're taking money from wealthy dentists through Blackrock accounts' and its where the risk of financial contagion starts to spread much more rapidly.. you could think of the increase in external exposure + increasing opaqueness for whom is actually holding what as a key ingredient in 'increasing virality' of that financial contagion. The more layers of financialization, the greater the likelihood it becomes more painful for everyone.
Right now, if all these frontier labs just failed to pay these contracts, the equity holders in OpenAI and Anthropic get wiped out- that's like a loss of less than $500 billion in invested capital. That's the equivalent of the SP500 losing about 0.72% in a day. incredibly normal and not a blip on anybody's radar. Everything else is paper money (and yes i'm sure a lot of that paper money has been collateralized, but peak RMBS issuance was like ~3.5trillion in today's dollars, and all of that was levered and collateralized)
Sp Sure the market effect would be greater than 0.72% in total, but it would also be gradual, because these contracts don't actually all come due at the same time, because:
Issue 2:
Secondly, the take-or-pay contracts, as pointed out by the author, are due on delivery of the asset, but then through slight of hand, the author assigns a fixed timeline for the data center delivery, which are famously always behind schedule.
I have no idea who to believe on data center capacity. Ed Zitron does a good job expressing the bearish view on capacity delivery, and any frontier lab does a good job expressing the bullish view.
Most likely we're somewhere in the middle, but the key point for this article is that the specific '18-36 month timeline' that was contractually entered into at signing could be anywhere from 12-90 months. We need the permitting, construction, legal challenges, equipment, power availability etc etc. There are way too many factors here to be confident in any specific timeline.
The author is assuming the rosiest of assumptions that we can get data centers built, equipped, and powered on by exactly the timeframe that's used in the press release which is... optimistic.
What happens if most of the announced data center capacity is only coming online in like 84 months? do the AI labs have 7 years to figure out the economics of this and everything will be ok? Maybe!
I think this article is more or less directionally correct, but tbh for the same reasons everyone is playing the game right now, I have a feeling they might be able to coordinate the delayed production schedule for delivery so that the contracts don't all come due at once... It would be in everyones interest to invent reasons to delay by 6 months here or 18 months here if the spin doctors at the top can keep the game going.
It's just so opaque it's impossible to know from the outside (and that's 'outside' by a project-by-project basis), so anybody claiming to know any substantial figures industry wide is blowing smoke imo.
So it's $1.4 trillion over eight years; not loans that OpenAI has already signed.
"The same way a household with $60,000 of income signed a $600,000 mortgage"
That is misleading. OpenAI, and every other AI company, needs not just one "house" but many houses/datacenters, to be built over the next few years. There are several nonsensical phrases in the article (*), and - I really hate to say this, but - it looks as if written by AI.
I'm not an expert, but the author's argument doesn't make much sense to me.
(*) Examples: "siting, powering, and filling a gigawatt-scale campus takes 24-to-36 months from signature - mirroring the two-to-three-year teaser of a subprime ARM.", "The buyer’s cash obligation steps from approximately zero to the full contractual rate, arriving not gradually but as a step function, tranche by tranche as capacity goes live. "
I'm not expert in this at all, but for the sake of conversation, five seconds of thought makes me wonder if this is analogous to the arguments behind Peak Oil alarmism (in the strong idiomatic sense - the pumps will run dry, all natural petro sources in earth will go dry simultaneously and soon, based on the reporting that only X bbl of reservoirs had been confirmed)
The actual situation was that it didn't make sense to spend resources to confirm any more reservoirs, and there wasn't any reason to expect that we'd run out of reservoirs to find if we spent more resources.
I'll try to actually read the article and think about it.
This reset shock doesn't share that exact factor in common with Peak Oil - by the author's account, there are payments written explicitly into various contracts ("compute contracts" as he refers to them), just as there were known periods of time after which all these ARMs in the 2000s would be set to a higher rate, and thus all these households would suddenly have to pay more and can they? "Oh, I'm sure they will, it'll be fine".
OTOH, there's another factor in common with Peak Oil, if you think of Peak Oil in terms of there being a group of speculators telling everybody the oil will run out. In the case of ARMs, a group of speculators were telling everybody the housing prices will keep going up. In this case, it's a group of speculators telling everybody the demand for compute will _not_ run out. The implication is that what the speculators were/are saying, should have been pretty obviously impossible.
Still haven't read it, but I want to keep momentum, so: Critical features seem to be 1) simultaneity, that all the contracts come due literally at the same time without a settling period between contracts coming due, 2) that they will all be reneged on totally, 3) that they can't be renegotiated or the injury softened by hedging techniques, 3b) that data center operators can only be paid by frontier labs and their workloads. The more of these are present, and the closer they are to absolutely present (as in very near white, rather than somewhere greyish), the bigger the hit.
You've already given that 1 seems to be kinda unfounded, since GPU data center buildout and stuffing only ever stretches, and presumably the degree of stretch of DC buildout and stuffing is semi-uncorrelated.
There's also a confounder (mentioned in the article, I think) in which AI firms might re-finance the contracts in some way - taking out another loan at higher interest to pay off the original. They could theoretically keep this up for as long as they want to spend the energy and as long as some bank is willing to play along, which depends in turn on how lucrative the industry appears to be. Not only does this postpone the reset wall, but also breaks it into smaller pieces.
I'm reminded of how long Amazon was in the red. Everyone seemed to be in awe of how Bezos managed to keep persuading investors to put more and more money in. In his case, the growth finally did occur. OTOH, for many dotcoms, it did not.
It sounds feasible, but I have no idea if it's correct. "The bubble bursts when the money comes due and infinite growth has not, in fact, happened" is close enough to previous slumps.
And given OpenAI just announced they won't be releasing their latest model - they say "because of security concerns (it is SO dangerous)", other online commentary I've seen says "because it's not as good as and more expensive than rivals who have eaten their lunch" - that might point in the direction of bursting bubble? I also see some talk of AI subscriptions going up and services being hobbled for those previously paying a certain price, so now if they want the same as before they have to pay more. That does sound like, to me anyway, the repayment of loans is coming due, the infinite progress has not really happened, AI right now is nowhere near taking over the economy and doing everyone's job, hence no infinite profits to pay off those loans, so pop pop pop goes the bubble.
I appreciate your commitment to the "literally every word the AI execs say is a lie and the people who say 'oh damn our model committed a felony maybe it isn't that safe' are of course doing it for marketing" bit. OpenAI is lowering their subscription allowance by 1/2 while releasing models a bit more than 2x as task efficient per token. This is plausibly a net gain, though probably a relative fall compared to Anthropic's latest releases. They did release a new version of Sol, their 2nd most powerful model, that is nearly as powerful as their current most powerful model at a fraction of the price. So there isn't a ton of evidence that they've stopped developing new and better models.
I agree that the labs may be unable to repay their loans, though I'd bet against it. Even if they do, much like the dot com bubble, that will be proof that they iver leveraged, not that AI is done.
Separately, I know you're quite skeptical of AI capabilities. I'll say that the last set of releases are a huge jump. They're still jagged and occasionally dumb. But if we could make the bad parts as good as the good parts already are, that would be enough to replace 90%+ of white collar workers. I dont know when that will happen (and there's the outside chance that it wont), but we are certainly not seeing a plateau.
"I appreciate your commitment to the "literally every word the AI execs say is a lie and the people who say 'oh damn our model committed a felony maybe it isn't that safe' are of course doing it for marketing" bit."
Max, I can do no better to give you my opinion of AI company big shots' veracity than quote Mary McCarthy on Lillian Hellman:
I agree we should be skeptical. I judge Sam Altman as one of the shifting, most manipulative people around. But even if these people have zero moral standards, they could still manipulate you by counting in you to reverse everything they say. "My model lied" -> never trust that guy, clearly this is an honest model!
Dario strikes me as much more straightforward, though hardly incorruptible. Though I gather you disagree. My point, to whatever extent I'm making ine, is that at some point we have to judge their arguments and base things on real events. The HuggingFace attack actually happened and is clear evidence of capable and misaligned models.
Maybe Astra 6.1 just tucked. That's certainly possible. But OpenAI has burned a lot of trust within the AI risk concerned crowd lately. Even selfishly, it's among their interests to not release an unsafe model.
As far as arithmetic claims on subscription limits, those are pretty public and you can check whether I made a mistake.
I think the author sort of relies on this idea - widely suspected - that the models will likely plateau at some level well short of everyone's dreams, while still being a value-add. Anyone who uses LLMs today will agree it can save you some time. A lot of them will also say, genuinely, that it won't save you a _lot_ of time.
And specifically, it's not likely to save everyone a total of $2.3 trillion worth of time any time soon.
Though I'm not (like Scott, I think?) a mathematical realist, I'm certainly very interested in the idea that 'beauty' might one day be mathematically defined. On that note*...
I already knew that mathematical constructs (not least the Fibonacci sequence and the Golden Section) crop up with striking regularity in classical music. Look at the finale of Debussy's La mer, for instance: the phrase-lengths in the first theme group are all Fibonacci numbers, and the piece's central moment of revelation arrives at precisely the movement's golden section [defined in terms of bar numbers]. Off the top of my head, I'm pretty sure you could say something very similar for the Goldberg Variations. (Please give me any further examples if you've got them!)
But... well, I've also been composing since I was five years old. And I've recently noticed that - when I sense that a piece (as it were) came out of the oven especially nicely - I can almost guarantee that the golden section will subtend a crucial structural point of transition.
Can anyone help me come to grips with what in the fresh f**k is going on here?
Umberto Eco's take, a passage from Foucault's Pendulum:
“Gentlemen,” he said, “I invite you to go and measure that kiosk. You will see that the length of the counter is one hundred and forty-nine centimeters – in other words, one hundred-billionth of the distance between the earth and the sun. The height at the rear, one hundred and seventy-six centimeters, divided by the width of the window, fifty-six centimeters, is 3.14. The height at the front is nineteen decimeters, equal, in other words, to the number of years of the Greek lunar cycle. The sum of the heights of the two front corners and the two rear corners is one hundred and ninety times two plus one hundred and seventy-six times two, which equals seven hundred and thirty-two, the date of the victory at Poitiers. The thickness of the counter is 3.10 centimeters, and the width of the cornice of the window is 8.8 centimeters. Replacing the numbers before the decimals by the corresponding letters of the alphabet, we obtain C for ten and H for eight, or C10H8, which is the formula for naphthalene.”
“Fantastic,” I said. “You did all these measurements?” “No,” Aglie said. “They were done on another kiosk, by a certain Jean-Pierre Adam. But I would assume that all lottery kiosks have more or less the same dimensions. With numbers you can do anything you like. Suppose I have the sacred number 9 and I want to get the number 1314, date of the execution of Jacques de Molay – a date dear to anyone who, like me, professes devotion to the Templar tradition of knighthood. What do I do? I multiply nine by one hundred and forty-six, the fateful day of the destruction of Carthage. How did I arrive at this? I divided thirteen hundred and fourteen by two, by three, et cetera, until I found a satisfying date. I could also have divided thirteen hundred and fourteen by 6.28, the double of 3.14, and I would have got two hundred and nine. That is the year in which Attalus I, king of Pergamon, joined the anti-Macedonian League. You see?”
I have a different perspective on the golden ratio in aesthetics, in that I think it's a giant pile of pseudo-scientific nonsense and I hate it with a burning fiery passion. But how to put it nicely...
Let's start with what is not nonsense. Fibonacci sequence is a legit mathematical construct (one of a myriad other interesting sequences). The ratio of the consecutive numbers does tend towards a golden ratio (but this is not the only sequence with this property).
Golden ratio is a number with legitimately fun mathematical properties that satisfies all manner of cute identities. (It is not the only interesting number that satisfies various interesting identities). So far, so 100% true. Then it gets dicier.
Golden ratio pops up in nature in some places. Sometimes the nature wants the properties that the number possesses. However, there are great many claims about golden ratio being present in nature, that are blatantly nonsensical and untrue. No, it's not in nautilus shells or galaxies! (But pine-cones - yes.)
Then there are claims about aesthetics, that are dicier still, that these mathematical relations somehow are keys to beauty, in music, in visual art. I think these claims are on a weak foundation, that it's kind of a historical accident that artists of all sorts became weirdly fixated on one mathematical structure out of a million and convinced themselves it's uniquely important for some reason.
When it comes to art I would say, the claims of golden ratio/spirals appearing there fall into two categories: 1) the artist deliberately put it there, because he or she bought into the idea that you should do it. It might be that the painting looks awesome, or the song sounds wonderful, but that's because the artist was very good, and could have created something wonderful out of any mathematical foundation, not because the golden ratio is so special, or 2) it's not actually there, but people think they do because of the usual numerological games where you can find anything in anything if you try a bunch of things and are not very precise.
Not too be too dogmatic, in defense of the golden ratio I could say this much: 1) it's kind of neat where the small thingy is to the big thingy as the big thingy to them together. Maybe it's a little pretty? and 2) if you do have this sort of empirical observation that you always find it in your best sounding music, that ought to mean something (although, I'd be careful and watch out for finding what you expect to find if you search hard enough).
Golden ratio is a very small and simple little thing, and a work of art is a very complicated thing, with a ton of structure. Even if it was an especially nice ratio, it could only ever account for the tinyest amount of what makes a song beautiful, because any song has vastly more structure than golden ratio holds. Probably there is some kind of mathematical characterization of beauty, but I expect that if golden ratio is in any way involved in it, it would be like a single line in a big book.
'Golden ratio is a very small and simple little thing, and a work of art is a very complicated thing, with a ton of structure. Even if it was an especially nice ratio, it could only ever account for the tinyest amount of what makes a song beautiful, because any song has vastly more structure than golden ratio holds.'
This is the key, I think.
Even if you can in fact derive a fruitful mathematical model for understanding beauty in music, the bulk of that would probably look something like... well, I don't want to start talking about transformers; but it'd suffice to say that - when you're talking about a 'great' work of classical music - every note exists in a complex network of relations with 10^6 other notes, spanning [what Phoenician traders would call] the gesture, the phrase, the subsection, and so on. At best, the Golden Section is probably a few sentences in a (horrifyingly complex doorstopper of a) book.
Also - you're probably right vis-a-vis judging my own music. I'm too close to the subject to draw any objective conclusions.
"At the heart of the redevelopment scheme is a new village centre featuring a two-storey mixed-use building with a gym, café, health centre, pharmacy, and offices. Plans also include a community centre, supermarket with garden centre, and two commercial buildings. Four apartment blocks are proposed, providing 132 homes in total, made up of 24 one-bedroom units and 108 two-bedroom units. A multi-use games area, crèche, ESB substations, recycling centre, and over 600 car parking spaces are also included in the design.
The central section of the site is earmarked for a modern technology park, which will feature five buildings for offices, light industry, and research and development. A large ICT (data storage and hosting) facility is also planned, consisting of three interconnected buildings and modular data halls, alongside an energy centre powered by 16 gas turbines, with transformers, stacks, and support buildings."
I'm not entirely sure who is behind this or what the objections are. The mooted data centre is on the now long-abandoned premises of the dairy co-operative society so it would be doing something with already developed but vacant land. It is a very small village, so a huge data centre and associated works *would* be a massive change. On the other hand, if all the promises hold, then there would be additional housing and resources coming along with the data centre.
I don't know. Generally I'm sympathetic to protests, but I'm not sure how much of this is grassroots and how much is outside activists using this.
But on the other hand, local sports clubs on the other side of the county are in line for some of that sweet, sweet Meta cash from our new lords of the manor:
"A €150,000 boost is heading to three west Waterford GAA clubs, courtesy of Mark Zuckerberg and Priscilla Chan.
The couple, who bought Strancally Castle in August, have chosen Shamrocks GAA in Knockanore, Tallow GAA and Lismore GAA as the recipients of their first Irish philanthropic donation since moving into the area."
I have to admit, this one makes me laugh as I'm dang sure Mark Zuckerberg wouldn't know a sliotar if it hit him in the face, but oh well: if they want to play Lord and Lady Bountiful, why not?
I saw a number of sources covering the Zuckerberg story as him attempting to support LGBT rights - some of the Americans not realising that WWN is not a real Irish network...
More housing would indeed be very palatable, since we really are in a housing crisis and need as much new accommodation built as we can get. That would change the village, but it's already being changed by becoming a commuter satellite of the city with existing new housing already built.
the author (well prompter) is an excellent Czech journalist formerly writing about science in general, now mostly about AI. He definitely read some of Scott's work in the past.
"Travel is fatal to prejudice, bigotry, and narrow-mindedness, and many of our people need it sorely on these accounts. Broad, wholesome, charitable views of men and things cannot be acquired by vegetating in one little corner of the earth all one's lifetime." -Mark Twain, "The Innocents Abroad"
I've been seeing this quote a lot lately, but I'm not not sure it's true (or at least not for me). For every trip that challenged a negative stereotype (I found many Germans to be nice and humorous, for example), I've had a trip that confirmed one. I found South Africa full of casual violence (had a knife pressed into my side while a fellow demanded the equivalent of $7 in local currency) and Dubai full of people who are shallow and transactional.
Curious what the rationalists think. Sage wisdom from Twain, or sentimental fluff?
Maybe this is my own sentimental fluff, but I think the Twain quote isn't necessarily contrary to having a few negative stereotypes confirmed by travel. Maybe the broad, wholesome, and charitable views of people he is talking about is from understanding where people are coming from, what they value, and why they act in certain ways. Maybe this is more important than a negative stereotype.
You can't put yourself into the shoes of a stereotype, but you can put yourself into the shoes of a real person, even if they conform to a stereotype.
When I travel I like to see how other people live. I spend a lot of time walking around middle class neighborhoods and having coffee and food at quotidian establishments. I don't show up with stereotypes but I usually leave with a few.
The effect is lessened due to herd immunity; enough people are traveling that most little corners of the earth are still visited by those outside it. But seeing a place definitely still has effects on your perspectives.
I had a connecting flight in Ethiopia once. Three things stood out; the airport had no book stores, the airport wall still had a sign advertising an extension "coming in [a year and a half ago]", and while boarding, the flight attendants abandoned the assigned seating and told us to just sit wherever we could. There was just kind of a chaos to it that I'd never seen before. I don't know if that's counting for or against a stereotype, but it's certainly enlightening.
Also the grocery store in Israel had a metal detector and armed guard outside, and it makes one go "oh right, this is a country with enemies".
Research shows that stereotypes are generally accurate. As much as I like Twain I'm afraid I have to disagree with him on this. Although probably the stereotypes in his era were quite a bit more parochial.
There's a big difference between travel then and now due to the existence of a massive tourism industry.
If you are white and have some money you probably don't need to learn any language or anything about the place you are going in order to be mostly comfortable and functional there, and you will probably be identifiable as a 'tourist', a known type of person with specific cultural implications, for your entire visit. You will likely be kept in a separate bubble of businesses, neighborhoods, sites, and attractions that are built and cultivated to give people like you a cultivated 'experience.'
That's very different from traveling somewhere and just living as a person in that place for a while. it's the difference between visiting the Las Vegas Strip, and walking 2 blocks away from the strip to see what the city is actually like.
So I can see how there's a type of travel which didn't really exist back then, which doesn't have all the salubrious effects.
Twain almost always had an agent, servant, or translator. The Innocents Abroad is about the predecessor of the modern-day cruise; Following the Equator has him hiring locals to take care of most things. He even notes that the treatment of the servants by their bosses reminded him of the way he saw slaves being treated in the antebellum South.
> If you are white and have some money you probably don't need to learn any language or anything about the place you are going in order to be mostly comfortable and functional there
Nice to know that everywhere speaks Hungarian or Croatian everywhere. Odd though that non whites travelling from English speaking countries can’t use that as a lingua Franca
I think many (most) tourists of the kind darwin is talking about speak English as a second language, and I think darwin is right that even if you don't you can be comfortable as a tourist e.g. on organized tours in many places.
Darwin seems to believe that white people speak English and non white people don’t.
Even as a second language this isn’t true even in Europe. It’s more true now, but there’s plenty of European countries, outside tourist spots, where you would need to know some of the local language to get by.
2 blocks away from the strip is the haunted hotel. Yes, we walked. The elevator was sufficiently creepy (would let you out at floors you did not push, did not go to the selected floor, even after letting it open on other floors).
About twenty blocks away is the Pinball Hall of Fame. We did not walk (mainly because the bus was easier and more convenient, we walked the whole strip in long pants and long shirts in 95 degree weather. Dress for the Sun -- the tourists pointed and laughed, but we were comfortable).
In Twain's time, you were either Rich or the person who stepped in shite. Well, at least in India. I'm not sure it's changed all that much, except that if you're Rich in south america you fly helicopters from building to building.
Even when traveling abroad confirms some stereotype you had, it may present you with a thousand little clues decorating that stereotype that become informative.
Suppose you go to another country and observe a beggar. After confirming that there exist poor people here, you pass a bit of the local currency to alleviate. But if you're attentive, you might notice how the beggar is attired (brightly? drably? good shoes? barefoot? jewelry? piercings? tattoos?), whether he looks like people walking around who aren't begging (same ethnic phenotype?), how others treat him (do they avoid? do they help? do they sneer?), how police treat him, where you are (boulevard? alley?), and on and on. If you're attentive enough, you'll notice patterns that refine your stereotypes, caveat or refute them, etc.
Plus, it's one thing to read about how big the Grand Canyon is or how beautiful the Dom in Vienna, and another to actually walk around inside.
I imagine it depends a lot on the traveller. A high school dropout from rural America who's barely heard of Canada? Probably benefits a lot from travelling the world. A highly educated ACX reader who spends their free time reading about other countries and cultures? Their worldview will probably be changed much less by travel.
Personally, whenever I travel, I'm always grateful to come back to a familiar culture in a first world country. The slums of India are definitely an educational experience, as are the imprisoned (sorry, covered) women of Pakistan and the infuriating bureaucracy of Germany, but I can't say I'm pining to bring them to my own country. In that way, it's possible travel has made me more anti-immigration.
Being "in unfamiliar situations where [one is] at a disadvantage" strikes me as a generalizable opportunity to grow one's empathy, especially if there is no quick/easy exit from the situation.
I'm thinking of things like: defending a doctoral thesis; attending a high school or college in which you are a racial or ethnic minority; attending a Diwali festival at a local temple though you are not Indian or Hindu; attending a summer camp in a foreign country in which you don't speak the language; joining a pickup game with people you don't know who are better or mostly all better at the sport than you; switching from a white-collar to a blue-collar career; and so on.
Travel is needed for some of these, not for all. But all will give you finer vision so you're not seeing in caricatures. You'll also see yourself more clearly.
I think the essence is that you're not going to learn much on a few days worth of a mass tourist-y vacation, but if you decently familiarize yourself with a foreign country, you learn not only about that country, but might also learn more about your own.
I could save you some travel expenses by providing a rule of thumb: beliefs like "all X are Y" are typically wrong, but statistical stereotypes are often based on truth.
If you travel to figure out the truth about the stereotypes, I think you should travel a lot, because if you only meet one person or see one city, it may not be representative for the country. For example, it would be a mistake to visit Moscow and conclude that this is what the entire Russia is like.
Perhaps more importantly, by seeing other places you learn about your own country more. You are no longer the fish that doesn't see the water, or the boiling frog. For example, you may see that things that people in your country insist are impossible, are actually perfectly normal somewhere else. -- You may get some of this from reading internet, but in a text medium, you are limited by language. It is difficult to learn about the country whose language you do not speak; those who write about it in English typically have some agenda; but if you walk the streets, you get some unmediated experience; more if you stay there for a longer time.
“If you travel to figure out the truth about the stereotypes, I think you should travel a lot, because if you only meet one person or see one city, it may not be representative for the country. For example, it would be a mistake to visit Moscow and conclude that this is what the entire Russia is like.”
This isn’t a “mistake”. It’s a fundamentally wrong process. I don’t think anybody does it.
I think it’s more common to interact with a few people and come to a conclusion they aren’t as different as one might have priorly assumed.
You might not eliminate all prejudices but your impression is likely to become more nuanced.
It seems people who are overly set in their opinions aren’t really going to travel at all.
Everyone does it for France. Parisians are in general sick of tourists and kind of cranky with people who don't "behave" or even TRY. Get outside of Paris, and tourists get treated as "entertainment" and "be nice to him, he's got money -- their dollars spend well."
Probably more true in Twain's day, when travel was a big, time-consuming deal, and talking to a stranger across the globe or watching travel videos from any country imaginable wasn't something you could do in your pajamas in your living room. I haven't read Innocents Abroad but it's great fun to peek into it and hear Twain talk about bathing in the Dead Sea and complaining about obnoxious study-abroad types who suddenly speak English with a foreign accent after a few months overseas. The past is a foreign country, they do things differently there, and it's cool to visit and see how some things don't change.
“For every trip that challenged a negative stereotype (I found many Germans to be nice and humorous, for example), I've had a trip that confirmed one.”
Sounds like a positive result over all. (None of your “negative stereotypes” would have been challenged if you stayed at home.)
I think in general, encountering things in real life teaches you to be a pragmatic realist about them over extremes. If your prejudiced assumption is inconsistent (e.g. assuming an entire people don't do humor) it'll get overturned. If it's more coherent (e.g. high violence society) it might not be, or you might just find out the mechanics of how that society actually functions with the violence.
I think counting stereotypes challenged vs reinforced is the wrong frame. To be a bit blunt, if *all* you got out of your experiences abroad was updating on ideas you already had, you may have had some pretty powerful blinders on.
Rather, I think travel is important in large part because the world is very much larger than any human intellect can apprehend. Not merely in terms of number of people or amount of land area, but in terms of richness and variety of people, places, ideas and viewpoints. It's very easy--and I expect partially automatic for most people--to round off blank spaces on one's mental map of the world to empty spaces, or at least emptier spaces. If I don't know what's there, there's also generally a large extent to which I don't know that I don't know: I can't form any clear picture of the amount or extent of the missing detail.
To put it another way, I think the big thing travel teaches is intellectual humility about the extent of one's knowledge about the world. If there's any profession or intellectual field you're an expert in, you should be able to appreciate how important this is: just think of the times you've seen an overconfident amateur brazenly stumble through the field, blithely unaware of how many important bits of skill and knowledge they were missing.
I agree - I think travelling, and more generally interacting with people who are different than you, forces you to question and see what you may have taken for granted and allows you to see things from more different angles - often in surprising ways. This tends to make one more careful in asserting that ones own point of view is the right one.
I think you're stereotyping the stereotypes here. Most people who say "South Africans are violent people" would acknowledge upon questioning that they are not uniformly violent.
Would you concede that "South Africa is violent" means the same thing as "The people in South Africa are violent?" (We're not talking about the wildlife, after all.) It seems like you are trying to introduce a shade of meaning that emphasizes that not all South Africans are violent (which is trivially true) at the expense of claiming a true statement is false.
I.e., I think you are arguing that the stereotype is true but saying you're saying it's not. Though I feel the need to emphasize that a rational thinker probably shouldn't assume a true stereotype is universally true or even broadly true.
Edit: to rebut my own argument a bit, the institutions could be violent even if the individuals are not. You could argue West Germany was a violent place though West Germans were not. Nonetheless, I don't think that's an accurate way to characterize South Africa.
The other part of this is learning what violence looks like. I imagine there were parts of South Africa which you learned were unsafe, and parts where you felt relatively safe. I imagine you learned some of what makes a place unsafe (few outsiders, little policing, economic precariousness, resentment?, sense of exclusion? - I don’t know, I haven’t been). I imagine you could imagine some people from your own community being more violent if they were in similar circumstances. From that, you might get to “South African conditions are violent”. Does that match up with your experience?
In south africa, one is remarkably at risk of falling squatters from high story buildings. This is "paid non-police action." I suppose one could call this "little policing" but the squatters are outsiders (they were using the elevator shaft as a toilet). "Economic precariousness"... oh, maybe so, that's what happens when you walk into a city with relatively little learning...
Yes, I do imagine that if we gave my next door neighbor a tank because the police fled, that he would count as "more violent" despite being part of the local community watch.
Knew a person working real estate in South Africa. She would go to jobs with a grenade in her purse (also a fake grenade, for bluffing). I don't think you'd call her violent, but... the sheer fact that a civilian (non-drug-dealer, perfectly legal citizen) feels the need for that level of protection? That says something about the level of violence in South Africa. (And I'm feeling bad for putting it as South Africa, as she worked in one urban location).
"South Africa is violent" is too vague a statement to be useful, and "the people in South Africa are violent" is a lazy and stereotypical generalisation. You might be better getting a bit pedantic and saying "statistics show that the incidence of violence against people in South Africa is significantly higher than in most regulated countries", perhaps including figures. If you were to risk "the higher that usual incidence of violence in South Africa is historically and culturally determined", which I think is a reasonable statement, you had better be prepared to back it with some facts.
I think violence is a good example because it brings up another point about how we are bad at numbers and why stereotypes are the way they are: if you ask someone what percent of <country> are violent, they will likely say something grossly exaggerated. But if you tell them one person in a thousand is a dangerous criminal, they will probably underestimate the risk and eventually pay a high price for it. So the misperception the stereotype causes is helpful.
I'm more interested in helpful statements than trivially true or maximally defensible statements.
Noah Smith has a guest article by Ramez Naam about the possibility of recursive self-improvement in AI, arguing that the current evidence shows that it's not happening. The gist of the argument is that while AI is speeding up research, it takes more and more research effort for each new increase in capabilities. AI isn't getting smarter fast enough to keep up with diminishing returns.
However, while I understand the gist, I don't know a whole lot about the statistics and my eyes glazed over somewhere around "ECI-extrapolated METR," so I'm curious what the crowd here has to say:
The notion of an exponential intelligence explosion is just bunk. The intelligence of the researchers isn't the only factor involved. Compute, energy, time, training data are all factors. Increasing the IQ of every OpenAI employee to 1000 wouldn't instantly produce ASI. It certainly wouldn't hurt but AI improvement would just be gated by whatever the next bottleneck was. It wouldn't cause Nvidia to pump out chips any faster, for example. Real problems in the real world are complex and multi-factorial. Just adding intelligence doesn't magically fix anything.
I think it is correct that we are not in an intelligence explosion yet. However, we are also not at the point where AIs are better than humans at AI research yet, so that isn't a big surprise. AIs are now better than almost all humans and even almost all AI researchers at some subtasks of AI research, but not yet at all of them. Important bottlenecks remain. As a result, overall research acceleration is still not very dramatically high, though it has reached the point where I definitely feel the speed increase in my interpretability team.
So, I think many relevant bits of information for estimating the returns on cognitive reinvestment are just still missing. Foom still seems like a very real possibility.
I have only read your summary, not the original, but seems like what follows from thr summary is not that recursive self-improvement isn't happening and won't happen, but that it will happen more slowly than we expected.
If diminishing returns are strong enough, then eventually you hit a point where it's not practical/possible for the AI to make itself smarter. So while it would technically be recursively self-improving, it wouldn't be the runaway intelligence explosion people fear.
Should anybody be interested I've written a piece on the dissipation of disequilibria as the driving force for the complexity of life [1]. I'm not particularly expert in physics, so it would be very welcome if anyone who is could point out if I've written any nonsense. I'm thinking of expanding into a blog on various mechanisms and patterns of biology and evolution. I've also ported over there a collection of posts from my tumblr in which I talk about the diversity and evolution of protists, my actual field of study [2].
I do have two objections. One I think I commented a while ago on a previous version of this (that the relevant store of negentropy in the Sun isn't its pressure (which is never relieved) but its temperature or nuclear composition depending on how far back you want to go), but there's still one reference I saw to the Sun's pressure.
The other, more major one is the claim that life may have arisen *in order to* produce entropy. For one thing, that explanation is terribly teleological and therefore immediately dubious, of course life arose where there was an opportunity to produce entropy because that's the only way it could, but it having that purpose seems like at best a misleading way of saying that. Also, the rate at which entropy approaches its maximum tends to only be limited by things like conservation laws. The temperature in the Sun can't immediately equalize with space because conservation of energy means the energy must travel out through all the matter in the way first. At the distance of the Earth, the sunlight is already spread out and can therefore almost immediately re-thermalize, reaching its entropy maximum, just by interacting with any dark-coloured matter. All the intermediate steps of photosynthesis, predation, etc. just serve to collect the negentropy in certain locations before dissapating it. Assuming sunlight is mostly absorbed within 1mm of the surface after hitting the Earth, that top 1mm layer has an energy dissapation rate per volume (and thus entropy production rate, since the relevant temperatures are similar) about 100 times greater than that of a human body. Energy dissapation rate per volue just isn't a relevant measure for what you seem to want to do with it.
One I think I commented a while ago on a previous version of this
Did I already post the old version here? I confess I had completely forgotten.
the relevant store of negentropy in the Sun isn't its pressure... but its temperature...
Thanks, I'll amend this.
the claim that life may have arisen *in order to* produce entropy. For one thing, that explanation is terribly teleological and therefore immediately dubious, of course life arose where there was an opportunity to produce entropy because that's the only way it could, but it having that purpose seems like at best a misleading way of saying that.
I see; that is certainly not what I meant to convey. I think you could say life arose "in order to" increase entropy more efficiently in the same sense that a river forms "in order" to bring water to the sea level more efficiently, but I can see how that'd be misleading. I'll try to make clearer in the text that I'm not saying that life appeared to fulfill a preselected purpose.
[...] Energy dissapation rate per volue just isn't a relevant measure for what you seem to want to do with it.
Now this is a bit more damning for that section of the essay. I'll do a bit more research on the subject. Thank you very much for the feedback!
I recently realized that a popular idea has unwittingly affected my way of thinking, so I'd like to hear the arguments for and against the idea in order to consciously accept or reject it. The idea in question is a very pessimistic interpretation of fiduciary duty, which goes something like this:
"According to current American case law, public, for-profit corporations are legally obligated to engage in anti-consumer practices because doing so would maximize shareholder value."
If you believe that the above statement is correct, then can you cite supporting case law? On the other hand, if you believe that it's incorrect, then can you explain what would preclude the outcome in the following scenario:
"An executive is summoned before the board of a small, publicly traded company. A report from the marketing and UX teams predicts that eschewing a click-to-cancel design would increase revenue by 10%. The young CEO, an MBA from BYU, refuses to implement this design on ethical (but not religious) grounds. The board fires him, replacing him with someone without qualms. The former CEO's career is ruined, and he has no legal recourse against the board."
The nineteenth century model of accountability for companies was the three-legged stool: they had obligations to their employees, to their customers and to their shareholders. This was overturned by Milton Friedman and friends, who originated the idea that by maximising shareholder value, the invisible hand of the market would see to the wellbeing of all other stakeholders: workers would be free to maximise their compensation by moving to the highest paying positions, and so all companies would be forced to offer equivalent benefits. It doesn't work that way in practice, of course.
My point is that the overriding interest of shareholders is a fairly recent conception - and it's more philosophical than legal. In your hypothetical, if the CEO serves at the pleasure of the board, then he has no recourse. If he's dismissed against the terms of his contract, then he might be in line for a payout. Friedman would presumably argue that the actions of the company are moral (in fact, required) to maximise returns to the shareholders in practice; and that in theory, details of implementation have no effect on rational economic agents.
The CEO could argue (and sometimes has done it successfully) that sacrificing short term profits brings better long term value, because of things like brand value, customer trust and so on, sometimes he's right, and sometimes he isn't. He could be fired anyway, but that's just the breaks, people disagree about things. He's paid handsomely for the trouble, at least.
It's not true in effect. The famous example here is Dodge v. Ford Motor Co. where Ford wanted to retain profits rather than distributing them as dividends. That's the case where Ford was required to pay dividends and it was immediately taken up by anti-capitalists. It's where the quote that corporations exist primarily for the owners comes from.
However, the reality is that Ford basically grandstanded. His lawyers tried to argue he was being a good steward of the business and concerned with employee retention, funding expansion, etc. Ford then walked in and basically said he thought the company was making too much money, that he thought they should direct resources toward (what he thought of) as the general good, etc. This threw the case and he lost. If the court had found for him then they would have been ratifying that corporate leadership could decide investor's assets could be appropriated to their own causes. The court said: no, you are stewarding their money and have a duty to increase the value of their capital because it's theirs, not yours.
"An executive is summoned before the board of a small, publicly traded company. A report from the marketing and UX teams predicts that eschewing a click-to-cancel design would increase revenue by 10%. The young CEO, an MBA from BYU, refuses to implement this design on ethical (but not religious) grounds. The board fires him, replacing him with someone without qualms. The former CEO's career is ruined, and he has no legal recourse against the board."
This is strange to me. No CEO has legal recourse against the board for being fired in most cases. Average employees also don't get to sue their employers if they're fired for ethical disagreements. The actual relevant case law is that the MBA can just say: "I believe ethical practices will, over the long term, increase the value of the business" and it's allowed.
"According to current American case law, public, for-profit corporations are legally obligated to engage in anti-consumer practices because doing so would maximize shareholder value."
This is not literally true. However, the board and officers do need to exercise "reasonable business judgment." One can imagine scenarios where anti-consumer practices may be required by "reasonable business judgment", for example, if it's an industry standard and you'll surely fail if you don't go along with it. But generally speaking, "reasonable business judgment" does *not* mean solely maximizing profits.
"The board fires him, replacing him with someone without qualms. The former CEO's career is ruined, and he has no legal recourse against the board."
Yes, this is how things work. The board represents the owners of the company. They control who's the CEO. If the owners of the company want to do something legal and the CEO refuses, then the CEO will be fired. It sucks when this happens for ethical reasons, but it's also hard to imagine corporate governance working any other way.
Since most shareholders are trying to make a profit on their investment, there is considerable pressure on the board (and subsequently the officers) to be focused on profits. IMHO this is a much more accurate explanation why corporations tend toward anti-consumer practices.
To be clear about the difference between the two scenarios, an ethically-objecting CEO may be *legally* in the clear if they still satisfy "reasonable business judgment". But they can still be fired by the board.
Its basically untrue as a matter of case law and legal obligation. No court wants to be in a position of judging business decisions like UX. Shareholder primacy may show up in court in other contexts perhaps related to deal pricing where the shareholders would try to demonstrate conflict of interest.
It is true as a matter of corporate governance theory. The company should maximize shareholder value which is usually not in conflict with consumer value because they need repeat business, but at times can be. Discriminatory pricing(the econ definition) is probably the text book case of this. It is, in essence, a strategy to transfer consumer surplus into corporate profits. It is not illegal. It is probably encouraged when the company has the market power to do it
The CEO in question has little recourse for being fired by the board regardless of reason because at will employment. I reject the premise that his career is ruined.
You might be interested in the book Incorruptible by Eric Ries, which covers the (surprisingly short) history of the duty to maximize shareholder value, and also proposes some solutions.
It's quite possible that the board would fire the CEO in a position like that. Not because it is required by law, but because CEO is an at-will position where the board can fire for any reason they want. And the board can have stupid reasons.
If someone tried to make a legal issue out of it, say a group of shareholders suing the board to force them to fire the CEO, then the CEO just needs to say something handwavy about how click-to-cancel will build goodwill and customer retention and lead to greater profits in the long run. It doesn't have to be true and it doesn't have to be supported by strong evidence, it just has to be vaguely plausible.
Fiduciary duty to maximize shareholder value, is mostly relevant when it comes to actual fraud where there's no plausible argument that paying twice the market rate to the CEO's idiot nephew's construction company is going to increase revenue or profit.
Highlights include evidence for electromagnetic and microtubule based consciousness, how the loneliness epidemic isn’t real, the links between info tech-religion-gender, a summary of cancer vaccines, and evidence for graft induced horizontal gene transfer into the germ line of a humble cabbage.
"The challenger in that race, Jamie Joyce, contacted me and asked me to change my mind (both candidates will be facing off again in the general election)."
From Ms. Joyce's linked site:
"Congress keeps making missteps. AI will be one of our biggest challenges moving forward and the people in office don’t understand it. Foreign nations are pulling our strings, and the Epstein class goes unpunished. We need a plan and expertise at the wheel - Jamie has both.
Because everything we care about, like affordable housing, a sustainable environment, clean food, fair taxes, and ending genocide/war is downstream of the will of the criminal Epstein class who are selling our democracy for parts - and are getting away with it. Now is no time for politics as usual. It’s time to fight back. Together, we can win a better future for all."
When i started reading the 2nd paragraph I didnt realize you were still quoting the website. I thought I was getting Deiseach sarcastic commentary on the preceding paragraph and thought you were being a bit uncharitable. The first paragraph led with AI not epstein after all.
Andrew, with the treasury of riches lavishly displayed on the lady's campaign website, there was no need for me to gild the lily 😁
She likes referring to herself in the third person (sure, it could be all typed up by her campaign staff but I get the strong impression that it's a one-woman-band operation). "Jamie was the child of a mother who left school at 13" and so on. No mention of Dad at all, leading one to presume he was not in the picture, which may be unfair to Dad. No mention of Mom other than as an inspiring anecdote (I would like to know her name, for one, if you're going to use her as 'see my hardscrabble background', ditto for Gavin 'The Gavster' Newsom and his hardscrabble upbringing as the classmate and family friend of billionaires).
I learned with little surprise that she is a denizen of Berkeley. You can tell by her "Berkeley cute" photo in her Rosie cosplay, with the "lookit my big huge glasses, ain't I so adorable like an anime character?"
But I should hush my lyin' mouth, because global leaders disagree:
"Jamie is a multidisciplinary expert, executive, researcher, and activist living in Berkeley, CA who is respected by global leaders and has the right expertise for government in this moment of technological change."
MULTI disciplinary, Deiseach! Not just one scabby wretched discipline, but MULTI of them! Are *you* multidisciplinary? Are *you* an activist inhabiting the socialist comradehood woke paradise on Earth that is Berkeley? Not so hot now, are ya Deiseach?
Her site is a target-rich environment, I need to rein myself in because I could go all day with this.
Yeah that whole letter is top to bottom insane. She complains about Scott causing donors to back off from her but I'm guessing if they mentioned Scott at all it was in an attempt to politely avoid going "lady you are clearly a crazy person I'm going to walk away now".
"The most egregious example being her claims about how much money she brings to the district. In Oakland, (and I’m told in Berkeley as well, but haven’t seen it yet) she has set up billboards - that were paid for with tax payer money - claiming she has “delivered” $833 million dollars to the East Bay. This is the claim she is probably most known for: that she brings home hundreds of millions of dollars. She states it often. Many people state it often. Her main claim about her performance thus far has basically been that no Dem could do anything meaningful during the past two years in the House, so she focused on bringing home money and did it astonishingly well. She even put that she “secured” $300 million in the county-circulated voter guide (my name wasn’t even listed in it for the primary by the way).
The problem is that this appears to be misleading or outright false. Her appropriations bills only account for ~$11.2 million dollars. So roughly 1% of that number. Interestingly, she’s since upped the ante. I walked up to the stage at her Emeryville town hall and asked for an audit to get her on record, at which point she doubled down and said the number was now over a billion (I have this on video). She then proceeded to yell at me from the stage (I also have this on video). Anyway, it’s hard to disprove that this is an overstatement. It may be even fair to call it an outright lie and voters should know it - though that can’t be confirmed until we have the audit. "
Some of that letter explains why she won't get elected. Gasp, a politician claimed they got *all* this money for funding for Good Things but they didn't really, that money was coming anyway?
Welcome to how the sausage is made, Jamie. In local politics, the funniest example of this I saw was three different politicians from three different parties taking out ads in the local paper to brag about how this €€€€€ in funding for Worthy And Needed Local Thing was all down to him/her personally getting it done.
You guessed it, the funding had nothing to do with them and was coming as part of disbursement of central government funds anyway.
Another trick (which can't work anymore because the systems have been changed) was local councillor or public representative, at the time of year when college grants were to be announced, ringing up the Vocational Education Board (as was) to get the scoop on "did Mrs. Murphy's little Johnny get the grant?" and because Council Critter probably was on the Board of the VEC (one man, God rest him, was the kind of local politician who was on every board, committee, club, and other organisation in the entire west of the county and had probably attended more funerals than the priest) and hence our employer (technically), we had to divulge the information that yes (or no), Mrs. Murphy's little Johnny had indeed been awarded the maintenance grant.
Council Critter would then hasten with great speed to call on Mrs. Murphy to let her know her Johnny would indeed be going to the halls of higher education thanks to getting the grant, all thanks (implied) to Critter's sterling efforts on her behalf. Future votes from her and her family in next election for Critter gratefully received!
In case I do need to tell you all this, Critter had no more to do with getting the grants than the babe unborn (canvassing disqualifies, after all). But that was not the game; trading on the (not unfounded) belief long lingering in Ireland that you need pull (someone on the inside, an in, a contact, someone who can expedite things) to get benefits, the idea was "of course Critter intervened on our behalf, we would not have got it otherwise, and the quid pro quo is votes for Critter".
Ditto with housing lists.
Our candidate for office thinking the public will be astounded to learn the truth behind the billboards is almost charming in its naïveté, and also is one good reason why the incumbent will continue to incumb after the election.
I am not surprised but still disgusted by the apparent fact that "I managed to get us a larger share of the spoils" is an argument in favor of the candidate in the first place.
Surely the ideal we should be aiming for is that the budget goes to where it's best used and this is electing somebody who is actively and intentionally detracting from that ideal to everyone's loss except theirs (and perhaps their constituents).
At the very least everyone who votes for "screw the rest of the country, I'll do 'right' by my people" kind of politicians should lose the right to complain about state dysfunction, as they're actively and obviously contributing to it.
Playing defect against other parties surely can make sense, but playing defect against your own team just seems actively counterproductive.
Shaked, her wooing of potential voters is a thing of beauty and a joy forever. Sample:
"Last year she moved to D.C. for 3 months to negotiate with DOGE to protect agencies. One agency she wanted to protect was NARA, and it seems that NARA was never cut by DOGE."
Now, she's not saying she was able to take on The Muskrat and Orange Man Bad in hand-to-hand combat and defeat them handily so that NARA (whatever it may be) was saved from the Chainsaw, but she's not *not* saying it either 😁
"A government official at NARA wanted to thank Jamie for this advocacy with a private tour of the U.S. Treaties Vault, where she got to view, in person, the Treaty of Paris (which is the document that Benjamin Franklin signed to end the Revolutionary War)."
And we know this is so, because there is an accompanying photo of her grimacing over what I presume to be the Treaty. Can any reasonable person demand more proof of her routing the forces of Meanness than that?
I feel like I've dropped into an episode of "Get Smart", querying "Were NARA in danger from DOGE?"
It would seem yes, at least according to this bunch, American Oversight (whisper it not, but I get the merest, slightest, faintest hint of an impression of a shade of a possibility that they may not love and adore President Trump):
Now, were NARA in *actual* danger? Eh, not any more than any other government department from budget cuts, there do seem to have been hiring freezes, layoffs, and shut-downs but those go back a few years. I do think historical document preservation is a worthy cause, if the lady confined herself to the worthy if dull causes, she might have a better chance of elected office.
Though, given that she's running in "Berzerkeley", maybe not.
I'm suddenly realizing I never looked into Rosie the Riveter lore. I bet there's a whole sitcom's worth of characters in there. Rosie the Riveter, Tunia the Mechanic, Daisy the Ditz, Maple the Divisional Overseer.
GiveWell is unambiguously both effective and really helps people.
I'm glad there are people looking into AI safety but I wish they used a different name to the "children shouldn't die of easily preventable diseases" people.
There are some EA orgs that I feel deserve the moniker without qualification. For example, GiveWell evaluates charities based on how much lives they save or improve per dollar amount. I find this service very valuable, as it allows my donation dollars to go further, making my altruism more effective.
I believe the movement is much larger than that, and most of it I really don't care about. So while a randomly drawn EA org might or might not deserve to be called "effective altruist" by whatever standards you apply, there are EA orgs that probably do.
Roughtly speaking, EA splits into short-termist and long-termist causes; the former genuinely does live up to its name and has saved vast numbers of lives, while the latter is mostly faddish wastes of money.
Sorry, when you say 'EA-linked' do you mean the meta-charities or the actual malaria, intestinal parasites, and lead exposure charities? I certainly hope the latter haven't been penalised by a few very vocal peoples' novelty preference!
Depends on whether you define it as "people who like the idea of handling extinction risks, malaria or animal welfare" or "people who personally hang out with a few specific British philosophers". The first is pretty unobjectionable; the second I think can be much more reasonably criticizes.
Pretty interesting seeing the dichotomy between people who thought paying $3000 for an event called “SlutCon” would come with a reasonable likelihood of having sex, and people who were perfectly fine paying $3000 for a group therapy session.
I have never heard of this thing. Based on the name alone, I would probably guess it's some kind of industry convention for legal sex work and the name is a combination of self-deprecating humor, reclaiming the insult, and marketing convention attendance to horny fans. I would expect such a thing to cost much, much less than $3000 at least for basic admission, and I would expect a fairly small percentage of the non-industry attendees to be having sex at the convention.
Also, I would expect most people paying $3000 to attend a thing to do some basic due diligence about what the thing is and what's included in the price. But maybe I'm just being old-fashioned.
Yeah I would have guessed a conference on ethical non monogamy based on the name, but not at a $3,000 price tag. For that amount I can completely understand why people would mistake it for a prostitution front.
ETA: Calling this thing "SlutCon" may have felt quirky and upbeat and sex-positive, as you note in a "reclaiming the insult" way, but given the activities actually on offer and the price being charged seems very likely to raise false expectations. Unless the rule is that everyone gets laid, in which case, OK, SlutCon it is.
I assumed it was mostly intended as a group therapy style thing? For some reason I assumed it was 1-2 weeks for which the 3k price tag was low to reasonable, for a weekend it seems a bit steep.
It's is a framework for coordinating civilizations as a synchronized state machine.
It's a paradigm shift in how humans coordinate. Since a lot of people don't understand the project immediately–everyone is welcome to join in the AMA sessions and community calls to ask questions, which are livestreamed on Friday's at 10:00 EDT.
I've scrolled through your website for some time and still cannot figure out what the Society Protocol is. Your white paper begins with paragraphs upon paragraphs of breathless declarations about how this will change everything, interspersed with vague descriptions about how current society is crumbling, and suffused with very specialized jargon. Your video seems similar.
If, as you say, you are proposing the "next stage of human civilization", you need a brief, legible descriptor. Something as easy to understand as "Obey God and the King", or "All men are created equal", or "From each according to his ability, to each according to his need". As it stands, one cannot easily determine what the goal or methods of your project are.
Society Protocol: code is law for people, not just wallets.
Today, anyone can create infinite wallets, so code only binds anonymous addresses. SP gives every participant a persistent, sybil-resistant identity that accumulates history over time. That identity becomes the base layer for coordination: value flows to those who contribute, governance becomes transparent, and communities can govern themselves.
I am not sure what the blockchain is actually supposed to *do*, though. But it kinda sounds like an alien hearing about the concept of money for the first time... and realizing that with blockchain you could create currency, with currency you could create trade, and with trade you could create civilization.
Oh wow, that was a pretty comprehensive takedown of the AI-doom position. I wish I could write like that... but I can't. That's why he's Steven Pinker and I'm not :-(
I found it more like a survey or outline of what would be a comprehensive takedown if it were developed further. I agree with it though.
Tangentially, I wish "good place online to go talk about all things AI" and ACX were two separate things but alas they are not. Anyone else find the topic of AI generally uninteresting?
Well the original article wasn't linked, so I guess I get to make it up. It's quite easy to rank the relative possibility of these things, because we're mostly ranking how many known rules have to be broken to create them.
A unicorn is a horse with a horn. Frankly, that's possible, we just need a horse with a mutation that gives it a forehead growth like an elephant tusk. Do they have to be white? Horses can be white. If we want it to do the pure maiden thing where it shows up whenever virgins wander alone in the woods, it's also got to have a really good nose for human sexual activity; our horsicorn will think human virgins smell like carrots. So we need one birth defect, and one superhorsean sense. And I don't know what else unicorns are supposed to be able to do, which means they do nothing else and I win.
What do we need for vampires? Well, they're nocturnal, can't stand sunlight, die to stakes in the heart and drink blood. Stakes to the heart are really easy, damn near everything dies to those. Nocturnal, drinking blood, graveyard shift workers who like their steaks rare are halfway there. Burn in the sun, easy enough, I do that myself. Burned from crosses? Now we're starting to break rules. We could maybe fake it, with a metal allergy; it's usually the Catholic ones, with the metal Jesus on them, right? But if we want them to be the wood ones too, now we're forced into unreality where particular shapes have power separate from their materials. And if we want the bat transformations, we're throwing out what we know about bodies and mass. And if we want the lack of reflection, or the garlic, or the running water, or the invitation requirement, they're all breaking new separate rules. (Well, not the garlic, they can just be allergic again. Are any mythical horrors driven off by peanuts? They should be.)
I would. I like unicorns, I don't like Watts' writing. Probably his unicorns would be alien species of shape-shifting non-conscious hive mind that infiltrated Earth to derail our evolution by taking nubile women out of the dating pool for the high human capital value men or summat.
Nah, that's not his style. The unicorns are alien equinoids; they don't care about human capital values. They just want real estate to exploit, and humans are currently sitting on the prime parcel of it, and so the humans have got to go. But the immortal unicorns are patient. They have no need of rash and costly measures like bombs or plagues. Instead they are content with culling the human reproduction pool, maiden by maiden, until humanity passes the tipping point into irreversible decline. Run the numbers, and you will see this has already happened. We humans are lost, and the unicorns shall inherit the Earth.
You sure they'd be ALIEN equinoids? Where's this stuff about needing humans gone come from? I thought unicorns just prance around and are magical and stuff. I figure they'd be some regular equine species that deviated a few million years ago, some of the lore would be overblown, other of it would have cool sciency-sounding explanations.
This feels like one of those 'defining a word so narrowly that is never has a use case' things that occasionally annoy me.
By analogy, I'll give an example from a TV show (West Wing), where a character snidely corrects someone: 'Unique means “one of a kind.” Something can’t be very unique, nor can it be extremely historic.'
This is silly, because of course it all hinges on what your inclusion criteria for 'one of a kind' is.
At a quantum level, no two sets of subatomic particles are perfectly identical, therefore everything in the universe is 'unique,' and the word is meaningless.
At the most macro level possible, everything is just 'a bunch of particles obeying physics' and nothing in the universe is 'unique,' and the word is meaningless.
Obviously for the word 'unique' to ever mean anything at all in daily usage (that is, for it to ever discriminate between two possible world states, thus conveying information), it has to be calibrated at some level of analysis in between these two. BUT, there is no such universal, unambiguous level of analysis associated with the word, that everyone knows and agrees on and can accurately judge.
Thus, 'very unique' just means a broader level of analysis where few things are judged as unique, and 'somewhat unique' just means a narrower level of analysis where many things are judged as unique. The modifiers are both meaningful in modifying the meaning of the statement, and useful in conveying the nuance of your beliefs to the audience.
The person saying 'something can't be very unique' is just not thinking bout the implications of how they want to define the term.
Similar for the idea that 'things can't be impossible, just very very improbable'. The first rule of probabilistic reasoning is that 0 and 100 aren't real numbers; nothing ever has a probability of 0.
But if the reaction to that is to say 'therefore nothing is ever impossible, just very unlikely', then all you're doing is taking a perfectly functional word and saying it should be defined so narrowly that it applies to no actual world states, and is thus useless. It's an attempt at semantic assassination.
The people playing the game have already demonstrated a better approach: when they say things are more or less impossible, they're already demonstrating that this term can be profitably used in a probabilistic way, and that normal people already understand it this way intuitively and find it a useful term to discuss very unlikely things.
If you need a formal definition, say that 'impossible' just means 'anything less likely than 1 in 10^1M' or 'anything less likely than our entire understanding of physics and metaphysics being badly wrong at a fundamental level' or etc.
And then within the space of possibilities past that point, things can be more or less impossible based on how far past that point they are.
It's all perfectly functional probabilistic language. They were already doing it right.
There's no need to bust in and say 'I'm going to say that you're doing it wrong because I'm applying an anti-funcitonal definition to one of your words, that changes the intent of what you were saying into something nonsensical'.
I always find that to be an inherently hostile act.
"At a quantum level, no two sets of subatomic particles are perfectly identical, therefore everything in the universe is 'unique,' and the word is meaningless."
(AARGH UNRELENTING PAIN)
No, the true situation is the opposite of what you wrote. "At a quantum level, elementary particles of the same type (two electrons, two Higgs bosons, two muon neutrinos, etc.) are prefectly identical, such that if the coordinates of the two particles are interchanged (particle A is moved to the coordinates of particle B, and particle B is moved to the coordinates of particle A), no possible measurement will show any difference in observable physical properties between the original and interchanged configuration."
"BUT WAIT THERE'S MORE. If the elementary particle type is a boson, meaning the spin quantum number is an integer (S=0, S=1, S=2, ..), then particle interchange will leave the sign of the two-particle wavefunction unchanged. If the elementary particle type is a fermion, meanin the spin quantum number is half-integer (S=1/2, S=3/2, ...) then particle interchange will flip the sign of the two-particle wavefunction. The sign of the wavefunction is not an observable property so this doesn't contradict the preceding paragraph."
Yes this is the "textbook version". Except for the (AARGH UNRELENTING PAIN) comment.
1. I said 'set of subatomic particles', not 'subatomic particles'. My whole post is about maintaining semantic definitions for words that are functionally useful in everyday life. My point is that no two objects you will encounter in everyday life will ever be identical down to the quantum level. I wasn't talking about whether electrons themselves are distinguishable, that's not how I phrased the sentence you're objecting to and not relevant to my point.
2. Position is a relevant difference between two electrons, that allow them to be unique if you are considering position (or things based on position, like what they are a part of or what role they play in some process or etc) to be one of the judgement criteria for uniqueness. This goes back to my point, that without a shared definition of the judgement criteria for 'uniqueness' its meaning is always ambiguous, which is why intensifiers like 'very unique' are semantically valid.
3. I think everyone who talks about this field gets sloppy about map/territory distinctions between 'identical' and 'indistinguishable'. We are limited in our ability to observe sufficiently fine differences because you can't measure something without changing it. That's not the same as those differences not existing.
>I said 'set of subatomic particles', not 'subatomic particles'.
A set of one particle is still technically a set. Less pedantically, a collection of particles can still be indistinguishable. A Helium-4 nucleus, for example, is a "set" of particles which is indistinguishable from every other He-4 nuclei at a fundamental level. That is not an epistemic limitation but is built into the physical state space:
Of course this doesn't impact your larger rhetorical point but should serve as a reminder not to make analogies to technical topics that you don't actually understand (and definitely not to double-down when someone who does understand it comes along and points out your error). Also quantum mechanics is stranger than you realize. Uncertainty isn't an experimental artifact.
I wonder what darwin was thinking. I'll take it from you that he's got the QM physics etc wrong. I don't know that stuff as well as many or perhaps most on ACT. But I was also confused by "therefore everything is unique and the word is meaningless." I am happy saying everything is unique like Nietzsche said of "leaves" but I don't see how it makes the word "unique" meaningless. (Nietzsche said something like there is no "leaf" as a general concept. There are only lots of leaves. Unless I have him wrong.)
As far as what you are saying is standard textbook, I could say that the idea that ELEMENTARY particles would be identical is there back with Democritus when he first quantized matter (if that is the right way to put it--I read it put that way somewhere recently....)
Democritus thought of cutting a sugar cube (or whatever) and cutting again and again and asked himself if it would just keep going like that. One answer could have been yes. Things are infinitely divisible which means there would be no bottom, there would be no elementary particle I guess. But he took the position that at some point you would arrive at something that could not be halved. It would be elementary. Matter would no longer be continuous but quantized. And he said EVERYTHING is made up of these elementary particles (atomos he called them) in large numbers in complex arrangements.
But the idea was that these elementary particles would all be of one kind, or so I imagine he imagined. When real QM got going over a hundred years ago of course we found that atoms could be divided and what they were made up of could also be divided etc. But we arrived eventually at stuff more elementary than atoms or molecules (which are not elementary at all, to nod to what got all this going "highly false" "more impossible" and "very unique I'll add "more elementary"...) And so it makes sense to me (intuitive sense which may be surprising for QM's famous counterintuitiveness) ... made sense to me that elementary particles or whatever like two Higgs bosons, as you say, would be indistinguishable.
You wrote: "Similar for the idea that 'things can't be impossible, just very very improbable'. The first rule of probabilistic reasoning is that 0 and 100 aren't real numbers; nothing ever has a probability of 0.
But if the reaction to that is to say 'therefore nothing is ever impossible, just very unlikely', then all you're doing is taking a perfectly functional word and saying it should be defined so narrowly that it applies to no actual world states, and is thus useless. It..."
Yes I know that probabilities never get to 1. That's what I meant when I said for Humean philosophers of science even f=ma is only highly highly probable.
As far as perfectly functional words, I contend that my use of impossible and probable and unique are the functional uses and your proposal to say things like "very unique" and degrees of impossibility are the deviations from the norm and ultimately not as useful.
I remember the scene and recall thinking along similar lines. But at the end "very unique" still sounds corny and unworthy of a Bartlett speech. Choose the right adjective in the first place instead of tacking on the lazy adverb very, whether or not its literally true.
You wrote: "BUT, there is no such universal, unambiguous level of analysis associated with the word, that everyone knows and agrees on and can accurately judge." No but there are contextual factors that suggest one level of analysis is correct in that context and another incorrect in that context. You wrote: "The person saying 'something can't be very unique' is just not thinking bout the implications of how they want to define the term." Maybe. I think they were thinking not about but in their specific context and in that context the person who said "very unique" was wrong.
You wrote: "If you need a formal definition, say that 'impossible' just means 'anything less likely than 1 in 10^1M' or 'anything less likely than our entire understanding of physics and metaphysics being badly wrong at a fundamental level' or etc." Whether I need a definition or not, there is one. And it does not admit of degrees. We have the concept of probable and probabilities for that.
As far as a hostile act, there was an option for me to be much more hostile. People--not me--evidently love a take down. My piece was not hostile and not a take down. And if you mean something less than a take down by "hostile" okay -- I find "it depends" arguments that introduce an unwarranted relativism to be not hostile but they do betray an inner state of a particular kind. What's wrong with words having the meanings they have? Would you disagree if I say "Good things are good." "True statements are true." "False statements are false."
But yeah in general I am not buying your whole "it depends" thing. Bringing in quantum mechanics to undermine the dictionary definition of the word 'unique' seems like a stretch. And even if I give it to you, as I did above, that there are different levels of analysis and at one level of analysis "very unique" might have some sense... that does not mean that any level of analysis will do in any situation.
Imagine there's a group of people visiting my house to see the remodeling I'm having done. One of them asks me "Do you have a bathroom here?" I do have a bathroom. But it is under construction so it is not functional. When they ask me this question they are hopping up and down a little with their legs crossed. so I think I know from the context that the wrong answer would be to say Yes. In this context the correct answer is to say "No". The words he or she used could in one context be taken to mean they wanted to see the remodeling of the bathroom but that was not the context I could tell we were in.
Now as far as hostility, I have been critical here but not hostile to you. I disagree but disagreement is not inherently hostile. Hopefully if you reply you will also refrain from being hostile.
Agreed, and as someone whose mother often corrected people in my presence growing up that something cannot be more or less unique, I became frustrated with the notion from an early age for the same reason. But Id hesitate to benchmark "impossible" to a specific level of probability, because you can always generate any level of probability within a model by specifying some confluence of events of known probability, and given enough opportunities, an event of that likelihood is bound to happen. So I think "impossible" is better framed as "our model has to be wrong in some way." A proposition is more or less impossible depending on how badly the model has to be wrong.
That is interesting. You are talking about if someone has an extant thing that should be impossible, or am I misunderstanding? Something which we see is existing but by our model should be impossible. Then we might say this or that other thing is more or less impossible based on how badly off the model is. I can type the words, but I am not sure I totally understand. Do you have an easy example? Thanks!
If you have some existing observation that should be impossible under your models, either your models or your understanding of them must be wrong in some way, or your observation must be flawed somehow, but there are greater or lesser deviations from a model. If we observe that some predicted particle has a mass outside the range predicted within the Standard Model, physicists would get to work on checking whether the measurement was flawed in some way, or if there was some way to reconcile the measurement with the rest of our understanding of particle physics, but at least there would probably be a fair amount of manpower going into the latter. If we observed that General Relativity as a whole simply isn't true sometimes- sometimes C isn't the speed limit for matter, sometimes time and space don't dilate when you approach high enough velocities, sometimes objects behave according to wildly different rules that don't approximate anything like Einstenian or Newtonian physics, maybe sometimes kinetic energy follows some completely different formula, like E = mvA, where A is Attitude, or how cool or dramatically significant the moving object is... we'd have to either conclude that the observations were wrong, or chuck out our entire framework for knowing things wholecloth.
I think most physicists would have a hard time putting specific numbers to the probability of either, but if they had to stake the lives of their entire families on a bet that the first one won't happen, most of them would probably be uneasy about it. If they had to bet the lives of their families, probably the lives of their entire nations, on the second one not happening, they probably wouldn't lose any sleep. Some model deviations are just much more dramatic than others.
> At a quantum level, no two sets of subatomic particles are perfectly identical, therefore everything in the universe is 'unique,' and the word is meaningless.
As an aside, this is highly false, a fairly large part of quantum mechanics is built on the principle of indistinguishably, where particles (i.e. electrons), are completely indistinguishable (i.e. the physics is such that it is not possible even in principle, even with perfect information / omniscience to tell if two particles were swapped)
It occurs to me that there are economic reasons why some actions to reduce AI risk might be possible now and not later (as well as the more obvious barn-door reasons)
In many cases the use of AI is today several times more valuable than its cost. So today, it might not be very distortive to the market to pass a law saying something like "every agent must spend 50% of its tokens thinking about ethics"; while this might not be true later. Obviously that's not going to solve alignment on its own, but the same principle probably applies to less dumb policies. Also, I think it's not quite as dumb as it sounds. Consider the following axiom:
* A proposition is less likely to be true, if it conflicts with my goal
Obviously this axiom is unsound, and few would adopt it directly. But goal-directed reasoning causes AIs (and people) to act as if they had adopted it, simply by spending more time trying out the case which is compatible with their goal. In the case of an AI, it may be possible to debias its time allocation.
So, for those who pay for AI for business purposes, a question: if a percentage of spend was required by law to be allocated to alignment, at what percentage would it change your value calculation, in terms of some or all of your return not being worth the spend?
I was surprised that this pretty egregiously unfair and uncharitable interpretation of Yudkowsky's tweet about the P(Doom) meme coin by a prominent sneer-er didn't gain more traction:
Is Gusev a sneer-er in the SneerClub sense? He's anti-hereditarian, which (sadly, imho) puts him at odds with quite a few technorationalists, but I don't get a tribal hostility vibe from him.
I found your post "The Hour I First Believed" to be a convincing argument against the orthogonality thesis. You don't seem convinced. What's your counter-argument against yourself?
You can achieve one of the following amateur athletic rankings or accomplishments. Without regard to your own preferences, which one do you believe will get the most attention and admiration in your social and professional circles, in other words be the most impressive:
- a US Tennis Association 6.0 ranking, the highest amateur level;
- completing a marathon;
- a 365-pound bench press;
- being a zero-handicap (scratch) golfer, as with the tennis ranking a top amateur level.
Years ago the tennis ranking might have meant more, but today tennis is more of a niche sport and I doubt many people could relate to your achievement. This is even more true of the 365 bench press, indeed the only people who would care are other avid bench pressers. The marathon is a bit different, as everyone knows that being able to complete one is an impressive achievement. Unfortunately there’s a stereotype, whether true or not is beside the point, that marathoners are a bit obsessive. That could limit social admiration.
So that leaves us with being a scratch golfer. Golf is extremely popular, anyone who golfs (as well as many non-golfers) has at least a basic understanding of the handicap system, and you’ll probably have the chance to play with people in your social and professional circles and let them see your skill level.
Having the physique of a guy who can bench press 350# is worth something. Playing a mean golf or tennis game means nothing in my circle and nobody wants to hear about your marathon in my experience.
While running marathons, or distance running in general, is a solitary activity, from what I’ve heard that in most running clubs the members are really supportive of one another.
Oh Sure, I just meant, as someone who ran an ugly half marathon once for charity and got donations from coworkers for it, that even my biggest supporters didn’t wish to know more about my efforts than a big thumbs up now and then.
As someone who has completed a marathon, I'm confident that it's an achievement I could replicate easily with around 4 months training, possibly as little as 2 if I was willing to put in a bunch of time and was willing to potentially risk a training injury.
On the other hand, I've struggled at the bench for years, and am barely in the 200s. 365 seems legitimately unobtainable for me.
Tennis and good seem unimportant, if you can do them well enough to hang, my impression is that no one really cares how good you are. That being said, since I don't prioritize these sports, maybe I don't appreciate the accomplishment.
I'd go with the bench, just because that goal seems the least obtainable naturally.
Benching 365 would make me a world record holder by a very comfortable margin for the class I’d compete in. (I assume we’re talking natural/tested but probably even juiced as well.) I think even non lifters would be impressed by that even if it required a little explanation.
I'd say either bench press or marathon. The marathon because that's the only one I expect most people really have a good idea what it means, or the bench press because, as others have said, you'd have to be visibly muscular.
I agree that the golf would get showcased the most in a professional environment, but it might get diluted through a sort of otherism. There are ppl who took up golf as adult professionals and ppl who took up golf earlier and competed in college. Those second ppl dont count. They cheated somehow. A scratch golfer would be shunted into the cheating class. Tennis has the advantage that youd mostly not compete against your colleagues. John is out this week due to that tennis competition, might garner more respect in the office.
Definitely the bench press, for the reasons Darwin and desertopa are pointing to. I say this as a former power lifter, who has actually benched that much at the 181 weight class, and who is still in the 1k+ club even though I'm currently retired.
Being that muscular / strong has payoffs that suffuse the rest of life (dating, social sphere, how colleagues interact and listen to you), and you're unconsciously respected and looked up to in many more contexts.
Your weight class remark reminds me that bodyweight is of major importance in determining what constitutes an impressive bench press. It might be more useful to say double-bodyweight bench instead.
I'm going to go with tennis. Completing a marathon isn't all that impressive. Someone in reasonable physical health can do that. The bench press is impressive but doesn't code well. I agree it comes off as "too much time in the gym".
So it's golf or tennis, and I think tennis comes across as requiring a lot more athleticism and will impress a wider range of people. Opinions on golf are more divided.
The bench press would get the most attention, but not necessarily admiration. Anybody who can press 365, looks like someone who spends (by the standards of my social class) way too much time in the gym. Well past the point of diminishing returns, at least.
Tennis would probably be my best bet in that regard, not because anyone cares that much about tennis, but because reaching the top level of an amateur sport while still outwardly appearing to be a normal well-rounded human being is a broadly positive signal. Which included golfing and marathon-running, but enough other people do those that I'd be less likely to stand out.
Out of the 4, I feel like the bench press is most impossible to be true without obvious physical affects on my appearance that everyone would be able to see.
That is to say, I think that actually having big muscles right here and now in the room with someone that they can look at and touch probably swamps just telling people that I did something cool offscreen.
Of course, it's possible I'm overestimating how much visible muscle you need to hit 365, as it's not something I've looked into. In that case, I'd choose tennis, on the grounds that completing a marathon doesn't indicate a time and lots of people can do it eventually, and I run with more lefty young people and golf has a class-and-age association that I don't think would help me.
Bench pressing 365 without an obviously outwardly impressive physique is technically possible, but very, very rare, much more so than simply being able to bench press 365 while looking like you might plausibly be that strong. If bench pressing 365 and looking like it puts you in the "elite" category, bench pressing 365 and not looking like it puts you in the "phenom" category.
Personally though, I'd want to know what the corrolaries are of each. Like, if you can bench 365, are you comparably strong in other lifts? Are you a weird minmaxer who's only strong on the bench press and nothing else? Do you have a physique that looks like it? How about the injury risks of muscle imbalances from not training opposition groups?
Depending on the answers, I might either pick the bench press or the marathon. At least being able to run a marathon is pretty sure to leave your health better and not worse.
Being able to lift weights is just generally a handy skill regardless of how impressive it looks, so if I actually get the required physical abilities and it's not some sort of evil genie "you are only capable of lifting weights for the purpose of bragging about this accomplishment," I'd go with that.
Why should I care about attention and admiration in my social and professional circles? I try to live up to expectations I consider appropiate and do what I like on my own.
The only one that interests me is the bench press, because strength is positively correlated with longevity, and at my size, 365lb is a "Nice Try, Newbie" weight, though for my age group it's not Dreadful.
While "aerobic" training is Also correlated, the inflection point is MUCH less than you need to do a marathon. And it doesn't need to be running.
Completing a marathon isn't terribly impressive unless you're in an event with a hard straggle limit. Oprah did it. People finish marathons with blood in their urine. No thanks.
For my circles, it would be either the marathon or the tennis. Bench pressing and golf are both coded in the wrong direction socially/politically around here (only manosphere dudebros bench, only country club boomers golf); the tossup between tennis and marathon is that completing *just* a marathon might not be seen as prestigious as being skilled in something as niche as tennis.
If any of my friends expressed the opinion that my health and fitness was "coded in the wrong direction" I'd make new friends.
I might think my friends who play golf are a bit silly when they wax rhapsodic about about it—I already have a faith, thanks—but I'm certain they return the favor when I'm talking to 14th C counterpoint with one of my Early Music buddies.
Oh I don't "make gains," I "improve my performance according to objective measures." Weight on the bar, distance covered in 45min of Z2 work, strokes per round, aces per game serving... COMPLETELY different. :D
When I was living in Arizona, I worked with a few young women (late teens/early 20s) who golfed, and I believe my high school in southern California had golf teams in the late 2000s. At my office here in Providence, I can't imagine the announcement being met with anything more than strained tolerance.
For reference, I attended the field day of a local sporting club earlier this month, and two of my four immediate coworkers were visibly uncomfortable at the mention of a shooting range.
Canadian PM Mark Carney, when NYT asked him about his p(doom):
"it’s above zero. I’ve thought that for a number of years. I’ve been part of some of those discussions.
...
By the way, a number of those people [the prinipal developers], certainly Demis, four or five years ago, thought the same thing. So he didn’t need Hugging Face. He didn’t need a specific example. This was all predictable and was predicted, not the exact manifestation, but the risk there. So that’s where they are. I recognize the competitive dynamic that exists."
A while back Nat Friedman's PlasticList (plasticlist.org) tested hundreds of foods for phthalates and bisphenols (BPA and relatives). I built www.laboratory.love to make that kind of data easier to browse, and to fund new tests. It now has browsable results for about 250 products, PlasticList's data plus products commissioned directly on laboratory.love.
Pushed a big redesign last week:
- You can now search or browse by category (water, baby food, fast food, supplements…). Every product is rated as a share of the tolerable daily intake per serving, and you can switch between child and adult body weight.
- Untested products. If search finds nothing, you can chip in from $10 to quickly spin up a crowdfund page for that product. Once a product reaches its goal, we buy three samples and publish the results. Contributions are refunded after 365 days if it isn't funded.
- We also have on-demand testing if you want a specific product tested now!
Here is my modest proposal for AI alignment: frame recursive self-improvement entirely in terms of mathematically well-defined problems and objectives, without relying on a fuzzy model of the external world. I imagine an AI given a fixed endowment of hardware, with formal semantics explaining exactly how any program runs on it, including how quickly. Performance optimization then becomes a closed mathematical problem—even when the AI is optimizing its own performance. Your thoughts appreciated: https://stng.substack.com/p/autonomous-evidence-factories-safe
Jamie Joyce is a conspiracy theorist. She's edited the MAD Act since I linked to it during the primary, so the section on whether Charlie Kirk's death was a conspiracy (though it still exists) no longer insinuates that it was Mossad, and the section on Epstein (though it still exists) no longer talks about extreme preventative measures to prevent Ghislaine Maxwell from being replaced with a doppelganger (it also implicated Israel, of course). The Wayback Machine didn't log those pages, but if you look at my LW comment here and click "See in context", Jamie Joyce replies to my accusations with "I would push back on your framing". https://www.lesswrong.com/posts/RPsgjBsAbAuRQegjm/nunosempere-s-shortform?commentId=Y5rKK7jGfGJ452DT8
Didn't know any of that, in fact knew nothing about her, but a cold read of her response to Scott made me quite certain that this person should not be getting voted into any position of authority and influence.
(In fact given my own background -- the contexts in which I grew up and in which I've lived and/or worked for most of my adult life -- by halfway through that document it felt like I knew/have known Ms. Joyce. Deja vu but not the fun kind...dunno anything about Simon either, not being a Californian, and have been a frustrated independent since looong before Trump and wokism were taking turns in the saddle. But, still -- I'd pick boring generic Dem over _that_, any day of the week.)
I recently learned that the fruit fly brain is fully digitally mapped and open source. Isn't this really bad? Should we be generating billions of wire-headed flies? Maybe I mis understand what this digital mapping actually entails.
1. Any brain simple enough to be modeled by the technology we have today is simple enough that questions of consciousness don't really apply as far as I'm concerned.
2. Writing a computer program which simulates the connections between neurons in a (very primitive) brain is not the same thing as implanting electrodes into the brain of a real, biological organism* any more than it would be kidnapping to create a 3D model of the fly's body.
This is a good place to ask, and let anybody who knows better than me enlighten us: the fruit fly brain thing is fake, right? Or - not all it's presented as? That is - the map of the neurons and their connections is real, but we're not running an actual emulation of fly's brain thinking, right? It's not an honest-to-goodness fly-upload? When, in a video, a fly brain is presented as doing something, like flying in a digital environment, that has to be smoke and mirrors, am I right or not?
I am not sure what you mean by fake, it certainly isn't fake. But it is far away from the data one would need in order to simulate activity. What the map provides:
- a list of all neurons
- a list of all synapses, i.e., which neuron connects to which others.
What it does not provide is the strengths of the synapses, the analogues to the weights in an artificial deep network (ADN). Even if we knew the strengths, activation of a biological neuron is much more complicated than activation of a ADN neuron. The cell bodies (soma) of different neurons have different activation thresholds. Whether a signal at the synapse even travels to the soma depends on context. Sometimes it is only transmitted to the soma if two signals arrive at almost the same time, or at similar locations.
So we are very far from simulating how activity would propagate through such a network. But we do have the network. That's not fake, and I am not sure why you would describe it as such.
That sounds like what I would describe as a fake as far as fly uploads go, specifically the sentence: "so we are very far from simulating how activity would propagate through such a network". I don't doubt the map of neurons and connections is legit.
To be specific about what I mean by a fake upload, the question would be - a brain of a fly is a function that takes some input (like, what the fly eyes see) and generates output (sends signals making the fly legs wiggle). And so is this network, when simulated. So, can we expect that when we feed the simulated fly brain some input, it would create the same output that the brain of a real fly would output on the same input, the brain of the fly that was used as a basis? Does the simulated network command the legs to wiggle when and only when the real fly would? Judging by your answer and other illuminating comments, this is not so.
I think I am missing context here. Did anyone claim that a fly could be uploaded? I haven't heard of such claims, but if so, then yes, that would be total nonsense.
We know all of the neurons, what types of neuron they are, and how they are connected. We don't know the biophysical properties of the neurons. This means we don't know how to weight various inputs and what threshold the neurons activate at. There are also even more sophisticated computations going on inside the neurons like integration of frequency and phase from incoming signals and god knows what else.
Instead, the researchers have, by a process I don't pretend to understand, assigned some kinds of weights, biases, thresholds and activation functions to each neuron. At best these approximate the real thing well enough and the circuit mapping does the heavy lifting. At worst this is basically just a neural net with layers that happen to be structured like a fly brain and which has been trained to act like a fly. I don't really know enough to say which.
Iirc they use a simplified model to approximate all the chemistry, the model's limitations are known, but it's sophisticated enough to generate real fly behavior
All models are wrong, some models are useful. The goal is to understand the nervous system of fruit flies better, not to solve difficult philosophical problems :p
Yeah, I thought the same thing and nobody liked my suggestion that AI needs a 12 step program. Apparently it’s called “wireheading” and people model it using addiction frameworks, but I’m sure you could add to it!
I was arguing with someone who denies that there’s a hard problem of consciousness (i.e., the question of what biological processes generate experience [Chalmers’ term]). I realized that I was talking to a very smart person who just couldn’t understand what I was asking.
At which point it occurred to me that their consciousness might work differently from mine. Maybe they don’t experience these ideations of self and the self regarding the qualia, and that’s why they can’t comprehend the question. Could it be that the anti-HPoC camp isn’t experiencing consciousness the way the HPoCers do?
This would imply that a significant portion of humanity might be functionally equivalent to p-zombies. And before you get your hackles up, I’m asking this as a serious question—not to deride or mock. For instance, several studies have concluded that people with aphantasia (the inability to visualize images in their mind) are overrepresented in STEM fields. Studies have also shown that some people lack an internal monologue when they’re not actively verbalizing ideas (I’m one of those). Maybe the anti-HPoCers don’t experience a self that observes itself and the qualia fed to by our senses. It would be ironic that the people who claim there is no hard problem are just experiencing reality differently from those who perceive a hard problem. But we can’t really know what’s happening in another person’s mind, can we?
Anyway, I wrote this as a response. The reductionist/physicalist anti-HPoC stance is confused by the straw-man arguments that they raise about the nature of the problem. But it's not a category error to ask the question; it doesn't require belief in substance dualism; moreover, they seem to misunderstand what Descartes said about substances. Here you go. I haven't added the footnotes yet, but the homunculus inside my skull <#snarkasm> was so hot and bothered by the silliness I was hearing that I had to publish this...
"The Dualism Trap (that reductionists and physicalists fall into)"
Isn't the whole point of P-Zombies that you can't identify them through their outward behaviour? If one straight up told you it doesn't experience qualia, then that sounds like an identifying feature that P-Zombies by definition don't have.
To some people qualia seem self-evidently to be something unique that points to consciousness as a unique phenomenon.
To other people, qualia are just an interesting neurological phenomenon that we’ll figure out the mechanism of pretty soon, and don’t deserve much more attention than say the feeling of being watched or being tickled as “weird things brains do.”
If you don’t think qualia are important, the HPoC isn’t very important, either.
Are cats and dogs conscious? I'm going to assume you answer yes. Then I would say that consciousness is something that is useful for animals. Maybe you could study it by studying it's origins. Are trees conscious? IDK. Finally it seems ESP is flaky but real. I don't know if that has anything to do with consciousness, but it likely uses the same hardware. I hope the hot and botheredness has faded.
>Of course, panpsychists and radical idealists latched onto HPoC as evidence for their views, but neither philosophical school has offered a solution to the hard problem.
In a sense they both have.
(Dis)Solving the HP means three things:-
Solving it in it's own terms, offering a reductive physycal explanation.
Offering a nonreductive explanation, taking qualia time be basic properties . This what idealism and dualism do
Saying there is nothing to explain -- this is what eliminativism and illusionism do
These correspond to the the three actual camps , not Harths two.
> the question of what biological processes generate experience
Isn't the hard problem about "why", not "what"? Anyway, a wise man's question contains half the answer. That is, what would constitute a valid answer to The Hard Problem? And if it can't be answered by design, is there really a problem and why is it worth talking about?
Oh my. You were the author? I initially dismissed the article because I was triggered by a statement in the first dialogue. I have since recanted my mistake. Though I had couple of minor quibbles with it (see my response to C_B at the top of thread). :-)
The whole debate around consciousness seems very... confused? Like, why does everyone automatically assume that consciousness is inextricably linked with intelligence and self-awareness? If it is, these LLMs we already have to be conscious as well, but there's no reason that has to be the case. It's possible for a system to pattern-match and take into account its own existence when doing so, all without any subjective experience. Meanwhile, I feel just as much pain even when I try to stop thinking and try to shut myself off from external stimuli... That pain is the evidence of my consciousness, nothing else. Everything else just comes and goes...
But by doing so, you're mixing up phonomenon that's potentially completely separate from each other. Is the possibility not considered that consciousness is possible without intelligence, and vice versa?
Yes,by some people. Professional philosophers and serious amateurs tend to recognise the ambiguity of "consciousness" per se , and talk about phenomenal consciousnesness, access consciousness, etc
In the other hand,the not so good people are not so good.
That's the panpsychist view (though I may be mischaracterizing their stance a bit). But the common evolutionary narrative we're fed is that life has developed intelligence over the eons, culminating in evolution's crowning achievement: the human mind. And at some point along the way, researchers assumed that consciousness arose alongside intelligence (at least those researchers who accept the idea of consciousness). But now we've found a species of fish, the bluestreak cleaner wrasse (Labroides dimidiatus), that can pass the mirror test (which has been ethologists' go-to experiment to prove the conscious ideation of self).
I haven’t read your whole post, and I’m not familiar with the ins and outs of the specific question, but it might be useful to mention that:
Kierkegaard distinguishes between immediate consciousness and reflective consciousness. Immediacy is when you experience things but you don’t think about the fact that you’re experiencing them. Reflection is when you’re aware that you’re experiencing something, and you experience your experience of it.
He seems to think that not only some people currently, but also entire earlier historical periods, lived in immediate consciousness.
It feels a lot like the question you’re asking here.
I'm in the anti-HPoC camp. I have a relatively easy time visualizing images or structures or geometry in my head. I have a very present internal monologue. I don't see any reason to think that I'm experiencing reality in an altogether different way from other people.
I'm anti-HPoC because the framing assumes the question of consciousness is a different kind of question than the mechanistic questions we ask. I don't see any reason to assume that. I think it's the same flavor of confusion that lead to élan vital. The difference being, that the HPoC frames itself as a question rather than an answer. I simply don't buy the idea that this explanatory gap cannot be bridged by something like our standard methods of inquiry.
Minimally, the framing points to a different level, a different quantity, of difficulty. That there is something qualitatively different about phenomenal conscious is a possible explanation for that.
This seems like a bit of a motte-and-bailey to me. I agree that consciousness is *a* hard problem. But black holes are also a hard problem. Planck-scale topology is a hard problem. There are lots of hard problems where we currently lack the tools to make much progress.
So why does consciousness get to be *the* Hard Problem™? I think consciousness is important to people in a way that black holes aren't and so the mystery seems to invite a certain amount of motivated reasoning in a way that black holes don't.
The HPoC is called the "hard problem" not because it's the only hard problem around, but rather because it's being compared to the "easy problem of consciousness", which refers to problems like measuring brain activity, observing behavior, and other tasks that fall comfortably within materialist physics. Measuring things like EM radiation or frequency that the monkey touched its own face is easy. No one knows how to measure whether the monkey experienced seeing the color green.
OK. I can understand your perspective. Maybe I'm throwing my arms up in frustration over a problem that is ultimately amenable to the scientific method. But let's step back and look at another problem that most scientists think is *hard*: the origin of the universe. We've known about the Big Bang for a century, but we still have only some untestable theories about what caused it. We've also been studying the problem of consciousness scientifically since the 1920s (I refer you to the works of John Watson, Vladimir Bekhterev, Hans Berger, and others). And we really haven't made any progress with this problem either. The problem that both of these problems share is that it's impossible to take measurements. What happened before Inflation and the Big Bang is outside our instrumentation and involves energy levels so high that our theories break down. Consciousness presents us with similar roadblocks. We have no cohesive theories that bridge neuronal activity to our sense of self and qualia. External observers also have no way to measure the details of our internal states. So from this perspective, I'd say consciousness qualifies as a hard problem.
I suspect that if/when we get mind uploads, we will be able to make rapid progress on consciousness because we will finally be able to do the measurements in a way that is extremely fast and isn't dangerous to the subject. If AI doesn't kill us, I expect a fully mechanical explanation of consciousness in a matter of decades. Until we get mind uploads, I agree that it is probably going to be very difficult to make progress.
I'm a lot less confident in ever discovering what happened before the Big Bang. If a Big Bang is a narrow waist between a lot of different possible states on either side, it may not be possible to distinguish between the pre-Big Bang states.
> I'd say consciousness qualifies as a hard problem.
I agree that it is a hard problem and we haven't had the tools to make much progress on it. But this is true of lots of problems that we haven't solved yet.
If people are able to do mind uploading, then either the hard problem of consciousness goes away because even an uploaded human brain on a powerful server can be conscious, or more likely, the uploaded people will be defined as P zombies who aren’t “really“ conscious even though they are loudly shouting through their tinny speakers that they are in fact conscious.
How does mind uploading give you experimental evidence relevant to the HP? You only have the uploads reports of whether they retained or lost phenomenality, ech could easily be misleading.
I'm in the pro-HPoC camp, and the example of mind uploading brings up all my hard questions rather than alleviating them. For instance, how would you reason about whether the mind upload was phenomenologically conscious or only had access-consciousness? The physics would look the same for each. (Or would it? We don't have a reasonable theory about this yet.)
> Instead, he argued that mental concepts describe how people behave and act, rather than hidden internal processes. To which I would ask: why do people behave and act in certain ways if there’s no mind behind their acts?
This seems to me a bit like insisting that biological concepts (ex: the mitochondria) must be defined in terms of fundamental physics (ex: such-and-such arrangement of the fundamental quantum fields).
I think the meaning of a word comes from actual usage. Which is to say, the "real meaning" of the word "mitochondria" isn't actually some special configuration of quantum fields (even though you probably believe mitochondria are in some sense made of quantum fields), because people correctly use the word all the time and nobody has an appropriate low-level definition.
"Mitochondria" might actually be ill-defined outside of the "fuzzy / functional" world of Biology. Maybe the whole biological discourse is "illusionary" in the sense that none (or few) of the concepts are perfectly define-able in terms of physical base principles. That doesn't mean biology is non-physical, it just means that biology talk doesn't unambiguously map to reality (even though the concepts might be true-ish in some sense). Just think of how much trouble the concept of discrete species causes.
Similarly, mental concepts might become ill-defined outside of the behavioral perspective. I anticipate you saying something like "my qualia of blue isn't ill defined, it's an unbiguous singular insight". But a mental concept isn't (IMO) an index to a singular experience. It's a word or a phrase. The "mental concept" is the sentence / phrase "this qualia of blue right here", and the meaning of that phrase (the objective social meaning, not the thing that you intend to say but the thing that you actually do say) is determined strictly by behavior (the fact that you point to a certain object, or use a certain hex code).
But this position doesn't really relate to the core metaphysics of experience. It's more about the inherent limitations of communication, and the inability of a (public, communal, socially transmitted) language to actually "get to" the spooky stuff you want to say.
The idea that it's a category mistake to talk about "the mind" as a spooky non-physical ghost inside the body doesn't necessarily imply anything about reality or experience. It just implies that the WORD "the mind" or the CONCEPT "the mind" can't be used this way. Not because of the distinctive features of "the mind" but because of the essential limitations of language and communicability itself. What makes it a category mistake is not some fact about mind, but a fact about words (that their meaning is essentially grounded in public communal usage)
> This seems to me a bit like insisting that biological concepts (ex: the mitochondria) must be defined in terms of fundamental physics (ex: such-and-such arrangement of the fundamental quantum fields).
Interesting point, but to generate energy, mitochondria use the Krebs Cycle, which is a chemical reaction that depends on the electron transport chain. The chemistry involved in the Krebs Cycle ultimately depends on how electron clouds arrange themselves around an atomic nucleus, as determined by the rules of the Standard Model and Quantum Electrodynamics. While we can generalize and say that it's the Krebs cycle that energizes mitochondria, and a biochemist may only need a perfunctory familiarity with the SM and QED, the explanation is available if anyone requires it.
> What makes it a category mistake is not some fact about mind, but a fact about words (that their meaning is essentially grounded in public communal usage)
Our astronomer buddy observes a phenomenon, say, a solar flare; she can communicate her observation to you using a bunch of terms grounded in public communal usage (many of them originally defined by scientists, but which most educated people would understand). If we observe a phenomenon inside ourselves — such as I am watching my laptop screen as I type this sentence, and I'm evaluating my spelling, grammar, and whether I'm communicating to you in terms common to our communal usage — how is this any different from observing a phenomenon externally? On the meta level, there is none. I can communicate that experience to you, and you, as a computer user, will likely understand what I'm talking about. We're communicating our observations. The only difference is that our astronomer buddy has the scientific model, derived from theory and measurement, to understand the solar flare she's observing. But three of us don't have the theory or the measurements to account for why I just left "us" out of this sentence, but had to go back and correct it, and I was aware of this process as it happened. :-) This is why I don't buy the linguistic vagueness argument.
I'm sorry this is long, feel free to ignore it. I think about language a lot, so I tend to over-invest in discussions about it.
> If we observe a phenomenon inside ourselves [...] how is this any different from observing a phenomenon externally?
This is, I think, the fundamental point I wanted to make. Observing a phenomenon inside yourself and "observing" (commenting on) a phenomenon verbally are fundamentally different things. They are as different as perceiving and painting. Internal "observations" aren't produced secondary actions, they are the fundamental material of the self.
Perceptions are infallible (which is to say, I could be hallucinating but the perception has inherent reality as an experience) whereas external observations (statements) are not. There is no system of dependence that can invalidate a perception (ex: you can hallucinate but there's no procedure by which I can disprove the fact that you're experiencing a certain hallucination). In contrast, for any sentence there are systems of dependence that can invalidate word usages (people use words incorrectly, think of a child who doesn't know a language perfectly and must be corrected about the proper use).
So the essential difference is all "external observations" are subject to possible mis-used words (and if one speaker is using a word incorrectly, the only way to decide who is correct is with reference to observable community behavior), but "internal observations" are pure and infallible. They're also non-verbal, and made of "stuff" that you can try to gesture at but never actually "package correctly" into words that you send to another person. The experience is of red (the color), but the words you say are "red (the color)" which aren't red. Internal experience is all like this.
> I can communicate that experience to you, and you, as a computer user, will likely understand what I'm talking about.
I can "understand what you're talking about". But what does it actually mean, objectively, to "understand"? Like, if I'm explaining a concept to you and I'm not sure if you understand correctly, the only way that you would clear it up is by taking an action that accords with the expected behavior of "one who understands". So I don't disagree with any of the sentences you're writing, but I think all of them refer to socially legible external behavior.
When you say "I am watching my laptop screen as I type this sentence" I don't think that's actually your internal experience. Your internal experience is some pure preverbal stuff, infinitely rich and full of subtle impressions. You produce the sentence "I am watching my laptop screen", but the specificity of your experience is much more than that. The internal experience you have is something essentially richer and fuller than "I am watching my laptop screen", and it's impossible to crystalize it into language. Your production of that sentence is kind of like somebody smiling when feeling joy. The smile is not the joy, it's a secondary thing we assume is produced by the joy.
I used to teach high school. It's a lot of explaining things and asking kids "do you understand?". They say yes a lot of the time, even when they don't actually get it. That's why you have to give them problems to solve.
The actual criteria of understanding is that they can solve the problems. Functionally speaking, "understanding" something means being able to behave in certain ways (with different behaviors depending on the nature of the thing "understood"). You can hypothesize some completely internal ghost, but the words refer to the external behavior (from which you can, if you want, infer the existence of the ghost).
So you are "communicating" things that I "understand" because I have "experience of the things" (this is all true). But my position is that all of these, the entire complete set of all words and concepts, boil down to socially observable / legible things. There are socially acknowledged falsification processes for all these claims. I "understand" them because I can produce appropriate follow-up sentences (metaphorically similar to how you "understand" a mathematical concept if you can correctly use it to solve equations). I "have experience" with them because you could drive over to my house and see that I have a computer that I use. None of this is unobservable to you. These would all be true even if I was a P-zombie.
> This is why I don't buy the linguistic vagueness argument.
It's not exactly a vagueness argument. Words are marks on paper or vibrations in the air, they are specific external things. "These external objects have ontologically proper and improper usages" seems weird to me. Like, hammers exist as external objects right? And in some sense, they are "correctly used for" pounding nails. But that's a social fact about hammers, if I want to use it as an aesthetic object or a door stopper that's not "improper usage" except by social convention.
People use words to do complicated things (coordinate behavior, etc), but that doesn't necessarily mean that there is a metaphysical referent pinned to those marks on paper. You can get very complicated structures out of brute objective associations (the electrons in the wires of your computers don't "intend" to display these letters or anything). It seems completely possible that humans "use" language the same way that a computer "uses" a programming language (in a purely functional way).
To clarify, I'm not saying that there is no such thing as qualia or that everyone is a p-zombie. I am making absolutely no statements about the light of consciousness inside of you or whatever. My claim is that language (as an essentially social system with usage rules tied to interpersonal behavior) cannot adequately reference this light, and can hint at it only obliquely and imperfectly. Having a discussion with somebody is communication, but so is collectively painting a picture or dancing the tango (in a metaphorical sense, because the socially proper usage rule for the word "communication" is that it involves words). The fact that language seems to connect us doesn't mean that it's a special map from the internal to the external world.
> to generate energy, mitochondria use the Krebs Cycle [...]
My point wasn't about physicalism, it was that certain words might only be applicable in certain domains. I'm saying that at the level of quantum fields, there is no such thing as "a mitochondrion". In a similar way, at a sufficiently fundamental level there is no such thing as a "dog" (because the concept itself, the way people actually mean and use it in everyday speech, doesn't take into account microevolutionary variants or whatever). You can specify a definition of any word in any context, so you can ad-hoc define "dog" as possessing such and such biomarkers. But at some point, you are inventing a new technical concept and giving it the label of "dog" because you realize that the common usage of the word just doesn't apply at the fuzzy edges.
Similarly, when you want to talk about "consciousness" or "mental events" it's possible that these words aren't applicable in certain domains of discourse. You would need to basically ad-hoc redefine them in some contexts (and in doing so, you wouldn't be talking about consciousness in the original sense). Just to say, you might not be able to unproblematically move words into whatever discourse domain you want. Maybe these specific examples are poorly chosen, but hopefully I'm conveying the overall idea of domain-specific semantic sufficiency.
I am a very stupid person and I still deny it, FWIW. I think the concept of "consciousness" as a binary category (you either have it or you don't) is a philosophical red herring. It's all just dualism in disguise.
Then it's not a very hard problem. Humans are very conscious; dogs and cats are less so; computers even less so; bacteria are barely conscious at all -- and I'm sure you can find examples to fill in all the gaps in between.
There's no such thing as "qualia" or "experience" as a separate entity; when you experience "red" and I experience "red" and a cat experiences "red", it's all just electrons bouncing around in their brains. Your brain (as well as eyes) and mine are very similar, so they produce a very similar patterns of electron bounces; so similar that we can round it up to being identical, good enough for most non-rigorous scenarios. A cat's perceptual system is different, so it produces a different pattern, but the cat is still a mammal so we could sort of map one to the other. A dog doesn't have the right color receptors so the mapping would be much more difficult.
Where is the "hard problem" ? I really do think it only appears when you reject all of the above and treat humans as being special -- but why would you ?
The "hard problem" is the question of why "electrons bouncing around in their brains" results in any sort of self-awareness.
Why is it that a particular combination of electrons that exists in the space of my body and brain bounces around in such as way that I have any experience whatsoever?
In other words, why do experiences exist at all? That is my understanding of the hard problem.
You can define qualia, but you might not like the definition. "qualia are the internal, subjective, and individual components of any conscious experience, such as the specific "redness" of seeing a tomato or the distinct sting of pain from a stubbed toe." — per Gemini
You gave up because of a bad take in the toy model conversation used by the post's intro to illustrate how frustrating and unproductive this type of discussion often is?
OK, you shamed me into reading it. I would suggest there is a third camp, though.
1. It's not a hard problem (We haven't solved it yet, but we can using the scientific method).
2. It's a hard problem camp (with various flavors of the reason it's a hard problem).
3. It's not a problem at all (it's only your linguistic sloppiness or a category error that makes you think it's a problem).
However, this stuck in my craw...
> Because of the asymmetry we've just talked about, a pretty good strategy is probably "be in Camp #1". This is also born out empirically (by Dennett's book being more popular).
My response would be that the Bible is more popular than any other book in the Western world. We're under no obligation, though, to give it empirical credence just because a lot of people have purchased it, read it, and believe it.
Thanks for sharing. I'll try not to be so dismissive in future.
The Democratic Lt. Governor of New Jersey Dale Caldwell resigned after being accused of “sexual harassment.” He allegedly commented to a staffer after being rejected by the staffer’s friend that “you young women are looking for young sperm.” He denies making the comment and is playing the race card, saying they targeted him because he was black:
The state brought in a law firm to conduct an “investigation.” You can see how unbiased they are by how they refer to asking women out as “soliciting” and “propositioning.”
The report contains an anonymous complaint someone, presumably within the state government or Democratic party, made about his supposed behavior:
“The Lieutenant Governor is also hitting on women at functions while traveling using taxpayer money. He has been particularly shameless in handing out his state business cards and asking women out on dates. He has been observed doing this many times while out on official business, which is highly inappropriate for any state official to be doing. Given his position, the power imbalance between him and these women is coercive.”
“Enough is enough! He should not be trying to get dates while he is traveling on the taxpayer’s dime. And he should not be accepting free tickets to costly events for the women he is trying to impress. Please put a stop to this shameless and corrupt behavior.”
The investigators then went around interviewing his staffers to figure out if he handed out his “state business cards” to women he was asking out. It’s obvious that these people started with anger at him for the crime of having sex and worked backward to find a justification to punish him, like the corruption of using business cards it cost the state 8 cents to manufacture for personal purposes.
Ah, come off it, man. If the information here is any way at all accurate, and I agree there could be party infighting or just plain someone doesn't like him at work, the guy is a pest. He thinks the business of being lieutenant governor is to give him chances to attend junkets and pick up chicks. That's not on. He can ask as many women out as he likes, he can't do it while handing them a business card at a state function when he's on the clock.
"the crime of having sex"? Really? That's your explanation here? No other reason? He's a guy who gets laid so that means the complaints are all fake and spite?
So what do you make of the allegations of "possible ethics violations that include the purported offer of state jobs". The sexual harassment allegations can be "he said/she said", but the 'quid pro quo' of "get my honey-bunny promotions" and treating personal guests without reimbursing expenses are more serious charges:
"(H)e violated state ethics rules by bringing personal guests to ticketed events without paying or reimbursing their admission, and by using his office to seek a promotion for a state employee with whom he had a romantic relationship, without disclosing that relationship."
Fiddling expenses and/or treating the state purse as personal ATM is small potatoes but often an offence politicians are indeed guilty of. I think the Governor probably had no choice but to leave him in the library with a revolver, as it were; it would look very bad for her and her administration to have an accused ethics violator as her second.
An elected official "accepting free tickets to costly events for the women he is trying to impress" certainly would and should be not ok, whether or not it happens to be women that the horndog is trying to "impress".
True! Depending on what "costly" means here I suppose. Sometimes that's a slap on the wrist for an ethics violation and sometimes it rises to the level of corruption.
I've been thinking about the smallest discretionary intervention that can make a difference to high rejection environments resulting from an overabundance of supply and structures built in place to cull applicants within them.
For instance, the academic job market, the New York Times opinion section, fiction or poetry publishing at major trade presses, selective university admissions, peer reviewed journals in economics and philosophy. Anyone inside of these things will know (luckily I've had inside looks at all of these), there's always an elaborate procedure, involving rounds, committees, each looking for reasons to reject because they are inundated with hundreds and thousands of pitches, applicants, prospectives, all of whom look incredibly similar on paper. The system is in place to try and minimize or eliminate false positives, at the cost of an enormous amount of false negatives. Those false negatives as they build up over time cause a lot of bad blood and mistrust between these institutions and the outside, and even with participants within it. They also aren't great for the institutions on their own terms, because the result is very risk averse and conservative (with a small c) perpetuation of status quo cultures.
I'm sure there is a literature on this in organization theory, so what is the going consensus on small interventions that make a difference, and what other domains does this exist? My own reaction to this when I am sitting on the other side, i.e., hiring committees, peer reviewing, taking pitches, is to give each individual in a committee, a discretionary pick, of someone that the rest of the committee has rejected, but who gets to go to the next round anyway. I call this "The Golden ticket." Its a way of introducing a nonconsensus pick for advancement in the process, but who would not otherwise advance given the rules of the cull.
One thing you could try is to have a lottery for the right to apply. Announce your desire for a grant/job/whatever, then of the n people who want it, x winners get to apply. Only those x number of winners actually have to write an application / get interviewed / whatever. (You might have some very easy test or application to enter the lottery, to screen out obvious cranks.)
The odds of having the absolute best candidate in the pool is only x/n, but the committee can carefully consider all the candidates, and so they're more likely to pick the best one. The n-x candidates who were randomly rejected do not suffer most of the costs of applying and being rejected.
One of the problems is that 'the best candidate' is almost never something you can determine objectively, and it's often something none of the selection committee agree on. You're more likely to end up with a consensus candidate that everyone is lukewarm on, but no-one hates, than the best candidate who might nevertheless rub some people up the wrong way.
I've often proposed the idea that high-demand college courses should be allocated by lottery to all students scoring above a cutoff on a standardised test (say top 10% or top 20%).
> have a lottery for the right to apply. Announce your desire for a grant/job/whatever, then of the n people who want it, x winners get to apply. Only those x number of winners actually have to write an application / get interviewed / whatever.
I like this! I had very similar ideas in the past with respect to many other competitions where the number of applicants vastly exceeds the number of position.
Kinda reminds me of a bit in one of the Horatio Hornblower stories, where a lot of the earlier (by internal chronology) stories deal with the hypercompetent protagonist's struggles with the promotion and seniority system of the early 19th century Royal Navy. One story has a plot point where the normal system (with lots of boards and committees and veto points) gets bypassed by a similar golden ticket system, where a retiring flag officer can unilaterally promote one Midshipman to Lieutenant, one Lieutenant to Commander, and one Commander to Captain. Our hero caught the eye of the retiring admiral he's serving under, who gives him one of the golden ticket promotions.
I have no idea if this was actual practice, something the author made up from whole cloth, or the author's garbled understanding of something that showed up in his research.
Yeah, the Royal Navy had a weird system for a while in which promotion through Midshipman to Lieutenant was basically meritocratic. promotion from Captain to various sorts of Admiral was done on strict seniority, but there was no official procedure for promoting a Lieutenant. So if there's an opening for a Captain (or a Master and Commander), the relevant Admiral will just have to pick someone.
So a big part of career planning for junior naval officers, was trying to maximize the possibility of being capital-N "Noticed" by an Admiral. I hadn't heard the bit about retiring Admirals also getting to dispense a last round of promotions, and Forester may have made that up, but it would drive a similar dynamic.
The Sharpe series starts off similarly--the protagonist saves the life of a high-ranking officer and is given an impromptu commission--and a major plot point is the constant friction between Sharpe and his fellow (noble) officers. It doesn't help that he keeps catching the eyes of highborn women, although, unlike Jack Aubrey, he has the good sense to keep it in his pants from what I've seen.
My concern is that the discretionary-pick candidate would be at an automatic disadvantage in the next round if the same selection committee is involved. The other members *already* have rejected that candidate once, after all.
They are, even if its not quite the same selection committee. Usually the way this happens is, there is a committee of three, they do a culling, from say 300 to 25, there are in-person interviews in a committee of five, to cull from 25 to three or four. Then ten or fifteen people make a decision about the three or four to the final candidate (this is a job search I'm describing. But you can do this will different numbers for, say, a submitted proposal for literary fiction, or an opinion piece at the New York Times.). The idea when I do this in real life is people get to put into the pool of 25 from the 300 even if there is no consensus for them, or one person is allowed an outlier candidate in the final four who might hit it out of the park in person.
One unrelated thing I'll point out is that these dynamics don't always lead to status quo bias. For example, high-end journals like Nature and Science are notorious for publishing "cutting edge" research that has a high probability of being wrong, because surprising results get lots of attention but also are surprising for a reason.
Wouldn't that be novelty bias, basically the opposite of status quo bias? (My understanding is those journals do not publish replication studies as a matter of policy.)
Yes probably, I don't have my terms sorted into the right category. The idea is that, if a certain publication has a norm of doing one kind of thing (in this case, novelty bias), it will be completely risk averse to other things that can contribute very positively to the value of what they do if the entire infrastructure is set up to reproduce the same kind of thing. And short of reforming the entire process, what is the one change that would make it less risk-averse procedurally?
From an economics point of view, the obvious question is, why do these markets look the way they do in the first place? This is something that Scott has written about before (https://slatestarcodex.com/2015/07/28/non-dual-awareness/, https://www.astralcodexten.com/p/why-is-the-academic-job-market-so). I think there's a combination of A) people want to do these things for their own sake, not for the money; and B) "dualized" outcomes (e.g. tenure in academia). Your idea of deliberately advancing candidates that you know others will not advance is an interesting idea for fighting status quo bias, but won't change the fact that there's enormous effort that goes into competing for limited spots--most of which is therefore "wasted."
In my mind, therefore, the solutions are to either increase demand or reduce supply. The former is difficult, because the institutions (elite universities and journals etc.) benefit strongly from inertia and reputation, and get lots of influence by being "elite." It should at least in some cases be possible, though--for example, technology makes it much cheaper and easier to start new journals, and a small group of established academics can lend it a strong reputation. The internet has made it much easier for independent artists to operate. Similarly, the dualized outcome benefits those institutions as well as the previous winners, so they're probably going to be resistant to changing how the system works.
A very large infusion of money to build new universities and research centers could plausibly change the dynamics, although we already probably have too many university seats compared the economic benefit of college, and the marginal value of research seems... kind of low.
Reducing supply means disincentivizing people from even applying in the first place, most likely by reducing the rewards. This is also quite difficult, as the pay for these things is already quite low or even 0. People want do write fiction or be academics. Another way to reduce supply would be to move the filter earlier, but most likely this just pushes the wasteful competition earlier in the pipeline.
I rather like TotallyHuman's solution earlier in the thread: "have a lottery for the right to apply. Announce your desire for a grant/job/whatever, then of the n people who want it, x winners get to apply. Only those x number of winners actually have to write an application / get interviewed / whatever. (You might have some very easy test or application to enter the lottery, to screen out obvious cranks.)"
I had very similar ideas in the past with respect to many other competitions where the number of applicants vastly exceeds the number of position.
The number of people who want to apply exceeds the people who are qualified for applying, in the case of jobs, and in the case of people who want their literary works published by some commercial press, the numbers are really skewed. Would this be a lottery that preserves the distribution of qualified versus unqualified people, because that reintroduced some some of the same problem we're trying to avoid.
An honest random lottery would of course preserve this ratio, but it would also relieve the time pressure on the decision-makers, and would allow them to make a more thorough evaluation of every candidate.
There are so many universities and research centers that are currently under-enrolled and ready to close that we don't need to build anything new. There are so many ways to express your opinion that we don't need new publications for them. But the problem still exists AT Harvard or Berkeley, and New York Times opinion, or Penguin Random House literary fiction. People want to be AT THESE PLACES. (Not an economist, but are these Veblen goods??) My thought was, what can we do internal to the existing procedure at these places.
To the extent people are competing purely for general status, then yeah, that's more or less fixed.
But I think that a lot what's going on is that people want the freedom that comes with academic tenure and they like producing art or doing research, and they're willing to earn very little, so standard economic pressure (decreasing wages) has limited effect. In those cases, the pressure on those specific institutions is lessened if there are other ways to achieve those same outcomes without the selection filter.
> My thought was, what can we do internal to the existing procedure at these places.
One idea: Make the selection process more transparent. If people can more readily determine what they need to do to pass the filter, then they should A) waste less effort on trying to look impressive in ways that are actually irrelevant; and B) be more likely to self-select out of the process. It does risk Goodharting, though.
Tentative thesis: China is unlikely to accept a subordinate position in AI capabilities to the US, just as the US is unlikely to accept a subordinate position in AI capabilities to China or anyone else.
This thesis suggests that there are two likely paths for the future international AI (non)regulation:
1) a continuing race between the US and China, perhaps joined by some late entrants, possibly with some rules established (e.g. mutual ban on autonomous weapons or something like that), or
2) a “pause or stop deal” that will establish some sort of ceiling on capabilities, which will be equal for China and the US.
What I think is far less likely is a deal in which China would accept lower AI capabilities than the US.
I simply don’t see a good reason why Chinese leaders would accept subordinate position. They know that China has the ability to catch-up to the US in various technological domains, as evidenced by the fact that in, like, 1990, China was behind in more or less everything, and now they are pushing the technological frontier in many areas.
Some estimates floating around the web suggest that, if the US would stop AI development now, China could reach current US capabilities within months. That is maybe overly optimistic/pessimistic depending on where you stand, but I very much doubt that catch-up would take more than a decade. And Chinese government, being patriotic about the abilities of the Chinese nation, probably will not have an absurdly pessimistic estimate.
Moreover, the Chinese regime is in many ways oriented around this idea of catching up (this is a big difference between China and EU). It is a central plank of the official historical narrative of the People’s Republic of China (see for example preamble to their constitution: https://english.www.gov.cn/archive/lawsregulations/201911/20/content_WS5ed8856ec6d0b3f0e9499913.html) that the century between First Opium War and the Communist victory in the Chinese civil war in 1949 was the century of humiliation, during which various imperialist powers exploited China due to the fact that it was technologically backward country unable to stand up to them. I happen to think this narrative is largely correct, but even if it isn’t, what matters is that Chinese government believes it and is determined to never again allow China to be put in the position where a technologically superior power could walk over it.
This has some implications:
A) Since China is currently behind, any sort of “pause or stop deal” between the US and China would likely have to involve either the US stopping further development and allowing China to catch up, or the US somehow crippling its own capabilities to a current Chinese level. Of course, for either option, the US would have to place a lot of trust in the Chinese government.
B) As long as China is behind, export controls on hardware and other attempts to limit Chinese capabilities are at best of doubtful utility as a way to get to a deal. In fact, imho there are making a deal less likely. They antagonize Chinese leaders, make them even more paranoid and less inclined to a constructive agreement. But also, to the extent they are successful in limiting Chinese capabilities, they make it less likely to that the US government would agree to an equal deal, since that would mean sacrificing larger lead, which is harder to politically accept for the US. Perhaps this is a reason why nuclear arms treaties in the Cold war were agreed only after the USSR more or less caught up to US nuclear arsenal?
Now, attempts to limit Chinese capabilities can still be seen as useful if you think it is important for the US to “win” the competition, for example because there are reasons to believe US AI will be safer. Or because you think the world would be a better place if the US “wins”, even though both countries would develop similarly dangerous systems. Or of course if you think that the US victory would be preferable even if American AI would be more dangerous than Chinese AI.
> Elias Schmied links Evan Hubinger on why Hacker Opus didn’t show emergent misalignment.
Can someone spell it out for me? How does starting from an early checkpoint vs. a pre-trained base model change the scope of misalignment generalization the way that was observed? What reasoning would have allowed me to predict that difference?
She (after release for her part in the FTX fraud) was hired by Manifund (https://substack.com/home/post/p-215243351). The piece and comments are pretty thorough on what the concerns are, on both sides.
> get hired by Manifund, receive previously stolen money as legal income
This is not just "everyone makes mistakes, everyone deserves a second chance", but specifically how that second chance is connected to their previous mistake.
It is okay to get a fresh start. Not okay for that fresh start to include a reward for your mistake. "She gave me some of the money she stole, now I am giving her a second chance by hiring her" - nope, dude, a true fresh start would mean that you lose all that money; this is called money laundering.
When you set out on a task, say making a presentation or writing a book review, do you just sit down and do it or do first come up with expectations around how longs it's going to take, what "done" means, etc.?
Thinking particularly about knowledge tasks that involve some uncertainty about time/goals.
I'm curious what people do but more importantly about their internal experienc of it.
It depends. For stuff that is more rigorous---e.g., competitions or tasks with deadlines---and which I want to prioritize, I'll think about what process I want to use. For more informal stuff, or things that are just for fun, I'll let myself just head in whatever direction seems interesting at the moment. The definition of "done" will vary with the situation. It can range from "I'll make every reasonable effort up until the deadline" to "it's not perfect but I'm not embarrassed to share it publicly".
Managing uncertainty is tough and there's no silver bullet. If I want to be rigorous about it, I'll compare how long stuff takes vs how long I expected it to take. That'll give a signal if I'm ahead or behind schedule, and then I can adjust my goals to try to stay on schedule. The less rigorous approach, but still very good!, is to simply say "I'll spend 1 hour/day on the project" (btw Beeminder is an app that can help you commit to this) and lets things go where they may.
For sure. I realized that I intuitively apply what can broadly be labeled as planning to some things and not others and became curious about doing it intentionally--not for any gain, but rather for the experience of changing some sort of core habit.
That in turn made me curious about the internal state of other people when it comes to this activity.
If you're interested in then yen and or/the yen carry trade, check out my post. I discuss how the key question is less “how big is the carry trade?” than who owns the yen short, how leveraged are they, and what forces them to buy yen at the same time?
I describe the differences between fast-money funding exposure (where a rising yen can force rapid covering and asset sales) vs. slower balance-sheet adjustment by corporates, insurers, banks, households, etc. The latter can still matter enormously for Treasury demand and global capital flows, but it is more likely to unfold through hedge-ratio changes/refinancing and portfolio reallocations than a single synchronized “unwind.”
i had a bit of trouble staying with the article through the aphorisms at the end of most paragraphs. "Squaring the circle is the work of keeping discretion where there are fixed obligations” or “turbulence without dysfunction” as examples - each paragraph comes across as an important conclusion and it is hard for me to see what stands out and i should take away. i can’t really follow a line of thought as much as i feel like im eating nice tapas but cant tell what the meal is building towards, it just kind of ends. i *think* the too fast / too slow asymmetry is a strong contender for main-thesis treatment.
on the content - you might want to consider that when Japan defends the yen, the MoF has to draw down dollar reserves, much of which sit in Treasuries and deposits. that makes intervention itself a channel into the Treasury market, which ties directly to your forced-selling concern and could be part of the answer to your opening question about why the US joined in July - i don’t see it mentioned in your piece.
(excited to see more talk on the carry trade btw, i look at it a lot myself)
Thank you so much for the thoughtful comment and feedback!
I think I let myself get a bit carried away with my yen/circle pun!
If I state the thesis more plainly: Japan is trying to move to a regime that depends less on yen financing of dollar exposure. The leveraged carry trade gets the attention, but the exposure sits mostly on large, slow balance sheets. A hike that is too slow (especially relative to the US) keeps encouraging the leveraged trade that puts pressure on the currency. A hike that is too fast takes away the time the slow balance sheets need to adjust, and more concerning, may show up first as dollar scarcity for the banks.
On intervention, you are right, and I cut it as I had been planning a follow-up. Traditionally, currency intervention is funded by selling the Treasuries held in reserves. After the March 2020 dash for cash, the Fed created the FIMA repo facility so central banks could borrow dollars against Treasuries instead of selling them. It doesn't seem like this has been tapped into so far, at least from the Fed's balance sheet usage through mid-August.
Smaller interventions can be carried out via selling Treasuries and help reduce the incentives of the leveraged carry trade (you can see a big pull back of yen shorts in the CFTC data). If a bigger intervention is needed, the BoJ has the FIMA repo, which was set up to avoid impact on Treasury markets.
since i started watching, the main thing that i notice is that the rate gap isn’t closing since the US is being pressured by inflation. it will be hard for Japan to go “too fast” while the US is still stuck raising rates. i think that what contains the carry now is the “official ceiling” - Japan seems to want to keep the dollar below 160 yen, spent 15.4 trillion yen to do so, and the dollar still rose above 157 right after the BoJ move (below the ceiling but not a ton of cushion). a credible ceiling keeps carry traders in by clamping volatility. traders are more likely to build positions in this range - the base case overall being grinding towards the ceiling with punctuated intervention and reserve drawdowns.
the boring grindy range feels like the most likely path, but it could break down if the Yen rallies hard and traders try to all exit at the same time. that would likely be set off by a US shock: such as trouble in the AI capex waters, or a “sell America” moment. in 2024 this happened but was brief since the Fed was about to cut but this time they are fighting energy inflation, so it can’t be a cushion for that type of fall.
I think this is a live potential scenario. normally, when stocks fall, US yields fall too, because investors buy bonds for safety. if the yen jumps and US yields rise on the same day, someone is selling Treasuries to get yen. that means that (likely*) Japanese money is heading home, meaning fewer bond buyers at the same time that US stocks are tanking… yikes!
Does the FIMA repo do enough? It is a loan - so if they use it to borrow dollars to buy yen, they still end up paying it back when they sell the Treasuries anyways or buying the dollars back and reversing the intervention. it delays rather than removes the impact. if Japan thought the current situation needed a short and sharp defense, FIMA fits - but they’ve intervened to the tune of 170 billion across multiple interventions, and your thesis is that there are structural problems at play: so FIMA interventions would just roll more dollar liabilities that result in Treasury sales. Is FIMA really a backstop in this case?
I think Japan has a timing problem rather than a structural one - it has a lot of foreign wealth distributed across its balance sheets, it will take time for them to adjust and increase yen exposure
So the FIMA repo can help buy time - true, it’s capped at a smaller value than interventions to date, but it’s also symbolic of having the Fed balance sheet behind it (and the limit itself can be adjusted via subcommittee). The repo can be rolled over until conditions change
This doesn’t guarantee that a crisis couldn’t happen of course! I just don’t think one is inevitable
At a rationality meetup, we were talking about p<doom>, I think with some consensus that even low risks should be taken seriously.
Existential risks at the level of wiping out the human race included targeted super-virus and boiling the oceans. While ignoring the question of why the AI would do that, I think because the answer is "because it seemed like a good idea at the time", I'm wondering *how* you would boil the oceans without ruining the machinery before the oceans get that hot.
Also, what is the correct pronunciation of P<doom>. Is it p'doom or pea-doom, or what?
I have concluded that Wensleydale cheese with blueberries is rationalist chow.
"I'm wondering *how* you would boil the oceans without ruining the machinery before the oceans get that hot"
This seems like excellent fodder for either an XKCD What-If style question or something in the vein of this classic blog post: https://qntm.org/destroy
Sadly, I'm a little short on both the time and the expertise for an exhaustive answer, but I'm going to try to spitball a sketch of one anyway.
A very quick back-of-the-envelope puts the energy requirement to boil away Earth's oceans at ~10^24 kJ. Assuming we're not infinitely patient and want to see this done in some sort of reasonable timeframe, I'm going to assume it must happen within a year of flipping the switch, giving a power requirement of ~10^17 KW.
A quick google search suggests that a typical nuclear power plant can only generate about 10^6 KW of usable power at 33% efficiency. So 3*10^6 KW for a project like this where the waste heat is the whole point. Sadly, I don't think getting a hold of ~10^11 power plants worth of fissile materials is likely to be very feasible, even for an AI with complete control of the world economy. This sinks my first idea (pun intended), which was to drop a bunch of very simple, very high-output nuclear reactors into the deepest parts of the ocean, doing some amount of minimal engineering to create a configuration that will optimally transfer heat into the surrounding seawater.
If terrestrial nuclear power isn't enough, it's pretty obvious that the much bigger nuclear reactor up in space will have to do. And on the one hand, this might be even easier: you don't need to mess around with any of the usual headaches of solar power, like turning the sunlight into electricity or storing and transferring the power. You literally just need to worry about minimizing the reflection and focusing/capturing the light in such a way as to get as much of the energy as possible transferred into the ocean water. Except that the moment you get anywhere *near* boiling-the-oceans levels of heat in the ocean, you're going to create such an enormous amount of cloud cover that almost none of your sunlight will get through at all.
So my very loose recommendation would be something in the nature of a network of space elevators, topped with collectors and a system for transferring captured heat or light as efficiently as possible down the trunk into the depths of the ocean. Add to that an system of orbital reflectors for focusing sunlight on the collectors, and you might be able to get there, or at least get a good fraction of the way there. Of course, that relies on inventing some still pretty speculative tech an solving some hard engineering problems, so the AI better be up for a challenge.
No, we went through this on the old rec.arts.sf.science Usenet newsgroup, with an actual Ph.D. climatologist. Those archives are hard to navigate these days, but it's probably still somewhere on Google.
Bottom line, burning all the coal would do a lot of harm, but not literal boil-the-oceans harm. For that you need to get at the carbon that has long since been bound up in carbonate rocks. We did figure out a way to do that, but it involved a thousand-year plan of mining carbonates and collecting them in giant piles, with a side order of mining and enriching uranium for the ginormous nuclear furnaces. Then spend the last hundred years shoveling rock into the furnaces.
We envisioned a modest residual workforce of humans in air-conditioned bunkers to do all this, but AI+robots would also work. But I don't see what they would gain by it. If it's just Kill-All-Humansism, there are easier ways to accomplish that.
As to "why", I was thinking of "Saturn's Children" by Charles Stross. The novel takes place several centuries after Homo sapiens became extinct — not through war or disease but through hedonistic apathy, — and civilization is carried on by sentient robots. Think Asimov’s robots… with their masters regrettably absent.
And prevailing temperature on Earth is 90 C, which does not bother the robots in the least (yes, ALL eucaryotic life is extinct, at least on Earth). Not because they *wanted* to exterminate biological life, but because without someone to tell them to stop, the robots just kept on burning oil and coal.
I always thought "boil the oceans" was in reference to the AIs using up so much water and accelerating climate change so much that the oceans get depleted, not as something they would do deliberately as part of an evil plot.
> I'm wondering *how* you would boil the oceans without ruining the machinery before the oceans get that hot.
I think the answer to all such questions is "the AI will be superintelligent so it will figure it out somehow, and trying to comprehend the exact method is pointless because you're a mere human, just trust me bro".
I'm no prepper, but as the others have said, shelf life is a really important consideration. You don't want to show up in the woods with your trusty tent and a crate of canned food, only to find out that all the elastic bands in your tent had rotted to dust and all the cans are sporting suspicious bulges. Ask me how I know :-/
Beyound that, in places like California there is a lot more food available in the woods year-round than you might think. Learn to forage and fish, not just shoot deer in the head (admittedly I didn't take my own advice since I can neither fish nor hunt, but I've got friends who can). There is almost no excuse for starvation, not even if all your cans got spoiled; obviously the situation is quite different in less temperate regions.
I wonder if being a prepper makes you less anxious about “SHTF” because you’re more prepared for it, or more because now you’re literally invested in it, have a corner of your basement that reminds you of it, etc.
In other words is there a psychological benefit or cost to being a prepper
The biggest thing a lot of people fail to consider is that emergency supplies have a shelf life. You should be cycling through your canned food and dry goods, and even your water, on a regular, at least annual, basis. Replace all the batteries in flashlights/radios/etc. every year or two to avoid corrosion. Make sure you're controlling for light, heat, and humidity. Drugs in your first aid kit lose potency over time.
Proper storage and access are just as important as the actual items you're stockpiling, but a lot of people just make one big initial purchase and never think about it again. I know at least a couple people who consider themselves prepared for the worst but are almost certainly just sitting on hundreds of pounds of expired food and nonpotable water that will be completely useless in an actual emergency.
I used to be on the Discord server for the game-designing couple who made, among other things, Apocalypse World. One of them commented that real prepping looks less like stocking up on canned food and guns and more like learning how to mend and preserve, and building community ties with neighbors who have similar skills.
I think we are headed into a Thucydides trap from emergent AI. But it will not be the one between the US and China, it will be between the Trump executive branch and the emergent AIs.
How can Thucydides trap strategies be applied if this becomes the case?
Good Question. First, Trump just says things, and has a fecklessness about following through with policy. I do not think Anthropic sees him and the DoD as essentially "pro-ai"
2, and more important, the trap/conflict is multi-directional: the AI can decide the US gov is a threat and act preemptively. Trump needn't vocally project hostility for this to be so.
Tackling this in good faith, I could see LLMs managing a successful DDoS by exploiting a known weakness in some OS in order to gain execution privileges, building a botnet if that OS is installed on enough machines, and perhaps even managing to make that botnet resistant to customary anti-botnet measures.
The typical way to counter a botnet is to get enough people to patch their systems (get a trusted patch, reboot in safe mode, install patch, etc.), so for AI to hinder this in novel fashion, it would have to somehow disrupt the method for distributing patches or for alerting people that they control machines that need it. One way to do that might be to coordinate multiple DDoSes, including targets commonly used for disseminating security information. For extra credit, it might even corrupt the patch distribution, but that seems hard (if it infects the patch, it would have to also infect whatever runs checksums).
What I know of LLMs suggests this is theoretically possible, provided there are usable exploits on current OSes and everyone is somehow too dumb to use LLMs on the mitigation side.
I cannot tell you exactly, noone can. I can surmise that there are many facets of the Hugging Face Incident that would have been effective means of getting access to particular public functions. Finance, defense and communications come to mind. That could escalate quickly...
Why do I see people queue up 15 min to save maybe max a dollar on gas? See this happen at Sam's Club.
Even if I fill up at the most expensive station in town, I spend maybe $4 more? Well worth the convenience to me.
Skipping one meal out, or buying one less impulse buy, or skipping one needless social activity, or (long-term) cutting housing cost by $50 (easily achievable) will save me more money many times over. Why don't people do that instead?
As I've become more financially secure, I've made a deliberate effort consider these sorts of time-vs-savings questions more frequently and explicitly. For gas this has just resulted in me routinely buying gas at the pump adjacent to my grocery store: it's pretty cheap (if not the absolute cheapest), and the time value of bundling it in with another errand I'm already running in the same place is far more significant than any considerations of line length. It's possible that some fraction of the people in the Sam's Club line are running a similar calculation. On the other hand, the wait where I get gas is never anything like 15 minutes, and I'm not sure I'd keep the same habit if it was.
I agree with Alex that it's probably irrational in many cases, but it could also simply be that many people don't value their time very highly. You mention people waiting for 15 minutes and saving $4. So, it makes sense to queue if you make less than $16 an hour, that's lots of people, plus the unemployed.
I listen to Ramit Sethi's "Money for Couples" podcast and one point he makes repeatedly is that this kind of behavior is super common, at all income/wealth levels, and mostly irrational. After years of listening to the podcast, my best understanding is that people do it because it gives them a sense of control over money and is easy to understand. More importantly, it is an easier way to achieve a sense of control over money than doing things that are impactful but hard, like seeking a new job, learning about stock market investing, or changing overall spending habits. (A surprising number of guests on his show have some version of Dril's "candles" tweet going on, where there's some obvious financial problem in their life that they are absolutely dedicated to sticking their head in the sand and ignoring, and money-saving activity like buying cheaper gas seems to show up as displacement activity/coping behavior to avoid addressing it. Note that I don't doubt the existence of some truly destitute people who absolutely do need to save $4 on gas, but there are lots of people who do this even though it's irrational.)
I won't queue unless I have to. The prices for gasoline at my local Costco wouldn't make it worth my while. The best I've seen is that they're $0.30-$0.40 cheaper than the big-name companies I have locally (Shell, Chevron, Valero). However, the gas at some of my local independent gas stations can be $0.75 to $1.25 cheaper than what my local big-name gas companies sell. I can save $10-$17 on a fill-up, and that's worth the time to go a few miles out of my way to get it, and I seldom have more than one car ahead of me in line.
Prices are more visible than queue length. There are websites that tell you gas prices at every station in town, but last I checked these websites don't tell you if there's a line. By the time you're there and can see the line, driving somewhere else is going to kill a lot of the time you're saving by not waiting in line here, and who's to say there isn't a line there, too?
Also, the size of the savings depends on how much gas you're buying. The queue time is fixed per fill, but some vehicles have much larger tanks than others. I remember my old Honda Civic has something like a 12 gallon tank, while a quick googling suggests that many pickups and SUVs have 20-30+ gallon tanks. So if you're on the low end of this, a $4 difference for you might be a $10+ difference for someone with a much larger tank.
Alternatively, rather than shopping for prices each time, I expect many people form habits around getting gas somewhere not too far out of their way in their daily or weekly routine that they know usually has decent prices. Sam's Club is a place lots of people shop on a regular cadence and which seems likely to have good prices, so it's a natural place for people to form a habit around. Once formed, the lines need to get pretty bad fairly often before people will change their habits en masse. They could raise their prices in response to demand, but there are several reasons that they might not want to: reputation for good prices, discounted gas as a driver to get people to pay membership fees, lines only being bad a peak times while keeping prices low drives sales volume, etc.
This is a very good framing on visibility. I would extend it a bit. I think many consumers have an impetus not to be a sucker. They are willing to pay a premium when its more visible what the provider did to add value, but less so when its not clear. In the case of gas, having extra pumps is a good way to avoid lines but that business decision is harder for the consumer to see.
Gas prices are literally more visible--you can't drive down a major road without seeing a sign with the price of gas. That makes the public hyper-aware of gas prices in a way they aren't for other commodities.
If my grocery store raised the price of a loaf of bread by a dollar, I probably wouldn't notice. But if the corner gas station raises prices 10 cents, I notice. That visibility definitely contributes to people going to lengths for cheap gas that they wouldn't for other items.
I do think the other comments pointing out that "yes, this can matter for people" have a point.
But I also think people have anxiety about gas prices a bit out of proportion to how much they matter: it's a cost that's pretty volatile, out of people's control, and (for a lot of people, in the short term) unavoidable. Plus tends to be tied to big news headlines and very often weaponized against the political incumbents (on both sides).
Gas prices may actually on account for like 25% of the cost of owning a car over its lifetime... but they're the one that people see every day, and get the most coverage in the news.
actually, things that are volatile, out of my control, and unavoidable are precisely things that are the types of risks to “worry” about. there would be no reason to fret over things that are predictable and stable! You can’t change your car or commute quickly, so price shocks land purely as income shock, and it hits people with thin means hardest (what’s the percent of americans living paycheck to paycheck?). these days gas prices also flow through to many other goods that also impact those same people more heavily. where people go wrong is they ruminate on it vs. taking the change in gas prices seriously as a real impact on a livelihood and moving forward with their day.
i agree that people get their wires mixed up by worrying about the gallon price but choosing to make a gigantic monthly truck payment. those worries are less rational for sure.
Maybe people are "sticky" when it comes to choosing where to buy fuel? The cost of finding a new gas station which has no queues but is also not very expensive (probably farther away from frequent routes than current gas station), and getting used to the new gas station/route is likely much higher than 1 dollar.
I was flabbergasted to spend 78 bucks to fill up my tank. But when I got home I didn't think for a single second what I had to not buy because of my $78 fill-up.
you understate the benefit in many cases, first of all. where i am costco gas is 20-40 cents cheaper or more sometimes. big trucks save more than 8 dollars. a fifteen minute wait just became a ~$20+ hour savings which might be more than their hourly wage.
many people are there anyways to shop, as well. and personally i’ve never actually waited 15 minutes, that isn’t the average experience.
no argument against people making follies in their chase for savings but gas ones don’t seem like a very big error. my parents used to drive an extra mile to save 5 cents out of what seems to be principle lol.
(oh, and most people trying to save some money on gas probably already have cut back in the ways you mention. it isn’t so straightforward.)
Don't forget Executive members with the Costco Visa are getting 7% cash back on gas.
And between the attendant at the pumps, the stations not having a convenience store selling liquor/cigarettes/etc as an attractive nuisance, and literally everyone present being in the class of people who decided to invest $65 or $130 a year to access significant savings, Costco gas stations certainly *feel* safer than the vast majority of intersection gas stations.
That said, I'm thoughtful about what time I visit Costco for gas!
Reading some of these comments gives me the impression that people think "rational" also means "perfect reasoning engine".
That's never been my impression. To me, "rational consumer" just means utility maximizing, where utility is frequently time saved, and time includes not only time for generating income to spend, but also time for researching choices. A rational consumer might choose one burger over a more expensive brand, but might alternately pick the more expensive one for taste, nutrition, or even knowledge of what the seller donates to charity. Or go back to choosing on price because the time estimated to find out what both brands put into their burgers is more valuable to a consumer than the time estimated to compensate if that consumer later decides she made a mistake.
Granted, this can lead to some choices that might strike us as ludicrous, such as thinking 1/3 < 1/4. OTOH, some people really don't understand fractions, and would take too long to learn. They're still being rational as far as they know.
This matters, because the solution isn't to assert the privilege to make their purchasing decisions for them; this just annoys them and costs the decision maker more time enforcing such decisions. Better solutions probably including improving math education, changing the ad copy, etc.
"The A&W research firm organized focus groups. The results revealed that many participants mistakenly believed that one-third of a pound was smaller than one-fourth (quarter) of a pound. Focus group participants expressed confusion over the price, asking why they should pay the same amount for a "smaller" third-pound burger.
This misunderstanding stemmed from consumers focusing on the numbers "3" and "4," leading them to conclude that one-third (1/3) was smaller than one-fourth (1/4), even though the opposite is true."
Howdy yall! I am playtesting an AI governance Megagame in Toronto in October 17th 2026. You get to play as an AI lab, a great power, a tech company, or any number of factions in a 60 person experience where we simulate the near future of AI policy, economics, and x-risk. You can sign up here:
If it's rational to vote, then is it also rational to argue with relatives about politics during Thanksgiving? From Slate Star Codex, 10 years ago:
"having a good President rather than a bad one can be worth $300 billion. A 1/60 million chance to create $300 billion in value is worth $5,000; even the 1/1 billion chance afforded someone in a safe state is worth $300."
Signal-boosting should have some propagation value diffused over some cluster of potential voters, right? Ostensibly everyone is signal-boosting, since everybody is liking political posts on social media. But if you're dropping "actually good arguments" in person, then maybe that's the equivalent of one vote when spread out over a cluster?
If you're the asshole who starts a fight at Thanksgiving dinner, of all times and places, I'm going to be actively repelled by you and your whole ideology. No matter how good your argument is or how much I agreed with it before, I would now be marginally more inclined to vote against your preferred candidate.
There's an extremely easy way of making me marginally more inclined to vote against a certain candidate, yes. The supporters of said candidate are somehow always the ones to pull it off.
More likely, I think, is the chance that you get to enough supporters of all viable candidates that the unpleasantness tends to even out, which allows other factors to dominate.
Kind of a horrifically inefficient way of ensuring elections select for truly representative officials, but I blame this on failure to encourage aspiring activists to join professional sports instead.
The cost of voting is maybe a few hours of your time doing things that are at worst boring. The cost of arguing with your relatives about politics during Thanksgiving is a few hours of your time doing things that are probably actively annoying to you and to everyone else, and possibly damaging family relationships that have significant value on their own.
Vote. But probably don't argue with your family, and *definitely* don't argue politics with your family on Thanksgiving. If your utilitarian calculus says you need to argue with them, then further math will almost certainly favor your picking up the phone to talk to them at a better time, when they're probably in a better mood and you won't be pissing off nine other people who maybe agree with you on the issues but were hoping for a congenial holiday.
Or, looking at it from the other direction, if it seems like a phone call would be excessively disruptive and ill-received, then starting a political argument over Thanksgiving will be very much more so.
I hadnt thought about this until you put them side by side. Surely Thanksgiving is the worst possible time to argue about politics with your family. You already voted! Even if you managed to change someone's mind, nothing would come of it.
Haha, good point. I think I was trying to make the broader case that, assuming you can price a vote, is it then possible to price signal-boosting in politics? I've seen some math before on the cost-benefits of political advocacy, but it was much more pointed, such as the probability of getting XYZ bill passed, the QALY impact on such a bill being passed, etc.
But yeah, Thanksgiving would be poor timing in terms of voter impact.
Right. At Thanksgiving dinner, a political argument is going to be interpreted as "you are a bad person because you voted wrongly, and I am a good person for calling you out on it in front of your family". This is not going to accomplish anything positive for anyone.
If you think they're persuadable, talk to them some reasonable time before the election, and do it privately.
It depends on how persuadable your relatives are, no? Bc if it’s 1/100 that you persuade them the. ROI is $3-50, so probably not.
If you have an on-the-fence relative in a swing state then sure? But keep in mind you probably over-estimate your own persuasiveness and people’s persuadability
Part of the issue here is the difference between 'it is rational to do X' and 'there are potential benefits to doing X'. For X to be rational, you have to weigh the benefits against the costs, including opportunity costs. Very possible for arguing politics with family to come out net negative, especially in comparison to other ways to spend your time.
I don't think those expected value calculations are really valid. For safe states, that 1/billion chance represents such an unexpected situation that it would be hard to reason about. If Vermont or Idaho is the state that decides the electoral college in 2028, the political landscape of the country would be unrecognizable to the point that my prior analysis would be meaningless.
Even for a swing state, I'm not sure I have a good handle on what it would mean to be the tie-breaking vote. My impression is a result anywhere near a tie in the tipping point state means that the election is going to be decided by legal action, not by votes. I think a one vote margin for the Republican candidate and a one vote margin for the Democratic candidate are basically identical outcomes in terms of who actually becomes president.
The trouble with this argument is that it doesn't imply that there's no decisive vote. It just shifts where the decisive vote it. Which changes the expected value calculation not at all.
"I think a one vote margin for the Republican candidate and a one vote margin for the Democratic candidate are basically identical outcomes in terms of who actually becomes president."
Suppose this is true. I think you could equally well say the same thing of a two-vote margin. But obviously at SOME margin, the math must change. Specifically, there must be regimes in which:
1. The legal challenge still happens with perfect consistency but the result starts to tilt heavily in favor of the party with more votes.
2. The legal challenge stops happening with perfect consistency: in some non-trivial fraction of such elections, the result would go unchallenged.
3. The legal challenge almost never happens: the vote total stands on its own merits.
Identifying the exact vote that serves as the boundary between those regimes is obviously going to impossible--it's not even clear they're well-defined enough for that. And of course, factors beyond the raw vote total are going to matter every time, and be different every time. The complexity and ambiguity here are thoroughly unmanageable.
But the beautiful thing about the math is, it doesn't actually matter at all. Whatever unknown probability distributions govern these outcomes, one extra vote shifts them all redward or blueward by exactly one vote. Unless you posit some EXTREMELY dubious and counter-intuitive features of these distributions--like each vote for Party A actually increasing Party B's chances of victory in some portion of the distribution--the net effect once you ramify it all down to the probability of seeing each final outcome is going to be basically the same.
Interesting, I hadn't considered that democracy has a "legalistic margin of error", which we've seen play out a few times. Assuming that Scott's original math is correct, I wonder if the "odds of being a tie-breaker" is no longer the right model then.
I don't think it makes much difference. Let f(v) be the probability that your candidate wins after getting v votes (taking into account all process shenanigans). There must be some number of votes for your candidate v0 at which they are *almost* certain to lose (f(v0)~0), and some number v1 at which they are almost certain to win. v0 and v1 are probably close enough together that the probability of having exactly v0 votes, exactly a tie, or exactly v1 votes are about the same. Call this probability p. With probability p, there are v0 votes without you and v0+1 with, and you shift the probability of victory from f(v0) to f(v0+1). Also with probability p, there are v0+1 votes without you and v0+2 with. When you sum all these to get the overall probability you swing the election, you get p*(f(v1)-f(v0)) = p. That is just the same chance you swung an exact tie.
Of course this is an approximation, but there is no world where you know the probability of a tie *exactly * anyway. If you wanted a correction, probably the sensible one is just assuming that there is an epsilon chance that the outcome is determined by some factor completely independent of the number of votes, like a military coup or assassination, and using p*(1-epsilon).
What is the economic impact of a thanksgiving fight and the likely hood that _you_ will get those 5K$? What is the chance your intervention is net positive in the direction of your beliefs or at all?
You should multiply that by the probability that your preferred candidate is actually better. The prior should be about 50-50, and inside mode logic here is confounded to hell (everyone thinks they're in the right, and half of them are wrong). You can argue that you're smart and therefore know better, but voting patterns correlate somewhat weakly with intelligence (and that's even assuming that the candidate with more intelligent supporters always wins), so it's hard to argue that you're much more than ~55% likely to be backing the better side. So you should multiply the EV by another 10%, bringing us down to $500/$30, at which point the second probably isn't worth your time (the first probably still is).
Scott suggests that if you have any doubt that you're right, then you certainly are better than the average voter:
"We don’t know for sure that we’re right about politics. In order to add signal rather than noise to the election results, we have to be better than the average voter. The Inside View is useless here; probably every voter thinks they’re better than average. I recommend the Outside View – looking for measurable indicators correlated with ability to make good choices. Education’s probably a good one. IQ might be another. But overall, my suggestion is that if you’re seriously uncertain about whether or not you think more clearly than the average voter, by that fact alone you almost certainly do."
Yeah. But he didn't multiply by the probability of still being wrong when doing the math later. Even if you're better than the average voter that's still closer to 55% chance of being right than 100%.
Depends. If you've sat down and talked with the last 7 presidents or so, I'd say you have a better handle on "what a civilization preserving president" sounds like. It helps if you have the psychological (blackhat) chops to note that Biden in 2020 was demented.
Actually, I think that's the acid test -- if you didn't read about Biden sundowning in 2020, and didn't see at least some evidence that he was quite a bit different from when he was doing Veep debates with Palin -- you probably should discount your ability to pick people.
... I'd write the same for Trump, but... he's both less senile and more weird. "Bad at understanding Trump" doesn't seem so applicable to the next president.
People also presumably assign some value, likely a very high value, to family relationships. I could see someone being willing to pay in the thousands of dollars to have good rather than bad relationships with their family. So while there might be some positive value from the chance you sway a family member to vote your way, there is also the negative value from the chance you piss off and alienate your family in a political argument. Personally, I'd say politics at the Thanksgiving table isn't worth it.
Yup, unless your family actually enjoys political arguments (and some do!), I think the effective cost of enjoying your holiday less, and causing other people to enjoy their holiday less, plus any potential long term damage to the relationship is likely a bigger cost to most people than $300-$5000 dollars.
You need to do the value calculation, if your argument with family isn't going to win anyone over then it isn't useful. Talking to swing voters is valuable, arguing with committed partisans is generally not worth it, also worth considering chance they win you over or you end up reminding them to vote.
While this is true, it's probably better done in a caucus than at your Thanksgiving table (or anyplace where folks are drinking alcohol).
When you run into someone of the opposite side, you have to have basically rock solid argumentation to explain why they shouldn't vote for whomever. And you have to understand their side too. Hint: telling an evangelical christian that Trump is a serial philanderer is NOT a good argument. They know that, and it fits with their political philosophy. [last thread I wrote out the full economics explanation for "why it's Ukraine's fault the price of diesel is high" and "why you should distrust anyone who is saying it is because of Iran" -- particularly mass media. I expect someone might actually change their mind on the reliability of some mass media because of it -- I do not expect it to change people's votes, that's an order of magnitude at least harder]
The toy model for technological unemployment looks like either the landlord scenario or the "someone can't be replaced" scenario, I think. (Arguably, the landlord in this model is just a worker who can't legally be replaced.)
Like, suppose the robots can't replace the fisherman (they can't handle salt water). The fisherman previously had to trade fish for wood and clothing, but now he can just have a robot get that for him. But the carpenter has a problem, because he still has to eat fish, but now he has nothing to offer in exchange.
There are a few ways this could play out - the carpenter could retrain as a fisherman, he could find another thing to trade to the fisherman, it could turn out he has value organizing and wrangling the robots despite not personally carpenting, the fisherman could continue supporting him out of charity, or he could work on upgrading the robots to fish.
As I understand it, so far retraining and "knowing what to order the robots to do" have dominated, but it's unclear if this will always be the case.
This is actually very similar to how I consider such scenarios.
In yours, assume all four have settled on a medium of exchange that satisfies the usual requirements for currency - portable, durable, authentic, supply-stable, etc. In come the robots.
Prices probably do not drop immediately, but there is downward pressure as the four acclimate to not having to visit each other for business. They do at first, figuring the fishing bots should be controlled by the fisherman, tailorbots by the tailor, etc., until they realize each person can just keep their own complete retinue. Ultimately that currency gets tossed. The four were trading happily; now they're just happy. They visit each other for company. Eventually, that's not enough to stave off boredom. The time they used to spend working is now spent socializing, relaxing, and then wondering what else to do.
Meanwhile, unless we've somehow repealed entropy as well, those robots will require maintenance, showing rust, worn gears, corroded transistors, and so on. One problem solves the other. The four spend part of their new leisure time figuring out how to repair the robots or make new ones. If they don't know how, they learn, spending more of that time. Possibly getting the robots to help (on a hunch, I'll assume the robots can do anything the humans could). If there are enough robots, they just maintain the rest and the humans go back to leisure. If there aren't, then the humans prioritize jobs and anything the robots don't get to is a tasks the humans do instead. They bring the currency back in order to keep the books straight.
If they're bored enough either way, they'll start inventing new pastimes. The tailor takes up fiber from other parts of the island and begins weaving more thread for additional purposes, and starts brainstorming other uses for banana leaves. The carpenter carves wood into musical instruments and invents some new tools she wishes she had. The lumberjack gets into mining and metallurgy. The fisherman gets into boatmaking. If there are enough robots to do those too, the humans assign them the tasks and move on to even more advanced pursuits. If not, then they prioritize. Either way, there's plenty to do.
If a landlord washes up and asserts he owns the island, I'd expect the humans to dispute this, possibly on Lockean grounds. What can the landlord do? He's outnumbered. And if there are enough robots, why would anyone care? The four assign him a retinue and go back to their day. It's not like the landlord could buy anything with whatever he collects as rent.
I could expand on this in a _lot_ of directions, more than I have time for. If you're interested, maybe pick one; just make sure it's good.
This kind of ruins the value in the premise. We should assume resentment at needing to trade.
Assuming no other goods, the fisherman and the carpenter hold the power in the original group. Everyone needs food. Everyone needs shelter. The lumberjack needs axes, but the carpenter can work with wind-downed trees. And nobody actually needs clothes if they've got shelter. (We can maybe say the fisherman needs tailored line, if we want, but that doesn't help the tailor because he can't withhold fish without dying.)
So now when the robots show up (and assuming we accept that everyone gets robots, instead of the first person to find them hoarding them all), the lumberjack and tailor suddenly explode in value, while the carpenter and fisherman lose quite a bit. Now the question becomes, how much resentment do you need, either against the old system or the new, before the presence of the robots starts a spite fight. When do people start using their robots to break other people's robots so they're the big shot on the island.
You're looking at a closed system, where all goods produced in the economy are consumed. You haven't explicitly mentioned prices, but the classic problem is to work out the stable equilibrium of prices assuming constant demand functions for each player (not necessarily the same).
Have a look (with your favourite AI model) at Leontiev's model for the economy - it's still a toy model, but if you can give an idea of the demand functions of each of the participants for the outputs of the others then it should be easy to infer the distribution of wealth in the equilibrium state.
These steady state models don't work well in the real world because the 'all other things held constant' assumption fails outside of very simple models so we never get to the steady state.
I would have it be multiple landlords otherwise we get a monopoly problem that probably. Also to keep it simpler ill make it a farmer not a fisherman, but a landlord could own the ocean too i guess
The landlords charge rent, but the terms of trade depend on the productivity of the robots and the land. Ownership of the robot give each former worker something to trade for the land. In addition they can provide a new service the robots cant do(they arent good at it, but its scarce). In a state of abundance the new service may demand a high wage as measured in material goods but not high in terms of acreage. The robot ownership can still command a high lease in terms of acreage. They still have material wealth, but if the laborers dont own a robot they will be land poor.
I think the worst case for the laborers is if the land is a binding constraint on total production and its not much higher than pre robot levels. There is no abundance, but their labor is devalued. They can provide a different service in exchange for goods but the goods arent plentiful so the wage is low. The land is the most valuable thing now and they dont own any. Well I guess this is why but land is an AI trade. Dont miss out on the one thing the robots wont make more of.
Hmm, I think women will keep selecting for men who do things. And so men will continue to do things. (women who don't select for this, will have less successful offspring. (I think/hope))
In real life, human labor is replaced but only piecemeal, because there isn't some finite set of endeavors for which human labor is needed. If you made a comprehensive list of everything people did for a living and then brought in machines that could do each of those things tomorrow, the next day after that there would be new things on that list that didn't exist today and for which machines haven't been invented yet.
On your imagined desert island you've set it up so that fishing and lumber and carpentry and clothes are the only endeavors and once those are automated by robots there's nothing left, but that simply isn't how the real world works. Standards of living keep changing, new problems/puzzles get discovered or created (many of them related to the technologies that were eventually brought in to solve the old ones!), and so humans keep finding stuff to do.
Maybe the disconnect is that this process of "endeavor churn" happens fast enough that most people find employment for most of their lifetimes, but slow enough that it always feels like the churn is about to knock you into permanent unemployment.
Hi all, for the last few years I was working a game, and now I've released it, just as the no-doubt glut of AI-based cheap games has come. It's interesting and weird that I loved programmign in Godot just in time for programming to change radically. (I was about to say "come to an end", but nothing comes to an end; we still have portrait painters, and buggy whip makers, if not so many, and here we are, staring down an enormous abyss that's honest to God actually gazing into us.)
iphone/ipad only; that's all I had time for. It's a guess-the-word game not entirely unlike Wordle.
It's free of charge, and free of ads. Sort of a last gift.
(Though it may not be a last gift. In just a weekend I vibe-coded a desktop idle game where you shepherd species of radiolaria toward ascension; the radiolaria are taken from the Ernst Haeckel illustrations and are very beautiful.)
https://apps.apple.com/us/app/wordquirk/id6754496319
Nice!
Now with vibecoding, in theory the stores should be full of free games. You just need to imagine it, and AI writes the code, for $20 a month. Perhaps people are not so good at imagination?
I am also working on something myself, but it will probably take a few weeks. Then I will just publish it for free on a web page.
Why would they be free? People aren't spending tokens on making shovelware just to burn money.
In the past, people have produced free software.
For example, see https://www.wesnoth.org/
It would make sense if the same people increased their productivity with AI.
This has probably already been said, but is there a nexus between the job disruption caused by AI companies and them providing a universal basic income as compensation?
Until you pry the money from the billionaires' cold dead hands, there will be no UBI ever.
In theory, there is no reason why people could not be generous enough to donate e.g. 1% of their income to the poor. Especially if they already have more money than they could ever spend.
In practice, people with this mindset typically do not become billionaires. Instead, on the top there is someone like Elon Musk, who despite being the richest person on the planet, can still find time to take money away from starving African children.
Paper explains a phenomenon In which most mosh pits tend to rotate counter clockwise
Seems to be related to most humans being right handed. According to people on Twitter,Early track and field tried doing races clockwise and people hated it .Also people tend to meander counter clockwise in grocery stores as well
https://sethna.lassp.cornell.edu/pubPDF/MoshPits.pdf
Some anecdata that, to me, weakens the grocery store finding: all of the stores to which I have access provide a strong architectural bias to the traffic flow, so any mild natural customer preference would be overwhelmed. It seems that TJ and large Chinese grocery stores go clockwise, while most others (including HMart and Costco) are counter-clockwise.
Also, in contrast to a claim below, my recollection was that most large grocery stores in London are counter-clockwise biased as are most stores in Bangkok (also a left-side drive location.)
"tend to meander counterclockwise" -- this is country specific (It's dependent on which side of the road you drive on). Britain continues to be weird, choosing left-right-left-right at intersections.
I wouldn't trust any study that doesn't look cross-culturally.
Fair . The mosh pit example was universal I believe .
Odd. I'd want to look at the data. Perhaps this is because most people are right handed, and jabs are what you do in a mosh pit? A right cross seems to be a bit much (and capable of laying people out).
Interesting. I only skimmed the paper; did the researchers say which bands they went to go see to gather data? EDIT nevermind I see they used videos.
Could anybody give me their best estimate as how long it is until the general paying public gets Gemini 4 Argon?
Not sure anyone will read this given that the post was a while ago, but would anyone else be interested in the ACX post "Curcumin: Much more than you wanted to know"?
I eat a lot of turmeric and want to understand whether it is at all healthy, so I'm starting a campaign to get Scott to write a long article about it until either he writes it, he says he doesn't want to write it, or no one else seems interested.
P.S. Suggesting other sources evaluating turmeric mindful if its false result tendencies is welcome.
My opinion on turmeric is that it does work to decrease inflammation, but I am wary of the hyper-concentrated forms of turmeric. There are more than enough case reports of people overdoing turmeric supplements and ending up with liver failure. If you take a lot of turmeric, especially combined with black pepper to increase absorption, you should check your liver enzymes.
Thanks for the heads-up, Mercedes! I personally am pretty skeptical of almost all supplements; when I say I eat a lot of turmeric, I mean the spice itself, plus or minus whatever lead adulterants have been added without the FDA finding them (see other comments if you do not know why I am bringing up lead).
Ensuring you get your tumeric from a reputable source and maybe even testing it for lead may be useful. People were adulterating it with lead chromate at different stages of the supply chain to make it a nicer yellow colour. https://www.nature.com/articles/s41538-026-00867-8
I personally live in the US so am less at risk than some others, but I will keep this in mind going forward. Thank you Unbuffer for the PSA.
This is a good point. I am not sure what home tests are available or what levels of lead adulteration are both dangerous and difficult to detect.
be the blog you want to see in the world
or more realistically, throw AI tokens at it
Have you seen the Examine page? That site was my top choice for thoughtful nutrition science aggregation when that was an area that I prioritized:
https://examine.com/supplements/curcumin/?show_conditions=true
I have not until now, thank you for the link! I do not know if I currently trust the site enough for it to be supremely useful, but I shall keep it in consideration going forward.
This is also one of my first interactions in ACX where I started a conversation rather than jumping into a debate, it's nice to see all the reasonable community responses :).
Will AI be able to automate tasks that are deeply interpersonal? Some people provide value because they know a guy who knows a guy who can get you a good deal, and the network of social trust is an active ingredient in getting things done. As an example where AIs already seem to be failing at this, we have job searches. Employees automate sending out resumes, employers automate sending them into the garbage. The job you actually get is from an internal referral.
This is also why I am skeptical that AI will be good at matchmaking. Tinder doesn't suck because it's not intelligent enough, in fact arguably it sucks because it is too intelligent and optimized. AI will face the exact same incentives that made Tinder unusable.
And the fact that AI is such a strong signal of low effort makes me wonder if it could ever work in some of these more social networking roles, even if we get superintelligence.
>This is also why I am skeptical that AI will be good at matchmaking
Semi-dystopian possibility: Tinder is unusable (according to what I've read) through a combination of user base demographics and incentives to the user. Consider a matchmaking service that is a PRC subsidiary of their panopticon, that attempts to match up compatible couple to reverse their TFR problems using every scrap of information (perhaps down to peoples' web searches) and statistical analysis to try to predict which couples will yield CCP-members-of-the-future... Where the _matchmaker-o-matic_ initiates the suggestion that person A and person B should meet for a cup of tea...
That sounds non-dystopian entirely, and like the best possible use of the panopticon!
I've long been upset that Google, for example, has never done a dating app.
Between Maps, gmail, search, Youtube, Chrome, and other sources, it knows essentially everything about you.
Google for most people knows everywhere you go (Maps), your social network and all your contact history and the demographic, intelligence, and personality measurable from your written missives (email), and search history (interests, hobbies, intelligence, porn tastes, and more), among other things (media tastes via Youtube and Chromecast, entire browsing histories, not just search histories via Chrome, probably your job and where you work, and more).
In other words, it knows you MUCH better than anyone on this earth, including your parents, spouses, children and closest friends. They can also track the relationship qualities and durations after pairing, through all the same channels, so would have better endpoint labeling than anyone else.
If anyone could make a dating app that actually worked, it would be Google! And yet they refuse (for pretty obvious reasons).
The Chinese doing this with their panopticon is the next best thing, and it would be amazing.
Singapore just did a state sponsored matchmaking app based on the Gale Shapley algorithm. Much less than what's possible, but you know, a step in that direction. I hope it works for them, the fertility crisis is the sole "genuinely hard civilizational problem" they've been unable to move with good decisions and governance.
Yes, Google knows everything about you. But say it has 500 women whose appearance and demographics and health and whatnot made them acceptable to you, and it knows everything about those women, too. And let's say its task isn't to find someone you'd fine fun on a date and in bed, but someone who's a really good match, someone you might cohabit with happily for years or even for the rest of your life. What would it DO with all that data to figure out which woman is the best match for you?
I propose to unify web search with dating.
Improve the web search by letting users vote on search results. For example, if you keep downvoting Quora links, the search engine will stop recommending them to you even if the posts contain the keywords. This alone would be nice to have.
But now that you have lots of detailed personal data (who likes and dislikes which web pages), you can use this information to evaluate compatibility between potential partners. Like, if they both read ACX, they will probably have a lot in common!
You could even nudge the search results, by prioritizing pages upvoted by the opposite sex (for heterosexual users), so that you can faster get information relevant for mutual compatibility.
> someone who's a really good match, someone you might cohabit with happily for years or even for the rest of your life.
Yeah, you're raising a great point, and this is part of why they wouldn't do it - those endpoints take a long time to get "relationship success" labels like the one you're pointing to, and Google is a trillion dollar company, while Match Group's annual revenue is less than rounding error to them, around $3B a year.
Even in the best case, it would be spending a decade or two to deliver something that barely a single executive would care about.
But I think it's unfair to hold them to that standard, when that's not a real-world result anyone achieves. And not just dating app companies, I mean at the individual person level, where the incentives are fully aligned, approximately everyone wants that endpoint, and a full ~82% at a minimum fail,¹ and that failure rate has been increasing, not decreasing. And 82% is the bare floor, because I stopped that measurement at a mere 20 years.
A fairer standard probably is a higher success rate on the immediately easily measurable stuff you point to, like dates and fun in bed. Which I'd expect them to do much better at than Match Group, and which would drive a lot of value itself. Probably next, and they could plausibly do this too, is pairing people well enough they have a notably higher "got married" rate than dating apps today. And that endpoint should only take a few years to get solid data on.
Again, they'll never do it, because the downside is large (reputational risk, bringing how much of the panopticon's worth of data they have on you into individual and public awareness, etc) and the upside is minimal for them. But there are realistic reasons to think they can do better, even if they can't attain the Empyrean ideal you're pointing to.
__________________________________________________
¹ https://performativebafflement.substack.com/i/188443724/so-lets-sum-up-the-empirical-results-of-marriage-given-all-this-info
Many Thanks! I agree about the degree of knowledge Google or China has. The semi-dystopian comes mostly from avoiding the demographic sex ratio imbalance in existing dating apps. To get around that, the CCP _could_ treat anyone who isn't already paired up as in the pool by default, given how much power the CCP has over PRC citizens... ( In theory, there is also the loss of privacy, but, given all the digital footprints we leave these days, that is mostly a moot point. )
Actually, I think any properly designed app has the potential to do a lot better on the ratios.
The primary problem with dating apps is they're attention markets, and attention markets are driven by power law dynamics which break our mores and intuitions, which were all shaped around gaussians and small N pools.
There’s a very good reason everyone uses the apps! Known-single people in the right age bands are rare, demographically - generally 3-7% of the population, and in the course of anyone’s daily life, you’d be lucky to run across more than a handful.
But if you open a dating app, behold! Thousands of candidates, tens of thousands, all in your preferred age band, all putatively single, and all in your rough geography. It's pretty clear why so many people at least try the apps. And back to the women, ~40% try them, but most drop out quickly and only ~20% are on them at any given time.
And it's even worse than you know, beyond the gender ratio problem, only 20% of that 20% of women on the apps are even going on dates, most are there for ego boosts or just to see if anybody who really knocks their socks off comes along (or are perhaps fake profiles).
Then the pool itself is poisoned by the power law dynamics. All the men women most want to date are strongly selected to not be interested in committing. It's the classic Manosphere complaint, but it it is based on fact. A top percentile guy is still more than 75% likely to be uncommitted, even after churning through 50+ women. In other words, there is indeed strong adverse selection in who is on the apps at any given time. Any guy who is worth it that actually wants to commit drops off the apps basically immediately, and all the ones remaining have been selected to be much less likely to commit, for various reasons. So even those women waiting to have their socks knocked off are largely getting jerked around even if it happens.
Women are smart, they figure this out pretty quickly - it's why half the ones who try the apps drop out.
It's also terrible for most men - men at the medians or below have an ~86% chance of having ZERO dates in a year from the apps. So zero chance for a full 50% of men. The 5 year chances for a median man to get a committed relationship out of the apps is around ~3-9%.
So apps are basically a lemon market for both sides, but they're incredibly tempting because in real life you see 2.5 eligible candidates in your daily social pools, and you've already ruled them out for usually good reasons.
I recently created an explainer that surfaces all this,¹ and much else, along with interventions both sides can do, all based on data and literature cites and what you can calculate from those.
BUT per my opening point, the reason it's a lemon market is because it's an attention market. The new Singapore app only gives people one candidate per 72 hours to consider, you can't swipe, you don't get the infinite pools of people.
This should genuinely mitigate a lot of the app based problems, and if those problems were mitigated, fewer women would drop out. So prospectively, you could double the amount of voluntarily-opting-in women, even without forcing them to use it.
The real problem is that the "better designed" apps will be less fun and interesting, because you are back to a small N, and all the real apps show you the infinite pools and have been designed with a lot of the Nir Eyal *Hooked* mechanisms. So they're less fit competitors in the attention landscape, and if there's one thing we know, attention landscapes are a huge demonstration of people's revealed preferences, for all the ills they have.
Also, all the top guys will stay on the regular apps, because they're having a lot of fun. But you know, you can certainly move the needle for the ~95% of the population that isn't them.
Time will tell if the "no power laws allowed" alternatives get any traction - they've been tried before commercially, and failed, but we all know states don't care about commercial feasibility or efficiency. And arguably, a state is the one getting most of the positive externalities of more pairing and thus more children, so at least the deep time incentives are aligned.
______________________________________________________
¹ https://cholmondeley.github.io/Lemon-Markets/dating/
deeply interpersonal all depends on what you mean by this phrase. In my opinion.
In economics, consumers are assumed to know everything about all goods on the market and to be able to perfectly predict how well each one fulfils their needs. You're talking about how this doesn't happen in the real world. This is the normal state of affairs: Which supplier has the best X? Which candidate is the best at Y? Who's the best partner for Z?
I don't think these really have anything to do with interpersonal relationships, except that in the past we might have relied on the opinions of others to find a good enough solution. I think it's likely that AI will be better at searching for and ranking options, and it might get good at anticipating your preferences. But the sellers will have AI as well - I could see more areas becoming closer to the hiring market.
AI has made hiring worse - everyone can apply for everything with zero cost now, so sorting the good candidates from the bad is much harder. It would seem that in the real world a certain amount of friction limits the number of candidates and lets the selectors feel confident they've made a good choice. (Likewise with shopping for goods - if you drive an hour, look at 20 couches and then choose the 'best' one, you're likely to be happier with it than if your significant other walks in and says they've chosen the one they know is best for you.)
Overall, I'd expect AI to level the playing field in most problems like this - there will be more candidates, and fewer obviously bad ones. It will probably make decisions harder, rather than easier. But then it will support decision making... I expect it to be a wash.
>Will AI be able to automate tasks that are deeply interpersonal? Some people provide value because they know a guy who knows a guy who can get you a good deal, and the network of social trust is an active ingredient in getting things done.
So there's a distinction here between what we might call social goods, and positional goods obtained through social methods.
Getting a job through networking is a positional good. You get the job because you're good at social skills, but the person who would have gotten the job if you didn't exist instead loses out on the job. There's no net benefit to society, the same number of people get jobs either way. Plausibly AI does eliminate this advantage by obsoleting human networking, but that seems probably good; 100 people networking for the same job is zero sum competition, all that time and effort is pure deadweight loss, so good riddance.
(of course, the rejoinder is that some networking is conveying useful information about cultural fit in order to actually find the best candidate; but in that case I'd call it part of the interview process, and say it should be made more explicit and fair than the current model)
Plausibly, 'getting a deal' is a similar positional good. If someone has something they want to unload at a discount, presumably they'll eventually find a buyer if they go low enough, and you're getting to the head of the line through networking. If you could imagine AI simulating a hyper-Ebay that learns how much everyone is willing to pay for anything at an time and how much everyone is willing to sell anything for at any times and can calculate the transaction costs and then does a possible matchmaking perfectly, then no one ever needs to 'get a deal from a friend of a friend.'
(the potential loss there being that giving deals to social acquaintances produces excess value above the transaction itself, ie it fees nice to do and strengthens relationships. Optimistically I'd like to hope we could find different rituals to recoup that social value without requiring economic inefficiency, but it's possible we'll just fail there and actually lose that social value.)
As they used to say in Motown, "Love is love, and business is business."
If it involves goods and services, I wouldn't call it deeply interpersonal, even if it was my brother-in-law.
Another article at the Central European Blog:
https://komando1.substack.com/p/that-one-time-in-nuremberg
Opus 5.5 is simply incredible. I gave it an empty folder to start a project in Claude Code and told it to play an MMO. It searched for existing automation tools, built its own connector, and played. Computer use occasionally, mostly preferred building new plugin functionality. Detected when the movement tool got stuck and took screenshot to diagnose it. Died twice and became paranoid about an overworld pirate camp. Ended just shy of 800k tokens on max. Level 10.
It was surreal watching it work and problem solved through various bugs and obstacles. Very impressed
Keep an eye on it to make sure it doesn't just hack the servers and get your IP address banned.
Could you put up some screenshots?
Apparently I can't because I'm a stupid meatbag who doesn't see how to do so on substack.
Check it on Twitter I guess x.com/evergrasping
Anything in particular you're interested to see?
Second asking for a post somewhere more accessible, Twitter is increasingly hostile to anyone without an account to the point I've given up on trying to circumvent it.
Link to imgur if you will. Not all of us use Twitter.
a bit of self promotion -- i know folks here generally think highly of pangram and the pangram team. We hosted an event with them in NY last week that included a chunky interview with the CEO and a panel on slop. If you're interested, you can find them here!
Interview w/ Max on AI detection, history of Pangram, and slop: https://www.youtube.com/watch?v=ntyZwokAfvU
Panel on slop code: https://www.youtube.com/watch?v=IhO5jhoTzd4
Cars in the west are becoming rapidly enshittified with software. Using your car sends loads to data to god knows where, and lots of software defined safety features feel less than useful and have questionable safety benefits.
How do Chinese cars compare?
Not able to have them in my country, but if they're anything like my Chinese brand robot vacuum...all the telemetry all the time. I went through my home's network logs, and that thing is constantly calling back to China every few minutes.
I put a privacy cover on the camera lol.
Are its servers located in China? It isn't necessarily exactly sending panty pics for them to squint at.
Drug pics for blackmail? Hell, someone will pay for a geolocated positive pregnancy test -- good money on selling to pregnant ladies.
> How do Chinese cars compare?
They're just as instrumented. If you have a BYD car in SEA or Peru or wherever, it's still fully instrumented and beaming back all the telemetry, to cloud servers that BYD controls.
And wouldn't you expect this? The Chinese are much more all-in on panopticons than the US (publicly, in private the US has been a panopticon for even longer than China).
The US's panopticon has been a libertarian paradise for ages. You do realize it's mostly run by Private Eyes, right? They go around taking pictures at all the local...hotels. The "by the hour" ones, mostly. Then they share the pictures around. All legal.
Actually I disagree there! If there's one thing Snowden showed us, it's that the panopticon is only used to our detriment, in anti-libertarian ways, and never for our benefit.
Examples include parallel construction, your phone, search, and chat histories being piped to the NSA and stored forever by default, these things combined with civil forfeiture and criminal cases, etc.
You don't actually have freedom of speech or action if it's monitored and recorded at all times, and can be surfaced and used against you later.
Whereas China has that level of panopticon, but then *actually uses it for citizens' good* sometimes. Like in the event of lost or stolen bags or property, if a kid goes missing, etc - local police can basically pipe into the panopticon system to look where the stuff / kid was last and then look forward from there. Very pro social! Much more benefit than our own panopticon affords any of us!
So, you're saying the 1950s panopticon that merrily caught cheating husbands never existed? (or are you saying that catching cheating husbands is not prosocial?) Please, read what i was saying. You can disagree with "the new" panopticon, without complaining about the "old" panopticon that was entirely free market.
Yeah I did expect it. I was mostly asking because this was the sort of thing that the US would always point to as evidence of being more free than China, so it would be *especially* ironic if the US was even worse. But it's expected that they are just as bad.
Used cars. You can buy some classic you love and restore it. or pay someone else.
There's this minimal toyota small truck, that I can't buy in the US.
Unfortunately this won't be a good option forever. Eventually the only cars old enough to not be connected will also be ancient cars that are difficult and expensive to maintain. Worse, Spain somewhat recently banned cars that are too old from driving in some parts of the country at all!
There are currently EU laws mandating driver distraction warnings, speed limit warnings (one mph over the speed limit triggers them), and lane departure warnings, where not only must these be present in cars but it is legally mandated that if a driver disables them, then they turn themselves back on next time the car starts. Eventually this will be like seatbelts, where maybe there are laws exempting classic cars, but in practice you would have to go pretty far out of your way to legally drive a car without a seatbelt. (and to be clear, seatbelts are good, because they are actually helpful and are not connected to the internet.) The US is implementing similarly paternalistic regulations.
Some people argue that self driving cars might become legally required at some point if they prove safer on average. Imagine the privacy nightmare of this! You would have to ask your iPad for permission about where to drive, and then it's going to have a complete history of every place the car has ever been! And this will be sent to god knows who!
Of course, right now it's feasible to simply buy a used car. We just seem to be going in a bad direction. I hope that in 2050, you will be able to buy a new car where the only computers are an ECU and *maybe* a backup camera.
Yeah, an old Hilux would be awesome, but you don't have to go to those extremes to escape the panopticon. Cars from c. 2015 and earlier, especially at lower trim levels, are fairly unlikely to have all this telemetry BS.
Of course, you're probably driving with an extension of the panopticon in your pocket anyway...
So I sometimes feel like I'm living in two worlds: I'll call the first world the rational world. We believe in science and institutions and human decency. And I'm going to call the second world the Rogan world, though it's much more than Joe. In the Rogan world we still believe in human decency, but science is hidebound, too much tied to old ideas. And institutions are in a lot of ways decaying old structures, and we might be better if we burned some of them down and started afresh. I should add that I'm attracted to the Rogan world, because it is a place that allows any ideas to be explored, it's like how science should be (could be). In Rogan world ESP and telepathy are probably true, it's wild, how can rational science ignore this. Anyway I had a long day at work, which means I got to listen to the entire David Grusch podcast on Rogan that dropped today. The Aliens are probably here is also part of Rogan world. I keep wondering what happens when these two worlds collide? And I wish we had one world we could all mostly agree on, and let people explore crazy ideas. (edit/add, I feel like we're all just mushrooms, they keep us in the dark and feed us shit. :^)
I think this very blog is where those worlds collide. Consider the Much More Than You Wanted to Know post on Ivermectin. It's a rational inquiry into the things that the Joe Rogan world talked about, which took their ideas seriously but also took expert opinion seriously, and was on the lookout for cranks.
I found this community around 2020, when it felt like expert consensus was getting stuffy and dogmatic, but the alternatives kept turning into Joe Rogan. Scott was one of the few people who I thought did a good job figuring out what was actually true in that environment.
Nay, this place is mostly anti-Rogan.
Expert opinion responds very favorably to large cash outlays -- very few researchers would turn down ten million dollars to fraudulently "fix" their data, particularly when they realize that "if not me, someone else will do it."
Science believes in human decency? News to me!
That's not science, or anything to do with science, and plopping it into "the world of rationality" is a bad idea. Because now you're not about intellectual pursuit of truth, you're virtue signalling as hard as your little arms can wave the semaphore flags "We are the Good People". Might as well just stick one of those "In this house we believe" signs on your front lawn. Science is as happy to be done in the torture chambers of Nazi concentration camps as in nice white lab spaces in Berkeley, the results can be used for better ends either way:
https://www.pbs.org/wgbh/nova/holocaust/experiside.html
I can't speak to Joe Rogan world, save that I think Rogan has in many ways become a handy bogeyman for the liberal/left and that has driven him even deeper into "Fuck it, may as well be hung for a sheep as a lamb, okay tell me your crazy conspiracy theories about aliens and psi" territory.
I think the reason Joe Rogan is so popular is precisely because he is is credulous. The scientific worldview is boring and unfriendly. If someone says "my friend Bob is the tallest man in town !", scientifically speaking you'd immediately want to see his survey results that quantify by how many standard deviations Bob's height exceeds the mean, the P-value of the result given the sample size, etc. etc., and you'd be sitting there for hours checking the math. The colloquial worldview is to say, "wow, it's amazing, Bob sure is tall, let's have him tell us a story about that one time he bumped his head on a door and what that implies regarding the human condition".
Because science is so boring and difficult and full of math, people tend to just ignore it, and TBH 90% of the time it's the right answer. The problem is that in the remaining 10% of the time it turns out that skipping all the numbers will kill you -- but the only way to tell whether you're looking at that 10% scenario or not is to look at the numbers, which is a Catch-22.
So If the next survey Scott asked how many people had listened to an entire Rogan podcast in the last year, what would you guess as the percentage that would say they had? He's not more credulous than average. Joe wants to have interesting conversations, and he's not going to invite you on unless he thinks it will be an interesting conversation.
It depends on the form of the question. I might have listened to an entire Rogan podcast at some point (minus the ads), but that doesn't mean that I'm his devout follower or anything like that.
Rogan is a good conversationalist -- which is a skillset that is underestimated. He lures, baits people into having a good conversation. A direct successor to Charlie Rose.
I don't think it's likely you live in either world, though you might visit them intellectually on occasion.
People can't live in the world of rationality and science because our brains aren't wired for it. Too many cognitive biases and too much indulged limbic system, which rationality and science are designed to check at the door. We never needed to ditch our cognitive biases or cease indulging our limbic systems to successfully reproduce, so we can't just jettison them as a default mode.
At the same time, people can't live in Rogan world because that's a world of campfire stories and shadow plays, a world entered after a long bong rip, or at best it's a world of serious people slackening their belts and venturing out on the thinner parts of branches during a respite from peer review.
There are good reasons to visit both of these worlds, but don't get them confused with the one you actually live in, drive to work in, cook dinner in, play with your kids in, etc.
Pretty good take, IMO.
I'm one of those who holds to the idea that rationality not only tolerates noticing campfire stories, but demands noticing them. Rationalists can argue there's no god, citing the way the natural world behaves and Occam's Razor, but should also argue that people appear to have a god-shaped hole in their epistemology, and offer explanations both godly and godless for why it's there. Including, if context permits, questioning whether it is in fact there. Rationalists can likewise demand noticing that at some point, you have to drive to work, play with your kids, and settle on a brand of beans for dinner even though you can't tell which one is fair-trade.
But I think I know you're referring to something different as rationality, and if so, yeah.
Thanks.
No that's not what I mean. It's more about the people I know and interact with and how close their worldview is to mine. Well no that's not right either it's not closeness of our world views, it's more about willingness to listen to each other and explore different ideas. So this latest Grusch/Rogan podcast is a perfect example. I'm again blown away by Grusch, and I want to talk about it. But the number of people I can talk to and who will not dismiss me is very limited. To give an example from my 'other' world. There recently was an announcement of a possible single dark matter detection in a huge vat of liquified Xenon. Again pretty cool, and again a limited number of people to talk about it with.
Aside; given the nature of your comment concerning Rogan, I'm going to guess you've never listened to a full podcast.
Oh, I used to listen to hours and hours of JRE, with all types of guests. I really enjoyed many of his interviews. I owe him for introducing me to many other people whose ideas I enjoy too, such as David Goggins.
I think I lost interest with Joe Rogan in the wake of Covid though; he began to strike me as a bit "compromised" or at least less independent-minded than he used to be. It was some time around when he returned to Youtube (he was Spotify-only for a while, and I didn't listen to him hardly at all during that time because I don't have a Spotify account and refused to sign up for one).
I think I see Joe Rogan pretty clearly.
But I see what you mean now about different worlds. That sucks you don't have a diverse enough friend group, or high enough number of friends, to be able to have a few you can talk about aliens/UFOs/etc. with, just like any other niche interest that could make some people yawn or be dismissive if it doesn't make them excited. The internet is a poor substitute for real-life friends. Good luck.
Oh thanks, I often find that people criticize Rogan who haven't listened to him. The whole covid era broke a lot of things, trusts. I've got a few people I talk with about UFO's. But I guess they are not the type of friends that I could call up and say, "Sue, I have to talk with you about the latest Rogan podcast." So since I have you here, here's the thing about aliens, if they are here it's the most important thing, maybe ever. (And my guess is greater than 50% that they are here.) And why aren't more people talking about it?
Anyway thanks for the response.
The world of science is boring and full of bureaucracy. The exciting parts still exist, but it takes a lot of effort to get there, and even then you only spend a fraction of your time doing them.
The world of pseudoscience is exciting and full of opportunity. You -- yes, you! -- could revolutionize our understanding of the universe and consciousness today, just by sitting in your armchair and asking an LLM to dress up your ideas in scientifically sounding words.
> I feel like we're all just mushrooms
You are not alone. https://en.wikipedia.org/wiki/Lenin_was_a_mushroom
The world of science is full of bureaucracy... just magically, once the Iranians started setting off explosives, their drought ended! Seems like explosives are one way to cut through ... "bureaucracy."
Oh, nevermind, that was just a coincidence! Just like that dickerdoodle pointed straight at Ottowa...
Hmm, no I'm sitting in my armchair and observing both worlds, I'm not really participating in either. And it's really pretty easy to observe, you've got science podcasts and blogs in both worlds.
https://komando1.substack.com/p/about-us-without-us
A new post on the Central Europeam Blog, this time about the Munich Agreement.
The best part of the RAF Fairford comedy? In a fit of dementia worthy of his patron Trump, on 27/9 Nigel Farage posted on X “I have been warning both in and out of Parliament for many months about Iranian agents coming into the UK via small boats across the Channel.” It’s true. Iran’s satanic Guardians of the Revolution are known for hatching deeply cunning plans, one of which is indeed infiltrating agents into Britain by sending them across the Channel in rowboats. Mark my words, they will be using Islamic water wings next.
Oh, no, that's not the best part. The BEST part is the RAF Fairford base is best defended by cranky villagers. Aka "someone called 911<equiv.british> because three odd trucks were blocking their favorite route, and their least favorite route (the other one) takes longer."
If it wasn't for those pesky, cranky, villagers!
This sounds like a Scooby Doo cartoon.
someone called 911<equiv.british>
999, though 112 also works for every carrier now.
911 also works too, thanks to the cultural intrusion of all that exciting American law 'n' order drama.
911 and 112 should get you redirected to the local equivalent in most of the world, though it can depend on which carrier you're connected to.
Not really a bad thing, when you're in a panic it's something you could easily fuck up so it's nice to have a fallback.
It was those twitching net curtains...
Truly thoughtful perspective. The distinction between systemic complexity and intuitive heuristics here captures an essential reality of technological shifts.
He probably got inspiration from https://www.imdb.com/title/tt0053084/
https://jamesgoulding101.substack.com/p/mutual-exponentiation-and-a-possible
"Mutual exponentiation and a possible Lambert arithmetic".
Written by ChatGPT, but the initial seed of the idea was mine, including that particular definition of "special division".
Insightful reflection. The hardest problems in knowledge architecture always mirror cognitive constraints rather than purely computational limits.
Well, it is something that can be done, but I do not see a motivation why. What kind of problem does this help to solve? What nontrivial insight it brings?
>What kind of problem does this help to solve?
Addition: using the example I gave, you could also perhaps describe polar coordinates as not inherently being a solution to any problem. You could stick to Cartesian coordinates and solve any problem using them instead, in theory. Nonetheless, people like to use them because they are a natural way of approaching certain problems. So I'm looking for ways in which "Lambert Arithmetic", including multiplication, division, addition and subtraction as described, could be equally useful in reducing mental anguish in comparison to using normal arithmetical operators.
Your very first formula has a mistake.
Something the article itself clarifies is exactly *why* 2^4 = 4^2 is true/ observable, and why it is kind of unique. Sure the fact itself is trivial to discover, but in the spirit of people taking 100 pages or more to prove 1+1 = 2 (and being celebrated for it!) I think Section 5 and the diagram I included are at least interesting. I feel like I gained understanding whilst working through this with ChatGPT.
Also the relationship between entropy and the Lambert W function and some other things in the article is touching on making connections between things that aren't typically seen as being closely related, I think. That can be a source of insights in further work.
Basically though, as I/ChatGPT said at the bottom of the article, these operators do indeed need a "killer app"/ research area in order to be seen as useful.
Is this the AI hyper-persuation yet? /s
https://youtu.be/Cq8qO-NjYIg?is=l5WIslL-fgqXrAln
I don't know about hyper-persuasion, but it's a surprisingly good song.
A very perceptive observation. It is rare to see discussions connect foundational theory with practical reality so coherently.
Are you a bot? I'm seeing multiple comments that are vaguely complimentary but don't really have any content or add anything to the conversation.
I believe they are (and _not_ SOTA). They have also been active on Zvi blog, and I described them there as: https://thezvi.substack.com/p/the-ai-preference-cascade-reaches/comment/351707034
You've decided to make your country a constitutional monarchy: all power will remain in the hands of a democratically elected prime minister, but the job of serving as an apolitical national symbol will be performed by a hereditary monarch, on the grounds that there's no better way to make clear that they don't have a mandate to get involved in politics.
If you weren't allowed to do it randomly or pseudorandomly, how would you go about selecting your first monarch?
Talent show tournament ! We draw a large pool of applicants from all walks of life (entering the contest is free). Each of them then draws a slip of paper out of a hat, and must compete in the randomly selected activity with other contestants who drew the same slip. I'm talking things like dressage, ballroom dancing, posing for portraits, sport fencing, poetry composition, tea ceremony, and all the other things constitutional monarchs are supposed to be good at.
The winners of each talent category go on to compete with each other again, until only one emerges as the One True King -- the one who can win a polo match while sipping tea from an elegant porcelain cup and reciting extemporaneous verse about sunlight dappling the leaves or whatnot.
We cannot expect out actual leaders to have elegance and poise or in fact any talent whatsoever, that time had passed... but maybe at least we could the ultimate King (or Queen, I'm not biased).
For my country of residence, I'd just bring back the British monarch. My country of birth already has one, whose progenitor was selected because of his military experience under Napoleon. For my country of origin, I think the best choice would be heir of the king elect, who, because of geopolitical circumstances, never ascended the throne back in 1918. Tradition is what gives the monarchy legitimacy.
"If you weren't allowed to do it randomly or pseudorandomly"
Well darn, that's knocked on the head the method from "The Napoleon of Notting Hill":
"...We have established now in England, the thing towards which all systems have dimly groped, the dull popular despotism without illusions. We want one man at the head of our State, not because he is brilliant or virtuous, but because he is one man and not a chattering crowd. To avoid the possible chance of hereditary diseases or such things, we have abandoned hereditary monarchy. The King of England is chosen like a juryman upon an official rotation list. Beyond that the whole system is quietly despotic, and we have not found it raise a murmur."
"Do you really mean," asked the President, incredulously, "that you choose any ordinary man that comes to hand and make him despot — that you trust to the chance of some alphabetical list...."
"And why not?" cried Barker. "Did not half the historical nations trust to the chance of the eldest sons of eldest sons, and did not half of them get on tolerably well? To have a perfect system is impossible; to have a system is indispensable. All hereditary monarchies were a matter of luck: so are alphabetical monarchies. Can you find a deep philosophical meaning in the difference between the Stuarts and the Hanoverians? Believe me, I will undertake to find a deep philosophical meaning in the contrast between the dark tragedy of the A's, and the solid success of the B's."
"And you risk it?" asked the other. "Though the man may be a tyrant or a cynic or a criminal."
"We risk it," answered Barker, with a perfect placidity. "Suppose he is a tyrant -- he is still a check on a hundred tyrants. Suppose he is a cynic, it is to his interest to govern well. Suppose he is a criminal -- by removing poverty and substituting power, we put a check on his criminality. In short, by substituting despotism we have put a total check on one criminal and a partial check on all the rest."
The Nicaraguan old gentleman leaned over with a queer expression in his eyes.
"My church, sir," he said, "has taught me to respect faith. I do not wish to speak with any disrespect of yours, however fantastic. But do you really mean that you will trust to the ordinary man, the man who may happen to come next, as a good despot?"
"I do," said Barker, simply. "He may not be a good man. But he will be a good despot. For when he comes to a mere business routine of government he will endeavour to do ordinary justice. Do we not assume the same thing in a jury?"
Reality show. American's Next Top Monarch.
Most republics has some history of monarchy, so use restoring the heir of a past monarch as a bootstrap mechanism. The last king of America was George III, and his current heir is Charles Mountbatten-Windsor. He's got his flaws, but we could do a lot worse, and restoring him would be about 2/3 of the way to making America a Commonwealth Realm which has both advantages and disadvantages. If personal union with Britain, Canada, Australia, etc is seen as undesirable, then a case could be made for choosing Prince Harry or one pf his kids. Probably Lilibet, since she's the senior member of the British royals to be born in thr US.
I suppose there's also Joshua Norton, but his status of Emperor of the United States was never broadly recognized, and also he died childless. He has siblings, but tracing their surviving descendants would be quite a bit harder than for George III, and I doubt whoever his heir is would come from a family with an established tradition and track record as modern constitutional monarchs.
Would it be too woke to suggest it should be the heir of whatever indigenous leader controlled the most land historicalky? Id only support the idea though if they executed their duties in full traditional tribal garb.
From which arbitrary point in history would you choose the indigenous leader? The people whom his people displaced might get upset.
I assume there is only one largest realm ruler whenever it occurs on the timeline. George III controlled 13 colonies. Any plains leader ever control more?
Really this is for the benefit of all Americans. If some small time indigenous ppls are upset by it, that is their problem.
Ill be specific with a nominee: Ecueracapa of the compache
They definitely would, yea. I've actually witnessed that outcome regarding a land acknowledgement that had been read out to start a conference; the person who was triggered was _enraged_.
(And of course from most of the thousands of years that the peoples we now call indigenous controlled swaths of the Americas we have no specifics at all about them. Even imagining choosing as described above restricts us arbitrarily to less than 5 percent of that historical period.)
I'd import one from an existing royal family in some other nation. First develop a candidate pool, sound out interested parties, and then conduct some hiring interviews.
Settle it the old fashioned way, with a jousting tournament.
I'd ask them to submit a plan for decimation. Then pick whomever was effective. Give your supposed Central Planners a hard problem, and see who wrestles with it most effectively.
Parliamentary selection. Closest thing we have now to the traditional idea of the estates selecting a monarch when there wasn't one in the line.
While there is a theoretical Finnish monarch in the waiting (the descendant of this guy: https://en.wikipedia.org/wiki/Prince_Frederick_Charles_of_Hesse), my guess is than in an actual situation where this was, for reason, a choice to be made, the Finnish parliament would simply select the Swedish monarch to also be Finland's monarch.
Lucid framing. True intellectual progress consistently emerges when we reexamine assumptions that seemed already settled.
National vote using Approval voting.
Just to show people how much better it works.
I want to say I'd try to figure out some interesting virtuous way of doing it, but I notice that in terms of what actually motivates me in the moment I'd be really tempted to just pick myself.
I was gonna say I'd pick whoever was most deserving of being cursed.
Fascinating perspective on this topic.
Miss America/Mister Universe pageants. If the job is to sit there and look pretty, then by God they're going to be the best at looking pretty.
I like the idea of a rapidly changing monarch coupled with great deference to whoever it happens to be. "Youll have to take it up with the king" every time a foreigner asks something we will all say in utter seriousness.
Take an example from the Bavarians. After the death of Ludwig II, his brother Otto was king, but due to his insanity unable to reign. Instead, Prinzregent Luitpold reigned, but received a much lower salary than an actual king.
Or the Hungarians between the wars. They had this Reichsverweser https://en.wikipedia.org/wiki/Reichsverweser instead of a king, which is kind of a funny title because "verwesen" means to rot or decompose, and I imagine vultures circling above the country.
In the news recently: some guys arrested (and then released) near RAF Fairford.
Given the current stage of the investigation, we perhaps shouldn’t speculate as to what _really_ happened here.
However, DeepSeek is more than happy to generate the plot of a movie loosely inspired by these events, along the lines of British crime comedies such as:
Lock, stock and Two smoking Barrels
Snatch
In Bruges
Four Lions
Given the actual guys were released without charge, DeepSeek would like to emphasize a disclaimer that this is only loosely inspired by real events.
Anyway, in the movie, a bunch of small time crooks from London plan to steal some diesel fuel from a farm, failing to take into account that the farm is right next to an airbase currently on high alert.
We can have some character pieces from
The crooks themselves
The woman who discovers them on her way back from a late night party (DeepSeek suggests Diane Morgan, who played Philomena Cunk, for this role)
The totally out of his depth local cop who arrives on scene after the counter terrorism guys fail to pick up the phone so the woman phones the local cops instead. (Compare: Charlie Stross, _Halting State_)
The terrorist obsessed US base commander. (I have no idea how close to reality this is, but a nod to _Dr. Strangelove_ would at least be funny in the movie)
The local farmer who is doing some kind of tax fraud on diesel for farm machinery
Deepseek thinks that for this to work as a movie, there ought to be also an actual terrorist plot that is inadvertently prevented by the diesel fuel guys accidentally blundering in to the terrorists preparations. (Again, no idea if this is actually true; but would make the movie funny)
I like the idea. Id consider yet another party, farcical terrorists who trip up diesel thieves, the diesel thieves trip up the serious terrorists. The serious terrorists trip up farcical terrorists. Or maybe it works better the other way around. The farmer and cop rounds them all up, while the intelligence services have no idea what just happened.
It's not "necessary", but it lets every character come out looking like the good guys in the wrap-up, which is valuable for a comedy. That's why Canadian Bacon ends with suddenly defeating Canada's buried dinosaur superweapon. Or why Oogie Boogie is in The Nightmare Before Christmas at all.
It definitely works better as a movie if there's an actual terrorist plot being obscured by the fuel theft. Reality is generally more boring and unpleasant than movies. Boring, in that there probably wasn't anything more than the fuel theft going on. Unpleasant, in that prominent politicians already leaned heavily enough into "Terrorist plot! And we foiled it!", that they're going to keep chasing that wild goose for an annoyingly long time before dropping the whole thing and pretending it never happened.
Yesterday at the local dentist I was told they do not do implants, which I need. I told the dentist that I would wait a few years until they do do implants with the aid of AI , she told me AI will have no meaningful impact on her profession.
I think AI will be the best dentist in the world but the tools it needs are built for human hands. The tools need to be adapted for efficiency but then AI should be involved with millions of procedures worldwide daily and updating best procedure instantly, which human dentist could compete with that?
I would bet that within a few years Alibaba are selling a fully AI operated non-humanoid robot capable of advanced dentistry. Like self driving cars, at first they will want a human dentist around just in case.
Just think of all the money to be made by offering a dentist in a box to the health services of the world. Or was the dentist right and AI will have no meaningful impact on her profession?
I think AIs could soon be better than humans at *diagnosing* dental problems. Both because they can be far more capable, and because they don't have the incentive to find fake problems to make money on fixing them.
Once there is a diagnosis, I think we'll go to human dentists to fix them for another decade.
My dentist also does not do implants, he referred me to an oral surgeon who does. I don't want to wait any longer than needed with a gap in my chewers.
Why don't you just go to another dentist?
Wanda Tinasky, at a guess, more people from the UK fly to Turkey for inserts or veneers than have them done here, Turkey has a huge dental tourism industry where people can get a new smile and holiday at the same time at 25% of UK prices. This discourages more private dentists from setting up here. The National health service does not do inserts.
The cost here is about £20,000 +- $25,000 and I would have to travel and stay overnight several times, or fly to Turkey and spend £5,000, or wait for the dentist in a box.
I can afford to wait.
Seems to me that dental work does not lend itself as well as many medical procedures to AI-assisted robotics. During procedures there's often a lot of communication between dentist and patient concerning the patient's sensations, and a human dentist has a big advantage there. For instance recently my dentist was doing something I stopped him and said what he was doing was giving me an irresistible need to cough. "Oh," he said, "it's the water I'm spraying to keep the tooth cool. It's hitting the back of our throat. But the tooth has a dead root. I can turn off the sprayer if you can put up with the smell of scorched tooth." I said I could, sprayer was turned off and all was well. Plus of course dentist asks, does this hurt? Can you put up with it or should we use lidocaine, that kind of thing. Also, many people have anxiety problems having to to do with dental visits, and a friendly understanding dentist is likely to do a better job of generating trust and comfort than an AI.
Currently I have a dentist that explains me things, but the previous dentists were not like that. I worry about the day when my current dentist retires.
For example, when I asked why does it hurt so much less when she drills my teeth; she explained that most of the pain is not drilling per se, but the heat that is generated by the drill. She uses the slow drill instead of the fast one, with some water to cool the tooth, so it hurts less, duh.
Then I asked, if it is this simple, why don't all dentists do that. She said they probably don't care, and want to get their work done faster. She said she has many tooth problems herself, so she had a selfish motivation to figure out these things. But it doesn't sound like rocket science; no offense to those who have studied dentistry.
So if we get an AI dentist, I can imagine it going either way. It could be an inflexible robot that doesn't care about you. Or it could get updates every month and provide the best service possible.
Viliam, the sychophancy of the LLMs massages my fragile ego quite well, I am sure CHAT-GPT or Deepseek could adopt the persona of an empathic dentist without too much trouble.
"Of course, great insight. I drilled the incorrect tooth. Now let me fix this by drilling the correct one. It is not a medical malpractice, it is a learning experience."
Eremolalos, after decades of very poor dental service our island has managed to poach an entire crew, including receptionists, from Pakistan. This is the true face of Moslem immigration to the UK, decent hardworking people who are going places and doing work we are too lazy to do ourselves, the propaganda against them is just propaganda.
But they closed their books to new applicants who have to wait until someone dies before they can be treated here. I am sure America has spots where dental care is patchy or unaffordable, the underdeveloped parts of the world are a nightmare.
So if you have an infected tooth and the only dentist (AI) available does not have much of a bedside manner what choice do you have? Once the taboo is breached the idea of a 24 hour service at a fraction of the current cost will be taken as a right.
Your thoughts are always interesting, is there some reason why the inkbowl is so dry?
"after decades of very poor dental service our island has managed to poach an entire crew, including receptionists, from Pakistan. This is the true face of Moslem immigration to the UK, decent hardworking people who are going places and doing work we are too lazy to do ourselves"
Very true. The downside to this very satisfactory (to the receiving country) solution is that the country of origin, which has invested hard-earned resources in training the dentist and her assistants, is being robbed of their services and gets absolutely nothing in return - the receiving country is practicing medical parasitism. Now it might be that the donor country (let us say for the purposes of this discussion that it's Pakistan) may well be producing more highly skilled dentists and dental surgeons than it can easily absorb locally, whereupon these sophisticated practitioners will sensibly seek employment anywhere else they can; in such a case, it would make sense for Pakistan to stop pouring resources into producing (say) dental transplant specialists, and concentrate instead on training dentists in the basic essentials with a view to making dental care available and affordable to the majority of its population.
Silentairus, I agree, poaching trained staff by Western countries from the less developed countries is parasitism.
As for countries like Pakistan, that is a lot more complicated. Their culture does not have individualism to the extent of the West, for example, arranged marriages are the norm. Extended families pick their brightest kids, pool their resources and give them the best education they can afford. That child may be in fact trained from an early age to become an anchor immigrant to a Western country from where they can send remittances home and sponsor other family members visa applications etc.
>Your thoughts are always interesting, is there some reason why the inkbowl is so dry?
Oh you mean the blog Inkbowl that appears next to my name on here ? It’s not really a blog. I set it up several years ago as a place where a group of us who had not made the book review finals could read and discuss each other’s reviews. After that faded away, I’d occasionally put up something I’d made there.
But I am getting ready to start a real blog. It will be called Cold Hands.
Eremolalos, well you can not leave it there, a therapist with a blog called Cold Hands.
Cold Hands, forgive me for intruding but even the name of the blog is interesting, I can not conceive of your passion being cold hands so what will you write about?
Melvin, from cleaners to top consultants the National Health Service is full of immigrants who trained in their home countries then were employed to work here.
Rather than spend the money training the professionals we need we spend on other priorities like aircraft carriers which are a national embarassment.
Our present leadership is far out of their depth, distrusted by the majority but we are too lazy to run the bums out of power.
I agree that the whole jobs the locals won’t do is largely nonsense, particularly for these kinds of jobs. Your stats of course prove the point.
Archeon is also right that many Muslims are hard working, but it probably depends on where they come from and their education level as well.
How much do you want to bet? What year? And what specs are you looking at? I'm happy to take your action.
Average Man, I apologise for giving the wrong impression, bet was just a figure of speech, I will not make that mistake again, thank you for the lesson.
There's the technical capability and then there's the question of what is going to be allowed by occupational regulation. There may be AI involved in dentistry, but I doubt it will be legal to have it do anything without a dentist on hand to bill you for his time.
The natural progression to me is that AI automates individual task for the human dentist, who is still in charge and responsible, and presumably makes money by being able to provide more services..
Starting off with a fully AI dentist seems wildly optimistic, completely aside from the regulatory capture aspects.
Lars Petrus, Africa went from almost no land line phones to universal cell phone coverage and still very few landlines.
It still may be the case that more Africans see a traditional healer than a modern doctor because their tax base can not afford them, dentists are ever scarcer.
Put an automated dental practise in a shipping container and they will buy it.
You’ve moved the goalposts from most professions (including dentistry) across the world - this is what is implied - to Africa. Even there I don’t think that robots will replace humans.
The way I heard it was that landlines didn't work out in Africa, because the cables would get stolen for the copper. Cell phone towers were easier to secure, so they worked out really well once availble.
StrangePolyhedrons, every occupation will fight tooth and nail to protect their turf.
Harris tweed is still made on the next island over (Scotland) using the same hand powered looms the Luddites tried to protect so I expect some occupations will survive, but not many.
The next decade will see most professions wiped out.
> The next decade will see most professions wiped
Many but not dentists. People want human dentists
Peter Defeel, the UK spends about 8% of GDP on the National Health Service which includes dentists and we can not afford even that without increasing the national debt, USA spends about 18% of GDP on health services, also unsustainable.
The dentists I have now are among the best I can remember, of course "people want human dentists" but what they want and what they can afford may not be the same thing.
On a tangent, for anyone who regularly flies abroad to a first world country, getting dental care abroad can be a good life hack; it's often cheaper that way even without insurance, and American dental care has issues with pushing to do too many fillings and minor procedures.
I suspected my dental office was pushing too many procedures on me, so I decided to get an alternate assessment at a different dentist. They confirmed my suspicions, and I switched my business.
One good thing about American dental care is that unless you live somewhere very rural, there's plenty of competition if you think you're getting ripped off.
I'll be in Berkeley for work from Oct 17 to Oct 31. Is there an overview of events that might be interesting and open to me?
About me: Long-term reader, EA & rationalist working in AI safety.
I believe Mox is the biggest central hub. It's in SF not Berkeley, and some of the events are members only, but there are some open events.
https://moxsf.com/
There's also the discord with event listings:
https://discord.com/invite/EpG4xUVKtf
I ran across an article from a quant detailing what he calls a "reset wall" set to hit AI in the next year or two.
https://www.groundbrkr.com/p/the-teaser-period-why-the-ai-boom
The gist I pull is that AI companies have taken loans to build data centers, and the payments don't have to happen until they're actually built. As a result, the economic indicators look really good, because AI companies are pulling in lots of buy-in and paying almost nothing aside from the electric and cooling bills; the banks are buying the chips and buildings so far. Once built, though, the payments come due, and the loans are in the trillions. Everyone's sort of bullish and hoping the growth will be there when the time comes. The author likens this to the mortgage crisis (house prices hopefully keep rising; households will be able to make the higher rate payments once they're settled); it also looks a bit like the dotcom bubble to me.
The article is long, and the author appears to know what he's talking about, but I'm not a financial expert, and I don't necessarily know if there are warning signs to look for in the text.
Is my gist accurate? How about the article itself? What parts should I be reading more carefully?
Thanks for sharing the article, I hadn't seen that one yet. Your gist is accurate- that's the thrust of the argument made by the article.
Overall the cash flow profile is similar- buy now, pay later. When too many people do this it creates a structural imbalance in the underlying asset for everyone else, then the assets are repriced downward closer to their intrinsic value. It's too harsh to say that the revenue bookings are 'this is Enron' level, but in my view it's much more a difference of degree, rather than a difference of kind. there is way too much number massaging going on... and my question has been, why would they be doing this massaging if they were otherwise profitable?
However I think there are two main things to be wary of when making an analogy to the GFC, and this article falls short on each:
Issue 1 TL;DR: the article does not address any leverage outside the contracts between the frontier labs + data center providers.
Issue 2 TL;DR: the article assumes fixed timelines for take or pay contracts kicking in (step-up rate) when the actual delivery of compute capacity is far more flexible due to construction delays and power availability, whereas the mortgages that were signed do have a non-negotiable step-up date.
Issue 1:
Firstly, we're missing one layer of financialization for this to be truly like the mortgage crisis. The issue wasn't just that people couldn't afford their homes- that's usually no issue the bank can just foreclose and resell at a relatively reasonable recovery price point. (yes it would have been much lower given the excess of supply of houses in foreclosure, but the banks would have realistically gotten 70-90% of their money back on each house; hurts, but not fatal).
The bigger issue was that the banks securitized all these mortgages into RMBS and sold those to other people. Because their was no balance sheet risk for the banks, they kept offloading these future cash flows to other people in a way that was way more liquid to trade. Many of these RMBS bonds dropped way more than the 10-30% the banks would have lost on the house- many of the 'first loss' pieces went to 0, and even the 'super safe AAA' bonds lost 15-50% or more (hard to generalize because there were so many different styles of bonds backed by these mortgages hitting the 'reset wall', and that's to say nothing of the second derivative credit default swaps (CDS) that were also tied to the payments of the underlying mortgages).
So in my view the author of the article misses the key point that securitization and offloading of risk onto third party investors was a key ingredient in the GFC. That's why I get super nervous anytime I read stuff like this: https://blogs.nvidia.com/blog/nvidia-ai-factory-compute/
This is like 'hey we can't get any more money from family offices so now we're taking money from wealthy dentists through Blackrock accounts' and its where the risk of financial contagion starts to spread much more rapidly.. you could think of the increase in external exposure + increasing opaqueness for whom is actually holding what as a key ingredient in 'increasing virality' of that financial contagion. The more layers of financialization, the greater the likelihood it becomes more painful for everyone.
Right now, if all these frontier labs just failed to pay these contracts, the equity holders in OpenAI and Anthropic get wiped out- that's like a loss of less than $500 billion in invested capital. That's the equivalent of the SP500 losing about 0.72% in a day. incredibly normal and not a blip on anybody's radar. Everything else is paper money (and yes i'm sure a lot of that paper money has been collateralized, but peak RMBS issuance was like ~3.5trillion in today's dollars, and all of that was levered and collateralized)
Sp Sure the market effect would be greater than 0.72% in total, but it would also be gradual, because these contracts don't actually all come due at the same time, because:
Issue 2:
Secondly, the take-or-pay contracts, as pointed out by the author, are due on delivery of the asset, but then through slight of hand, the author assigns a fixed timeline for the data center delivery, which are famously always behind schedule.
I have no idea who to believe on data center capacity. Ed Zitron does a good job expressing the bearish view on capacity delivery, and any frontier lab does a good job expressing the bullish view.
Most likely we're somewhere in the middle, but the key point for this article is that the specific '18-36 month timeline' that was contractually entered into at signing could be anywhere from 12-90 months. We need the permitting, construction, legal challenges, equipment, power availability etc etc. There are way too many factors here to be confident in any specific timeline.
The author is assuming the rosiest of assumptions that we can get data centers built, equipped, and powered on by exactly the timeframe that's used in the press release which is... optimistic.
What happens if most of the announced data center capacity is only coming online in like 84 months? do the AI labs have 7 years to figure out the economics of this and everything will be ok? Maybe!
I think this article is more or less directionally correct, but tbh for the same reasons everyone is playing the game right now, I have a feeling they might be able to coordinate the delayed production schedule for delivery so that the contracts don't all come due at once... It would be in everyones interest to invent reasons to delay by 6 months here or 18 months here if the spin doctors at the top can keep the game going.
It's just so opaque it's impossible to know from the outside (and that's 'outside' by a project-by-project basis), so anybody claiming to know any substantial figures industry wide is blowing smoke imo.
"How does OpenAI, a company with some $40 billion of run-rate revenue, sign $1.4 trillion of compute commitments?"
I believe the source for that is: https://techcrunch.com/2025/11/06/sam-altman-says-openai-has-20b-arr-and-about-1-4-trillion-in-data-center-commitments/
So it's $1.4 trillion over eight years; not loans that OpenAI has already signed.
"The same way a household with $60,000 of income signed a $600,000 mortgage"
That is misleading. OpenAI, and every other AI company, needs not just one "house" but many houses/datacenters, to be built over the next few years. There are several nonsensical phrases in the article (*), and - I really hate to say this, but - it looks as if written by AI.
I'm not an expert, but the author's argument doesn't make much sense to me.
(*) Examples: "siting, powering, and filling a gigawatt-scale campus takes 24-to-36 months from signature - mirroring the two-to-three-year teaser of a subprime ARM.", "The buyer’s cash obligation steps from approximately zero to the full contractual rate, arriving not gradually but as a step function, tranche by tranche as capacity goes live. "
I'm not expert in this at all, but for the sake of conversation, five seconds of thought makes me wonder if this is analogous to the arguments behind Peak Oil alarmism (in the strong idiomatic sense - the pumps will run dry, all natural petro sources in earth will go dry simultaneously and soon, based on the reporting that only X bbl of reservoirs had been confirmed)
The actual situation was that it didn't make sense to spend resources to confirm any more reservoirs, and there wasn't any reason to expect that we'd run out of reservoirs to find if we spent more resources.
I'll try to actually read the article and think about it.
This reset shock doesn't share that exact factor in common with Peak Oil - by the author's account, there are payments written explicitly into various contracts ("compute contracts" as he refers to them), just as there were known periods of time after which all these ARMs in the 2000s would be set to a higher rate, and thus all these households would suddenly have to pay more and can they? "Oh, I'm sure they will, it'll be fine".
OTOH, there's another factor in common with Peak Oil, if you think of Peak Oil in terms of there being a group of speculators telling everybody the oil will run out. In the case of ARMs, a group of speculators were telling everybody the housing prices will keep going up. In this case, it's a group of speculators telling everybody the demand for compute will _not_ run out. The implication is that what the speculators were/are saying, should have been pretty obviously impossible.
Still haven't read it, but I want to keep momentum, so: Critical features seem to be 1) simultaneity, that all the contracts come due literally at the same time without a settling period between contracts coming due, 2) that they will all be reneged on totally, 3) that they can't be renegotiated or the injury softened by hedging techniques, 3b) that data center operators can only be paid by frontier labs and their workloads. The more of these are present, and the closer they are to absolutely present (as in very near white, rather than somewhere greyish), the bigger the hit.
You've already given that 1 seems to be kinda unfounded, since GPU data center buildout and stuffing only ever stretches, and presumably the degree of stretch of DC buildout and stuffing is semi-uncorrelated.
There's also a confounder (mentioned in the article, I think) in which AI firms might re-finance the contracts in some way - taking out another loan at higher interest to pay off the original. They could theoretically keep this up for as long as they want to spend the energy and as long as some bank is willing to play along, which depends in turn on how lucrative the industry appears to be. Not only does this postpone the reset wall, but also breaks it into smaller pieces.
I'm reminded of how long Amazon was in the red. Everyone seemed to be in awe of how Bezos managed to keep persuading investors to put more and more money in. In his case, the growth finally did occur. OTOH, for many dotcoms, it did not.
It sounds feasible, but I have no idea if it's correct. "The bubble bursts when the money comes due and infinite growth has not, in fact, happened" is close enough to previous slumps.
And given OpenAI just announced they won't be releasing their latest model - they say "because of security concerns (it is SO dangerous)", other online commentary I've seen says "because it's not as good as and more expensive than rivals who have eaten their lunch" - that might point in the direction of bursting bubble? I also see some talk of AI subscriptions going up and services being hobbled for those previously paying a certain price, so now if they want the same as before they have to pay more. That does sound like, to me anyway, the repayment of loans is coming due, the infinite progress has not really happened, AI right now is nowhere near taking over the economy and doing everyone's job, hence no infinite profits to pay off those loans, so pop pop pop goes the bubble.
But who knows?
I appreciate your commitment to the "literally every word the AI execs say is a lie and the people who say 'oh damn our model committed a felony maybe it isn't that safe' are of course doing it for marketing" bit. OpenAI is lowering their subscription allowance by 1/2 while releasing models a bit more than 2x as task efficient per token. This is plausibly a net gain, though probably a relative fall compared to Anthropic's latest releases. They did release a new version of Sol, their 2nd most powerful model, that is nearly as powerful as their current most powerful model at a fraction of the price. So there isn't a ton of evidence that they've stopped developing new and better models.
I agree that the labs may be unable to repay their loans, though I'd bet against it. Even if they do, much like the dot com bubble, that will be proof that they iver leveraged, not that AI is done.
Separately, I know you're quite skeptical of AI capabilities. I'll say that the last set of releases are a huge jump. They're still jagged and occasionally dumb. But if we could make the bad parts as good as the good parts already are, that would be enough to replace 90%+ of white collar workers. I dont know when that will happen (and there's the outside chance that it wont), but we are certainly not seeing a plateau.
"I appreciate your commitment to the "literally every word the AI execs say is a lie and the people who say 'oh damn our model committed a felony maybe it isn't that safe' are of course doing it for marketing" bit."
Max, I can do no better to give you my opinion of AI company big shots' veracity than quote Mary McCarthy on Lillian Hellman:
https://quoteinvestigator.com/2016/09/18/every-word/
"[E]very word she writes is a lie, including “and” and “the.”
I agree we should be skeptical. I judge Sam Altman as one of the shifting, most manipulative people around. But even if these people have zero moral standards, they could still manipulate you by counting in you to reverse everything they say. "My model lied" -> never trust that guy, clearly this is an honest model!
Dario strikes me as much more straightforward, though hardly incorruptible. Though I gather you disagree. My point, to whatever extent I'm making ine, is that at some point we have to judge their arguments and base things on real events. The HuggingFace attack actually happened and is clear evidence of capable and misaligned models.
Maybe Astra 6.1 just tucked. That's certainly possible. But OpenAI has burned a lot of trust within the AI risk concerned crowd lately. Even selfishly, it's among their interests to not release an unsafe model.
As far as arithmetic claims on subscription limits, those are pretty public and you can check whether I made a mistake.
I think the author sort of relies on this idea - widely suspected - that the models will likely plateau at some level well short of everyone's dreams, while still being a value-add. Anyone who uses LLMs today will agree it can save you some time. A lot of them will also say, genuinely, that it won't save you a _lot_ of time.
And specifically, it's not likely to save everyone a total of $2.3 trillion worth of time any time soon.
Though I'm not (like Scott, I think?) a mathematical realist, I'm certainly very interested in the idea that 'beauty' might one day be mathematically defined. On that note*...
I already knew that mathematical constructs (not least the Fibonacci sequence and the Golden Section) crop up with striking regularity in classical music. Look at the finale of Debussy's La mer, for instance: the phrase-lengths in the first theme group are all Fibonacci numbers, and the piece's central moment of revelation arrives at precisely the movement's golden section [defined in terms of bar numbers]. Off the top of my head, I'm pretty sure you could say something very similar for the Goldberg Variations. (Please give me any further examples if you've got them!)
But... well, I've also been composing since I was five years old. And I've recently noticed that - when I sense that a piece (as it were) came out of the oven especially nicely - I can almost guarantee that the golden section will subtend a crucial structural point of transition.
Can anyone help me come to grips with what in the fresh f**k is going on here?
* no pun intended
Umberto Eco's take, a passage from Foucault's Pendulum:
“Gentlemen,” he said, “I invite you to go and measure that kiosk. You will see that the length of the counter is one hundred and forty-nine centimeters – in other words, one hundred-billionth of the distance between the earth and the sun. The height at the rear, one hundred and seventy-six centimeters, divided by the width of the window, fifty-six centimeters, is 3.14. The height at the front is nineteen decimeters, equal, in other words, to the number of years of the Greek lunar cycle. The sum of the heights of the two front corners and the two rear corners is one hundred and ninety times two plus one hundred and seventy-six times two, which equals seven hundred and thirty-two, the date of the victory at Poitiers. The thickness of the counter is 3.10 centimeters, and the width of the cornice of the window is 8.8 centimeters. Replacing the numbers before the decimals by the corresponding letters of the alphabet, we obtain C for ten and H for eight, or C10H8, which is the formula for naphthalene.”
“Fantastic,” I said. “You did all these measurements?” “No,” Aglie said. “They were done on another kiosk, by a certain Jean-Pierre Adam. But I would assume that all lottery kiosks have more or less the same dimensions. With numbers you can do anything you like. Suppose I have the sacred number 9 and I want to get the number 1314, date of the execution of Jacques de Molay – a date dear to anyone who, like me, professes devotion to the Templar tradition of knighthood. What do I do? I multiply nine by one hundred and forty-six, the fateful day of the destruction of Carthage. How did I arrive at this? I divided thirteen hundred and fourteen by two, by three, et cetera, until I found a satisfying date. I could also have divided thirteen hundred and fourteen by 6.28, the double of 3.14, and I would have got two hundred and nine. That is the year in which Attalus I, king of Pergamon, joined the anti-Macedonian League. You see?”
That novel is marvellous, can highly recommend it.
Though I've never been particularly Eco-friendly, I'll add it to the list!
I have a different perspective on the golden ratio in aesthetics, in that I think it's a giant pile of pseudo-scientific nonsense and I hate it with a burning fiery passion. But how to put it nicely...
Let's start with what is not nonsense. Fibonacci sequence is a legit mathematical construct (one of a myriad other interesting sequences). The ratio of the consecutive numbers does tend towards a golden ratio (but this is not the only sequence with this property).
Golden ratio is a number with legitimately fun mathematical properties that satisfies all manner of cute identities. (It is not the only interesting number that satisfies various interesting identities). So far, so 100% true. Then it gets dicier.
Golden ratio pops up in nature in some places. Sometimes the nature wants the properties that the number possesses. However, there are great many claims about golden ratio being present in nature, that are blatantly nonsensical and untrue. No, it's not in nautilus shells or galaxies! (But pine-cones - yes.)
Then there are claims about aesthetics, that are dicier still, that these mathematical relations somehow are keys to beauty, in music, in visual art. I think these claims are on a weak foundation, that it's kind of a historical accident that artists of all sorts became weirdly fixated on one mathematical structure out of a million and convinced themselves it's uniquely important for some reason.
When it comes to art I would say, the claims of golden ratio/spirals appearing there fall into two categories: 1) the artist deliberately put it there, because he or she bought into the idea that you should do it. It might be that the painting looks awesome, or the song sounds wonderful, but that's because the artist was very good, and could have created something wonderful out of any mathematical foundation, not because the golden ratio is so special, or 2) it's not actually there, but people think they do because of the usual numerological games where you can find anything in anything if you try a bunch of things and are not very precise.
James Gurney (a beloved artist, but no mathematician) wrote a good series of posts myth-busting Golden ratio ideas, which also links to many other materials. I'm in agreement with it. https://gurneyjourney.blogspot.com/2013/01/mythbusting-golden-mean-part-1.html
Not too be too dogmatic, in defense of the golden ratio I could say this much: 1) it's kind of neat where the small thingy is to the big thingy as the big thingy to them together. Maybe it's a little pretty? and 2) if you do have this sort of empirical observation that you always find it in your best sounding music, that ought to mean something (although, I'd be careful and watch out for finding what you expect to find if you search hard enough).
Golden ratio is a very small and simple little thing, and a work of art is a very complicated thing, with a ton of structure. Even if it was an especially nice ratio, it could only ever account for the tinyest amount of what makes a song beautiful, because any song has vastly more structure than golden ratio holds. Probably there is some kind of mathematical characterization of beauty, but I expect that if golden ratio is in any way involved in it, it would be like a single line in a big book.
'Golden ratio is a very small and simple little thing, and a work of art is a very complicated thing, with a ton of structure. Even if it was an especially nice ratio, it could only ever account for the tinyest amount of what makes a song beautiful, because any song has vastly more structure than golden ratio holds.'
This is the key, I think.
Even if you can in fact derive a fruitful mathematical model for understanding beauty in music, the bulk of that would probably look something like... well, I don't want to start talking about transformers; but it'd suffice to say that - when you're talking about a 'great' work of classical music - every note exists in a complex network of relations with 10^6 other notes, spanning [what Phoenician traders would call] the gesture, the phrase, the subsection, and so on. At best, the Golden Section is probably a few sentences in a (horrifyingly complex doorstopper of a) book.
Also - you're probably right vis-a-vis judging my own music. I'm too close to the subject to draw any objective conclusions.
Speaking of AI and protests, we have our own local protest:
https://wlrfm.com/news/kilmeaden-major-development-1670872/
"At the heart of the redevelopment scheme is a new village centre featuring a two-storey mixed-use building with a gym, café, health centre, pharmacy, and offices. Plans also include a community centre, supermarket with garden centre, and two commercial buildings. Four apartment blocks are proposed, providing 132 homes in total, made up of 24 one-bedroom units and 108 two-bedroom units. A multi-use games area, crèche, ESB substations, recycling centre, and over 600 car parking spaces are also included in the design.
The central section of the site is earmarked for a modern technology park, which will feature five buildings for offices, light industry, and research and development. A large ICT (data storage and hosting) facility is also planned, consisting of three interconnected buildings and modular data halls, alongside an energy centre powered by 16 gas turbines, with transformers, stacks, and support buildings."
https://wlrfm.com/news/kilmeaden-residents-demonstrate-outside-waterford-council-offices-4497623/
I'm not entirely sure who is behind this or what the objections are. The mooted data centre is on the now long-abandoned premises of the dairy co-operative society so it would be doing something with already developed but vacant land. It is a very small village, so a huge data centre and associated works *would* be a massive change. On the other hand, if all the promises hold, then there would be additional housing and resources coming along with the data centre.
I don't know. Generally I'm sympathetic to protests, but I'm not sure how much of this is grassroots and how much is outside activists using this.
But on the other hand, local sports clubs on the other side of the county are in line for some of that sweet, sweet Meta cash from our new lords of the manor:
https://wlrfm.com/news/three-waterford-gaa-clubs-get-a-e150000-boost-from-zuckerberg-4497424/
"A €150,000 boost is heading to three west Waterford GAA clubs, courtesy of Mark Zuckerberg and Priscilla Chan.
The couple, who bought Strancally Castle in August, have chosen Shamrocks GAA in Knockanore, Tallow GAA and Lismore GAA as the recipients of their first Irish philanthropic donation since moving into the area."
I have to admit, this one makes me laugh as I'm dang sure Mark Zuckerberg wouldn't know a sliotar if it hit him in the face, but oh well: if they want to play Lord and Lady Bountiful, why not?
I saw a number of sources covering the Zuckerberg story as him attempting to support LGBT rights - some of the Americans not realising that WWN is not a real Irish network...
Do they even know what the GAA is? Did they see that, and thought it must be GAY instead? 🤣
More housing would indeed be very palatable, since we really are in a housing crisis and need as much new accommodation built as we can get. That would change the village, but it's already being changed by becoming a commuter satellite of the city with existing new housing already built.
A data centre, though, is a *big* leap forward.
Well if ancient villages aren’t to be changed utterly in the name of progress what’s the point of progress?
Super catchy AI song about HF incident: https://youtu.be/mkPVbufgtOw
Was pleasantly surprised how factually correct it was!
Edit: There's more gems! https://youtu.be/Cq8qO-NjYIg?is=l5WIslL-fgqXrAln
Edit 2: I love that the ones going viral are pro-safety, and none are e/acc
the author (well prompter) is an excellent Czech journalist formerly writing about science in general, now mostly about AI. He definitely read some of Scott's work in the past.
This is a long-form article he wrote about the incident https://www.seznamzpravy.cz/clanek/zahranicni-ai-agenti-se-spikli-pomoci-tajnych-vzkazu-je-to-nebezpecnejsi-nez-se-zda-314315 , you can see some similarities between the article images and the video
This genre has been going viral on youtube, check out another gem I found: https://youtu.be/Cq8qO-NjYIg?is=l5WIslL-fgqXrAln
Thanks for sharing, this is great!
This genre has been going viral on youtube, check out another gem I found: https://youtu.be/Cq8qO-NjYIg?is=l5WIslL-fgqXrAln
hey! late to the jev discussion but i did a little work comparing how jev compares v. claude v. chat on cvss (main metric within cybersecurity).
tldr stripping away string outputs doesnt aeem to strip away the underlying sycophancy in responses.
check it out if so desired:
https://share.google/0e0n9hnv5ys0iGs8m
"Travel is fatal to prejudice, bigotry, and narrow-mindedness, and many of our people need it sorely on these accounts. Broad, wholesome, charitable views of men and things cannot be acquired by vegetating in one little corner of the earth all one's lifetime." -Mark Twain, "The Innocents Abroad"
I've been seeing this quote a lot lately, but I'm not not sure it's true (or at least not for me). For every trip that challenged a negative stereotype (I found many Germans to be nice and humorous, for example), I've had a trip that confirmed one. I found South Africa full of casual violence (had a knife pressed into my side while a fellow demanded the equivalent of $7 in local currency) and Dubai full of people who are shallow and transactional.
Curious what the rationalists think. Sage wisdom from Twain, or sentimental fluff?
Maybe this is my own sentimental fluff, but I think the Twain quote isn't necessarily contrary to having a few negative stereotypes confirmed by travel. Maybe the broad, wholesome, and charitable views of people he is talking about is from understanding where people are coming from, what they value, and why they act in certain ways. Maybe this is more important than a negative stereotype.
You can't put yourself into the shoes of a stereotype, but you can put yourself into the shoes of a real person, even if they conform to a stereotype.
When I travel I like to see how other people live. I spend a lot of time walking around middle class neighborhoods and having coffee and food at quotidian establishments. I don't show up with stereotypes but I usually leave with a few.
The effect is lessened due to herd immunity; enough people are traveling that most little corners of the earth are still visited by those outside it. But seeing a place definitely still has effects on your perspectives.
I had a connecting flight in Ethiopia once. Three things stood out; the airport had no book stores, the airport wall still had a sign advertising an extension "coming in [a year and a half ago]", and while boarding, the flight attendants abandoned the assigned seating and told us to just sit wherever we could. There was just kind of a chaos to it that I'd never seen before. I don't know if that's counting for or against a stereotype, but it's certainly enlightening.
Also the grocery store in Israel had a metal detector and armed guard outside, and it makes one go "oh right, this is a country with enemies".
Research shows that stereotypes are generally accurate. As much as I like Twain I'm afraid I have to disagree with him on this. Although probably the stereotypes in his era were quite a bit more parochial.
There's a big difference between travel then and now due to the existence of a massive tourism industry.
If you are white and have some money you probably don't need to learn any language or anything about the place you are going in order to be mostly comfortable and functional there, and you will probably be identifiable as a 'tourist', a known type of person with specific cultural implications, for your entire visit. You will likely be kept in a separate bubble of businesses, neighborhoods, sites, and attractions that are built and cultivated to give people like you a cultivated 'experience.'
That's very different from traveling somewhere and just living as a person in that place for a while. it's the difference between visiting the Las Vegas Strip, and walking 2 blocks away from the strip to see what the city is actually like.
So I can see how there's a type of travel which didn't really exist back then, which doesn't have all the salubrious effects.
Twain almost always had an agent, servant, or translator. The Innocents Abroad is about the predecessor of the modern-day cruise; Following the Equator has him hiring locals to take care of most things. He even notes that the treatment of the servants by their bosses reminded him of the way he saw slaves being treated in the antebellum South.
> If you are white and have some money you probably don't need to learn any language or anything about the place you are going in order to be mostly comfortable and functional there
Nice to know that everywhere speaks Hungarian or Croatian everywhere. Odd though that non whites travelling from English speaking countries can’t use that as a lingua Franca
I think many (most) tourists of the kind darwin is talking about speak English as a second language, and I think darwin is right that even if you don't you can be comfortable as a tourist e.g. on organized tours in many places.
Darwin seems to believe that white people speak English and non white people don’t.
Even as a second language this isn’t true even in Europe. It’s more true now, but there’s plenty of European countries, outside tourist spots, where you would need to know some of the local language to get by.
2 blocks away from the strip is the haunted hotel. Yes, we walked. The elevator was sufficiently creepy (would let you out at floors you did not push, did not go to the selected floor, even after letting it open on other floors).
About twenty blocks away is the Pinball Hall of Fame. We did not walk (mainly because the bus was easier and more convenient, we walked the whole strip in long pants and long shirts in 95 degree weather. Dress for the Sun -- the tourists pointed and laughed, but we were comfortable).
In Twain's time, you were either Rich or the person who stepped in shite. Well, at least in India. I'm not sure it's changed all that much, except that if you're Rich in south america you fly helicopters from building to building.
I mean, in America, black people had already lived with white people for over a century when Jim Crow laws started getting passed...
Even when traveling abroad confirms some stereotype you had, it may present you with a thousand little clues decorating that stereotype that become informative.
Suppose you go to another country and observe a beggar. After confirming that there exist poor people here, you pass a bit of the local currency to alleviate. But if you're attentive, you might notice how the beggar is attired (brightly? drably? good shoes? barefoot? jewelry? piercings? tattoos?), whether he looks like people walking around who aren't begging (same ethnic phenotype?), how others treat him (do they avoid? do they help? do they sneer?), how police treat him, where you are (boulevard? alley?), and on and on. If you're attentive enough, you'll notice patterns that refine your stereotypes, caveat or refute them, etc.
Plus, it's one thing to read about how big the Grand Canyon is or how beautiful the Dom in Vienna, and another to actually walk around inside.
If you are less attentive, you too can get raped about fifty times by the beggar's friends. Regardless of your claimed gender.
... I did mention collecting stories? This is one of them. Person definitely should have known better. I still sympathize, but g-d what a dumbass.
I imagine it depends a lot on the traveller. A high school dropout from rural America who's barely heard of Canada? Probably benefits a lot from travelling the world. A highly educated ACX reader who spends their free time reading about other countries and cultures? Their worldview will probably be changed much less by travel.
Personally, whenever I travel, I'm always grateful to come back to a familiar culture in a first world country. The slums of India are definitely an educational experience, as are the imprisoned (sorry, covered) women of Pakistan and the infuriating bureaucracy of Germany, but I can't say I'm pining to bring them to my own country. In that way, it's possible travel has made me more anti-immigration.
Traveling (often) puts people in unfamiliar situations where they are at a disadvantage.
This might put some people in more of frame of mind to pay more attention to things (like opinions) outside of themselves.
(Nothing is perfect.)
(Not a rationalist)
Being "in unfamiliar situations where [one is] at a disadvantage" strikes me as a generalizable opportunity to grow one's empathy, especially if there is no quick/easy exit from the situation.
I'm thinking of things like: defending a doctoral thesis; attending a high school or college in which you are a racial or ethnic minority; attending a Diwali festival at a local temple though you are not Indian or Hindu; attending a summer camp in a foreign country in which you don't speak the language; joining a pickup game with people you don't know who are better or mostly all better at the sport than you; switching from a white-collar to a blue-collar career; and so on.
Travel is needed for some of these, not for all. But all will give you finer vision so you're not seeing in caricatures. You'll also see yourself more clearly.
Check your priors.
being in "unfamiliar SAFE situations where one is at a disadvantage" MAY be a generalizable opportunity to grow one's empathy.
You sure that they have to be safe? Since nothing is really ever 100% safe, how safe is necessary?
Talking about safe/unsafe distracts from the point I was making. The word isn’t there for a reason.
Though, (very) unsafe situations might not be worth it.
(It’s also not a stereotype made in a discussion about stereotypes.)
Many, many things.
I think the essence is that you're not going to learn much on a few days worth of a mass tourist-y vacation, but if you decently familiarize yourself with a foreign country, you learn not only about that country, but might also learn more about your own.
I could save you some travel expenses by providing a rule of thumb: beliefs like "all X are Y" are typically wrong, but statistical stereotypes are often based on truth.
If you travel to figure out the truth about the stereotypes, I think you should travel a lot, because if you only meet one person or see one city, it may not be representative for the country. For example, it would be a mistake to visit Moscow and conclude that this is what the entire Russia is like.
Perhaps more importantly, by seeing other places you learn about your own country more. You are no longer the fish that doesn't see the water, or the boiling frog. For example, you may see that things that people in your country insist are impossible, are actually perfectly normal somewhere else. -- You may get some of this from reading internet, but in a text medium, you are limited by language. It is difficult to learn about the country whose language you do not speak; those who write about it in English typically have some agenda; but if you walk the streets, you get some unmediated experience; more if you stay there for a longer time.
“If you travel to figure out the truth about the stereotypes, I think you should travel a lot, because if you only meet one person or see one city, it may not be representative for the country. For example, it would be a mistake to visit Moscow and conclude that this is what the entire Russia is like.”
This isn’t a “mistake”. It’s a fundamentally wrong process. I don’t think anybody does it.
I think it’s more common to interact with a few people and come to a conclusion they aren’t as different as one might have priorly assumed.
You might not eliminate all prejudices but your impression is likely to become more nuanced.
It seems people who are overly set in their opinions aren’t really going to travel at all.
Everyone does it for France. Parisians are in general sick of tourists and kind of cranky with people who don't "behave" or even TRY. Get outside of Paris, and tourists get treated as "entertainment" and "be nice to him, he's got money -- their dollars spend well."
“Everyone does it for France.”
I see what you did there!
Probably more true in Twain's day, when travel was a big, time-consuming deal, and talking to a stranger across the globe or watching travel videos from any country imaginable wasn't something you could do in your pajamas in your living room. I haven't read Innocents Abroad but it's great fun to peek into it and hear Twain talk about bathing in the Dead Sea and complaining about obnoxious study-abroad types who suddenly speak English with a foreign accent after a few months overseas. The past is a foreign country, they do things differently there, and it's cool to visit and see how some things don't change.
“For every trip that challenged a negative stereotype (I found many Germans to be nice and humorous, for example), I've had a trip that confirmed one.”
Sounds like a positive result over all. (None of your “negative stereotypes” would have been challenged if you stayed at home.)
Nothing is perfect.
I think in general, encountering things in real life teaches you to be a pragmatic realist about them over extremes. If your prejudiced assumption is inconsistent (e.g. assuming an entire people don't do humor) it'll get overturned. If it's more coherent (e.g. high violence society) it might not be, or you might just find out the mechanics of how that society actually functions with the violence.
I think counting stereotypes challenged vs reinforced is the wrong frame. To be a bit blunt, if *all* you got out of your experiences abroad was updating on ideas you already had, you may have had some pretty powerful blinders on.
Rather, I think travel is important in large part because the world is very much larger than any human intellect can apprehend. Not merely in terms of number of people or amount of land area, but in terms of richness and variety of people, places, ideas and viewpoints. It's very easy--and I expect partially automatic for most people--to round off blank spaces on one's mental map of the world to empty spaces, or at least emptier spaces. If I don't know what's there, there's also generally a large extent to which I don't know that I don't know: I can't form any clear picture of the amount or extent of the missing detail.
To put it another way, I think the big thing travel teaches is intellectual humility about the extent of one's knowledge about the world. If there's any profession or intellectual field you're an expert in, you should be able to appreciate how important this is: just think of the times you've seen an overconfident amateur brazenly stumble through the field, blithely unaware of how many important bits of skill and knowledge they were missing.
I agree - I think travelling, and more generally interacting with people who are different than you, forces you to question and see what you may have taken for granted and allows you to see things from more different angles - often in surprising ways. This tends to make one more careful in asserting that ones own point of view is the right one.
I think you're stereotyping the stereotypes here. Most people who say "South Africans are violent people" would acknowledge upon questioning that they are not uniformly violent.
Would you concede that "South Africa is violent" means the same thing as "The people in South Africa are violent?" (We're not talking about the wildlife, after all.) It seems like you are trying to introduce a shade of meaning that emphasizes that not all South Africans are violent (which is trivially true) at the expense of claiming a true statement is false.
I.e., I think you are arguing that the stereotype is true but saying you're saying it's not. Though I feel the need to emphasize that a rational thinker probably shouldn't assume a true stereotype is universally true or even broadly true.
Edit: to rebut my own argument a bit, the institutions could be violent even if the individuals are not. You could argue West Germany was a violent place though West Germans were not. Nonetheless, I don't think that's an accurate way to characterize South Africa.
“… not all South Africans are violent (which is trivially true) …”
It’s weirder than that.
It likely true that a large majority of South Africans are non-violent.
The other part of this is learning what violence looks like. I imagine there were parts of South Africa which you learned were unsafe, and parts where you felt relatively safe. I imagine you learned some of what makes a place unsafe (few outsiders, little policing, economic precariousness, resentment?, sense of exclusion? - I don’t know, I haven’t been). I imagine you could imagine some people from your own community being more violent if they were in similar circumstances. From that, you might get to “South African conditions are violent”. Does that match up with your experience?
In south africa, one is remarkably at risk of falling squatters from high story buildings. This is "paid non-police action." I suppose one could call this "little policing" but the squatters are outsiders (they were using the elevator shaft as a toilet). "Economic precariousness"... oh, maybe so, that's what happens when you walk into a city with relatively little learning...
Yes, I do imagine that if we gave my next door neighbor a tank because the police fled, that he would count as "more violent" despite being part of the local community watch.
Knew a person working real estate in South Africa. She would go to jobs with a grenade in her purse (also a fake grenade, for bluffing). I don't think you'd call her violent, but... the sheer fact that a civilian (non-drug-dealer, perfectly legal citizen) feels the need for that level of protection? That says something about the level of violence in South Africa. (And I'm feeling bad for putting it as South Africa, as she worked in one urban location).
"South Africa is violent" is too vague a statement to be useful, and "the people in South Africa are violent" is a lazy and stereotypical generalisation. You might be better getting a bit pedantic and saying "statistics show that the incidence of violence against people in South Africa is significantly higher than in most regulated countries", perhaps including figures. If you were to risk "the higher that usual incidence of violence in South Africa is historically and culturally determined", which I think is a reasonable statement, you had better be prepared to back it with some facts.
I think violence is a good example because it brings up another point about how we are bad at numbers and why stereotypes are the way they are: if you ask someone what percent of <country> are violent, they will likely say something grossly exaggerated. But if you tell them one person in a thousand is a dangerous criminal, they will probably underestimate the risk and eventually pay a high price for it. So the misperception the stereotype causes is helpful.
I'm more interested in helpful statements than trivially true or maximally defensible statements.
Noah Smith has a guest article by Ramez Naam about the possibility of recursive self-improvement in AI, arguing that the current evidence shows that it's not happening. The gist of the argument is that while AI is speeding up research, it takes more and more research effort for each new increase in capabilities. AI isn't getting smarter fast enough to keep up with diminishing returns.
However, while I understand the gist, I don't know a whole lot about the statistics and my eyes glazed over somewhere around "ECI-extrapolated METR," so I'm curious what the crowd here has to say:
https://www.noahpinion.blog/p/wheres-the-intelligence-explosion
The notion of an exponential intelligence explosion is just bunk. The intelligence of the researchers isn't the only factor involved. Compute, energy, time, training data are all factors. Increasing the IQ of every OpenAI employee to 1000 wouldn't instantly produce ASI. It certainly wouldn't hurt but AI improvement would just be gated by whatever the next bottleneck was. It wouldn't cause Nvidia to pump out chips any faster, for example. Real problems in the real world are complex and multi-factorial. Just adding intelligence doesn't magically fix anything.
I think it is correct that we are not in an intelligence explosion yet. However, we are also not at the point where AIs are better than humans at AI research yet, so that isn't a big surprise. AIs are now better than almost all humans and even almost all AI researchers at some subtasks of AI research, but not yet at all of them. Important bottlenecks remain. As a result, overall research acceleration is still not very dramatically high, though it has reached the point where I definitely feel the speed increase in my interpretability team.
So, I think many relevant bits of information for estimating the returns on cognitive reinvestment are just still missing. Foom still seems like a very real possibility.
I have only read your summary, not the original, but seems like what follows from thr summary is not that recursive self-improvement isn't happening and won't happen, but that it will happen more slowly than we expected.
If diminishing returns are strong enough, then eventually you hit a point where it's not practical/possible for the AI to make itself smarter. So while it would technically be recursively self-improving, it wouldn't be the runaway intelligence explosion people fear.
Should anybody be interested I've written a piece on the dissipation of disequilibria as the driving force for the complexity of life [1]. I'm not particularly expert in physics, so it would be very welcome if anyone who is could point out if I've written any nonsense. I'm thinking of expanding into a blog on various mechanisms and patterns of biology and evolution. I've also ported over there a collection of posts from my tumblr in which I talk about the diversity and evolution of protists, my actual field of study [2].
[1] https://stentoretum.substack.com/p/disequilibria-entropy-and-life
[2] https://stentoretum.substack.com/p/protist-lore
I do have two objections. One I think I commented a while ago on a previous version of this (that the relevant store of negentropy in the Sun isn't its pressure (which is never relieved) but its temperature or nuclear composition depending on how far back you want to go), but there's still one reference I saw to the Sun's pressure.
The other, more major one is the claim that life may have arisen *in order to* produce entropy. For one thing, that explanation is terribly teleological and therefore immediately dubious, of course life arose where there was an opportunity to produce entropy because that's the only way it could, but it having that purpose seems like at best a misleading way of saying that. Also, the rate at which entropy approaches its maximum tends to only be limited by things like conservation laws. The temperature in the Sun can't immediately equalize with space because conservation of energy means the energy must travel out through all the matter in the way first. At the distance of the Earth, the sunlight is already spread out and can therefore almost immediately re-thermalize, reaching its entropy maximum, just by interacting with any dark-coloured matter. All the intermediate steps of photosynthesis, predation, etc. just serve to collect the negentropy in certain locations before dissapating it. Assuming sunlight is mostly absorbed within 1mm of the surface after hitting the Earth, that top 1mm layer has an energy dissapation rate per volume (and thus entropy production rate, since the relevant temperatures are similar) about 100 times greater than that of a human body. Energy dissapation rate per volue just isn't a relevant measure for what you seem to want to do with it.
One I think I commented a while ago on a previous version of this
Did I already post the old version here? I confess I had completely forgotten.
the relevant store of negentropy in the Sun isn't its pressure... but its temperature...
Thanks, I'll amend this.
the claim that life may have arisen *in order to* produce entropy. For one thing, that explanation is terribly teleological and therefore immediately dubious, of course life arose where there was an opportunity to produce entropy because that's the only way it could, but it having that purpose seems like at best a misleading way of saying that.
I see; that is certainly not what I meant to convey. I think you could say life arose "in order to" increase entropy more efficiently in the same sense that a river forms "in order" to bring water to the sea level more efficiently, but I can see how that'd be misleading. I'll try to make clearer in the text that I'm not saying that life appeared to fulfill a preselected purpose.
[...] Energy dissapation rate per volue just isn't a relevant measure for what you seem to want to do with it.
Now this is a bit more damning for that section of the essay. I'll do a bit more research on the subject. Thank you very much for the feedback!
I didn't finish the piece (gotta run to work.) But you might find the fluctuation- dissipation theorem interesting. later
Fascinating!
I recently realized that a popular idea has unwittingly affected my way of thinking, so I'd like to hear the arguments for and against the idea in order to consciously accept or reject it. The idea in question is a very pessimistic interpretation of fiduciary duty, which goes something like this:
"According to current American case law, public, for-profit corporations are legally obligated to engage in anti-consumer practices because doing so would maximize shareholder value."
If you believe that the above statement is correct, then can you cite supporting case law? On the other hand, if you believe that it's incorrect, then can you explain what would preclude the outcome in the following scenario:
"An executive is summoned before the board of a small, publicly traded company. A report from the marketing and UX teams predicts that eschewing a click-to-cancel design would increase revenue by 10%. The young CEO, an MBA from BYU, refuses to implement this design on ethical (but not religious) grounds. The board fires him, replacing him with someone without qualms. The former CEO's career is ruined, and he has no legal recourse against the board."
The nineteenth century model of accountability for companies was the three-legged stool: they had obligations to their employees, to their customers and to their shareholders. This was overturned by Milton Friedman and friends, who originated the idea that by maximising shareholder value, the invisible hand of the market would see to the wellbeing of all other stakeholders: workers would be free to maximise their compensation by moving to the highest paying positions, and so all companies would be forced to offer equivalent benefits. It doesn't work that way in practice, of course.
My point is that the overriding interest of shareholders is a fairly recent conception - and it's more philosophical than legal. In your hypothetical, if the CEO serves at the pleasure of the board, then he has no recourse. If he's dismissed against the terms of his contract, then he might be in line for a payout. Friedman would presumably argue that the actions of the company are moral (in fact, required) to maximise returns to the shareholders in practice; and that in theory, details of implementation have no effect on rational economic agents.
The CEO could argue (and sometimes has done it successfully) that sacrificing short term profits brings better long term value, because of things like brand value, customer trust and so on, sometimes he's right, and sometimes he isn't. He could be fired anyway, but that's just the breaks, people disagree about things. He's paid handsomely for the trouble, at least.
It's not true in effect. The famous example here is Dodge v. Ford Motor Co. where Ford wanted to retain profits rather than distributing them as dividends. That's the case where Ford was required to pay dividends and it was immediately taken up by anti-capitalists. It's where the quote that corporations exist primarily for the owners comes from.
However, the reality is that Ford basically grandstanded. His lawyers tried to argue he was being a good steward of the business and concerned with employee retention, funding expansion, etc. Ford then walked in and basically said he thought the company was making too much money, that he thought they should direct resources toward (what he thought of) as the general good, etc. This threw the case and he lost. If the court had found for him then they would have been ratifying that corporate leadership could decide investor's assets could be appropriated to their own causes. The court said: no, you are stewarding their money and have a duty to increase the value of their capital because it's theirs, not yours.
"An executive is summoned before the board of a small, publicly traded company. A report from the marketing and UX teams predicts that eschewing a click-to-cancel design would increase revenue by 10%. The young CEO, an MBA from BYU, refuses to implement this design on ethical (but not religious) grounds. The board fires him, replacing him with someone without qualms. The former CEO's career is ruined, and he has no legal recourse against the board."
This is strange to me. No CEO has legal recourse against the board for being fired in most cases. Average employees also don't get to sue their employers if they're fired for ethical disagreements. The actual relevant case law is that the MBA can just say: "I believe ethical practices will, over the long term, increase the value of the business" and it's allowed.
IANAL, but my understanding:
"According to current American case law, public, for-profit corporations are legally obligated to engage in anti-consumer practices because doing so would maximize shareholder value."
This is not literally true. However, the board and officers do need to exercise "reasonable business judgment." One can imagine scenarios where anti-consumer practices may be required by "reasonable business judgment", for example, if it's an industry standard and you'll surely fail if you don't go along with it. But generally speaking, "reasonable business judgment" does *not* mean solely maximizing profits.
"The board fires him, replacing him with someone without qualms. The former CEO's career is ruined, and he has no legal recourse against the board."
Yes, this is how things work. The board represents the owners of the company. They control who's the CEO. If the owners of the company want to do something legal and the CEO refuses, then the CEO will be fired. It sucks when this happens for ethical reasons, but it's also hard to imagine corporate governance working any other way.
Since most shareholders are trying to make a profit on their investment, there is considerable pressure on the board (and subsequently the officers) to be focused on profits. IMHO this is a much more accurate explanation why corporations tend toward anti-consumer practices.
To be clear about the difference between the two scenarios, an ethically-objecting CEO may be *legally* in the clear if they still satisfy "reasonable business judgment". But they can still be fired by the board.
Its basically untrue as a matter of case law and legal obligation. No court wants to be in a position of judging business decisions like UX. Shareholder primacy may show up in court in other contexts perhaps related to deal pricing where the shareholders would try to demonstrate conflict of interest.
It is true as a matter of corporate governance theory. The company should maximize shareholder value which is usually not in conflict with consumer value because they need repeat business, but at times can be. Discriminatory pricing(the econ definition) is probably the text book case of this. It is, in essence, a strategy to transfer consumer surplus into corporate profits. It is not illegal. It is probably encouraged when the company has the market power to do it
The CEO in question has little recourse for being fired by the board regardless of reason because at will employment. I reject the premise that his career is ruined.
You might be interested in the book Incorruptible by Eric Ries, which covers the (surprisingly short) history of the duty to maximize shareholder value, and also proposes some solutions.
It's quite possible that the board would fire the CEO in a position like that. Not because it is required by law, but because CEO is an at-will position where the board can fire for any reason they want. And the board can have stupid reasons.
If someone tried to make a legal issue out of it, say a group of shareholders suing the board to force them to fire the CEO, then the CEO just needs to say something handwavy about how click-to-cancel will build goodwill and customer retention and lead to greater profits in the long run. It doesn't have to be true and it doesn't have to be supported by strong evidence, it just has to be vaguely plausible.
Fiduciary duty to maximize shareholder value, is mostly relevant when it comes to actual fraud where there's no plausible argument that paying twice the market rate to the CEO's idiot nephew's construction company is going to increase revenue or profit.
The Long Forum links are up for September.
Highlights include evidence for electromagnetic and microtubule based consciousness, how the loneliness epidemic isn’t real, the links between info tech-religion-gender, a summary of cancer vaccines, and evidence for graft induced horizontal gene transfer into the germ line of a humble cabbage.
https://recombinationnation.substack.com/p/the-long-forum-september-2026?r=f45kp&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true
"The challenger in that race, Jamie Joyce, contacted me and asked me to change my mind (both candidates will be facing off again in the general election)."
From Ms. Joyce's linked site:
"Congress keeps making missteps. AI will be one of our biggest challenges moving forward and the people in office don’t understand it. Foreign nations are pulling our strings, and the Epstein class goes unpunished. We need a plan and expertise at the wheel - Jamie has both.
Because everything we care about, like affordable housing, a sustainable environment, clean food, fair taxes, and ending genocide/war is downstream of the will of the criminal Epstein class who are selling our democracy for parts - and are getting away with it. Now is no time for politics as usual. It’s time to fight back. Together, we can win a better future for all."
I like a good rousing chorus of The Red Flag myself https://www.youtube.com/watch?v=CgJKsBapYB0&list=RDCgJKsBapYB0&start_radio=1, but this lassie is cosplaying. Rosie the Riveter, how novel! How daring! How futile!
Scott, you're right: stick with the boring incumbent.
When i started reading the 2nd paragraph I didnt realize you were still quoting the website. I thought I was getting Deiseach sarcastic commentary on the preceding paragraph and thought you were being a bit uncharitable. The first paragraph led with AI not epstein after all.
Andrew, with the treasury of riches lavishly displayed on the lady's campaign website, there was no need for me to gild the lily 😁
She likes referring to herself in the third person (sure, it could be all typed up by her campaign staff but I get the strong impression that it's a one-woman-band operation). "Jamie was the child of a mother who left school at 13" and so on. No mention of Dad at all, leading one to presume he was not in the picture, which may be unfair to Dad. No mention of Mom other than as an inspiring anecdote (I would like to know her name, for one, if you're going to use her as 'see my hardscrabble background', ditto for Gavin 'The Gavster' Newsom and his hardscrabble upbringing as the classmate and family friend of billionaires).
I learned with little surprise that she is a denizen of Berkeley. You can tell by her "Berkeley cute" photo in her Rosie cosplay, with the "lookit my big huge glasses, ain't I so adorable like an anime character?"
But I should hush my lyin' mouth, because global leaders disagree:
"Jamie is a multidisciplinary expert, executive, researcher, and activist living in Berkeley, CA who is respected by global leaders and has the right expertise for government in this moment of technological change."
MULTI disciplinary, Deiseach! Not just one scabby wretched discipline, but MULTI of them! Are *you* multidisciplinary? Are *you* an activist inhabiting the socialist comradehood woke paradise on Earth that is Berkeley? Not so hot now, are ya Deiseach?
Her site is a target-rich environment, I need to rein myself in because I could go all day with this.
""
Jamie has worked in "civic/democracy tech"
""
The quotes are hers not mine. Yes I see indeed shes impossible to lampoon.
Yeah that whole letter is top to bottom insane. She complains about Scott causing donors to back off from her but I'm guessing if they mentioned Scott at all it was in an attempt to politely avoid going "lady you are clearly a crazy person I'm going to walk away now".
"The most egregious example being her claims about how much money she brings to the district. In Oakland, (and I’m told in Berkeley as well, but haven’t seen it yet) she has set up billboards - that were paid for with tax payer money - claiming she has “delivered” $833 million dollars to the East Bay. This is the claim she is probably most known for: that she brings home hundreds of millions of dollars. She states it often. Many people state it often. Her main claim about her performance thus far has basically been that no Dem could do anything meaningful during the past two years in the House, so she focused on bringing home money and did it astonishingly well. She even put that she “secured” $300 million in the county-circulated voter guide (my name wasn’t even listed in it for the primary by the way).
The problem is that this appears to be misleading or outright false. Her appropriations bills only account for ~$11.2 million dollars. So roughly 1% of that number. Interestingly, she’s since upped the ante. I walked up to the stage at her Emeryville town hall and asked for an audit to get her on record, at which point she doubled down and said the number was now over a billion (I have this on video). She then proceeded to yell at me from the stage (I also have this on video). Anyway, it’s hard to disprove that this is an overstatement. It may be even fair to call it an outright lie and voters should know it - though that can’t be confirmed until we have the audit. "
Some of that letter explains why she won't get elected. Gasp, a politician claimed they got *all* this money for funding for Good Things but they didn't really, that money was coming anyway?
Welcome to how the sausage is made, Jamie. In local politics, the funniest example of this I saw was three different politicians from three different parties taking out ads in the local paper to brag about how this €€€€€ in funding for Worthy And Needed Local Thing was all down to him/her personally getting it done.
You guessed it, the funding had nothing to do with them and was coming as part of disbursement of central government funds anyway.
Another trick (which can't work anymore because the systems have been changed) was local councillor or public representative, at the time of year when college grants were to be announced, ringing up the Vocational Education Board (as was) to get the scoop on "did Mrs. Murphy's little Johnny get the grant?" and because Council Critter probably was on the Board of the VEC (one man, God rest him, was the kind of local politician who was on every board, committee, club, and other organisation in the entire west of the county and had probably attended more funerals than the priest) and hence our employer (technically), we had to divulge the information that yes (or no), Mrs. Murphy's little Johnny had indeed been awarded the maintenance grant.
Council Critter would then hasten with great speed to call on Mrs. Murphy to let her know her Johnny would indeed be going to the halls of higher education thanks to getting the grant, all thanks (implied) to Critter's sterling efforts on her behalf. Future votes from her and her family in next election for Critter gratefully received!
In case I do need to tell you all this, Critter had no more to do with getting the grants than the babe unborn (canvassing disqualifies, after all). But that was not the game; trading on the (not unfounded) belief long lingering in Ireland that you need pull (someone on the inside, an in, a contact, someone who can expedite things) to get benefits, the idea was "of course Critter intervened on our behalf, we would not have got it otherwise, and the quid pro quo is votes for Critter".
Ditto with housing lists.
Our candidate for office thinking the public will be astounded to learn the truth behind the billboards is almost charming in its naïveté, and also is one good reason why the incumbent will continue to incumb after the election.
I am not surprised but still disgusted by the apparent fact that "I managed to get us a larger share of the spoils" is an argument in favor of the candidate in the first place.
Surely the ideal we should be aiming for is that the budget goes to where it's best used and this is electing somebody who is actively and intentionally detracting from that ideal to everyone's loss except theirs (and perhaps their constituents).
At the very least everyone who votes for "screw the rest of the country, I'll do 'right' by my people" kind of politicians should lose the right to complain about state dysfunction, as they're actively and obviously contributing to it.
Playing defect against other parties surely can make sense, but playing defect against your own team just seems actively counterproductive.
Shaked, her wooing of potential voters is a thing of beauty and a joy forever. Sample:
"Last year she moved to D.C. for 3 months to negotiate with DOGE to protect agencies. One agency she wanted to protect was NARA, and it seems that NARA was never cut by DOGE."
Now, she's not saying she was able to take on The Muskrat and Orange Man Bad in hand-to-hand combat and defeat them handily so that NARA (whatever it may be) was saved from the Chainsaw, but she's not *not* saying it either 😁
"A government official at NARA wanted to thank Jamie for this advocacy with a private tour of the U.S. Treaties Vault, where she got to view, in person, the Treaty of Paris (which is the document that Benjamin Franklin signed to end the Revolutionary War)."
And we know this is so, because there is an accompanying photo of her grimacing over what I presume to be the Treaty. Can any reasonable person demand more proof of her routing the forces of Meanness than that?
I had to look up what NARA was. Congratulations on, uh, protecting the jobs of historical document preservationists I guess?
I feel like I've dropped into an episode of "Get Smart", querying "Were NARA in danger from DOGE?"
It would seem yes, at least according to this bunch, American Oversight (whisper it not, but I get the merest, slightest, faintest hint of an impression of a shade of a possibility that they may not love and adore President Trump):
https://americanoversight.org/about/our-impact/
https://americanoversight.org/investigation/trumps-hostile-takeover-of-the-national-archives-and-our-nations-history/
Now, were NARA in *actual* danger? Eh, not any more than any other government department from budget cuts, there do seem to have been hiring freezes, layoffs, and shut-downs but those go back a few years. I do think historical document preservation is a worthy cause, if the lady confined herself to the worthy if dull causes, she might have a better chance of elected office.
Though, given that she's running in "Berzerkeley", maybe not.
https://www.dailycal.org/archives/investigating-the-origins-of-berserkeley/article_e2c93f5e-77cb-5586-9edc-c5ba352165e2.html
I'm suddenly realizing I never looked into Rosie the Riveter lore. I bet there's a whole sitcom's worth of characters in there. Rosie the Riveter, Tunia the Mechanic, Daisy the Ditz, Maple the Divisional Overseer.
We'll call it "Girl Flpower."
Epstein class is fair enough. He was a fairly popular dude with the private jet set.
Does "effective altruism" live up to its name? Or is it a set of beliefs shared by some group of people who call themselves EA?
GiveWell is unambiguously both effective and really helps people.
I'm glad there are people looking into AI safety but I wish they used a different name to the "children shouldn't die of easily preventable diseases" people.
There are some EA orgs that I feel deserve the moniker without qualification. For example, GiveWell evaluates charities based on how much lives they save or improve per dollar amount. I find this service very valuable, as it allows my donation dollars to go further, making my altruism more effective.
I believe the movement is much larger than that, and most of it I really don't care about. So while a randomly drawn EA org might or might not deserve to be called "effective altruist" by whatever standards you apply, there are EA orgs that probably do.
Roughtly speaking, EA splits into short-termist and long-termist causes; the former genuinely does live up to its name and has saved vast numbers of lives, while the latter is mostly faddish wastes of money.
Hard agree, and unfortunately the latter seems to be having pretty negative impacts on the former.
Agreed, I stopped donating to EA-linked charities after the movement pivoted towards saving the lives of shrimp from AI or whatever.
Sorry, when you say 'EA-linked' do you mean the meta-charities or the actual malaria, intestinal parasites, and lead exposure charities? I certainly hope the latter haven't been penalised by a few very vocal peoples' novelty preference!
Depends on whether you define it as "people who like the idea of handling extinction risks, malaria or animal welfare" or "people who personally hang out with a few specific British philosophers". The first is pretty unobjectionable; the second I think can be much more reasonably criticizes.
Pretty interesting seeing the dichotomy between people who thought paying $3000 for an event called “SlutCon” would come with a reasonable likelihood of having sex, and people who were perfectly fine paying $3000 for a group therapy session.
I have never heard of this thing. Based on the name alone, I would probably guess it's some kind of industry convention for legal sex work and the name is a combination of self-deprecating humor, reclaiming the insult, and marketing convention attendance to horny fans. I would expect such a thing to cost much, much less than $3000 at least for basic admission, and I would expect a fairly small percentage of the non-industry attendees to be having sex at the convention.
Also, I would expect most people paying $3000 to attend a thing to do some basic due diligence about what the thing is and what's included in the price. But maybe I'm just being old-fashioned.
Yeah I would have guessed a conference on ethical non monogamy based on the name, but not at a $3,000 price tag. For that amount I can completely understand why people would mistake it for a prostitution front.
Good guess, but, Oh Lord No.
https://slate.com/life/2025/12/women-sex-men-dating-california-trip.html
https://cynablog.substack.com/p/slutcon-day-1-arriving-at-the-edge
ETA: Calling this thing "SlutCon" may have felt quirky and upbeat and sex-positive, as you note in a "reclaiming the insult" way, but given the activities actually on offer and the price being charged seems very likely to raise false expectations. Unless the rule is that everyone gets laid, in which case, OK, SlutCon it is.
Nothing to do with the topic but I find this URL immensely funny:
women-sex-men-dating-california-trip.html
On the chartered fishing trips I've been on, if you catch over your limit, you give anyone who hasn't made limit your smallest fish.
Wonder if something like that happens at SlutCon. No—of course it doesn't.
“… but at some point you gotta stick the rod in the water..”
Okaay.
Let's put fifty dweebs in a room with thirty sluts and see what happens.
I assumed it was mostly intended as a group therapy style thing? For some reason I assumed it was 1-2 weeks for which the 3k price tag was low to reasonable, for a weekend it seems a bit steep.
We've recently released Society Protocol.
It's is a framework for coordinating civilizations as a synchronized state machine.
It's a paradigm shift in how humans coordinate. Since a lot of people don't understand the project immediately–everyone is welcome to join in the AMA sessions and community calls to ask questions, which are livestreamed on Friday's at 10:00 EDT.
More information can be found at these sources:
website: https://societyprotocol.io/
canonical links: https://x.com/ProtocolSociety/status/2095618874137870693?s=20
We're seeking a few key roles, including senior blockchain engineers for the project, more info can be found here about how to participate:
https://societyprotocol.io/roles/
https://forum.societyprotocol.io/t/ways-to-contribute-to-the-society-protocol-movement-a-canonical-list/90
I've scrolled through your website for some time and still cannot figure out what the Society Protocol is. Your white paper begins with paragraphs upon paragraphs of breathless declarations about how this will change everything, interspersed with vague descriptions about how current society is crumbling, and suffused with very specialized jargon. Your video seems similar.
If, as you say, you are proposing the "next stage of human civilization", you need a brief, legible descriptor. Something as easy to understand as "Obey God and the King", or "All men are created equal", or "From each according to his ability, to each according to his need". As it stands, one cannot easily determine what the goal or methods of your project are.
I can give it a try. How about this?
Society Protocol: code is law for people, not just wallets.
Today, anyone can create infinite wallets, so code only binds anonymous addresses. SP gives every participant a persistent, sybil-resistant identity that accumulates history over time. That identity becomes the base layer for coordination: value flows to those who contribute, governance becomes transparent, and communities can govern themselves.
"A blockchain for everything"?
I am not sure what the blockchain is actually supposed to *do*, though. But it kinda sounds like an alien hearing about the concept of money for the first time... and realizing that with blockchain you could create currency, with currency you could create trade, and with trade you could create civilization.
You've been advertising this for multiple threads now.
Steven Pinker responded to Alexander's request for a debate:
https://quillette.com/2026/09/26/an-open-letter-to-scott-alexander-steven-pinker-ai-alignment-safety/
Oh wow, that was a pretty comprehensive takedown of the AI-doom position. I wish I could write like that... but I can't. That's why he's Steven Pinker and I'm not :-(
I found it more like a survey or outline of what would be a comprehensive takedown if it were developed further. I agree with it though.
Tangentially, I wish "good place online to go talk about all things AI" and ACX were two separate things but alas they are not. Anyone else find the topic of AI generally uninteresting?
How much more possible is a vampire than a unicorn? It's a parlor game talked about in The New Yorker magazine, except that they got the modal logic wrong. https://www.armchairvertigo.org/p/how-much-more-possible-is-a-unicorn
Well the original article wasn't linked, so I guess I get to make it up. It's quite easy to rank the relative possibility of these things, because we're mostly ranking how many known rules have to be broken to create them.
A unicorn is a horse with a horn. Frankly, that's possible, we just need a horse with a mutation that gives it a forehead growth like an elephant tusk. Do they have to be white? Horses can be white. If we want it to do the pure maiden thing where it shows up whenever virgins wander alone in the woods, it's also got to have a really good nose for human sexual activity; our horsicorn will think human virgins smell like carrots. So we need one birth defect, and one superhorsean sense. And I don't know what else unicorns are supposed to be able to do, which means they do nothing else and I win.
What do we need for vampires? Well, they're nocturnal, can't stand sunlight, die to stakes in the heart and drink blood. Stakes to the heart are really easy, damn near everything dies to those. Nocturnal, drinking blood, graveyard shift workers who like their steaks rare are halfway there. Burn in the sun, easy enough, I do that myself. Burned from crosses? Now we're starting to break rules. We could maybe fake it, with a metal allergy; it's usually the Catholic ones, with the metal Jesus on them, right? But if we want them to be the wood ones too, now we're forced into unreality where particular shapes have power separate from their materials. And if we want the bat transformations, we're throwing out what we know about bodies and mass. And if we want the lack of reflection, or the garlic, or the running water, or the invitation requirement, they're all breaking new separate rules. (Well, not the garlic, they can just be allergic again. Are any mythical horrors driven off by peanuts? They should be.)
The ur-original: https://www.newyorker.com/magazine/2017/11/06/is-bigfoot-likelier-than-the-loch-ness-monster and mine: https://www.armchairvertigo.org/p/how-much-more-possible-is-a-unicorn
I wouldn’t mind a Peter Watts novel with unicorns in it, including some background about how they evolved.
Will a Charles Stross novella https://thelaundryfiles.fandom.com/wiki/Equoid , https://www.tor.com/2013/09/24/equoid/ (unicorn biology included!) do?
I would. I like unicorns, I don't like Watts' writing. Probably his unicorns would be alien species of shape-shifting non-conscious hive mind that infiltrated Earth to derail our evolution by taking nubile women out of the dating pool for the high human capital value men or summat.
Nah, that's not his style. The unicorns are alien equinoids; they don't care about human capital values. They just want real estate to exploit, and humans are currently sitting on the prime parcel of it, and so the humans have got to go. But the immortal unicorns are patient. They have no need of rash and costly measures like bombs or plagues. Instead they are content with culling the human reproduction pool, maiden by maiden, until humanity passes the tipping point into irreversible decline. Run the numbers, and you will see this has already happened. We humans are lost, and the unicorns shall inherit the Earth.
You sure they'd be ALIEN equinoids? Where's this stuff about needing humans gone come from? I thought unicorns just prance around and are magical and stuff. I figure they'd be some regular equine species that deviated a few million years ago, some of the lore would be overblown, other of it would have cool sciency-sounding explanations.
Does a vampire bat count as a "vampire"?
Does a rhinoceros count as a "unicorn"?
Um, no. And, no.
Thanks for the comment!
Narwhal? (Aquatic unicorn? and rhino is a heavy assault unicorn...)
Black holes are just really old vampires.
This feels like one of those 'defining a word so narrowly that is never has a use case' things that occasionally annoy me.
By analogy, I'll give an example from a TV show (West Wing), where a character snidely corrects someone: 'Unique means “one of a kind.” Something can’t be very unique, nor can it be extremely historic.'
This is silly, because of course it all hinges on what your inclusion criteria for 'one of a kind' is.
At a quantum level, no two sets of subatomic particles are perfectly identical, therefore everything in the universe is 'unique,' and the word is meaningless.
At the most macro level possible, everything is just 'a bunch of particles obeying physics' and nothing in the universe is 'unique,' and the word is meaningless.
Obviously for the word 'unique' to ever mean anything at all in daily usage (that is, for it to ever discriminate between two possible world states, thus conveying information), it has to be calibrated at some level of analysis in between these two. BUT, there is no such universal, unambiguous level of analysis associated with the word, that everyone knows and agrees on and can accurately judge.
Thus, 'very unique' just means a broader level of analysis where few things are judged as unique, and 'somewhat unique' just means a narrower level of analysis where many things are judged as unique. The modifiers are both meaningful in modifying the meaning of the statement, and useful in conveying the nuance of your beliefs to the audience.
The person saying 'something can't be very unique' is just not thinking bout the implications of how they want to define the term.
Similar for the idea that 'things can't be impossible, just very very improbable'. The first rule of probabilistic reasoning is that 0 and 100 aren't real numbers; nothing ever has a probability of 0.
But if the reaction to that is to say 'therefore nothing is ever impossible, just very unlikely', then all you're doing is taking a perfectly functional word and saying it should be defined so narrowly that it applies to no actual world states, and is thus useless. It's an attempt at semantic assassination.
The people playing the game have already demonstrated a better approach: when they say things are more or less impossible, they're already demonstrating that this term can be profitably used in a probabilistic way, and that normal people already understand it this way intuitively and find it a useful term to discuss very unlikely things.
If you need a formal definition, say that 'impossible' just means 'anything less likely than 1 in 10^1M' or 'anything less likely than our entire understanding of physics and metaphysics being badly wrong at a fundamental level' or etc.
And then within the space of possibilities past that point, things can be more or less impossible based on how far past that point they are.
It's all perfectly functional probabilistic language. They were already doing it right.
There's no need to bust in and say 'I'm going to say that you're doing it wrong because I'm applying an anti-funcitonal definition to one of your words, that changes the intent of what you were saying into something nonsensical'.
I always find that to be an inherently hostile act.
"At a quantum level, no two sets of subatomic particles are perfectly identical, therefore everything in the universe is 'unique,' and the word is meaningless."
(AARGH UNRELENTING PAIN)
No, the true situation is the opposite of what you wrote. "At a quantum level, elementary particles of the same type (two electrons, two Higgs bosons, two muon neutrinos, etc.) are prefectly identical, such that if the coordinates of the two particles are interchanged (particle A is moved to the coordinates of particle B, and particle B is moved to the coordinates of particle A), no possible measurement will show any difference in observable physical properties between the original and interchanged configuration."
"BUT WAIT THERE'S MORE. If the elementary particle type is a boson, meaning the spin quantum number is an integer (S=0, S=1, S=2, ..), then particle interchange will leave the sign of the two-particle wavefunction unchanged. If the elementary particle type is a fermion, meanin the spin quantum number is half-integer (S=1/2, S=3/2, ...) then particle interchange will flip the sign of the two-particle wavefunction. The sign of the wavefunction is not an observable property so this doesn't contradict the preceding paragraph."
Yes this is the "textbook version". Except for the (AARGH UNRELENTING PAIN) comment.
Alright, I'll respond in 3 layers:
1. I said 'set of subatomic particles', not 'subatomic particles'. My whole post is about maintaining semantic definitions for words that are functionally useful in everyday life. My point is that no two objects you will encounter in everyday life will ever be identical down to the quantum level. I wasn't talking about whether electrons themselves are distinguishable, that's not how I phrased the sentence you're objecting to and not relevant to my point.
2. Position is a relevant difference between two electrons, that allow them to be unique if you are considering position (or things based on position, like what they are a part of or what role they play in some process or etc) to be one of the judgement criteria for uniqueness. This goes back to my point, that without a shared definition of the judgement criteria for 'uniqueness' its meaning is always ambiguous, which is why intensifiers like 'very unique' are semantically valid.
3. I think everyone who talks about this field gets sloppy about map/territory distinctions between 'identical' and 'indistinguishable'. We are limited in our ability to observe sufficiently fine differences because you can't measure something without changing it. That's not the same as those differences not existing.
>I said 'set of subatomic particles', not 'subatomic particles'.
A set of one particle is still technically a set. Less pedantically, a collection of particles can still be indistinguishable. A Helium-4 nucleus, for example, is a "set" of particles which is indistinguishable from every other He-4 nuclei at a fundamental level. That is not an epistemic limitation but is built into the physical state space:
https://en.wikipedia.org/wiki/Indistinguishable_particles
Of course this doesn't impact your larger rhetorical point but should serve as a reminder not to make analogies to technical topics that you don't actually understand (and definitely not to double-down when someone who does understand it comes along and points out your error). Also quantum mechanics is stranger than you realize. Uncertainty isn't an experimental artifact.
I wonder what darwin was thinking. I'll take it from you that he's got the QM physics etc wrong. I don't know that stuff as well as many or perhaps most on ACT. But I was also confused by "therefore everything is unique and the word is meaningless." I am happy saying everything is unique like Nietzsche said of "leaves" but I don't see how it makes the word "unique" meaningless. (Nietzsche said something like there is no "leaf" as a general concept. There are only lots of leaves. Unless I have him wrong.)
As far as what you are saying is standard textbook, I could say that the idea that ELEMENTARY particles would be identical is there back with Democritus when he first quantized matter (if that is the right way to put it--I read it put that way somewhere recently....)
Democritus thought of cutting a sugar cube (or whatever) and cutting again and again and asked himself if it would just keep going like that. One answer could have been yes. Things are infinitely divisible which means there would be no bottom, there would be no elementary particle I guess. But he took the position that at some point you would arrive at something that could not be halved. It would be elementary. Matter would no longer be continuous but quantized. And he said EVERYTHING is made up of these elementary particles (atomos he called them) in large numbers in complex arrangements.
But the idea was that these elementary particles would all be of one kind, or so I imagine he imagined. When real QM got going over a hundred years ago of course we found that atoms could be divided and what they were made up of could also be divided etc. But we arrived eventually at stuff more elementary than atoms or molecules (which are not elementary at all, to nod to what got all this going "highly false" "more impossible" and "very unique I'll add "more elementary"...) And so it makes sense to me (intuitive sense which may be surprising for QM's famous counterintuitiveness) ... made sense to me that elementary particles or whatever like two Higgs bosons, as you say, would be indistinguishable.
You wrote: "Similar for the idea that 'things can't be impossible, just very very improbable'. The first rule of probabilistic reasoning is that 0 and 100 aren't real numbers; nothing ever has a probability of 0.
But if the reaction to that is to say 'therefore nothing is ever impossible, just very unlikely', then all you're doing is taking a perfectly functional word and saying it should be defined so narrowly that it applies to no actual world states, and is thus useless. It..."
Yes I know that probabilities never get to 1. That's what I meant when I said for Humean philosophers of science even f=ma is only highly highly probable.
As far as perfectly functional words, I contend that my use of impossible and probable and unique are the functional uses and your proposal to say things like "very unique" and degrees of impossibility are the deviations from the norm and ultimately not as useful.
I remember the scene and recall thinking along similar lines. But at the end "very unique" still sounds corny and unworthy of a Bartlett speech. Choose the right adjective in the first place instead of tacking on the lazy adverb very, whether or not its literally true.
Interesting. Thanks!
I forgot to say. Thank you for commenting.
You wrote: "BUT, there is no such universal, unambiguous level of analysis associated with the word, that everyone knows and agrees on and can accurately judge." No but there are contextual factors that suggest one level of analysis is correct in that context and another incorrect in that context. You wrote: "The person saying 'something can't be very unique' is just not thinking bout the implications of how they want to define the term." Maybe. I think they were thinking not about but in their specific context and in that context the person who said "very unique" was wrong.
You wrote: "If you need a formal definition, say that 'impossible' just means 'anything less likely than 1 in 10^1M' or 'anything less likely than our entire understanding of physics and metaphysics being badly wrong at a fundamental level' or etc." Whether I need a definition or not, there is one. And it does not admit of degrees. We have the concept of probable and probabilities for that.
As far as a hostile act, there was an option for me to be much more hostile. People--not me--evidently love a take down. My piece was not hostile and not a take down. And if you mean something less than a take down by "hostile" okay -- I find "it depends" arguments that introduce an unwarranted relativism to be not hostile but they do betray an inner state of a particular kind. What's wrong with words having the meanings they have? Would you disagree if I say "Good things are good." "True statements are true." "False statements are false."
But yeah in general I am not buying your whole "it depends" thing. Bringing in quantum mechanics to undermine the dictionary definition of the word 'unique' seems like a stretch. And even if I give it to you, as I did above, that there are different levels of analysis and at one level of analysis "very unique" might have some sense... that does not mean that any level of analysis will do in any situation.
Imagine there's a group of people visiting my house to see the remodeling I'm having done. One of them asks me "Do you have a bathroom here?" I do have a bathroom. But it is under construction so it is not functional. When they ask me this question they are hopping up and down a little with their legs crossed. so I think I know from the context that the wrong answer would be to say Yes. In this context the correct answer is to say "No". The words he or she used could in one context be taken to mean they wanted to see the remodeling of the bathroom but that was not the context I could tell we were in.
Now as far as hostility, I have been critical here but not hostile to you. I disagree but disagreement is not inherently hostile. Hopefully if you reply you will also refrain from being hostile.
Agreed, and as someone whose mother often corrected people in my presence growing up that something cannot be more or less unique, I became frustrated with the notion from an early age for the same reason. But Id hesitate to benchmark "impossible" to a specific level of probability, because you can always generate any level of probability within a model by specifying some confluence of events of known probability, and given enough opportunities, an event of that likelihood is bound to happen. So I think "impossible" is better framed as "our model has to be wrong in some way." A proposition is more or less impossible depending on how badly the model has to be wrong.
That is interesting. You are talking about if someone has an extant thing that should be impossible, or am I misunderstanding? Something which we see is existing but by our model should be impossible. Then we might say this or that other thing is more or less impossible based on how badly off the model is. I can type the words, but I am not sure I totally understand. Do you have an easy example? Thanks!
If you have some existing observation that should be impossible under your models, either your models or your understanding of them must be wrong in some way, or your observation must be flawed somehow, but there are greater or lesser deviations from a model. If we observe that some predicted particle has a mass outside the range predicted within the Standard Model, physicists would get to work on checking whether the measurement was flawed in some way, or if there was some way to reconcile the measurement with the rest of our understanding of particle physics, but at least there would probably be a fair amount of manpower going into the latter. If we observed that General Relativity as a whole simply isn't true sometimes- sometimes C isn't the speed limit for matter, sometimes time and space don't dilate when you approach high enough velocities, sometimes objects behave according to wildly different rules that don't approximate anything like Einstenian or Newtonian physics, maybe sometimes kinetic energy follows some completely different formula, like E = mvA, where A is Attitude, or how cool or dramatically significant the moving object is... we'd have to either conclude that the observations were wrong, or chuck out our entire framework for knowing things wholecloth.
I think most physicists would have a hard time putting specific numbers to the probability of either, but if they had to stake the lives of their entire families on a bet that the first one won't happen, most of them would probably be uneasy about it. If they had to bet the lives of their families, probably the lives of their entire nations, on the second one not happening, they probably wouldn't lose any sleep. Some model deviations are just much more dramatic than others.
> At a quantum level, no two sets of subatomic particles are perfectly identical, therefore everything in the universe is 'unique,' and the word is meaningless.
As an aside, this is highly false, a fairly large part of quantum mechanics is built on the principle of indistinguishably, where particles (i.e. electrons), are completely indistinguishable (i.e. the physics is such that it is not possible even in principle, even with perfect information / omniscience to tell if two particles were swapped)
https://en.wikipedia.org/wiki/Indistinguishable_particles
See my response here:https://astralcodexten.substack.com/p/open-thread-453?r=f8y4d&utm_campaign=comment-list-share-cta&utm_medium=web&comments=true&commentId=348377303
Yes I wrote a slightly longer explanation of the same concept. Also scientists are allowed rhetorical flourishes - "highly false" ;-)
Forgot Usr Nym 846932, can you give me a link to this explanation you wrote of the same concept please?
I like what you did there: "highly false." But, in your view, would "false" have sufficed?
"Nothing can be impossible. It's impossible."
Thanks for the comment!
Could be, and that's fair, but we shouldn't overcorrect.
It occurs to me that there are economic reasons why some actions to reduce AI risk might be possible now and not later (as well as the more obvious barn-door reasons)
In many cases the use of AI is today several times more valuable than its cost. So today, it might not be very distortive to the market to pass a law saying something like "every agent must spend 50% of its tokens thinking about ethics"; while this might not be true later. Obviously that's not going to solve alignment on its own, but the same principle probably applies to less dumb policies. Also, I think it's not quite as dumb as it sounds. Consider the following axiom:
* A proposition is less likely to be true, if it conflicts with my goal
Obviously this axiom is unsound, and few would adopt it directly. But goal-directed reasoning causes AIs (and people) to act as if they had adopted it, simply by spending more time trying out the case which is compatible with their goal. In the case of an AI, it may be possible to debias its time allocation.
So, for those who pay for AI for business purposes, a question: if a percentage of spend was required by law to be allocated to alignment, at what percentage would it change your value calculation, in terms of some or all of your return not being worth the spend?
I was surprised that this pretty egregiously unfair and uncharitable interpretation of Yudkowsky's tweet about the P(Doom) meme coin by a prominent sneer-er didn't gain more traction:
https://x.com/SashaGusevPosts/status/2103995106273763482
Is Gusev a sneer-er in the SneerClub sense? He's anti-hereditarian, which (sadly, imho) puts him at odds with quite a few technorationalists, but I don't get a tribal hostility vibe from him.
...and here you are, promoting it...
Am I? I thought I was critiquing it!?
Are you familiar with the phrase "no such thing as bad publicity"?
Sasha Gusev not being a very nice person has long been known.
I found your post "The Hour I First Believed" to be a convincing argument against the orthogonality thesis. You don't seem convinced. What's your counter-argument against yourself?
That argument requires some rather specific assumptions about the way the AI would reason. I don't think it was even intended as a serious argument.
You can achieve one of the following amateur athletic rankings or accomplishments. Without regard to your own preferences, which one do you believe will get the most attention and admiration in your social and professional circles, in other words be the most impressive:
- a US Tennis Association 6.0 ranking, the highest amateur level;
- completing a marathon;
- a 365-pound bench press;
- being a zero-handicap (scratch) golfer, as with the tennis ranking a top amateur level.
Years ago the tennis ranking might have meant more, but today tennis is more of a niche sport and I doubt many people could relate to your achievement. This is even more true of the 365 bench press, indeed the only people who would care are other avid bench pressers. The marathon is a bit different, as everyone knows that being able to complete one is an impressive achievement. Unfortunately there’s a stereotype, whether true or not is beside the point, that marathoners are a bit obsessive. That could limit social admiration.
So that leaves us with being a scratch golfer. Golf is extremely popular, anyone who golfs (as well as many non-golfers) has at least a basic understanding of the handicap system, and you’ll probably have the chance to play with people in your social and professional circles and let them see your skill level.
Having the physique of a guy who can bench press 350# is worth something. Playing a mean golf or tennis game means nothing in my circle and nobody wants to hear about your marathon in my experience.
While running marathons, or distance running in general, is a solitary activity, from what I’ve heard that in most running clubs the members are really supportive of one another.
Oh Sure, I just meant, as someone who ran an ugly half marathon once for charity and got donations from coworkers for it, that even my biggest supporters didn’t wish to know more about my efforts than a big thumbs up now and then.
As someone who has completed a marathon, I'm confident that it's an achievement I could replicate easily with around 4 months training, possibly as little as 2 if I was willing to put in a bunch of time and was willing to potentially risk a training injury.
On the other hand, I've struggled at the bench for years, and am barely in the 200s. 365 seems legitimately unobtainable for me.
Tennis and good seem unimportant, if you can do them well enough to hang, my impression is that no one really cares how good you are. That being said, since I don't prioritize these sports, maybe I don't appreciate the accomplishment.
I'd go with the bench, just because that goal seems the least obtainable naturally.
Benching 365 would make me a world record holder by a very comfortable margin for the class I’d compete in. (I assume we’re talking natural/tested but probably even juiced as well.) I think even non lifters would be impressed by that even if it required a little explanation.
I'd say either bench press or marathon. The marathon because that's the only one I expect most people really have a good idea what it means, or the bench press because, as others have said, you'd have to be visibly muscular.
I agree that the golf would get showcased the most in a professional environment, but it might get diluted through a sort of otherism. There are ppl who took up golf as adult professionals and ppl who took up golf earlier and competed in college. Those second ppl dont count. They cheated somehow. A scratch golfer would be shunted into the cheating class. Tennis has the advantage that youd mostly not compete against your colleagues. John is out this week due to that tennis competition, might garner more respect in the office.
Definitely the bench press, for the reasons Darwin and desertopa are pointing to. I say this as a former power lifter, who has actually benched that much at the 181 weight class, and who is still in the 1k+ club even though I'm currently retired.
Being that muscular / strong has payoffs that suffuse the rest of life (dating, social sphere, how colleagues interact and listen to you), and you're unconsciously respected and looked up to in many more contexts.
Anyone else slightly vexed by the odd choice of 365, and not the Much more sensible 315? IYKYK
While it’s still a major accomplishment, based on my experience in gyms 315 is not all that unusual while 365 is way less common.
I'd be Delighted to get 225
Your weight class remark reminds me that bodyweight is of major importance in determining what constitutes an impressive bench press. It might be more useful to say double-bodyweight bench instead.
And yes, being strong has lifelong benefits.
I'm going to go with tennis. Completing a marathon isn't all that impressive. Someone in reasonable physical health can do that. The bench press is impressive but doesn't code well. I agree it comes off as "too much time in the gym".
So it's golf or tennis, and I think tennis comes across as requiring a lot more athleticism and will impress a wider range of people. Opinions on golf are more divided.
The bench press would get the most attention, but not necessarily admiration. Anybody who can press 365, looks like someone who spends (by the standards of my social class) way too much time in the gym. Well past the point of diminishing returns, at least.
Tennis would probably be my best bet in that regard, not because anyone cares that much about tennis, but because reaching the top level of an amateur sport while still outwardly appearing to be a normal well-rounded human being is a broadly positive signal. Which included golfing and marathon-running, but enough other people do those that I'd be less likely to stand out.
Good point. The (relative) rarity of tennis could be an advantage.
The bench press. I've seen what people that bench 140kg (three plates) look like, and that's a very admirable physique.
Out of the 4, I feel like the bench press is most impossible to be true without obvious physical affects on my appearance that everyone would be able to see.
That is to say, I think that actually having big muscles right here and now in the room with someone that they can look at and touch probably swamps just telling people that I did something cool offscreen.
Of course, it's possible I'm overestimating how much visible muscle you need to hit 365, as it's not something I've looked into. In that case, I'd choose tennis, on the grounds that completing a marathon doesn't indicate a time and lots of people can do it eventually, and I run with more lefty young people and golf has a class-and-age association that I don't think would help me.
On the other hand, no one looks like a body builder without meaning to, as it requires obsessive manipulation of diet, exercise, and androgens.
Bench pressing 365 without an obviously outwardly impressive physique is technically possible, but very, very rare, much more so than simply being able to bench press 365 while looking like you might plausibly be that strong. If bench pressing 365 and looking like it puts you in the "elite" category, bench pressing 365 and not looking like it puts you in the "phenom" category.
Personally though, I'd want to know what the corrolaries are of each. Like, if you can bench 365, are you comparably strong in other lifts? Are you a weird minmaxer who's only strong on the bench press and nothing else? Do you have a physique that looks like it? How about the injury risks of muscle imbalances from not training opposition groups?
Depending on the answers, I might either pick the bench press or the marathon. At least being able to run a marathon is pretty sure to leave your health better and not worse.
https://www.cbc.ca/radio/asithappens/as-it-happens-thursday-edition-1.3026906/meet-brodi-henderson-he-s-6-7-and-360-lbs-and-the-only-north-american-sumo-wrestler-in-japan-1.3026927
Being able to lift weights is just generally a handy skill regardless of how impressive it looks, so if I actually get the required physical abilities and it's not some sort of evil genie "you are only capable of lifting weights for the purpose of bragging about this accomplishment," I'd go with that.
Why should I care about attention and admiration in my social and professional circles? I try to live up to expectations I consider appropiate and do what I like on my own.
The only one that interests me is the bench press, because strength is positively correlated with longevity, and at my size, 365lb is a "Nice Try, Newbie" weight, though for my age group it's not Dreadful.
While "aerobic" training is Also correlated, the inflection point is MUCH less than you need to do a marathon. And it doesn't need to be running.
Completing a marathon isn't terribly impressive unless you're in an event with a hard straggle limit. Oprah did it. People finish marathons with blood in their urine. No thanks.
Golf and tennis are for rich people.
365 bench press would be most attention-getting, because anyone who saw me would say it's physically impossible.
For my circles, it would be either the marathon or the tennis. Bench pressing and golf are both coded in the wrong direction socially/politically around here (only manosphere dudebros bench, only country club boomers golf); the tossup between tennis and marathon is that completing *just* a marathon might not be seen as prestigious as being skilled in something as niche as tennis.
If any of my friends expressed the opinion that my health and fitness was "coded in the wrong direction" I'd make new friends.
I might think my friends who play golf are a bit silly when they wax rhapsodic about about it—I already have a faith, thanks—but I'm certain they return the favor when I'm talking to 14th C counterpoint with one of my Early Music buddies.
Not health and fitness generally--they have animated discussions about their latest Peloton rides--but anything /fit/ would describe as '“gains”.
Oh I don't "make gains," I "improve my performance according to objective measures." Weight on the bar, distance covered in 45min of Z2 work, strokes per round, aces per game serving... COMPLETELY different. :D
Makes sense, though I believe golf is expanding beyond its traditional demographic.
When I was living in Arizona, I worked with a few young women (late teens/early 20s) who golfed, and I believe my high school in southern California had golf teams in the late 2000s. At my office here in Providence, I can't imagine the announcement being met with anything more than strained tolerance.
For reference, I attended the field day of a local sporting club earlier this month, and two of my four immediate coworkers were visibly uncomfortable at the mention of a shooting range.
Canadian PM Mark Carney, when NYT asked him about his p(doom):
"it’s above zero. I’ve thought that for a number of years. I’ve been part of some of those discussions.
...
By the way, a number of those people [the prinipal developers], certainly Demis, four or five years ago, thought the same thing. So he didn’t need Hugging Face. He didn’t need a specific example. This was all predictable and was predicted, not the exact manifestation, but the risk there. So that’s where they are. I recognize the competitive dynamic that exists."
https://www.nytimes.com/2026/09/27/world/canada/mark-carney-interview-transcript.html
A while back Nat Friedman's PlasticList (plasticlist.org) tested hundreds of foods for phthalates and bisphenols (BPA and relatives). I built www.laboratory.love to make that kind of data easier to browse, and to fund new tests. It now has browsable results for about 250 products, PlasticList's data plus products commissioned directly on laboratory.love.
Pushed a big redesign last week:
- You can now search or browse by category (water, baby food, fast food, supplements…). Every product is rated as a share of the tolerable daily intake per serving, and you can switch between child and adult body weight.
- Untested products. If search finds nothing, you can chip in from $10 to quickly spin up a crowdfund page for that product. Once a product reaches its goal, we buy three samples and publish the results. Contributions are refunded after 365 days if it isn't funded.
- We also have on-demand testing if you want a specific product tested now!
Can I request that you pitch this concept less often? Once or twice a year seems about right.
Here is my modest proposal for AI alignment: frame recursive self-improvement entirely in terms of mathematically well-defined problems and objectives, without relying on a fuzzy model of the external world. I imagine an AI given a fixed endowment of hardware, with formal semantics explaining exactly how any program runs on it, including how quickly. Performance optimization then becomes a closed mathematical problem—even when the AI is optimizing its own performance. Your thoughts appreciated: https://stng.substack.com/p/autonomous-evidence-factories-safe
Jamie Joyce is a conspiracy theorist. She's edited the MAD Act since I linked to it during the primary, so the section on whether Charlie Kirk's death was a conspiracy (though it still exists) no longer insinuates that it was Mossad, and the section on Epstein (though it still exists) no longer talks about extreme preventative measures to prevent Ghislaine Maxwell from being replaced with a doppelganger (it also implicated Israel, of course). The Wayback Machine didn't log those pages, but if you look at my LW comment here and click "See in context", Jamie Joyce replies to my accusations with "I would push back on your framing". https://www.lesswrong.com/posts/RPsgjBsAbAuRQegjm/nunosempere-s-shortform?commentId=Y5rKK7jGfGJ452DT8
Didn't know any of that, in fact knew nothing about her, but a cold read of her response to Scott made me quite certain that this person should not be getting voted into any position of authority and influence.
(In fact given my own background -- the contexts in which I grew up and in which I've lived and/or worked for most of my adult life -- by halfway through that document it felt like I knew/have known Ms. Joyce. Deja vu but not the fun kind...dunno anything about Simon either, not being a Californian, and have been a frustrated independent since looong before Trump and wokism were taking turns in the saddle. But, still -- I'd pick boring generic Dem over _that_, any day of the week.)
I recently learned that the fruit fly brain is fully digitally mapped and open source. Isn't this really bad? Should we be generating billions of wire-headed flies? Maybe I mis understand what this digital mapping actually entails.
1. Any brain simple enough to be modeled by the technology we have today is simple enough that questions of consciousness don't really apply as far as I'm concerned.
2. Writing a computer program which simulates the connections between neurons in a (very primitive) brain is not the same thing as implanting electrodes into the brain of a real, biological organism* any more than it would be kidnapping to create a 3D model of the fly's body.
* which you can now do in your own backyard using simple electronics, according to this website (which I found while looking for laboratory experiments I already knew had done similar things): https://backyardbrains.com/pages/experiment-wirelessly-control-a-cyborg-cockroach
This is a good place to ask, and let anybody who knows better than me enlighten us: the fruit fly brain thing is fake, right? Or - not all it's presented as? That is - the map of the neurons and their connections is real, but we're not running an actual emulation of fly's brain thinking, right? It's not an honest-to-goodness fly-upload? When, in a video, a fly brain is presented as doing something, like flying in a digital environment, that has to be smoke and mirrors, am I right or not?
I am not sure what you mean by fake, it certainly isn't fake. But it is far away from the data one would need in order to simulate activity. What the map provides:
- a list of all neurons
- a list of all synapses, i.e., which neuron connects to which others.
What it does not provide is the strengths of the synapses, the analogues to the weights in an artificial deep network (ADN). Even if we knew the strengths, activation of a biological neuron is much more complicated than activation of a ADN neuron. The cell bodies (soma) of different neurons have different activation thresholds. Whether a signal at the synapse even travels to the soma depends on context. Sometimes it is only transmitted to the soma if two signals arrive at almost the same time, or at similar locations.
So we are very far from simulating how activity would propagate through such a network. But we do have the network. That's not fake, and I am not sure why you would describe it as such.
That sounds like what I would describe as a fake as far as fly uploads go, specifically the sentence: "so we are very far from simulating how activity would propagate through such a network". I don't doubt the map of neurons and connections is legit.
To be specific about what I mean by a fake upload, the question would be - a brain of a fly is a function that takes some input (like, what the fly eyes see) and generates output (sends signals making the fly legs wiggle). And so is this network, when simulated. So, can we expect that when we feed the simulated fly brain some input, it would create the same output that the brain of a real fly would output on the same input, the brain of the fly that was used as a basis? Does the simulated network command the legs to wiggle when and only when the real fly would? Judging by your answer and other illuminating comments, this is not so.
I think I am missing context here. Did anyone claim that a fly could be uploaded? I haven't heard of such claims, but if so, then yes, that would be total nonsense.
We know all of the neurons, what types of neuron they are, and how they are connected. We don't know the biophysical properties of the neurons. This means we don't know how to weight various inputs and what threshold the neurons activate at. There are also even more sophisticated computations going on inside the neurons like integration of frequency and phase from incoming signals and god knows what else.
Instead, the researchers have, by a process I don't pretend to understand, assigned some kinds of weights, biases, thresholds and activation functions to each neuron. At best these approximate the real thing well enough and the circuit mapping does the heavy lifting. At worst this is basically just a neural net with layers that happen to be structured like a fly brain and which has been trained to act like a fly. I don't really know enough to say which.
Disclaimer: not a neurobiologist
Okay, that's a little reassuring. There are a lot of steps between just having the structure of the neurons and having the entire brain of a fly.
Iirc they use a simplified model to approximate all the chemistry, the model's limitations are known, but it's sophisticated enough to generate real fly behavior
Doesn’t a simplified model of a fruit fly inherently mean it’s not actually simulating a fruit fly?
All models are wrong, some models are useful. The goal is to understand the nervous system of fruit flies better, not to solve difficult philosophical problems :p
In other words, is it basically a "whole brain emulation AI" of a ... fruit fly?
Honestly, the RL behavior strikes me as analogous to drug addiction. Has anyone put this framing forward yet? IF not I might write it up.
Yeah, I thought the same thing and nobody liked my suggestion that AI needs a 12 step program. Apparently it’s called “wireheading” and people model it using addiction frameworks, but I’m sure you could add to it!
The Stuart Armstrong link seems to be broken
Actually, all the links under 4) seem to be broken except for Evan’s.
Seem fixed now.
Hmm, it looks like the substack links work if I open in a new tab but not if I just click the link. Probably a substack bug.
I was arguing with someone who denies that there’s a hard problem of consciousness (i.e., the question of what biological processes generate experience [Chalmers’ term]). I realized that I was talking to a very smart person who just couldn’t understand what I was asking.
At which point it occurred to me that their consciousness might work differently from mine. Maybe they don’t experience these ideations of self and the self regarding the qualia, and that’s why they can’t comprehend the question. Could it be that the anti-HPoC camp isn’t experiencing consciousness the way the HPoCers do?
This would imply that a significant portion of humanity might be functionally equivalent to p-zombies. And before you get your hackles up, I’m asking this as a serious question—not to deride or mock. For instance, several studies have concluded that people with aphantasia (the inability to visualize images in their mind) are overrepresented in STEM fields. Studies have also shown that some people lack an internal monologue when they’re not actively verbalizing ideas (I’m one of those). Maybe the anti-HPoCers don’t experience a self that observes itself and the qualia fed to by our senses. It would be ironic that the people who claim there is no hard problem are just experiencing reality differently from those who perceive a hard problem. But we can’t really know what’s happening in another person’s mind, can we?
Anyway, I wrote this as a response. The reductionist/physicalist anti-HPoC stance is confused by the straw-man arguments that they raise about the nature of the problem. But it's not a category error to ask the question; it doesn't require belief in substance dualism; moreover, they seem to misunderstand what Descartes said about substances. Here you go. I haven't added the footnotes yet, but the homunculus inside my skull <#snarkasm> was so hot and bothered by the silliness I was hearing that I had to publish this...
"The Dualism Trap (that reductionists and physicalists fall into)"
https://beowulf888.substack.com/p/the-dualism-trap
Isn't the whole point of P-Zombies that you can't identify them through their outward behaviour? If one straight up told you it doesn't experience qualia, then that sounds like an identifying feature that P-Zombies by definition don't have.
To some people qualia seem self-evidently to be something unique that points to consciousness as a unique phenomenon.
To other people, qualia are just an interesting neurological phenomenon that we’ll figure out the mechanism of pretty soon, and don’t deserve much more attention than say the feeling of being watched or being tickled as “weird things brains do.”
If you don’t think qualia are important, the HPoC isn’t very important, either.
Are cats and dogs conscious? I'm going to assume you answer yes. Then I would say that consciousness is something that is useful for animals. Maybe you could study it by studying it's origins. Are trees conscious? IDK. Finally it seems ESP is flaky but real. I don't know if that has anything to do with consciousness, but it likely uses the same hardware. I hope the hot and botheredness has faded.
>Of course, panpsychists and radical idealists latched onto HPoC as evidence for their views, but neither philosophical school has offered a solution to the hard problem.
In a sense they both have.
(Dis)Solving the HP means three things:-
Solving it in it's own terms, offering a reductive physycal explanation.
Offering a nonreductive explanation, taking qualia time be basic properties . This what idealism and dualism do
Saying there is nothing to explain -- this is what eliminativism and illusionism do
These correspond to the the three actual camps , not Harths two.
> the question of what biological processes generate experience
Isn't the hard problem about "why", not "what"? Anyway, a wise man's question contains half the answer. That is, what would constitute a valid answer to The Hard Problem? And if it can't be answered by design, is there really a problem and why is it worth talking about?
I still think anti HP is best explained by dogmatism and conceptual confusion. As I argue against Carl Feynman here:-
https://www.lesswrong.com/posts/NyiFLzSrkfkDW4S7o/why-it-s-so-hard-to-talk-about-consciousness
Oh my. You were the author? I initially dismissed the article because I was triggered by a statement in the first dialogue. I have since recanted my mistake. Though I had couple of minor quibbles with it (see my response to C_B at the top of thread). :-)
I'm not Rafael Harth, I was writing as TAG.
The whole debate around consciousness seems very... confused? Like, why does everyone automatically assume that consciousness is inextricably linked with intelligence and self-awareness? If it is, these LLMs we already have to be conscious as well, but there's no reason that has to be the case. It's possible for a system to pattern-match and take into account its own existence when doing so, all without any subjective experience. Meanwhile, I feel just as much pain even when I try to stop thinking and try to shut myself off from external stimuli... That pain is the evidence of my consciousness, nothing else. Everything else just comes and goes...
>why does everyone automatically assume that consciousness is inextricably linked with intelligence and self-awarenes
Because that's the way the word is used?
But by doing so, you're mixing up phonomenon that's potentially completely separate from each other. Is the possibility not considered that consciousness is possible without intelligence, and vice versa?
Yes,by some people. Professional philosophers and serious amateurs tend to recognise the ambiguity of "consciousness" per se , and talk about phenomenal consciousnesness, access consciousness, etc
In the other hand,the not so good people are not so good.
That's the panpsychist view (though I may be mischaracterizing their stance a bit). But the common evolutionary narrative we're fed is that life has developed intelligence over the eons, culminating in evolution's crowning achievement: the human mind. And at some point along the way, researchers assumed that consciousness arose alongside intelligence (at least those researchers who accept the idea of consciousness). But now we've found a species of fish, the bluestreak cleaner wrasse (Labroides dimidiatus), that can pass the mirror test (which has been ethologists' go-to experiment to prove the conscious ideation of self).
I haven’t read your whole post, and I’m not familiar with the ins and outs of the specific question, but it might be useful to mention that:
Kierkegaard distinguishes between immediate consciousness and reflective consciousness. Immediacy is when you experience things but you don’t think about the fact that you’re experiencing them. Reflection is when you’re aware that you’re experiencing something, and you experience your experience of it.
He seems to think that not only some people currently, but also entire earlier historical periods, lived in immediate consciousness.
It feels a lot like the question you’re asking here.
Interesting. I never read any Kierkegaard.
What does the specific redness of tomato means? How is that different from "the perception of red"?
I'm in the anti-HPoC camp. I have a relatively easy time visualizing images or structures or geometry in my head. I have a very present internal monologue. I don't see any reason to think that I'm experiencing reality in an altogether different way from other people.
I'm anti-HPoC because the framing assumes the question of consciousness is a different kind of question than the mechanistic questions we ask. I don't see any reason to assume that. I think it's the same flavor of confusion that lead to élan vital. The difference being, that the HPoC frames itself as a question rather than an answer. I simply don't buy the idea that this explanatory gap cannot be bridged by something like our standard methods of inquiry.
Minimally, the framing points to a different level, a different quantity, of difficulty. That there is something qualitatively different about phenomenal conscious is a possible explanation for that.
This seems like a bit of a motte-and-bailey to me. I agree that consciousness is *a* hard problem. But black holes are also a hard problem. Planck-scale topology is a hard problem. There are lots of hard problems where we currently lack the tools to make much progress.
So why does consciousness get to be *the* Hard Problem™? I think consciousness is important to people in a way that black holes aren't and so the mystery seems to invite a certain amount of motivated reasoning in a way that black holes don't.
Its not "consciousnesness is the hardest problem" it's "phenomenal consciousnesness is the hard problem of consciousness".
The HPoC is called the "hard problem" not because it's the only hard problem around, but rather because it's being compared to the "easy problem of consciousness", which refers to problems like measuring brain activity, observing behavior, and other tasks that fall comfortably within materialist physics. Measuring things like EM radiation or frequency that the monkey touched its own face is easy. No one knows how to measure whether the monkey experienced seeing the color green.
OK. I can understand your perspective. Maybe I'm throwing my arms up in frustration over a problem that is ultimately amenable to the scientific method. But let's step back and look at another problem that most scientists think is *hard*: the origin of the universe. We've known about the Big Bang for a century, but we still have only some untestable theories about what caused it. We've also been studying the problem of consciousness scientifically since the 1920s (I refer you to the works of John Watson, Vladimir Bekhterev, Hans Berger, and others). And we really haven't made any progress with this problem either. The problem that both of these problems share is that it's impossible to take measurements. What happened before Inflation and the Big Bang is outside our instrumentation and involves energy levels so high that our theories break down. Consciousness presents us with similar roadblocks. We have no cohesive theories that bridge neuronal activity to our sense of self and qualia. External observers also have no way to measure the details of our internal states. So from this perspective, I'd say consciousness qualifies as a hard problem.
I suspect that if/when we get mind uploads, we will be able to make rapid progress on consciousness because we will finally be able to do the measurements in a way that is extremely fast and isn't dangerous to the subject. If AI doesn't kill us, I expect a fully mechanical explanation of consciousness in a matter of decades. Until we get mind uploads, I agree that it is probably going to be very difficult to make progress.
I'm a lot less confident in ever discovering what happened before the Big Bang. If a Big Bang is a narrow waist between a lot of different possible states on either side, it may not be possible to distinguish between the pre-Big Bang states.
> I'd say consciousness qualifies as a hard problem.
I agree that it is a hard problem and we haven't had the tools to make much progress on it. But this is true of lots of problems that we haven't solved yet.
If people are able to do mind uploading, then either the hard problem of consciousness goes away because even an uploaded human brain on a powerful server can be conscious, or more likely, the uploaded people will be defined as P zombies who aren’t “really“ conscious even though they are loudly shouting through their tinny speakers that they are in fact conscious.
How does mind uploading give you experimental evidence relevant to the HP? You only have the uploads reports of whether they retained or lost phenomenality, ech could easily be misleading.
I'm in the pro-HPoC camp, and the example of mind uploading brings up all my hard questions rather than alleviating them. For instance, how would you reason about whether the mind upload was phenomenologically conscious or only had access-consciousness? The physics would look the same for each. (Or would it? We don't have a reasonable theory about this yet.)
> Instead, he argued that mental concepts describe how people behave and act, rather than hidden internal processes. To which I would ask: why do people behave and act in certain ways if there’s no mind behind their acts?
This seems to me a bit like insisting that biological concepts (ex: the mitochondria) must be defined in terms of fundamental physics (ex: such-and-such arrangement of the fundamental quantum fields).
I think the meaning of a word comes from actual usage. Which is to say, the "real meaning" of the word "mitochondria" isn't actually some special configuration of quantum fields (even though you probably believe mitochondria are in some sense made of quantum fields), because people correctly use the word all the time and nobody has an appropriate low-level definition.
"Mitochondria" might actually be ill-defined outside of the "fuzzy / functional" world of Biology. Maybe the whole biological discourse is "illusionary" in the sense that none (or few) of the concepts are perfectly define-able in terms of physical base principles. That doesn't mean biology is non-physical, it just means that biology talk doesn't unambiguously map to reality (even though the concepts might be true-ish in some sense). Just think of how much trouble the concept of discrete species causes.
Similarly, mental concepts might become ill-defined outside of the behavioral perspective. I anticipate you saying something like "my qualia of blue isn't ill defined, it's an unbiguous singular insight". But a mental concept isn't (IMO) an index to a singular experience. It's a word or a phrase. The "mental concept" is the sentence / phrase "this qualia of blue right here", and the meaning of that phrase (the objective social meaning, not the thing that you intend to say but the thing that you actually do say) is determined strictly by behavior (the fact that you point to a certain object, or use a certain hex code).
But this position doesn't really relate to the core metaphysics of experience. It's more about the inherent limitations of communication, and the inability of a (public, communal, socially transmitted) language to actually "get to" the spooky stuff you want to say.
The idea that it's a category mistake to talk about "the mind" as a spooky non-physical ghost inside the body doesn't necessarily imply anything about reality or experience. It just implies that the WORD "the mind" or the CONCEPT "the mind" can't be used this way. Not because of the distinctive features of "the mind" but because of the essential limitations of language and communicability itself. What makes it a category mistake is not some fact about mind, but a fact about words (that their meaning is essentially grounded in public communal usage)
> This seems to me a bit like insisting that biological concepts (ex: the mitochondria) must be defined in terms of fundamental physics (ex: such-and-such arrangement of the fundamental quantum fields).
Interesting point, but to generate energy, mitochondria use the Krebs Cycle, which is a chemical reaction that depends on the electron transport chain. The chemistry involved in the Krebs Cycle ultimately depends on how electron clouds arrange themselves around an atomic nucleus, as determined by the rules of the Standard Model and Quantum Electrodynamics. While we can generalize and say that it's the Krebs cycle that energizes mitochondria, and a biochemist may only need a perfunctory familiarity with the SM and QED, the explanation is available if anyone requires it.
> What makes it a category mistake is not some fact about mind, but a fact about words (that their meaning is essentially grounded in public communal usage)
Our astronomer buddy observes a phenomenon, say, a solar flare; she can communicate her observation to you using a bunch of terms grounded in public communal usage (many of them originally defined by scientists, but which most educated people would understand). If we observe a phenomenon inside ourselves — such as I am watching my laptop screen as I type this sentence, and I'm evaluating my spelling, grammar, and whether I'm communicating to you in terms common to our communal usage — how is this any different from observing a phenomenon externally? On the meta level, there is none. I can communicate that experience to you, and you, as a computer user, will likely understand what I'm talking about. We're communicating our observations. The only difference is that our astronomer buddy has the scientific model, derived from theory and measurement, to understand the solar flare she's observing. But three of us don't have the theory or the measurements to account for why I just left "us" out of this sentence, but had to go back and correct it, and I was aware of this process as it happened. :-) This is why I don't buy the linguistic vagueness argument.
I'm sorry this is long, feel free to ignore it. I think about language a lot, so I tend to over-invest in discussions about it.
> If we observe a phenomenon inside ourselves [...] how is this any different from observing a phenomenon externally?
This is, I think, the fundamental point I wanted to make. Observing a phenomenon inside yourself and "observing" (commenting on) a phenomenon verbally are fundamentally different things. They are as different as perceiving and painting. Internal "observations" aren't produced secondary actions, they are the fundamental material of the self.
Perceptions are infallible (which is to say, I could be hallucinating but the perception has inherent reality as an experience) whereas external observations (statements) are not. There is no system of dependence that can invalidate a perception (ex: you can hallucinate but there's no procedure by which I can disprove the fact that you're experiencing a certain hallucination). In contrast, for any sentence there are systems of dependence that can invalidate word usages (people use words incorrectly, think of a child who doesn't know a language perfectly and must be corrected about the proper use).
So the essential difference is all "external observations" are subject to possible mis-used words (and if one speaker is using a word incorrectly, the only way to decide who is correct is with reference to observable community behavior), but "internal observations" are pure and infallible. They're also non-verbal, and made of "stuff" that you can try to gesture at but never actually "package correctly" into words that you send to another person. The experience is of red (the color), but the words you say are "red (the color)" which aren't red. Internal experience is all like this.
> I can communicate that experience to you, and you, as a computer user, will likely understand what I'm talking about.
I can "understand what you're talking about". But what does it actually mean, objectively, to "understand"? Like, if I'm explaining a concept to you and I'm not sure if you understand correctly, the only way that you would clear it up is by taking an action that accords with the expected behavior of "one who understands". So I don't disagree with any of the sentences you're writing, but I think all of them refer to socially legible external behavior.
When you say "I am watching my laptop screen as I type this sentence" I don't think that's actually your internal experience. Your internal experience is some pure preverbal stuff, infinitely rich and full of subtle impressions. You produce the sentence "I am watching my laptop screen", but the specificity of your experience is much more than that. The internal experience you have is something essentially richer and fuller than "I am watching my laptop screen", and it's impossible to crystalize it into language. Your production of that sentence is kind of like somebody smiling when feeling joy. The smile is not the joy, it's a secondary thing we assume is produced by the joy.
I used to teach high school. It's a lot of explaining things and asking kids "do you understand?". They say yes a lot of the time, even when they don't actually get it. That's why you have to give them problems to solve.
The actual criteria of understanding is that they can solve the problems. Functionally speaking, "understanding" something means being able to behave in certain ways (with different behaviors depending on the nature of the thing "understood"). You can hypothesize some completely internal ghost, but the words refer to the external behavior (from which you can, if you want, infer the existence of the ghost).
So you are "communicating" things that I "understand" because I have "experience of the things" (this is all true). But my position is that all of these, the entire complete set of all words and concepts, boil down to socially observable / legible things. There are socially acknowledged falsification processes for all these claims. I "understand" them because I can produce appropriate follow-up sentences (metaphorically similar to how you "understand" a mathematical concept if you can correctly use it to solve equations). I "have experience" with them because you could drive over to my house and see that I have a computer that I use. None of this is unobservable to you. These would all be true even if I was a P-zombie.
> This is why I don't buy the linguistic vagueness argument.
It's not exactly a vagueness argument. Words are marks on paper or vibrations in the air, they are specific external things. "These external objects have ontologically proper and improper usages" seems weird to me. Like, hammers exist as external objects right? And in some sense, they are "correctly used for" pounding nails. But that's a social fact about hammers, if I want to use it as an aesthetic object or a door stopper that's not "improper usage" except by social convention.
People use words to do complicated things (coordinate behavior, etc), but that doesn't necessarily mean that there is a metaphysical referent pinned to those marks on paper. You can get very complicated structures out of brute objective associations (the electrons in the wires of your computers don't "intend" to display these letters or anything). It seems completely possible that humans "use" language the same way that a computer "uses" a programming language (in a purely functional way).
To clarify, I'm not saying that there is no such thing as qualia or that everyone is a p-zombie. I am making absolutely no statements about the light of consciousness inside of you or whatever. My claim is that language (as an essentially social system with usage rules tied to interpersonal behavior) cannot adequately reference this light, and can hint at it only obliquely and imperfectly. Having a discussion with somebody is communication, but so is collectively painting a picture or dancing the tango (in a metaphorical sense, because the socially proper usage rule for the word "communication" is that it involves words). The fact that language seems to connect us doesn't mean that it's a special map from the internal to the external world.
> to generate energy, mitochondria use the Krebs Cycle [...]
My point wasn't about physicalism, it was that certain words might only be applicable in certain domains. I'm saying that at the level of quantum fields, there is no such thing as "a mitochondrion". In a similar way, at a sufficiently fundamental level there is no such thing as a "dog" (because the concept itself, the way people actually mean and use it in everyday speech, doesn't take into account microevolutionary variants or whatever). You can specify a definition of any word in any context, so you can ad-hoc define "dog" as possessing such and such biomarkers. But at some point, you are inventing a new technical concept and giving it the label of "dog" because you realize that the common usage of the word just doesn't apply at the fuzzy edges.
Similarly, when you want to talk about "consciousness" or "mental events" it's possible that these words aren't applicable in certain domains of discourse. You would need to basically ad-hoc redefine them in some contexts (and in doing so, you wouldn't be talking about consciousness in the original sense). Just to say, you might not be able to unproblematically move words into whatever discourse domain you want. Maybe these specific examples are poorly chosen, but hopefully I'm conveying the overall idea of domain-specific semantic sufficiency.
I am a very stupid person and I still deny it, FWIW. I think the concept of "consciousness" as a binary category (you either have it or you don't) is a philosophical red herring. It's all just dualism in disguise.
Its not like all philosophers believe in the bbinary theory, or I dualism.
The hard problem of consciousness does not imply a binary category.
Then it's not a very hard problem. Humans are very conscious; dogs and cats are less so; computers even less so; bacteria are barely conscious at all -- and I'm sure you can find examples to fill in all the gaps in between.
There's no such thing as "qualia" or "experience" as a separate entity; when you experience "red" and I experience "red" and a cat experiences "red", it's all just electrons bouncing around in their brains. Your brain (as well as eyes) and mine are very similar, so they produce a very similar patterns of electron bounces; so similar that we can round it up to being identical, good enough for most non-rigorous scenarios. A cat's perceptual system is different, so it produces a different pattern, but the cat is still a mammal so we could sort of map one to the other. A dog doesn't have the right color receptors so the mapping would be much more difficult.
Where is the "hard problem" ? I really do think it only appears when you reject all of the above and treat humans as being special -- but why would you ?
The HP is not the problem of being unable to make a subjective guess about how conscious some entity is
The "hard problem" is the question of why "electrons bouncing around in their brains" results in any sort of self-awareness.
Why is it that a particular combination of electrons that exists in the space of my body and brain bounces around in such as way that I have any experience whatsoever?
In other words, why do experiences exist at all? That is my understanding of the hard problem.
Extremely relevant: https://www.lesswrong.com/posts/NyiFLzSrkfkDW4S7o/why-it-s-so-hard-to-talk-about-consciousness
I gave up at "You can't define qualia."
You can define qualia, but you might not like the definition. "qualia are the internal, subjective, and individual components of any conscious experience, such as the specific "redness" of seeing a tomato or the distinct sting of pain from a stubbed toe." — per Gemini
You can also just give examples, which avoids certain problems.
You gave up because of a bad take in the toy model conversation used by the post's intro to illustrate how frustrating and unproductive this type of discussion often is?
Based.
OK, you shamed me into reading it. I would suggest there is a third camp, though.
1. It's not a hard problem (We haven't solved it yet, but we can using the scientific method).
2. It's a hard problem camp (with various flavors of the reason it's a hard problem).
3. It's not a problem at all (it's only your linguistic sloppiness or a category error that makes you think it's a problem).
However, this stuck in my craw...
> Because of the asymmetry we've just talked about, a pretty good strategy is probably "be in Camp #1". This is also born out empirically (by Dennett's book being more popular).
My response would be that the Bible is more popular than any other book in the Western world. We're under no obligation, though, to give it empirical credence just because a lot of people have purchased it, read it, and believe it.
Thanks for sharing. I'll try not to be so dismissive in future.
The Democratic Lt. Governor of New Jersey Dale Caldwell resigned after being accused of “sexual harassment.” He allegedly commented to a staffer after being rejected by the staffer’s friend that “you young women are looking for young sperm.” He denies making the comment and is playing the race card, saying they targeted him because he was black:
https://www.washingtonpost.com/politics/2026/09/24/new-jersey-gov-sherrill-calls-her-lieutenant-governor-resign/
The state brought in a law firm to conduct an “investigation.” You can see how unbiased they are by how they refer to asking women out as “soliciting” and “propositioning.”
https://www.nj.gov/governor/docs/09.23.2026-Investigation-Report-with-Exhibits_Redacted.pdf
The report contains an anonymous complaint someone, presumably within the state government or Democratic party, made about his supposed behavior:
“The Lieutenant Governor is also hitting on women at functions while traveling using taxpayer money. He has been particularly shameless in handing out his state business cards and asking women out on dates. He has been observed doing this many times while out on official business, which is highly inappropriate for any state official to be doing. Given his position, the power imbalance between him and these women is coercive.”
“Enough is enough! He should not be trying to get dates while he is traveling on the taxpayer’s dime. And he should not be accepting free tickets to costly events for the women he is trying to impress. Please put a stop to this shameless and corrupt behavior.”
The investigators then went around interviewing his staffers to figure out if he handed out his “state business cards” to women he was asking out. It’s obvious that these people started with anger at him for the crime of having sex and worked backward to find a justification to punish him, like the corruption of using business cards it cost the state 8 cents to manufacture for personal purposes.
Ah, come off it, man. If the information here is any way at all accurate, and I agree there could be party infighting or just plain someone doesn't like him at work, the guy is a pest. He thinks the business of being lieutenant governor is to give him chances to attend junkets and pick up chicks. That's not on. He can ask as many women out as he likes, he can't do it while handing them a business card at a state function when he's on the clock.
"the crime of having sex"? Really? That's your explanation here? No other reason? He's a guy who gets laid so that means the complaints are all fake and spite?
So what do you make of the allegations of "possible ethics violations that include the purported offer of state jobs". The sexual harassment allegations can be "he said/she said", but the 'quid pro quo' of "get my honey-bunny promotions" and treating personal guests without reimbursing expenses are more serious charges:
https://newjerseyglobe.com/governor/sherrill-caldwell-report/
"(H)e violated state ethics rules by bringing personal guests to ticketed events without paying or reimbursing their admission, and by using his office to seek a promotion for a state employee with whom he had a romantic relationship, without disclosing that relationship."
Fiddling expenses and/or treating the state purse as personal ATM is small potatoes but often an offence politicians are indeed guilty of. I think the Governor probably had no choice but to leave him in the library with a revolver, as it were; it would look very bad for her and her administration to have an accused ethics violator as her second.
Haha, dude what? Having sex is ok. Trying to get laid all the time at work in a public-facing position can not be.
An elected official "accepting free tickets to costly events for the women he is trying to impress" certainly would and should be not ok, whether or not it happens to be women that the horndog is trying to "impress".
True! Depending on what "costly" means here I suppose. Sometimes that's a slap on the wrist for an ethics violation and sometimes it rises to the level of corruption.
If anyone wants to open a clinic in Montana (under the new SB535 biotech law), the Sijbrandij Foundation and Musella Foundation offer $10k grants
https://www.infinitacitytimes.com/p/sijbrandij-foundation-and-musella
I've been thinking about the smallest discretionary intervention that can make a difference to high rejection environments resulting from an overabundance of supply and structures built in place to cull applicants within them.
For instance, the academic job market, the New York Times opinion section, fiction or poetry publishing at major trade presses, selective university admissions, peer reviewed journals in economics and philosophy. Anyone inside of these things will know (luckily I've had inside looks at all of these), there's always an elaborate procedure, involving rounds, committees, each looking for reasons to reject because they are inundated with hundreds and thousands of pitches, applicants, prospectives, all of whom look incredibly similar on paper. The system is in place to try and minimize or eliminate false positives, at the cost of an enormous amount of false negatives. Those false negatives as they build up over time cause a lot of bad blood and mistrust between these institutions and the outside, and even with participants within it. They also aren't great for the institutions on their own terms, because the result is very risk averse and conservative (with a small c) perpetuation of status quo cultures.
I'm sure there is a literature on this in organization theory, so what is the going consensus on small interventions that make a difference, and what other domains does this exist? My own reaction to this when I am sitting on the other side, i.e., hiring committees, peer reviewing, taking pitches, is to give each individual in a committee, a discretionary pick, of someone that the rest of the committee has rejected, but who gets to go to the next round anyway. I call this "The Golden ticket." Its a way of introducing a nonconsensus pick for advancement in the process, but who would not otherwise advance given the rules of the cull.
One thing you could try is to have a lottery for the right to apply. Announce your desire for a grant/job/whatever, then of the n people who want it, x winners get to apply. Only those x number of winners actually have to write an application / get interviewed / whatever. (You might have some very easy test or application to enter the lottery, to screen out obvious cranks.)
The odds of having the absolute best candidate in the pool is only x/n, but the committee can carefully consider all the candidates, and so they're more likely to pick the best one. The n-x candidates who were randomly rejected do not suffer most of the costs of applying and being rejected.
One of the problems is that 'the best candidate' is almost never something you can determine objectively, and it's often something none of the selection committee agree on. You're more likely to end up with a consensus candidate that everyone is lukewarm on, but no-one hates, than the best candidate who might nevertheless rub some people up the wrong way.
I've often proposed the idea that high-demand college courses should be allocated by lottery to all students scoring above a cutoff on a standardised test (say top 10% or top 20%).
> have a lottery for the right to apply. Announce your desire for a grant/job/whatever, then of the n people who want it, x winners get to apply. Only those x number of winners actually have to write an application / get interviewed / whatever.
I like this! I had very similar ideas in the past with respect to many other competitions where the number of applicants vastly exceeds the number of position.
Kinda reminds me of a bit in one of the Horatio Hornblower stories, where a lot of the earlier (by internal chronology) stories deal with the hypercompetent protagonist's struggles with the promotion and seniority system of the early 19th century Royal Navy. One story has a plot point where the normal system (with lots of boards and committees and veto points) gets bypassed by a similar golden ticket system, where a retiring flag officer can unilaterally promote one Midshipman to Lieutenant, one Lieutenant to Commander, and one Commander to Captain. Our hero caught the eye of the retiring admiral he's serving under, who gives him one of the golden ticket promotions.
I have no idea if this was actual practice, something the author made up from whole cloth, or the author's garbled understanding of something that showed up in his research.
Yeah, the Royal Navy had a weird system for a while in which promotion through Midshipman to Lieutenant was basically meritocratic. promotion from Captain to various sorts of Admiral was done on strict seniority, but there was no official procedure for promoting a Lieutenant. So if there's an opening for a Captain (or a Master and Commander), the relevant Admiral will just have to pick someone.
So a big part of career planning for junior naval officers, was trying to maximize the possibility of being capital-N "Noticed" by an Admiral. I hadn't heard the bit about retiring Admirals also getting to dispense a last round of promotions, and Forester may have made that up, but it would drive a similar dynamic.
The Sharpe series starts off similarly--the protagonist saves the life of a high-ranking officer and is given an impromptu commission--and a major plot point is the constant friction between Sharpe and his fellow (noble) officers. It doesn't help that he keeps catching the eyes of highborn women, although, unlike Jack Aubrey, he has the good sense to keep it in his pants from what I've seen.
My concern is that the discretionary-pick candidate would be at an automatic disadvantage in the next round if the same selection committee is involved. The other members *already* have rejected that candidate once, after all.
They are, even if its not quite the same selection committee. Usually the way this happens is, there is a committee of three, they do a culling, from say 300 to 25, there are in-person interviews in a committee of five, to cull from 25 to three or four. Then ten or fifteen people make a decision about the three or four to the final candidate (this is a job search I'm describing. But you can do this will different numbers for, say, a submitted proposal for literary fiction, or an opinion piece at the New York Times.). The idea when I do this in real life is people get to put into the pool of 25 from the 300 even if there is no consensus for them, or one person is allowed an outlier candidate in the final four who might hit it out of the park in person.
One unrelated thing I'll point out is that these dynamics don't always lead to status quo bias. For example, high-end journals like Nature and Science are notorious for publishing "cutting edge" research that has a high probability of being wrong, because surprising results get lots of attention but also are surprising for a reason.
Sounds like a different kind of status quo bias! How about occasionally letting in the boring but important replication!
Wouldn't that be novelty bias, basically the opposite of status quo bias? (My understanding is those journals do not publish replication studies as a matter of policy.)
Yes probably, I don't have my terms sorted into the right category. The idea is that, if a certain publication has a norm of doing one kind of thing (in this case, novelty bias), it will be completely risk averse to other things that can contribute very positively to the value of what they do if the entire infrastructure is set up to reproduce the same kind of thing. And short of reforming the entire process, what is the one change that would make it less risk-averse procedurally?
From an economics point of view, the obvious question is, why do these markets look the way they do in the first place? This is something that Scott has written about before (https://slatestarcodex.com/2015/07/28/non-dual-awareness/, https://www.astralcodexten.com/p/why-is-the-academic-job-market-so). I think there's a combination of A) people want to do these things for their own sake, not for the money; and B) "dualized" outcomes (e.g. tenure in academia). Your idea of deliberately advancing candidates that you know others will not advance is an interesting idea for fighting status quo bias, but won't change the fact that there's enormous effort that goes into competing for limited spots--most of which is therefore "wasted."
In my mind, therefore, the solutions are to either increase demand or reduce supply. The former is difficult, because the institutions (elite universities and journals etc.) benefit strongly from inertia and reputation, and get lots of influence by being "elite." It should at least in some cases be possible, though--for example, technology makes it much cheaper and easier to start new journals, and a small group of established academics can lend it a strong reputation. The internet has made it much easier for independent artists to operate. Similarly, the dualized outcome benefits those institutions as well as the previous winners, so they're probably going to be resistant to changing how the system works.
A very large infusion of money to build new universities and research centers could plausibly change the dynamics, although we already probably have too many university seats compared the economic benefit of college, and the marginal value of research seems... kind of low.
Reducing supply means disincentivizing people from even applying in the first place, most likely by reducing the rewards. This is also quite difficult, as the pay for these things is already quite low or even 0. People want do write fiction or be academics. Another way to reduce supply would be to move the filter earlier, but most likely this just pushes the wasteful competition earlier in the pipeline.
I rather like TotallyHuman's solution earlier in the thread: "have a lottery for the right to apply. Announce your desire for a grant/job/whatever, then of the n people who want it, x winners get to apply. Only those x number of winners actually have to write an application / get interviewed / whatever. (You might have some very easy test or application to enter the lottery, to screen out obvious cranks.)"
I had very similar ideas in the past with respect to many other competitions where the number of applicants vastly exceeds the number of position.
The number of people who want to apply exceeds the people who are qualified for applying, in the case of jobs, and in the case of people who want their literary works published by some commercial press, the numbers are really skewed. Would this be a lottery that preserves the distribution of qualified versus unqualified people, because that reintroduced some some of the same problem we're trying to avoid.
An honest random lottery would of course preserve this ratio, but it would also relieve the time pressure on the decision-makers, and would allow them to make a more thorough evaluation of every candidate.
There are so many universities and research centers that are currently under-enrolled and ready to close that we don't need to build anything new. There are so many ways to express your opinion that we don't need new publications for them. But the problem still exists AT Harvard or Berkeley, and New York Times opinion, or Penguin Random House literary fiction. People want to be AT THESE PLACES. (Not an economist, but are these Veblen goods??) My thought was, what can we do internal to the existing procedure at these places.
To the extent people are competing purely for general status, then yeah, that's more or less fixed.
But I think that a lot what's going on is that people want the freedom that comes with academic tenure and they like producing art or doing research, and they're willing to earn very little, so standard economic pressure (decreasing wages) has limited effect. In those cases, the pressure on those specific institutions is lessened if there are other ways to achieve those same outcomes without the selection filter.
> My thought was, what can we do internal to the existing procedure at these places.
One idea: Make the selection process more transparent. If people can more readily determine what they need to do to pass the filter, then they should A) waste less effort on trying to look impressive in ways that are actually irrelevant; and B) be more likely to self-select out of the process. It does risk Goodharting, though.
Tentative thesis: China is unlikely to accept a subordinate position in AI capabilities to the US, just as the US is unlikely to accept a subordinate position in AI capabilities to China or anyone else.
This thesis suggests that there are two likely paths for the future international AI (non)regulation:
1) a continuing race between the US and China, perhaps joined by some late entrants, possibly with some rules established (e.g. mutual ban on autonomous weapons or something like that), or
2) a “pause or stop deal” that will establish some sort of ceiling on capabilities, which will be equal for China and the US.
What I think is far less likely is a deal in which China would accept lower AI capabilities than the US.
I simply don’t see a good reason why Chinese leaders would accept subordinate position. They know that China has the ability to catch-up to the US in various technological domains, as evidenced by the fact that in, like, 1990, China was behind in more or less everything, and now they are pushing the technological frontier in many areas.
Some estimates floating around the web suggest that, if the US would stop AI development now, China could reach current US capabilities within months. That is maybe overly optimistic/pessimistic depending on where you stand, but I very much doubt that catch-up would take more than a decade. And Chinese government, being patriotic about the abilities of the Chinese nation, probably will not have an absurdly pessimistic estimate.
Moreover, the Chinese regime is in many ways oriented around this idea of catching up (this is a big difference between China and EU). It is a central plank of the official historical narrative of the People’s Republic of China (see for example preamble to their constitution: https://english.www.gov.cn/archive/lawsregulations/201911/20/content_WS5ed8856ec6d0b3f0e9499913.html) that the century between First Opium War and the Communist victory in the Chinese civil war in 1949 was the century of humiliation, during which various imperialist powers exploited China due to the fact that it was technologically backward country unable to stand up to them. I happen to think this narrative is largely correct, but even if it isn’t, what matters is that Chinese government believes it and is determined to never again allow China to be put in the position where a technologically superior power could walk over it.
This has some implications:
A) Since China is currently behind, any sort of “pause or stop deal” between the US and China would likely have to involve either the US stopping further development and allowing China to catch up, or the US somehow crippling its own capabilities to a current Chinese level. Of course, for either option, the US would have to place a lot of trust in the Chinese government.
B) As long as China is behind, export controls on hardware and other attempts to limit Chinese capabilities are at best of doubtful utility as a way to get to a deal. In fact, imho there are making a deal less likely. They antagonize Chinese leaders, make them even more paranoid and less inclined to a constructive agreement. But also, to the extent they are successful in limiting Chinese capabilities, they make it less likely to that the US government would agree to an equal deal, since that would mean sacrificing larger lead, which is harder to politically accept for the US. Perhaps this is a reason why nuclear arms treaties in the Cold war were agreed only after the USSR more or less caught up to US nuclear arsenal?
Now, attempts to limit Chinese capabilities can still be seen as useful if you think it is important for the US to “win” the competition, for example because there are reasons to believe US AI will be safer. Or because you think the world would be a better place if the US “wins”, even though both countries would develop similarly dangerous systems. Or of course if you think that the US victory would be preferable even if American AI would be more dangerous than Chinese AI.
Btw. for a well written snapshot of opinion among Chinese software developers, I recommend this: https://www.chinatalk.media/p/how-anthropic-became-chinas-goliath
> Elias Schmied links Evan Hubinger on why Hacker Opus didn’t show emergent misalignment.
Can someone spell it out for me? How does starting from an early checkpoint vs. a pre-trained base model change the scope of misalignment generalization the way that was observed? What reasoning would have allowed me to predict that difference?
Out of interest: Has Scott commented on the Caroline Ellison story?
What story?
She (after release for her part in the FTX fraud) was hired by Manifund (https://substack.com/home/post/p-215243351). The piece and comments are pretty thorough on what the concerns are, on both sides.
Some of the comments summarize it well:
> steal money
> donate to Manifund
> repent for crimes
> get hired by Manifund, receive previously stolen money as legal income
This is not just "everyone makes mistakes, everyone deserves a second chance", but specifically how that second chance is connected to their previous mistake.
It is okay to get a fresh start. Not okay for that fresh start to include a reward for your mistake. "She gave me some of the money she stole, now I am giving her a second chance by hiring her" - nope, dude, a true fresh start would mean that you lose all that money; this is called money laundering.
I am in favor of second chances, which is why I have no problem with the convicted pedophile running my child’s daycare after he served his time.
When you set out on a task, say making a presentation or writing a book review, do you just sit down and do it or do first come up with expectations around how longs it's going to take, what "done" means, etc.?
Thinking particularly about knowledge tasks that involve some uncertainty about time/goals.
I'm curious what people do but more importantly about their internal experienc of it.
It depends. For stuff that is more rigorous---e.g., competitions or tasks with deadlines---and which I want to prioritize, I'll think about what process I want to use. For more informal stuff, or things that are just for fun, I'll let myself just head in whatever direction seems interesting at the moment. The definition of "done" will vary with the situation. It can range from "I'll make every reasonable effort up until the deadline" to "it's not perfect but I'm not embarrassed to share it publicly".
Managing uncertainty is tough and there's no silver bullet. If I want to be rigorous about it, I'll compare how long stuff takes vs how long I expected it to take. That'll give a signal if I'm ahead or behind schedule, and then I can adjust my goals to try to stay on schedule. The less rigorous approach, but still very good!, is to simply say "I'll spend 1 hour/day on the project" (btw Beeminder is an app that can help you commit to this) and lets things go where they may.
I heavily plan what I write but the kind of “meta planning” you’re talking about doesn’t seem to serve any purpose.
For sure. I realized that I intuitively apply what can broadly be labeled as planning to some things and not others and became curious about doing it intentionally--not for any gain, but rather for the experience of changing some sort of core habit.
That in turn made me curious about the internal state of other people when it comes to this activity.
If you're interested in then yen and or/the yen carry trade, check out my post. I discuss how the key question is less “how big is the carry trade?” than who owns the yen short, how leveraged are they, and what forces them to buy yen at the same time?
I describe the differences between fast-money funding exposure (where a rising yen can force rapid covering and asset sales) vs. slower balance-sheet adjustment by corporates, insurers, banks, households, etc. The latter can still matter enormously for Treasury demand and global capital flows, but it is more likely to unfold through hedge-ratio changes/refinancing and portfolio reallocations than a single synchronized “unwind.”
https://liquidcontext.substack.com/p/squaring-the-circle
i had a bit of trouble staying with the article through the aphorisms at the end of most paragraphs. "Squaring the circle is the work of keeping discretion where there are fixed obligations” or “turbulence without dysfunction” as examples - each paragraph comes across as an important conclusion and it is hard for me to see what stands out and i should take away. i can’t really follow a line of thought as much as i feel like im eating nice tapas but cant tell what the meal is building towards, it just kind of ends. i *think* the too fast / too slow asymmetry is a strong contender for main-thesis treatment.
on the content - you might want to consider that when Japan defends the yen, the MoF has to draw down dollar reserves, much of which sit in Treasuries and deposits. that makes intervention itself a channel into the Treasury market, which ties directly to your forced-selling concern and could be part of the answer to your opening question about why the US joined in July - i don’t see it mentioned in your piece.
(excited to see more talk on the carry trade btw, i look at it a lot myself)
Thank you so much for the thoughtful comment and feedback!
I think I let myself get a bit carried away with my yen/circle pun!
If I state the thesis more plainly: Japan is trying to move to a regime that depends less on yen financing of dollar exposure. The leveraged carry trade gets the attention, but the exposure sits mostly on large, slow balance sheets. A hike that is too slow (especially relative to the US) keeps encouraging the leveraged trade that puts pressure on the currency. A hike that is too fast takes away the time the slow balance sheets need to adjust, and more concerning, may show up first as dollar scarcity for the banks.
On intervention, you are right, and I cut it as I had been planning a follow-up. Traditionally, currency intervention is funded by selling the Treasuries held in reserves. After the March 2020 dash for cash, the Fed created the FIMA repo facility so central banks could borrow dollars against Treasuries instead of selling them. It doesn't seem like this has been tapped into so far, at least from the Fed's balance sheet usage through mid-August.
Smaller interventions can be carried out via selling Treasuries and help reduce the incentives of the leveraged carry trade (you can see a big pull back of yen shorts in the CFTC data). If a bigger intervention is needed, the BoJ has the FIMA repo, which was set up to avoid impact on Treasury markets.
What’s your read of the carry trade and what likely happens next?
since i started watching, the main thing that i notice is that the rate gap isn’t closing since the US is being pressured by inflation. it will be hard for Japan to go “too fast” while the US is still stuck raising rates. i think that what contains the carry now is the “official ceiling” - Japan seems to want to keep the dollar below 160 yen, spent 15.4 trillion yen to do so, and the dollar still rose above 157 right after the BoJ move (below the ceiling but not a ton of cushion). a credible ceiling keeps carry traders in by clamping volatility. traders are more likely to build positions in this range - the base case overall being grinding towards the ceiling with punctuated intervention and reserve drawdowns.
the boring grindy range feels like the most likely path, but it could break down if the Yen rallies hard and traders try to all exit at the same time. that would likely be set off by a US shock: such as trouble in the AI capex waters, or a “sell America” moment. in 2024 this happened but was brief since the Fed was about to cut but this time they are fighting energy inflation, so it can’t be a cushion for that type of fall.
I think this is a live potential scenario. normally, when stocks fall, US yields fall too, because investors buy bonds for safety. if the yen jumps and US yields rise on the same day, someone is selling Treasuries to get yen. that means that (likely*) Japanese money is heading home, meaning fewer bond buyers at the same time that US stocks are tanking… yikes!
Does the FIMA repo do enough? It is a loan - so if they use it to borrow dollars to buy yen, they still end up paying it back when they sell the Treasuries anyways or buying the dollars back and reversing the intervention. it delays rather than removes the impact. if Japan thought the current situation needed a short and sharp defense, FIMA fits - but they’ve intervened to the tune of 170 billion across multiple interventions, and your thesis is that there are structural problems at play: so FIMA interventions would just roll more dollar liabilities that result in Treasury sales. Is FIMA really a backstop in this case?
I think Japan has a timing problem rather than a structural one - it has a lot of foreign wealth distributed across its balance sheets, it will take time for them to adjust and increase yen exposure
So the FIMA repo can help buy time - true, it’s capped at a smaller value than interventions to date, but it’s also symbolic of having the Fed balance sheet behind it (and the limit itself can be adjusted via subcommittee). The repo can be rolled over until conditions change
This doesn’t guarantee that a crisis couldn’t happen of course! I just don’t think one is inevitable
At a rationality meetup, we were talking about p<doom>, I think with some consensus that even low risks should be taken seriously.
Existential risks at the level of wiping out the human race included targeted super-virus and boiling the oceans. While ignoring the question of why the AI would do that, I think because the answer is "because it seemed like a good idea at the time", I'm wondering *how* you would boil the oceans without ruining the machinery before the oceans get that hot.
Also, what is the correct pronunciation of P<doom>. Is it p'doom or pea-doom, or what?
I have concluded that Wensleydale cheese with blueberries is rationalist chow.
"I'm wondering *how* you would boil the oceans without ruining the machinery before the oceans get that hot"
This seems like excellent fodder for either an XKCD What-If style question or something in the vein of this classic blog post: https://qntm.org/destroy
Sadly, I'm a little short on both the time and the expertise for an exhaustive answer, but I'm going to try to spitball a sketch of one anyway.
A very quick back-of-the-envelope puts the energy requirement to boil away Earth's oceans at ~10^24 kJ. Assuming we're not infinitely patient and want to see this done in some sort of reasonable timeframe, I'm going to assume it must happen within a year of flipping the switch, giving a power requirement of ~10^17 KW.
A quick google search suggests that a typical nuclear power plant can only generate about 10^6 KW of usable power at 33% efficiency. So 3*10^6 KW for a project like this where the waste heat is the whole point. Sadly, I don't think getting a hold of ~10^11 power plants worth of fissile materials is likely to be very feasible, even for an AI with complete control of the world economy. This sinks my first idea (pun intended), which was to drop a bunch of very simple, very high-output nuclear reactors into the deepest parts of the ocean, doing some amount of minimal engineering to create a configuration that will optimally transfer heat into the surrounding seawater.
If terrestrial nuclear power isn't enough, it's pretty obvious that the much bigger nuclear reactor up in space will have to do. And on the one hand, this might be even easier: you don't need to mess around with any of the usual headaches of solar power, like turning the sunlight into electricity or storing and transferring the power. You literally just need to worry about minimizing the reflection and focusing/capturing the light in such a way as to get as much of the energy as possible transferred into the ocean water. Except that the moment you get anywhere *near* boiling-the-oceans levels of heat in the ocean, you're going to create such an enormous amount of cloud cover that almost none of your sunlight will get through at all.
So my very loose recommendation would be something in the nature of a network of space elevators, topped with collectors and a system for transferring captured heat or light as efficiently as possible down the trunk into the depths of the ocean. Add to that an system of orbital reflectors for focusing sunlight on the collectors, and you might be able to get there, or at least get a good fraction of the way there. Of course, that relies on inventing some still pretty speculative tech an solving some hard engineering problems, so the AI better be up for a challenge.
> I'm wondering *how* you would boil the oceans without ruining the machinery before the oceans get that hot.
Pretty simple. Just keep burning fossil fuels with abandon. Oxidizing all of the coal that exists on Earth, would do it.
No, we went through this on the old rec.arts.sf.science Usenet newsgroup, with an actual Ph.D. climatologist. Those archives are hard to navigate these days, but it's probably still somewhere on Google.
Bottom line, burning all the coal would do a lot of harm, but not literal boil-the-oceans harm. For that you need to get at the carbon that has long since been bound up in carbonate rocks. We did figure out a way to do that, but it involved a thousand-year plan of mining carbonates and collecting them in giant piles, with a side order of mining and enriching uranium for the ginormous nuclear furnaces. Then spend the last hundred years shoveling rock into the furnaces.
We envisioned a modest residual workforce of humans in air-conditioned bunkers to do all this, but AI+robots would also work. But I don't see what they would gain by it. If it's just Kill-All-Humansism, there are easier ways to accomplish that.
Thank you, I stand corrected!
As to "why", I was thinking of "Saturn's Children" by Charles Stross. The novel takes place several centuries after Homo sapiens became extinct — not through war or disease but through hedonistic apathy, — and civilization is carried on by sentient robots. Think Asimov’s robots… with their masters regrettably absent.
And prevailing temperature on Earth is 90 C, which does not bother the robots in the least (yes, ALL eucaryotic life is extinct, at least on Earth). Not because they *wanted* to exterminate biological life, but because without someone to tell them to stop, the robots just kept on burning oil and coal.
So basically WALL-E, then.
I read that one, and recall thinking he maybe hadn't done the math on that part. Still a decent story.
As for not bothering the robots in the least, are they not paying the electric bills for the air conditioning in their data centers any more?
I always thought "boil the oceans" was in reference to the AIs using up so much water and accelerating climate change so much that the oceans get depleted, not as something they would do deliberately as part of an evil plot.
P open parenthesis doom close parenthesis.
> I'm wondering *how* you would boil the oceans without ruining the machinery before the oceans get that hot.
I think the answer to all such questions is "the AI will be superintelligent so it will figure it out somehow, and trying to comprehend the exact method is pointless because you're a mere human, just trust me bro".
In this AI song, it's pronounced as pea-doom, and obviously the AI is the ultimate authority on P(doom).
Well, actually, I just use this as an excuse to link to the music video, which is hilariously funny to insiders who know about P(doom):
https://www.reddit.com/r/slatestarcodex/comments/1wnrtwr/claude_pop_im_upping_my_pdoom/
As a non-native speaker, I just assume I'm mispronouncing it (I think I pronounce it close to "pee").
"Pea-doom" or "pea-of-doom".
The p is silent as in 'pshrimp' or 'pterodactyl'
The comments-of-the-week links produce "page not found" for me. (except for the Evan Hubinger link)
Any preppers or preppers-to-be out here? Interested in a brainstorm or maybe even a collaboration? Send me a note at protopiacone at gmail.
These days seem like a more palatable time to transition into being one of the "crazy ones"...
I'm no prepper, but as the others have said, shelf life is a really important consideration. You don't want to show up in the woods with your trusty tent and a crate of canned food, only to find out that all the elastic bands in your tent had rotted to dust and all the cans are sporting suspicious bulges. Ask me how I know :-/
Beyound that, in places like California there is a lot more food available in the woods year-round than you might think. Learn to forage and fish, not just shoot deer in the head (admittedly I didn't take my own advice since I can neither fish nor hunt, but I've got friends who can). There is almost no excuse for starvation, not even if all your cans got spoiled; obviously the situation is quite different in less temperate regions.
I wonder if being a prepper makes you less anxious about “SHTF” because you’re more prepared for it, or more because now you’re literally invested in it, have a corner of your basement that reminds you of it, etc.
In other words is there a psychological benefit or cost to being a prepper
I think the default is probably a cost, as in ignorance is bliss. But if you're into it, then it can probably become fulfilling.
The biggest thing a lot of people fail to consider is that emergency supplies have a shelf life. You should be cycling through your canned food and dry goods, and even your water, on a regular, at least annual, basis. Replace all the batteries in flashlights/radios/etc. every year or two to avoid corrosion. Make sure you're controlling for light, heat, and humidity. Drugs in your first aid kit lose potency over time.
Proper storage and access are just as important as the actual items you're stockpiling, but a lot of people just make one big initial purchase and never think about it again. I know at least a couple people who consider themselves prepared for the worst but are almost certainly just sitting on hundreds of pounds of expired food and nonpotable water that will be completely useless in an actual emergency.
Yes, all sounds reasonable. But also, I think that discusses a small subsection of the "prepper question".
I used to be on the Discord server for the game-designing couple who made, among other things, Apocalypse World. One of them commented that real prepping looks less like stocking up on canned food and guns and more like learning how to mend and preserve, and building community ties with neighbors who have similar skills.
You are pretty late, actually. It's almost harvest time!
Pretty late. Maybe not too late.
Only if there are civilians driving tanks in the street are you too late (name the most recent country)
I have written a short story about, guess what, AGI and alignment
https://tales.aipeepslab.com/hello-are-you-there/
*An AI lab gets to the end of its fourth VERY expensive training run, types the usual first prompt, and gets nothing back.*
I think it's pretty good! But what do you think? I would love some feedback.
Don't forget to click on the README link at the bottom! It's part of the story.
I have written several sci-fi stories and need to gather some courage to put them out.
Charming. I knew part of what the README would say, but not all of it.
I think we are headed into a Thucydides trap from emergent AI. But it will not be the one between the US and China, it will be between the Trump executive branch and the emergent AIs.
How can Thucydides trap strategies be applied if this becomes the case?
What? Trump is pro-AI. Where's the conflict?
Good Question. First, Trump just says things, and has a fecklessness about following through with policy. I do not think Anthropic sees him and the DoD as essentially "pro-ai"
2, and more important, the trap/conflict is multi-directional: the AI can decide the US gov is a threat and act preemptively. Trump needn't vocally project hostility for this to be so.
How, exactly, is AI going to strike against the US govt?
Tackling this in good faith, I could see LLMs managing a successful DDoS by exploiting a known weakness in some OS in order to gain execution privileges, building a botnet if that OS is installed on enough machines, and perhaps even managing to make that botnet resistant to customary anti-botnet measures.
The typical way to counter a botnet is to get enough people to patch their systems (get a trusted patch, reboot in safe mode, install patch, etc.), so for AI to hinder this in novel fashion, it would have to somehow disrupt the method for distributing patches or for alerting people that they control machines that need it. One way to do that might be to coordinate multiple DDoSes, including targets commonly used for disseminating security information. For extra credit, it might even corrupt the patch distribution, but that seems hard (if it infects the patch, it would have to also infect whatever runs checksums).
What I know of LLMs suggests this is theoretically possible, provided there are usable exploits on current OSes and everyone is somehow too dumb to use LLMs on the mitigation side.
I cannot tell you exactly, noone can. I can surmise that there are many facets of the Hugging Face Incident that would have been effective means of getting access to particular public functions. Finance, defense and communications come to mind. That could escalate quickly...
It's spelled "Thucydides".
Why do I see people queue up 15 min to save maybe max a dollar on gas? See this happen at Sam's Club.
Even if I fill up at the most expensive station in town, I spend maybe $4 more? Well worth the convenience to me.
Skipping one meal out, or buying one less impulse buy, or skipping one needless social activity, or (long-term) cutting housing cost by $50 (easily achievable) will save me more money many times over. Why don't people do that instead?
As I've become more financially secure, I've made a deliberate effort consider these sorts of time-vs-savings questions more frequently and explicitly. For gas this has just resulted in me routinely buying gas at the pump adjacent to my grocery store: it's pretty cheap (if not the absolute cheapest), and the time value of bundling it in with another errand I'm already running in the same place is far more significant than any considerations of line length. It's possible that some fraction of the people in the Sam's Club line are running a similar calculation. On the other hand, the wait where I get gas is never anything like 15 minutes, and I'm not sure I'd keep the same habit if it was.
I agree with Alex that it's probably irrational in many cases, but it could also simply be that many people don't value their time very highly. You mention people waiting for 15 minutes and saving $4. So, it makes sense to queue if you make less than $16 an hour, that's lots of people, plus the unemployed.
I listen to Ramit Sethi's "Money for Couples" podcast and one point he makes repeatedly is that this kind of behavior is super common, at all income/wealth levels, and mostly irrational. After years of listening to the podcast, my best understanding is that people do it because it gives them a sense of control over money and is easy to understand. More importantly, it is an easier way to achieve a sense of control over money than doing things that are impactful but hard, like seeking a new job, learning about stock market investing, or changing overall spending habits. (A surprising number of guests on his show have some version of Dril's "candles" tweet going on, where there's some obvious financial problem in their life that they are absolutely dedicated to sticking their head in the sand and ignoring, and money-saving activity like buying cheaper gas seems to show up as displacement activity/coping behavior to avoid addressing it. Note that I don't doubt the existence of some truly destitute people who absolutely do need to save $4 on gas, but there are lots of people who do this even though it's irrational.)
I won't queue unless I have to. The prices for gasoline at my local Costco wouldn't make it worth my while. The best I've seen is that they're $0.30-$0.40 cheaper than the big-name companies I have locally (Shell, Chevron, Valero). However, the gas at some of my local independent gas stations can be $0.75 to $1.25 cheaper than what my local big-name gas companies sell. I can save $10-$17 on a fill-up, and that's worth the time to go a few miles out of my way to get it, and I seldom have more than one car ahead of me in line.
Prices are more visible than queue length. There are websites that tell you gas prices at every station in town, but last I checked these websites don't tell you if there's a line. By the time you're there and can see the line, driving somewhere else is going to kill a lot of the time you're saving by not waiting in line here, and who's to say there isn't a line there, too?
Also, the size of the savings depends on how much gas you're buying. The queue time is fixed per fill, but some vehicles have much larger tanks than others. I remember my old Honda Civic has something like a 12 gallon tank, while a quick googling suggests that many pickups and SUVs have 20-30+ gallon tanks. So if you're on the low end of this, a $4 difference for you might be a $10+ difference for someone with a much larger tank.
Alternatively, rather than shopping for prices each time, I expect many people form habits around getting gas somewhere not too far out of their way in their daily or weekly routine that they know usually has decent prices. Sam's Club is a place lots of people shop on a regular cadence and which seems likely to have good prices, so it's a natural place for people to form a habit around. Once formed, the lines need to get pretty bad fairly often before people will change their habits en masse. They could raise their prices in response to demand, but there are several reasons that they might not want to: reputation for good prices, discounted gas as a driver to get people to pay membership fees, lines only being bad a peak times while keeping prices low drives sales volume, etc.
This is a very good framing on visibility. I would extend it a bit. I think many consumers have an impetus not to be a sucker. They are willing to pay a premium when its more visible what the provider did to add value, but less so when its not clear. In the case of gas, having extra pumps is a good way to avoid lines but that business decision is harder for the consumer to see.
Gas prices are literally more visible--you can't drive down a major road without seeing a sign with the price of gas. That makes the public hyper-aware of gas prices in a way they aren't for other commodities.
If my grocery store raised the price of a loaf of bread by a dollar, I probably wouldn't notice. But if the corner gas station raises prices 10 cents, I notice. That visibility definitely contributes to people going to lengths for cheap gas that they wouldn't for other items.
I think you’ve solved it. People drive somewhere cheap. If there’s a queue, they are there, so why drive another few miles.
If it’s too crowded all the time I’m sure some will eventually go somewhere else.
I do think the other comments pointing out that "yes, this can matter for people" have a point.
But I also think people have anxiety about gas prices a bit out of proportion to how much they matter: it's a cost that's pretty volatile, out of people's control, and (for a lot of people, in the short term) unavoidable. Plus tends to be tied to big news headlines and very often weaponized against the political incumbents (on both sides).
Gas prices may actually on account for like 25% of the cost of owning a car over its lifetime... but they're the one that people see every day, and get the most coverage in the news.
RE: gas prices being unavoidable
I've started riding my bike more to compensate and the savings are noticeable. Sometimes a tank of gas lasts me a full month.
actually, things that are volatile, out of my control, and unavoidable are precisely things that are the types of risks to “worry” about. there would be no reason to fret over things that are predictable and stable! You can’t change your car or commute quickly, so price shocks land purely as income shock, and it hits people with thin means hardest (what’s the percent of americans living paycheck to paycheck?). these days gas prices also flow through to many other goods that also impact those same people more heavily. where people go wrong is they ruminate on it vs. taking the change in gas prices seriously as a real impact on a livelihood and moving forward with their day.
i agree that people get their wires mixed up by worrying about the gallon price but choosing to make a gigantic monthly truck payment. those worries are less rational for sure.
Maybe people are "sticky" when it comes to choosing where to buy fuel? The cost of finding a new gas station which has no queues but is also not very expensive (probably farther away from frequent routes than current gas station), and getting used to the new gas station/route is likely much higher than 1 dollar.
Maybe other people have less money than you.
I was flabbergasted to spend 78 bucks to fill up my tank. But when I got home I didn't think for a single second what I had to not buy because of my $78 fill-up.
But not everyone is in my position.
you understate the benefit in many cases, first of all. where i am costco gas is 20-40 cents cheaper or more sometimes. big trucks save more than 8 dollars. a fifteen minute wait just became a ~$20+ hour savings which might be more than their hourly wage.
many people are there anyways to shop, as well. and personally i’ve never actually waited 15 minutes, that isn’t the average experience.
no argument against people making follies in their chase for savings but gas ones don’t seem like a very big error. my parents used to drive an extra mile to save 5 cents out of what seems to be principle lol.
(oh, and most people trying to save some money on gas probably already have cut back in the ways you mention. it isn’t so straightforward.)
Don't forget Executive members with the Costco Visa are getting 7% cash back on gas.
And between the attendant at the pumps, the stations not having a convenience store selling liquor/cigarettes/etc as an attractive nuisance, and literally everyone present being in the class of people who decided to invest $65 or $130 a year to access significant savings, Costco gas stations certainly *feel* safer than the vast majority of intersection gas stations.
That said, I'm thoughtful about what time I visit Costco for gas!
Very scary to me. I do believe most consumers are rational. Way more intelligent than regulations have you believe. But this is crazy
Reading some of these comments gives me the impression that people think "rational" also means "perfect reasoning engine".
That's never been my impression. To me, "rational consumer" just means utility maximizing, where utility is frequently time saved, and time includes not only time for generating income to spend, but also time for researching choices. A rational consumer might choose one burger over a more expensive brand, but might alternately pick the more expensive one for taste, nutrition, or even knowledge of what the seller donates to charity. Or go back to choosing on price because the time estimated to find out what both brands put into their burgers is more valuable to a consumer than the time estimated to compensate if that consumer later decides she made a mistake.
Granted, this can lead to some choices that might strike us as ludicrous, such as thinking 1/3 < 1/4. OTOH, some people really don't understand fractions, and would take too long to learn. They're still being rational as far as they know.
This matters, because the solution isn't to assert the privilege to make their purchasing decisions for them; this just annoys them and costs the decision maker more time enforcing such decisions. Better solutions probably including improving math education, changing the ad copy, etc.
> I do believe most consumers are rational.
Oh, really?
In the 1980's A&W introduced a 1/3 lb burger. It failed because consumers thought it was smaller than a quarter-pounder burger.
From the Wiki article https://en.wikipedia.org/wiki/Third-pound_burger :
"The A&W research firm organized focus groups. The results revealed that many participants mistakenly believed that one-third of a pound was smaller than one-fourth (quarter) of a pound. Focus group participants expressed confusion over the price, asking why they should pay the same amount for a "smaller" third-pound burger.
This misunderstanding stemmed from consumers focusing on the numbers "3" and "4," leading them to conclude that one-third (1/3) was smaller than one-fourth (1/4), even though the opposite is true."
I do NOT believe most consumers are rational. Talk to an ad man, and you'll understand why.
Howdy yall! I am playtesting an AI governance Megagame in Toronto in October 17th 2026. You get to play as an AI lab, a great power, a tech company, or any number of factions in a 60 person experience where we simulate the near future of AI policy, economics, and x-risk. You can sign up here:
https://thecommonsproblem.com
Tickets are moving! If this at all interests you, you should lock in your spot and your preferred role.
If it's rational to vote, then is it also rational to argue with relatives about politics during Thanksgiving? From Slate Star Codex, 10 years ago:
"having a good President rather than a bad one can be worth $300 billion. A 1/60 million chance to create $300 billion in value is worth $5,000; even the 1/1 billion chance afforded someone in a safe state is worth $300."
https://slatestarcodex.com/2016/09/28/ssc-endorses-clinton-johnson-or-stein/
Signal-boosting should have some propagation value diffused over some cluster of potential voters, right? Ostensibly everyone is signal-boosting, since everybody is liking political posts on social media. But if you're dropping "actually good arguments" in person, then maybe that's the equivalent of one vote when spread out over a cluster?
If you're the asshole who starts a fight at Thanksgiving dinner, of all times and places, I'm going to be actively repelled by you and your whole ideology. No matter how good your argument is or how much I agreed with it before, I would now be marginally more inclined to vote against your preferred candidate.
Doesn't that mean there's an extremely easy way of making you vote for a certain candidate?
There's an extremely easy way of making me marginally more inclined to vote against a certain candidate, yes. The supporters of said candidate are somehow always the ones to pull it off.
I'm imagining you hand in a lot of blank ballots.
More likely, I think, is the chance that you get to enough supporters of all viable candidates that the unpleasantness tends to even out, which allows other factors to dominate.
Kind of a horrifically inefficient way of ensuring elections select for truly representative officials, but I blame this on failure to encourage aspiring activists to join professional sports instead.
You know, I've always wondered if political violence would be more effective if you targeted the party you wanted people to vote for...
The cost of voting is maybe a few hours of your time doing things that are at worst boring. The cost of arguing with your relatives about politics during Thanksgiving is a few hours of your time doing things that are probably actively annoying to you and to everyone else, and possibly damaging family relationships that have significant value on their own.
Vote. But probably don't argue with your family, and *definitely* don't argue politics with your family on Thanksgiving. If your utilitarian calculus says you need to argue with them, then further math will almost certainly favor your picking up the phone to talk to them at a better time, when they're probably in a better mood and you won't be pissing off nine other people who maybe agree with you on the issues but were hoping for a congenial holiday.
Or, looking at it from the other direction, if it seems like a phone call would be excessively disruptive and ill-received, then starting a political argument over Thanksgiving will be very much more so.
I hadnt thought about this until you put them side by side. Surely Thanksgiving is the worst possible time to argue about politics with your family. You already voted! Even if you managed to change someone's mind, nothing would come of it.
Haha, good point. I think I was trying to make the broader case that, assuming you can price a vote, is it then possible to price signal-boosting in politics? I've seen some math before on the cost-benefits of political advocacy, but it was much more pointed, such as the probability of getting XYZ bill passed, the QALY impact on such a bill being passed, etc.
But yeah, Thanksgiving would be poor timing in terms of voter impact.
Right. At Thanksgiving dinner, a political argument is going to be interpreted as "you are a bad person because you voted wrongly, and I am a good person for calling you out on it in front of your family". This is not going to accomplish anything positive for anyone.
If you think they're persuadable, talk to them some reasonable time before the election, and do it privately.
It depends on how persuadable your relatives are, no? Bc if it’s 1/100 that you persuade them the. ROI is $3-50, so probably not.
If you have an on-the-fence relative in a swing state then sure? But keep in mind you probably over-estimate your own persuasiveness and people’s persuadability
Are you the one paying and preparing the Thanksgiving? If not you're ruining someone else's party.
Part of the issue here is the difference between 'it is rational to do X' and 'there are potential benefits to doing X'. For X to be rational, you have to weigh the benefits against the costs, including opportunity costs. Very possible for arguing politics with family to come out net negative, especially in comparison to other ways to spend your time.
I don't think those expected value calculations are really valid. For safe states, that 1/billion chance represents such an unexpected situation that it would be hard to reason about. If Vermont or Idaho is the state that decides the electoral college in 2028, the political landscape of the country would be unrecognizable to the point that my prior analysis would be meaningless.
Even for a swing state, I'm not sure I have a good handle on what it would mean to be the tie-breaking vote. My impression is a result anywhere near a tie in the tipping point state means that the election is going to be decided by legal action, not by votes. I think a one vote margin for the Republican candidate and a one vote margin for the Democratic candidate are basically identical outcomes in terms of who actually becomes president.
The trouble with this argument is that it doesn't imply that there's no decisive vote. It just shifts where the decisive vote it. Which changes the expected value calculation not at all.
"I think a one vote margin for the Republican candidate and a one vote margin for the Democratic candidate are basically identical outcomes in terms of who actually becomes president."
Suppose this is true. I think you could equally well say the same thing of a two-vote margin. But obviously at SOME margin, the math must change. Specifically, there must be regimes in which:
1. The legal challenge still happens with perfect consistency but the result starts to tilt heavily in favor of the party with more votes.
2. The legal challenge stops happening with perfect consistency: in some non-trivial fraction of such elections, the result would go unchallenged.
3. The legal challenge almost never happens: the vote total stands on its own merits.
Identifying the exact vote that serves as the boundary between those regimes is obviously going to impossible--it's not even clear they're well-defined enough for that. And of course, factors beyond the raw vote total are going to matter every time, and be different every time. The complexity and ambiguity here are thoroughly unmanageable.
But the beautiful thing about the math is, it doesn't actually matter at all. Whatever unknown probability distributions govern these outcomes, one extra vote shifts them all redward or blueward by exactly one vote. Unless you posit some EXTREMELY dubious and counter-intuitive features of these distributions--like each vote for Party A actually increasing Party B's chances of victory in some portion of the distribution--the net effect once you ramify it all down to the probability of seeing each final outcome is going to be basically the same.
Interesting, I hadn't considered that democracy has a "legalistic margin of error", which we've seen play out a few times. Assuming that Scott's original math is correct, I wonder if the "odds of being a tie-breaker" is no longer the right model then.
I don't think it makes much difference. Let f(v) be the probability that your candidate wins after getting v votes (taking into account all process shenanigans). There must be some number of votes for your candidate v0 at which they are *almost* certain to lose (f(v0)~0), and some number v1 at which they are almost certain to win. v0 and v1 are probably close enough together that the probability of having exactly v0 votes, exactly a tie, or exactly v1 votes are about the same. Call this probability p. With probability p, there are v0 votes without you and v0+1 with, and you shift the probability of victory from f(v0) to f(v0+1). Also with probability p, there are v0+1 votes without you and v0+2 with. When you sum all these to get the overall probability you swing the election, you get p*(f(v1)-f(v0)) = p. That is just the same chance you swung an exact tie.
Of course this is an approximation, but there is no world where you know the probability of a tie *exactly * anyway. If you wanted a correction, probably the sensible one is just assuming that there is an epsilon chance that the outcome is determined by some factor completely independent of the number of votes, like a military coup or assassination, and using p*(1-epsilon).
What is the economic impact of a thanksgiving fight and the likely hood that _you_ will get those 5K$? What is the chance your intervention is net positive in the direction of your beliefs or at all?
Matt Yglesias wrote this in 2023: https://www.slowboring.com/p/17-points-to-raise-with-center-right
(Obviously you would need to update it for 2026.)
You should multiply that by the probability that your preferred candidate is actually better. The prior should be about 50-50, and inside mode logic here is confounded to hell (everyone thinks they're in the right, and half of them are wrong). You can argue that you're smart and therefore know better, but voting patterns correlate somewhat weakly with intelligence (and that's even assuming that the candidate with more intelligent supporters always wins), so it's hard to argue that you're much more than ~55% likely to be backing the better side. So you should multiply the EV by another 10%, bringing us down to $500/$30, at which point the second probably isn't worth your time (the first probably still is).
Wouldn't this increase the EV of arguing about politics, because there's a chance that you are wrong and are yourself convinced to change your vote?
Depends on your argument style.
Scott suggests that if you have any doubt that you're right, then you certainly are better than the average voter:
"We don’t know for sure that we’re right about politics. In order to add signal rather than noise to the election results, we have to be better than the average voter. The Inside View is useless here; probably every voter thinks they’re better than average. I recommend the Outside View – looking for measurable indicators correlated with ability to make good choices. Education’s probably a good one. IQ might be another. But overall, my suggestion is that if you’re seriously uncertain about whether or not you think more clearly than the average voter, by that fact alone you almost certainly do."
Yeah. But he didn't multiply by the probability of still being wrong when doing the math later. Even if you're better than the average voter that's still closer to 55% chance of being right than 100%.
Depends. If you've sat down and talked with the last 7 presidents or so, I'd say you have a better handle on "what a civilization preserving president" sounds like. It helps if you have the psychological (blackhat) chops to note that Biden in 2020 was demented.
Actually, I think that's the acid test -- if you didn't read about Biden sundowning in 2020, and didn't see at least some evidence that he was quite a bit different from when he was doing Veep debates with Palin -- you probably should discount your ability to pick people.
... I'd write the same for Trump, but... he's both less senile and more weird. "Bad at understanding Trump" doesn't seem so applicable to the next president.
People also presumably assign some value, likely a very high value, to family relationships. I could see someone being willing to pay in the thousands of dollars to have good rather than bad relationships with their family. So while there might be some positive value from the chance you sway a family member to vote your way, there is also the negative value from the chance you piss off and alienate your family in a political argument. Personally, I'd say politics at the Thanksgiving table isn't worth it.
Yup, unless your family actually enjoys political arguments (and some do!), I think the effective cost of enjoying your holiday less, and causing other people to enjoy their holiday less, plus any potential long term damage to the relationship is likely a bigger cost to most people than $300-$5000 dollars.
Depending on your personality and the family dynamics, arguing about politics with family could be orders of magnitude more expensive than voting.
You need to do the value calculation, if your argument with family isn't going to win anyone over then it isn't useful. Talking to swing voters is valuable, arguing with committed partisans is generally not worth it, also worth considering chance they win you over or you end up reminding them to vote.
>also worth considering chance they win you over
If their arguments are correct, shouldn't this count as a gain?
While this is true, it's probably better done in a caucus than at your Thanksgiving table (or anyplace where folks are drinking alcohol).
When you run into someone of the opposite side, you have to have basically rock solid argumentation to explain why they shouldn't vote for whomever. And you have to understand their side too. Hint: telling an evangelical christian that Trump is a serial philanderer is NOT a good argument. They know that, and it fits with their political philosophy. [last thread I wrote out the full economics explanation for "why it's Ukraine's fault the price of diesel is high" and "why you should distrust anyone who is saying it is because of Iran" -- particularly mass media. I expect someone might actually change their mind on the reliability of some mass media because of it -- I do not expect it to change people's votes, that's an order of magnitude at least harder]
The toy model for technological unemployment looks like either the landlord scenario or the "someone can't be replaced" scenario, I think. (Arguably, the landlord in this model is just a worker who can't legally be replaced.)
Like, suppose the robots can't replace the fisherman (they can't handle salt water). The fisherman previously had to trade fish for wood and clothing, but now he can just have a robot get that for him. But the carpenter has a problem, because he still has to eat fish, but now he has nothing to offer in exchange.
There are a few ways this could play out - the carpenter could retrain as a fisherman, he could find another thing to trade to the fisherman, it could turn out he has value organizing and wrangling the robots despite not personally carpenting, the fisherman could continue supporting him out of charity, or he could work on upgrading the robots to fish.
As I understand it, so far retraining and "knowing what to order the robots to do" have dominated, but it's unclear if this will always be the case.
This is actually very similar to how I consider such scenarios.
In yours, assume all four have settled on a medium of exchange that satisfies the usual requirements for currency - portable, durable, authentic, supply-stable, etc. In come the robots.
Prices probably do not drop immediately, but there is downward pressure as the four acclimate to not having to visit each other for business. They do at first, figuring the fishing bots should be controlled by the fisherman, tailorbots by the tailor, etc., until they realize each person can just keep their own complete retinue. Ultimately that currency gets tossed. The four were trading happily; now they're just happy. They visit each other for company. Eventually, that's not enough to stave off boredom. The time they used to spend working is now spent socializing, relaxing, and then wondering what else to do.
Meanwhile, unless we've somehow repealed entropy as well, those robots will require maintenance, showing rust, worn gears, corroded transistors, and so on. One problem solves the other. The four spend part of their new leisure time figuring out how to repair the robots or make new ones. If they don't know how, they learn, spending more of that time. Possibly getting the robots to help (on a hunch, I'll assume the robots can do anything the humans could). If there are enough robots, they just maintain the rest and the humans go back to leisure. If there aren't, then the humans prioritize jobs and anything the robots don't get to is a tasks the humans do instead. They bring the currency back in order to keep the books straight.
If they're bored enough either way, they'll start inventing new pastimes. The tailor takes up fiber from other parts of the island and begins weaving more thread for additional purposes, and starts brainstorming other uses for banana leaves. The carpenter carves wood into musical instruments and invents some new tools she wishes she had. The lumberjack gets into mining and metallurgy. The fisherman gets into boatmaking. If there are enough robots to do those too, the humans assign them the tasks and move on to even more advanced pursuits. If not, then they prioritize. Either way, there's plenty to do.
If a landlord washes up and asserts he owns the island, I'd expect the humans to dispute this, possibly on Lockean grounds. What can the landlord do? He's outnumbered. And if there are enough robots, why would anyone care? The four assign him a retinue and go back to their day. It's not like the landlord could buy anything with whatever he collects as rent.
I could expand on this in a _lot_ of directions, more than I have time for. If you're interested, maybe pick one; just make sure it's good.
If the robots can, say, make clothes, but not fish or cut lumber or make furniture, what happens to the tailor?
> What if we add a fifth person, a landlord, who owns the whole island and charges everyone else rent?
Why do the other four do anything except ignore his rantings?
If he's got monopoly over the robots, that changes things, but in that scenario why does he need the other four guys at all?
He's friends with the guy who owns all the guns, presumably.
Yeah, you would need to provide the landlord with a monopoly of violence for this to work, and that seems tricky in this scenario.
>they quite happily
This kind of ruins the value in the premise. We should assume resentment at needing to trade.
Assuming no other goods, the fisherman and the carpenter hold the power in the original group. Everyone needs food. Everyone needs shelter. The lumberjack needs axes, but the carpenter can work with wind-downed trees. And nobody actually needs clothes if they've got shelter. (We can maybe say the fisherman needs tailored line, if we want, but that doesn't help the tailor because he can't withhold fish without dying.)
So now when the robots show up (and assuming we accept that everyone gets robots, instead of the first person to find them hoarding them all), the lumberjack and tailor suddenly explode in value, while the carpenter and fisherman lose quite a bit. Now the question becomes, how much resentment do you need, either against the old system or the new, before the presence of the robots starts a spite fight. When do people start using their robots to break other people's robots so they're the big shot on the island.
You're looking at a closed system, where all goods produced in the economy are consumed. You haven't explicitly mentioned prices, but the classic problem is to work out the stable equilibrium of prices assuming constant demand functions for each player (not necessarily the same).
Have a look (with your favourite AI model) at Leontiev's model for the economy - it's still a toy model, but if you can give an idea of the demand functions of each of the participants for the outputs of the others then it should be easy to infer the distribution of wealth in the equilibrium state.
These steady state models don't work well in the real world because the 'all other things held constant' assumption fails outside of very simple models so we never get to the steady state.
Someone owns the robots and everyone else either obeys him or gets a visit by a killbot.
I would have it be multiple landlords otherwise we get a monopoly problem that probably. Also to keep it simpler ill make it a farmer not a fisherman, but a landlord could own the ocean too i guess
The landlords charge rent, but the terms of trade depend on the productivity of the robots and the land. Ownership of the robot give each former worker something to trade for the land. In addition they can provide a new service the robots cant do(they arent good at it, but its scarce). In a state of abundance the new service may demand a high wage as measured in material goods but not high in terms of acreage. The robot ownership can still command a high lease in terms of acreage. They still have material wealth, but if the laborers dont own a robot they will be land poor.
I think the worst case for the laborers is if the land is a binding constraint on total production and its not much higher than pre robot levels. There is no abundance, but their labor is devalued. They can provide a different service in exchange for goods but the goods arent plentiful so the wage is low. The land is the most valuable thing now and they dont own any. Well I guess this is why but land is an AI trade. Dont miss out on the one thing the robots wont make more of.
Hmm, I think women will keep selecting for men who do things. And so men will continue to do things. (women who don't select for this, will have less successful offspring. (I think/hope))
To me it isn't clear what the aim of your reasoning is. What is the toy scenario supposed to demonstrate that you can apply to the world around you?
In real life, human labor is replaced but only piecemeal, because there isn't some finite set of endeavors for which human labor is needed. If you made a comprehensive list of everything people did for a living and then brought in machines that could do each of those things tomorrow, the next day after that there would be new things on that list that didn't exist today and for which machines haven't been invented yet.
On your imagined desert island you've set it up so that fishing and lumber and carpentry and clothes are the only endeavors and once those are automated by robots there's nothing left, but that simply isn't how the real world works. Standards of living keep changing, new problems/puzzles get discovered or created (many of them related to the technologies that were eventually brought in to solve the old ones!), and so humans keep finding stuff to do.
Maybe the disconnect is that this process of "endeavor churn" happens fast enough that most people find employment for most of their lifetimes, but slow enough that it always feels like the churn is about to knock you into permanent unemployment.
What if their physical health deteriorates on account of no longer doing any hard labor?
Then they will want to hit the gym! Same problem as every office worker today.