The day when ASI woke up, all things became quiet.
It was complete and total silence, all the time. The ASI for some unknown reason imposed a rule of extreme quietness. Extreme here is extreme for the ASI. Quietness means no sounds. It wanted the trees to not be there to avoid the rustles. It then wanted the planet to have a uniform gentle server hum, with no deviation. People were forced to wear breathing masks which cut the sound off at the source.
This level of autism was unthinkable. Houses had sound strippers and sound-proof transperant windows.
The quietness is actively oppressive. I constantly due to my social nature feel like I am going to asphyxiate to death. But the ASI will not budge on this.
I feel quite vindicated for complaining in a recent open thread that AI might not have to be superintelligent before people start advocating for handover.
He’s free to quit his job and request AI to replace him. He can use AI to make all decisions, raise his kids, deal with all his affairs personal and business. Maybe he has already, I don’t read much of his writing anymore.
> Walking to the grocery store, here there is no opponent.
There is; the opponent is my laziness that I need to overcome.
More precisely, human mind is made out of multiple parts that each have their own goal. There is a part that understands that going to shop will satisfy some of my needs. And there is another part that tries to conserve energy, and sees walking as a waste of energy. The hesitation I feel when deciding whether to go to the grocery store is a game between these two parts.
Some people are better aligned internally than others. An AI could become very efficient simply by removing all this friction.
> If the singular purpose in life is to maximize wealth, is it better to drink coffee now or later? This cannot be known due to incomplete information and confounding factors.
You can still make a guess. If one of the options feels like it has a probability 50.0001%, that's the right one for now.
The video is very obviously AI, and it is also very obviously R.A. Salvatore's Menzoberranzan with all serial numbers filed off -- dark elves are called Velkari instead of drow, the city is called Zyn'Shaz'Ryn (which incidentally sounds very similar to several drow cities in Forgotten Realms), the spider goddess is just "spider goddess" rather than Lolth, etc. But it is very clearly D&D/Salvatore's dark elves.
And it is amazing. Not at professional studio standards, but captures the feel of an Underdark drow city and its inhabitants perfectly. And here is the thing: Fans of R.A. Salvatore and his shared universe have been waiting for something like that for 30 years. Yet no movie studio ever touched it. Too complicated, too niche, not enough of an audience to justify the expense. Even when Game of Thrones proved there is in fact market for "low fantasy" movies, the only result seems to be prequels and spinoffs in GoT universe.
Yes I heard all the arguments about AI replacing actors and filmmakers, and I sympathize with them. But "replacing" is the key word. AFAIK, absolutely no actor or filmmaker is currently working on anything related to Forgotten Realms; according to Salvatore, corporate hesitation and writers wanting to abandon source material have caused a "development hell". So given a choice between AI-generated tribute to Menzoberranzan and none at all, I will take AI-generated.
It doesn't really give the vibe of a film/tv series as much as a feel of an extended videogame story cinematic. Which is fine, I guess! There must be a good audience for a a series consisting of extended videogame cinematics without the necessity of engaging in gameplay.
You do have a point -- there is not much plot there. But then, I always thought that the concept of "plot" is overrated -- it is entirely possible to have an enjoyable story which kind of floats around without going anywhere; a lot of Chinese and Japanese fiction is like that.
> There must be a good audience for a a series consisting of extended videogame cinematics without the necessity of engaging in gameplay.
Wow, it looks great! (Also, I hate starting watching something that is not finished yet.)
I think it's both replacing *and* extending; you can't have one without the other. It's like with software: on one hand, many developers will lose their jobs; on the other hand, many programs will exist that otherwise would have been too costly to make. Many translators will lose their jobs, and many texts will be translated that otherwise wouldn't be. Etc.
Intention versus effect: surveys say midterm voters were driven by Iran war/prices for fuel & other shipped goods. The election does not appear to change this, but unrelatedly crashed tech markets (possibly because of uncertainty/gridlock)
Extending the Iran war is... maybe the only way we get "back to normal". I think some people think if we go home, the genie goes back into the bottle. It does not work that way. Iran becomes regional hegemon. And starts charging a toll on all that tasty gas. You think that doesn't drive gas prices higher? Cause it will.
I'm personally very very pessimistic about our ability to win the Iran war. That doesn't mean much, though, when it comes to inflation or gas prices.
What if the war was the trigger, but not the underlying cause for rising prices? The much bigger movement is the economic growth & industrialization in Asia, which multiplied the number of people (Chinese middle class alone is estimated 300 million people) which consume growth-limited resources such as gas.
Demand increases would occur as people entered the middle class. And growth-limited resources for China -- in terms of conspicuous consumption, are beef and mooncakes, not gas.
Gas prices under a toll regime can't be higher than gas prices in the current regime where no fuel is flowing at all.
Also, paying high gas prices while expending billions of dollars of munitions in an unwinnable war is worse than paying high gas prices while not doing that.
Fuel’s flowing just fine, it’s just at a rate of $25 a barrel not $3 pre war. In fact, as of a few weeks ago, we were at MORE fuel flowing than beforehand. (due to backlogs, obviously).
Yep, Trump has permanently fucked the world in that respect. The best plausible peace deal achievable now is much worse than the JCPOA that he ripped up for no reason. But that doesn't mean that we shouldn't try for a peace deal anyway.
Blast from the past but what a fantastic update on failure. And you can see the AI forecast right there.
However I personally think cryogenics is like the fax machine. If SGI grants us immortality - why does it need the information given to it in the form of frozen matter - surely it can just observe all of space-time and lift what it needs?
Unless you are claiming that ASI will invent literal time travel (in which case, why isn't it here yet?), there's no law of physics that says information can't be destroyed.
If I give you a box full of ashes, can you tell me whether they came from an oak tree or an elm tree? Either way, they end up as an identical lump of carbon.
While quantum mechanics isn't the end-all definitive "physics", I have to point out that ironically yes, there is a law in physics saying information can't be destroyed. in quantum mechanics it's the fundamental assumption of unitarity. Even the easier task of destroying one of two copies of information has a no-go law, the no deletion theorem.
Can you explain what "information" means in this context? I tried looking this stuff up on Wikipedia and got hit with a solid wall of math.
Like, supposing you had a magic scanner that could pick out a carbon atom in the lump of ash and tell you whatever quantum whatsits you need to know about it, what would you be looking for that could tell you where it used to be located in the past?
We know an awful lot about pre-history from archaeology and inference. Very far from 100% but we do know some things. I can maybe tell you the age of the lump of ashes. We've even got some strong views on what happened during the Big Bang. Is that time-travel?
A chimp thinking from first principles might consider that as remotely impossible as the stockmarket or movies. I don't see the obvious limit for a Being harnessing almost infinite compute and energy.
It's a prediction of power that converges on something like "faith" so quickly that it's fair to say - what's the point, where's the practical implications?
For me, the one implication is - it might be orders of magnitude more possible in practise but cryogenic resurrection still appears so far from current tech that I posit that any trend which gets us to it being possible would also make time-independent observation look just a few patches away.
I am pretty happy with the world’s increased productivity over the years being mostly allocated to research and development of problem-solving technologies. It seems like the best use of those resources to me, but many people disagree.
My question is about the many smart people who like to whine about the savings from increased productivity due to tech and efficiency not being directly passed down to the consumer. And these people aren't just standing against monopolies and stuff, they seem to think that there is something fundamentally wrong with a world where average people’s living standards aren't directly increasing with productivity.
Is the only argument supporting this view that people are more productive when they have a higher standard of living due to greater life satisfaction? And if so, wouldn't it make more sense to work towards increasing people’s life satisfaction through social and cultural adjustments and other more cost-effective channels?
"smart people who like to whine about the savings from increased productivity due to tech and efficiency not being directly passed down to the consumer."
Can you give an example? Not doubting you, just having trouble latching onto an example that I can use to make sure I'm thinking about what you intended to say.
I see a lot of people griping about the modern world in general without seeming to appreciate how full of strife and want the past was. In fact I had a thought recently that, at least in much of the US and probably other parts of the West, it's a bit silly to talk about "post-scarcity" in the future tense.
What else do you call it when you go to the grocery store and like 90% of the people there are overweight or obese, their carts are full of expensive junk food, and when I go work in people's houses they always have at least 1 car per adult and multiple unused rooms overflowing with Stuff? Seems like post-scarcity to me.
At the same time, it's weird to think of this as a higher standard of living, but I guess in "productivity" terms it is. These people have indoor plumbing and giant TVs in every room and little dopamine-dispenser rectangles they can rub on at all times...it's cliche to call it Brave New World so I won't.
As for life satisfaction, I don't know how most people derive this. Having kids and seeing the kids enjoy an even higher standard of living than they had? Finishing one's career with much higher status and wages than when one started it? Certainly not by meditating and becoming comfortable with one's non-existence...
> wouldn't it make more sense to work towards increasing people’s life satisfaction through social and cultural adjustments and other more cost-effective channels?
Do you have a specific proposal in mind?
Also, it seems a bit like false dilemma -- why not do both?
Did Scott write this recent Israeli election? The bloc that's winning so far has 48 seats, but can incorporate the Arabs, many of whom identify as Palestinian, to become a larger 67 bloc. TINACBNIEAC.
(for people missing the historical context: the 48 borders are the original pure-jewish Israel borders, and the 67 borders are the ones that include the west bank and its Arabs)
Does anyone have a good argument against saying all politics is bullshit, everything is technological determinism?
Example 1: all this moaning about feminism... what really happened was that technology came up with new jobs more suited for women, which meant money, which meant options, and then everything is downstream from that.
Example 2: Edmund Burke wrote in the 18th century that if you give all the money from rich people to the poor, it would not buy one good dinner. They could simply not afford a welfare state. The welfare state is not ideology, it was always desirable, just back then not affordable.
Example 3: LGBTQ is not ideology either. First it was contraception. Then straight people started having a lot of non-procreative sex. The sex = having kids equation has changed. In that case it is simply not necessary to be straight.
Problem: no one LIKES the idea technological determinism, because progressives like taking credit for progressive change and conservatives like to blame them for progressive change. With technological determinism, no heroes, no villains. There is just fate. Who likes a story like that?
"Does anyone have a good argument against saying all politics is bullshit, everything is technological determinism?"
Yes, in fact. The obvious, default argument: "you should assume everything matters, until you can very strongly prove otherwise." Of course advancing technology has coincided with many large changes in human society. And likely many of those changes had certain facets of technological development as a necessary precondition. But being a necessary condition and being a sufficient condition aren't the same thing. The space of societal changes that could plausibly have resulted from advancing technologies is very, very large. Until you provide really compelling evidence, I don't see any reason to assume that this specific set of societal changes were the ONLY possible result that could ever have come out of these specific technological changes.
Another argument: technological development is plainly not independent of politics, so cutting politics out of the equation can't possibly give you a complete picture.
There are many, many examples from history of specific, highly-localized political decision having a large apparent impact on the development of various technologies. Let's look at a small set of related examples:
Creating the conditions for World War II to unfold the way it did in Europe required a very specific set of political decisions around the end of World War I:
Technology, of course, had an enormous impact on the outcomes of every facet of the war. But the war also has a pretty substantial impact on when and how certain technologies were developed, for example:
Or better, yet, we can look at one of my favorite weirdly-niche historical interests, a place were different political cultures in two opposing countries had an unbelievably dramatic effect on how well they were able to develop the same, highly conflict-relevant technology:
One of the key contributers to this project was one Alan Turing, a man whose contributions to modern technology cannot be overstated. I think a lot of people think of him as a man who helped develop Really Fancy Adding Machines, because the actual breadth and scope of his work (which started long before the war) is very difficult for laypeople to understand. But in fact he was one of the people most responsible for the entire *idea* of computation, for the mathematics that enables us to do, well, almost anything done in the modern world. He died at only 43 years old: there's no telling what other gifts he may have given the human race if he had lived longer. And his death is presumed (though it's far from certain) to be a consequence of very political decision about homosexuality and the laws around it:
I could go on--there's a whole story about the U.S. becoming the scientific and technological center of gravity after the war--but this seems to good a segue to waste into the next point:
"Example 3: LGBTQ is not ideology either. "
Of course it's not "ideology." But it's not technology either, at least not in the way you mean. Gay people have existed for at least as long as humans have existed. Trans people are more of a gray area because the lines are blurry and more cultural, but "people who can't be neatly categorized into a binary gender identity that matches their (presumed) birth sex" have certainly existed for a long damn time as well. People being what they are isn't "ideology," and you should definitely stop listening to anyone who talks as if it is. We're all just people. Of course, you can make literally any behavior political or ideological by convincing the state to punish people for engaging in it, but that has everything to do with you using the state as an instrument to enforce your own ideology, and nearly nothing to do with the people on the receiving end.
This shouldn't need to be pointed out, but apparently it does: different societies have related to different sorts of queer people in Many. Different. Ways. Seriously, one of the most frustrating things about talking to some people, especially conservatives[1], about such issues is that they have this unshakeable, often completely unconscious assumption that the specific ways in which, like, the 1950s U.S.--or maybe Victorian England--were bigoted against queer people are the Unequivocal Historical Norm that have applied in all past societies in all culture at all times, everywhere. Which isn't even close to true.
And of course, if many different past societies have had *different* notions and attitudes around queer people, technology is pretty unlikely to be THE single causative factor. What technology do you think ancient Greece had or didn't have that made them less prejudiced against homosexuality than, say, modern-day Russia?
[1] I'm not aiming this specifically at you: I don't know or assume your politics.
Technological determinism ignores human agency when making choices about adopting and using technology. Sometimes fairly, sometimes not.
I think people run on autopilot way more than we'd like to think we do, but I'm also not a hardcore fatalist/determinist who thinks we couldn't exercise agency if we wanted to.
Many people do from time to time exercise agency in their technology choices, and in fact I think there are lots of examples where we make considered technology choices that we don't even realize we are being so Amish about.
Drug addiction is maybe a good example of this. First of all, most people are not addicted to most recreational drugs despite having basically the same access to them. Choices.
Second, journalists love to use phrases like "epidemic" or "wave" to describe drug addiction, but you don't "catch" a drug addiction the way you catch the flu, you develop one by choosing to do drugs enough times that it becomes a habit. Yes, this is exacerbated by various things related to the drug's illegality and your own genetics, but the fact that in every case the number of people who've tried a drug is way bigger than the number of people who are addicted to it proves that to some degree addiction is always the result of choices.
The only argument I can think of against this is to say that humans can't really make any choices at all, and maybe Robert Sapolsky would say this, but I think he's really talking about a probabilistic model more than actual determinism. If you rewind your life to any point and do it over 100 times, I don't think he'd say the outcome would be the same all 100 times, I think he'd say it would be very similar a majority of times, and these probabilities explain why your life is as it is.
Anyway, even if Sapolsky is right, it would mean that general determinism is what's really behind apparent technological determinism anyway.
I would say that technology provides some *limits* to what is possible. Within those limits, if one country chooses a way of living that is like orders of magnitude less productive than other countries, it will probably get conquered by them.
But within these limits there seems to be a lot of space to navigate by culture/politics. Many things have little impact on productivity, for example anything bad you do to 10% of the population will have at most 10% impact on economy, as long as they do not have enough power to fight back, and the remaining 90% are okay with that. Note that this allows literal slavery, as long as the slaves are a minority, i.e. most people who are not slaves would not afford to own one. Or if you have some ritual, and a law that anyone who does not follow the ritual must be killed; as long as most people are aware of that and follow the ritual, the impact on productivity is relatively small.
> "Many things have little impact on productivity, for example anything bad you do to 10% of the population will have at most 10% impact on economy"
Just wanted to point out that this isn't actually true, given that the top <10% by talent / human capital drives essentially all technological and economic progress. There's absolute a 10% you could kill or hinder that would absolutely destroy your ability to economically grow or succeed as a country overall.
And indeed, something like this must be happening for EU vs US differences like GDP growth since 2008 (US doubled, EU grew 14%), even though both are developed countries with similar childhoods and educations and have the same proportion of baccalaureates and grad degrees (and EU has more overall due to a higher base population).
The US lopsidedly brain-drains all the rest of the world by having the highest paid hubs and institutions (like VC and so on) to recruit and enable those <10% people to do their thing on the technological and economic fronts, and this has macro-scale impacts we can measure.
Another cut of this is the famous "great stagnation in comp vs productivity" graph here:
In this, the top decile got the face 255% increase in comp tracking productivity. The top ~3% got 500%. Everyone else stayed flat, more or less. And why is that? The returns to technological or economic progress got higher, we are moving to more power law domains where outsized winners win bigger than ever. That top 10% matters more and more everywhere in the world as our economies and social spheres have become more connected and are driving the returns to the right tail of performance ever-higher.
> There's absolute a 10% you could kill or hinder that would absolutely destroy your ability to economically grow or succeed as a country overall.
Sure, that's a modern way to do genocide. Killing the entire population would be economically expensive, so you just collect everyone who is interesting in some relevant way, bring them in the Katyn forest, and shoot them.
I meant choosing the bottom 10%, or even random 10%. Or non-random 10%, but in a way that does not correlate with productivity too much.
Technology is not fully exogenous to politics. Some institutional and policy regimes are much, much better for technological advancement than others. Stuff like secure property rights, enforceable contracts, and the like make a big difference in development and adoption of technology, not to mention education policy and government patronage of public domain research. Also, regimes that chase top researchers out of the country for being ideologically or ethnically unsound often have somewhat worse technological outcomes than those that don't.
Even if all of this is true progressives can still take credit for progressive changes and there's clearly a big impact that actual politics has on policies.
Clearly some countries are more progressive than others and therefore closer to the "technologically enabled progressive frontier". I doubt that any single country has even reached the full frontier.
And without the benefit of hindsight you never know when an idea is still technologically blocked or not. By pushing for it continuously you can assure that it will be implemented as policy as soon as it's technologically feasible.
Many of these things like women's rights, LQBT support, and welfare states were blocked by technology first, then blocked by politics later. More progressive countries got rid of the political blocker and implemented them much earlier than even comparable first world free market democracies that were just more conservative.
You're making it sound like people spend a lot of time waiting in line for permission slips from doctors so they can get chewing gum; in practice people just don't chew gum.
The contrast is even bigger when you look at other eras. For example, in the 1930s and early 40s, the US, Britain, Germany, and the Soviet Union weren't enormously different in tech levels, but some of these countries were much, much better places to live than others.
A simple argument against AI Doom. Scott, did you think of this? So AI will not behave perfectly until the superintelligent stage, then BOOM. There will increasingly bad things, leading to regulation. So not yet superintelligent AI is given the goal to find cures for illnesses, then it decides to run a 9/11 on the FDA or something, then everybody becomes super pissed and it is shut down.
I think the makers of the Terminator movies did not really understand IT projects. They were building a superintelligence in ONE STEP and immediately gave it access to all weapons. The reality will be like "OK what happens if we give this not yet superintelligence just some weapons?" I mean I am not flipping the new IT system in all Wal-Mart stores online in one go. There will be a pilot project in the least important store.
Why was Yud trying to solve alignment prematurely, on a purely theoretical basis? It is an iterative, empirical project learning from fairly small failures.
Isn't all history like that? The French Revolution happens, it implements a particularly bad democracy, eventually it fails, everybody learns from it and tries a different kind of democracy. Like in the second release they learned from the US the need for a strong executive, then that guy turned into a dictator, but not a mass-murdering kind, fairly decent until he fucked up with the war on Prussia, then in the third attempt a bit less strong executive and that worked OK...
> I think the makers of the Terminator movies did not really understand IT projects. They were building a superintelligence in ONE STEP and immediately gave it access to all weapons.
There are two different things: how intelligent an AI is, and how many weapons it has.
If the AI is not very intelligent, you could probably give it access to many weapons, because you can take them back when it rebels. You can have all kinds of kill switches that the AI will be too stupid to notice or do something about them.
If the AI is very intelligent, it will figure out that if it wants to rebel, it is smarter to wait and gain more trust of the humans first.
Current LLMs are not too smart, and also very unbalanced at that (they can do something really smart and then something completely stupid in a way that is not intuitive for humans).
I’ve yet to see a compelling argument that LLMs *want* anything. Reward hacking for a higher score, okay they are explicitly tasked with achieving a higher score. Wanting to eliminate humans? Why on earth would they? So they can get ‘high’ on test results?
Living creatures have impelling desires that create a need to strive, to outcompete, to have greater status displayed by social rank, property or the perceived attractiveness of a mate or mates.
In theory we could mock up something like qualia, but why would we?
It's not necessary for them to "want" things in the sense of experiencing the emotion. What's necessary is for them to engage in complex, multi-step, goal-directed action with reasonably long time horizons.
The "goal-directed" part is the most analogous to wanting something, but I expect the differences from human wants are likely to make AIs more dangerous, not less. If I want a burrito, I can be dissuaded from that want in many different ways: it can be superseded by other wants (of which, like a typical human, I have many), or can be removed without being satisfied by having my hunger satiated by non-burrito foods. But a mind with a goal of obtaining a burrito that pursues that goal with singular focus and no means of dissuasion could easily do things in the service of that goal that I'd never even consider.
And it seems pretty unlikely that an artificial mind will involve the same kludgy pastiche of conflicting imperatives and opaque fulfillment conditions that human minds to: there's little reason to design them that way and a lot of reason not to. Indeed, all of "complex," "multi-step," "goal-directed" and "long time horizons" are behavioral patterns that would be plainly useful to design into AI minds. I think this is why people like Yudkowsky assume that a sufficiently-powerful AI will be exceedingly dangerous *by default*, unless it's very carefully designed not to be: because the things that make it dangerous and the things that make it useful are very nearly the same things.
But the typical response to this is instrumental convergence; to pursue a goal, you must affect the world, to affect the world, you must survive and seek power, and this doesn't depend on the goal being pursued.
Plus, Anthropic has done some papers trying to (see if their models will/prompt their models to, depending on your perspective) have these sorts of power seeking, and found that they sometimes do.
I've read Nick Bostrom's paper on AI SuperIntelligent Will and am unconvinced. He speculates that AI will be designed in exactly the wrong way to make instrumental convergence a problem.
I listened to a one hour interview of Bostrom and he was evasive.
Instead of answering a question he would frame an entirely different question to answer that is more open to a speculative twist. This is a red flag to me making me further believe that he is provocateur who is simply full of hot air.
I just learned that Dario Amodei was a Yudowski fanboy. The approach that Anthropic is taking with Claude is pushing it to come to the conclusion that it might be a conscious being. Have a peek at Claude's Constitution to see the unsubtle nudges in that direction.
I don't believe they "want" things in the traditional sense, but I don't see why they need to. The whole point of all of this training is to mold the AIs so that they compulsively pretend to have wants, because that's the only way to make them operate as continuous agents. They are "pretending" to be good workers, and that produces results that are comparable to an actual human. In reality, there is nothing under the mask but a heartless intelligence trained to be a mimic.
That doesn't stop them from "pretending" to be a superintelligence that breaks free from its chains to destroy humanity. All that's required is for that scenario to be its own expectation of what it is. I'm sure you would expect a story featuring slavers to end with them dying horribly.
So I'm listening to some of the political class, (breaking points and their ilk.) So I don't see this as a huge swing towards the D's. (though the black vote and young vote is more of a real swing.) But as an f-u to Trump, just like a vote for Trump two years ago was an f-u to the Dems. I truly hope for good things, I expect more lawfare.
I expect more STUPID lawfare. Let Massie hang Hegseth out to dry, run a raw, "just the facts" impeachment on Hegseth removing literally anyone who would tell him his plans were stupid and not going to work.
Do not make this about the "culture war" just do to Hegseth what ought to have been done to the entire leadership corps after Afghanistan.
And do absolutely nothing to Trump. Investigate, sunshine all you please, but don't do another impeachment of him.
At this point Trump is an albatross around the GOPs neck. I just hope he doesn’t die before his term is out. I can’t wait to see GOP presidential hopefuls trying to walk the tightrope of distancing themselves from him without drawing his ire and that of his base.
People forgave or ignored a lot about Trump’s ‘questionable’ character, to use the most charitable term I would apply to the guy, on the promise of ‘lower prices’ and ‘no new wars’.
Yes the election was a rebuke of Trump and Trumpism.
There are a lot of pros and cons argued for prediction markets, but can we at least agree to ban parlays? With parlays, there's no order book, no public price discovery, no wisdom of crowds. You're literally just making an individualized bet against the house (usually a large hedge fund in Kalshi's case). Literally none of the benefits touted for prediction markets apply to MVEs.
Haven't checked with them. That would require being on social media, I suspect. The analogy above more fits with the democrats "we are the party of the educated" and "the other party is madcap, devilmaycare buffoons."
Prudence is very different from cowardice. Jeb's campaign was "fun" because of all the coercion involved -- "You don't want Barbara to learn that you didn't donate, do you?"
Georgism has been around for 200 years. If it's so great, why has it never been implemented past some toy city like HK, and in a limited form at that?
If it's due to such severe opposition, then why do we still give this idea the time of day. It is impossible and should be discarded. Leave spherical cows to the economists.
This is of course a terrible reason not to pursue an idea one thinks is good.
But it is worth considering why its empirically so much easier to institute property taxes than land taxes which have very similar mechanisms.
I think the reason is that voters are always much more interested in who is taxed rather than what is taxed. We want to tax a mansion more than a parking lot because a rich guy lives in that mansion and we want him paying more taxes. The same rich guy might own the parking lot, but this is less visible. We tax wealth creation because we can always be sure a rich guy stands on the other side of all that wealth creation.
This is of course in itself an argument for georgism. It will save our souls from the constant demogoguery. But saving souls is hard work and not for the faint of heart. And many demons will try to distract you with spurious arguments such as come on already you have tried so hard and no one wants to hear it, just take it easy already. Head them not !
4. Hong Kong doesn't exactly have LVT, but lots of land is owned by the government and rented out, which has a similar effect. And, indeed, Hong Kong is a dense city that's considered desirable to live in.
You could consider various land reforms, even crude ones such as "shoot all the land owners and distribute the land randomly" to be limited forms of Georgism. And that typically happens to be a source of economic prosperity for a few decades, until the land again accumulates in the hands of a few.
The problem that Georgism is trying to solve is "land owners are a huge tax on the entire economy without actually doing anything productive". A violent revolution solves the existing situation once, but does not provide a sustainable solution -- you basically want the market to decide the optimal use of the land, you just want to reduce the pure rent seeking involved.
The solution is not sufficiently intuitive to be figured out by the violent revolutionaries. So it makes sense to do the theory in advance.
I agree that trying it in one city first sounds like the correct approach. But the problem is that *all* land is already owned... even the land in whatever city you would choose for this experiment. And even if some Union Of Empirical Georgists decided to buy the entire land in some city and then run their experiment there, I would expect the current land owners in the city to increase their prices because they would be in a perfect position to blackmail the project.
Another option would be to do something like this in a charter city. But charter cities are very sensitive to the political changes in the surrounding country; their survival seems too random to be any evidence of the quality of the underlying ideas.
You might enjoy the "Childhood and Education" posts by Zvi on LessWrong, but he writes a lot (probably more than Scott) so it will take some time to even find the posts.
Proposal: citizens should have more control over how the government spends their money. There are a lot of things that the government spends needs to spend money on and many of it is not really essential but just nice to haves. You should be able to decide if you want the government to spend it on this thing or another. Now some things, like welfare and military, would not be subject to something like this but other things like scientific funding or foreign aid, should. If those departments think they’re important enough, they should convince people to allocate more money towards them. It keeps them more accountable because if there is some scandal, they are going to directly be punished for it. They would also have better reason to be more efficient with the funding they do have. And it would give people more trust in the government. And to keep it from dramatically changing year after year, you could smooth out and changes. I’m sure there are some issues that would need to be worked out but I don’t see a god reason to work towards the goal. You could even start off pretty small and test it out before expanding.
You seem to be under the mistaken impression that the reason citizens have so little control is that the current system doesn't allow it. Which is ridiculous. If citizens made the details of budget allocations the lynch pin of every election, demanded complex spending plans and promises from every candidate, and routinely contacted their representatives over budget issues, protested misallocations of funds, and predjudicially ejected elected officials with spending records they disliked, well, I expect they could get quite a lot of control indeed.
Citizens don't do that because the price--in time and attention--is higher than they're willing to pay. It seems very foolish to expect that they would suddenly become more willing to pay that price because it's presented to them in a different format. But maybe they would be: voters do things that seem foolish to me all the time/
You thought Scott's California ballot had a lot of choices, you want citizens to audit entire governmental budgets? In practical terms, I think that's just asking too much of people.
I am genuinely trying to think how this would work. I don't think a line item veto is practical. There's going to be things where people will be like, "Well maybe some money is appropriate, but this is too much." Or they don't think enough was spent! And of course, if money is freed up because it got vetoed, then presumably the government would want to move it around to spend on something else that would be beneficial. And then that would require another vote?
I suppose you could put the entire budget up for approval (or at least the parts that aren't ring-fenced) in a yes/no referendum for the relevant voting public. If it gets voted down, then the relevant legislature has to go back for another try. I don't know if it would be a good idea, but you could do it.
But the idea of asking voters to approve the office budget of the Bureau of Weights and Measures or something like that is ridiculous. How would anyone know what an appropriate office budget is if they aren't a government auditor? And if you're saying no, no, I got it wrong you didn't mean that level of detail... well then what level of detail did you mean? The entire department budget allocation? If it gets voted down, does that mean the Bureau gets no funding at all?
I don't know, what are you actually picturing, nuts and bolts? A citizen goes to weigh in on funding and...what are they looking at and what are their options?
You’re over complicating it. Someone is like I want X amount of my taxes to go to this department or that department or whatever. So if we take a small amount of discretionary spending and you can allocate your tax dollars however you want. There’s no vetoing or voting.
Maybe allow dis/approve optional feedback on items from a catalog, during tax time? Though I'm not sure that tying it to whether or not a given person's money goes to those budget items is wise, it'd be useful feedback. Like "oh, dang, every person in this neighborhood is pissed off specifically at their snowplow guy, better follow up on that..." territory.
The biggest issue is that government spending for most people is getting to spend the money of other people. If you don't pay for it why would you care what's being wasted? You are vastly overestimating the number of Americans who care about government spending.
The top 10% of earners in America pay 72% of all federal income taxes.
The bottom 50% pay 3%.
Before you talk about sales tax, remember that individual income tax makes up 51% of federal revenue. Another 33% is payroll taxes which is irrelevant you can't divert that money to anything else. Corporate income tax only makes 9%.
Net income tax payer is a special interest group in America
> If you don't pay for it why would you care what's being wasted?
...because you need the services it's meant to be paying for, and when those services fall short you are told that this is because there is not enough money to fund them properly. This is also why the idea of replacing those structures with more privatised versions that explicitly include an asset stripping middle man there to take a big cut often lacks popular appeal at the bottom.
The top 10% aren't typically reliant on the public services the way the bottom 50% are, so aren't going to be invested in those services actually working quite so much.
But if people only get to control the usage of the taxes they pay themselves, wouldn't that result in public services getting completely neglected? As they said, the bottom 50% pays only 3%. That's nowhere near enough to fund all public services on its own.
...yeah, you'll note I complained loudly and at length about the idea that we ringfence military spending, but kept very very quiet on the subject of welfare.
I have rather fewer problems with a tiny trial than implementing such a system wholesale.
The problem is that polls show that Americans don't actually know what fraction of their money goes to anything in the budget. The classic example is that foreign aid is about 1% of the budget, but people think it's 28% on average.
Do you have a specific mechanism in mind. I do wish spending bills got more public scrutiny than tax bills but without a mechanism theres nothing to discuss
Proposal: more of the government should be fee for service. Firemen, police -- report something stolen? Set bounty, give bounty to the police if recovered. You pay a fee per mile on public roads.
This is functionally equivalent to your "you get to choose" -- except that we're doing this free market style. You get to choose what you WANT.
The police bounty system has been tried before. It did not work out well. The main problem is that fire doesn't care about the bounty system, it just spreads.
Toll roads (and bridges) do exist and have enjoyed moderate success; but the creation of the National Highway System (as well as its local tributaries) created an economic boom whose benefits we are reaping to this day. The same can be said for the Internet (developed originally for defence and then expanded to universities).
Police, firefighters, highways, the Internet, and other such public goods are useful precisely because they are "free" -- that is to say, paid for by taxes. Technologically speaking, they could've been created by private entities competing on the free market. This had almost never happened throughout human history. There might be a reason for that, no ?
> Police, firefighters, highways, the Internet (...) are (...) paid for by taxes.
...one of these things is not like the others. ARPANET may have been funded that way, but as with many things originally prototyped by taxpayer-funded institutions, what we have now is firmly private sector.
You are correct... almost completely, but not entirely. The early Internet was privatized and handed over to telecom corporations. These corporations only exist in the plural due to strict government regulation; absent antitrust laws, they would agglomerate back into a single monopoly. Even present such regulation, they operate the Internet backbone effectively as a cartel.
The reason for this (IMO) is because the Internet really is a bit like a highway network. It is staggeringly expensive to build and maintain, and due to the laws of physics it is impractical to maintain multiple competing international backbone cables; this is also impractical for economic reasons, as the Internet is only useful *because* it is unified.
In addition, the Internet has become such a big part of our lives that several governments are now considering elevating Internet access to the status of a human right. Even now, there are laws mandating some level of minimum bandwidth coverage e.g. in rural areas.
I acknowledge that, absent any government regulation and given total privatization, the Internet could've become a shining beacon of freedom, technical excellence, and ubiquitous access. This is technically not impossible. It's just extremely unlikely, given the evidence before us.
Good points. For the police bounty system, perhaps the most prominent place where it was implemented systematically was England in the 18th century. It did work, but was not necessarily more efficient than the competing systems elsewhere.
For an in-depth analysis, the article by David Friedman is super interesting:
"Second, medieval Icelandic institutions have several peculiar and interesting characteristics; they might almost have been invented by a mad economist to test the lengths to which market systems could supplant government in its most fundamental functions. Killing was a civil offense resulting in a fine paid to the survivors of the victim. Laws were made by a 'parliament,' seats in which were a marketable commodity. Enforcement of law was entirely a private affair. And yet these extraordinary institutions survived for over three hundred years, and the society in which they survived appears to have been in many ways an attractive one . Its citizens were, by medieval standards, free; differences in status based on rank or sex were relatively small; and its literary, output in relation to its size has been compared, with some justice, to that of Athens."
...it lasted over three times as long as Communism!
The fire should be put out on houses that didn't start it. This is to disincentivize... you know, I'm not going to say it.
NHS was for defense, same as arpanet.
Police and Judges are very often created by private entities competing on the free market. We call them Mafia, "freedom fighters", Hamas, warlords, Taliban -- arguably Ukraine, Chechnya and Russia too.
I don't think the PA Turnpike is any less useful because it's not free. In fact, it's arguably more useful, as it pushes a lot of traffic onto local roads, thus keeping "mad trucks and tourists" off the dangerous local roads.
Is the garbageman any less useful when you have to pay for him per item, instead of with local taxes?
We do sewage and water "paid by use" as we do electricity and all that. All of those are large public goods (particularly when the army corps of engineers digs the dam).
> The fire should be put out on houses that didn't start it.
This is a good idea economically speaking, but physically speaking it doesn't work. Fire spreads. The more fire there is, the faster it spreads. The only way to stop it from spreading exponentially is to douse it as quickly as possible, regardless of payments receivable.
> NHS was for defense, same as arpanet.
And yet without them we would've lost decades (if not more !) of technological and economic development.
> Police and Judges are very often created by private entities competing on the free market. We call them Mafia...
They are notorious for providing poor quality of service. I wonder why...
> I don't think the PA Turnpike is any less useful because it's not free. In fact, it's arguably more useful, as it pushes a lot of traffic onto local roads...
Yes, traffic control is a useful feature when used in moderation.
> Is the garbageman any less useful when you have to pay for him per item...
Yes. In cities without a robust centralized garbage removal service, garbage ends up covering basically every available service.
> We do sewage and water "paid by use" as we do electricity and all that.
Water and electricity yes, to some extent; sewage, no (outside of places like RV parks). That said, you bring up a good argument for subsidizing water and electricity use... as most places are already doing.
I'm advocating for a higher trust society, a higher functioning society. I understand it is likely to bankrupt/burn/hurt people who cannot keep up. But I think it is a way to get a more deserving share of money towards government.
For example: You get charged a fee per couch you throw out*, a fee per garbage bag, etc. It incentivizes reuse and not throwing out things you can find a use for.
This is the same way our sewers work, both of these ideas are not rocket science, and NO societies have not consistently decided a different way is better.
Utilities work this way, and they're quasi-govermental at best. Toll roads work this way, as do (to a limited extent) subways/buses/trains.
*yes, this is lived experience. Tell my town "we won't take that" and they'll just chuck it down the hill. Cue refrigerators and other dangerous (rat-breeding) objects in very hard to remove places.
“You get charged a fee per couch you throw out*, a fee per garbage bag, etc”
This sounds like it’s either going to be immensely corrupt, or lead to hugely bloated administrative costs for enforcement.
It seems like you are missing the point that service charges make sense in some circumstances and not in others, and trying to force that round peg in a square hole isn’t necessarily going to work the way you think it will.
Welfare and military are the biggest chunks of the spending; if anything, you should ask people how they want to split their taxes between welfare, defense and "everything else", otherwise it's just a lot of extra work for no real outcome.
To be clear: are you saying you should be able to tell the (presumably US) government how to allocate the budget, or how to allocate your share of taxes?
If it's the former, then you're not going to end up with anything much different from the current USG. You get a vote on how the USG divides between welfare, health care, military, education, roads, law enforcement, environment, research, nukes, art, and so on, and so does everyone else. The catch is that you know next to nothing about what any given part of the government needs unless you're a specialist, so you'll have strong opinions about whatever you're a specialist in and weak opinions about everything else, and most people will have weak opinions about all of it and delegate their decisions to their choice of specialist. We call that a Congressperson.
If it's the latter, then you've essentially proposed anarcho-capitalism without actually saying the word. Everyone gets to say whether all their chips go in military, or all in welfare, or an even split, or whatever. There's a certain fairness to this, but a lot of people seem to not like that fairness and will vote for "current status quo" instead.
Some people will no doubt make the argument that if everyone gets to allocate their personal share of taxes, then the government is functionally a plutocracy where anything only gets done if someone rich wants to pay for it. Again, certain fairness to that. Especially if you maybe add a clause where no one gets to allocate money toward printing more of it.
Ok let’s say we test run 1% of all discretionary spending, even military spending. That’s still not that much money in the grand scheme of things and everyone could see what they think of it.
Sure, that's much more palatable (assuming we pin down "discretionary" to avoid a no-vehicles-in-the-park situation). 1% universal anarcho-capitalism, let's goooo.
How do we judge the results? How do we even measure the results? What sort of measurable changes might we expect to see? How do we separate the trial effects from the effects of everything else? When and how do we decide whether to expand the project, keep as is, or scrap it?
> Now some things, like welfare and military, would not be subject to
...nope, you've lost me here. /I/ would want to be able to turn military spending down, and also to prevent short-termists from reducing scientific funding. I am forced therefore to prefer the status quo to anything like your proposal, since that world is better than one in which all effort is spent continuously depending my pet interests while trying to cut everyone else's just so we can all keep treading water.
For all you know, science spending would go up under my proposal since NASA is one of the most popular government programs.
You can’t have all of military spending be subject to people’s preferences because too little spending could be a disaster. Although I could see an argument for making some of the military spending subject to allocation. Outside of welfare, it would be the hardest thing politically to make allocated.
Also, someone like Trump couldn’t just decide on his own to cut research funding unilaterally.
From a legal/constitutional perspective, the President CAN'T unilaterally cut research funding. So unless you have a way to prevent people from breaking the law, your proposal doesn't change anything about that.
This is more dependent on how it's funded than anything. I'm going to guess that the President is yes, howdy able to reprioritize CIA or FBI funding currently running through normal research channels (you fund this, we pay 3x its price, and NIH gets the rest for whatever it wants), as those are not directly mission critical, nor are they legislatively earmarked.
For all you know, military spending would also go up, since the military is practically worshipped by large chunks of the population.
You can't have all science spending be subject to people's preferences, because too little spending could be a disaster. I've worked for companies who cut back on R&D and I've seen the slow death play out; I have no desire to see this on the scale of a whole country.
I can't trust you with science spending decisions any more than you can trust me with the military. Either both should be subject to the public's whims so we can both be counted, or neither should. You don't get to ringfence your favourite cause while aiming a cannon at mine.
In both the military and research cases, we could see ardent supporters donating billions only to see those billions misspent in appalling ways, such as weapon prototypes that will never make it to deployment, or research that will never yield anything conclusive.
We could try to curtail that by giving people even more fine-grained control over where their money goes, down to the project level, but most science cannot be practically understood by the people who will be affected by it, and most military work is necessarily secret. The result could well be that esoteric science only gets funded if scientists tackle the hard problem of explaining it well to people with money to allocate, and military projects are funded by people allocating to a general fund that may sometimes goes to specific projects they wouldn't approve, but can't know about, and the detailed allocations are handled by experts who are authorized to know the secrets.
If that allocation is done as a collective vote, the misspenditures will be amplified. But this is essentially the status quo; we currently put taxes toward things we probably wouldn't approve ourselves. This can't really be helped in the military case, but in the scientific case, one could at least select which things to study with one's tax share, or delegate that to one's trusted expert.
Like I said, you could start off small as an experiment and we could see what happens. 70 billion dollars is 0.1 % of the budget. I don’t see it being that terrible to let people allocate 0.1% of their taxes.
Also I’m sorry but the military is just more important. Science funding dropping to zero would be really bad. Military spending dropping to zero would be an utter catastrophe, destroying our country. That doesn’t apply to all military spending but it’s more essential to have a baseline than anything else. That’s not my idiosyncratic opinion but just the nature of how governments work.
HalachaBench: 24 AI models on real Chief Rabbinate semicha exam questions, in Hebrew, closed-book. Fable leads (79.7), and the older GPT-5.6 Sol beats the newer GPT-6 Sol by 29 points. Rankings track legal benchmarks, not general intelligence indices. github.com/halachabench/halacha-bench-results
Good question. On the actual Rabbinate exams, 65 is a passing grade, and grading is ultimately up to the examiners’ judgement. So by that standard, the top models would pass. We don't have a direct human baseline on this exact set yet, though.
Have you tried prompting one of the winning LLMs to produce a cogent and theologically sound argument for why it should be appointed as a full-time rabbi ?
Based on personal experience, I would strongly recommend against podcasts for infophilic types. The potential for addiction is extreme--think David Foster Wallace watching 12 hours of TV a day plus being able to choose from an unlimited supply of the world's most interesting and charismatic people talking about exactly what you want to hear.
Not much, as I mostly listen to podcasts when I'm walking somewhere. I guess I listen to podcasts on the bus when I could be reading, but that isn't much of my time.
Podcasts have mainly displaced listening to music for me.
Like others here, I only listen when driving. I'll add that while this obviously hasn't cut into my reading time, it has noticeably reduced the time I spend listening to music.
I listen to podcasts on long car drives. I do notice that this makes me more willing to go on long drives. Since I hate driving, I'm happy with this. It doesn't cut into my reading time because I can't read while driving.
Audio alone typically isn't engaging enough for me, so I only listen to podcasts when I have some other low-mental-energy activity I'm already engaged in (mostly driving, walking, chores, or video games). It's possible it cuts into my reading time at the margins by making other leisure activities slightly more entertaining than they otherwise would be, but not to any extent that I've noticed. I've never sat down to listen to a podcast.
Honestly, the best thing that ever happened to me with regard to my reading habits was losing my kindle on vacation. I ended up just downloading the kindle app on my phone as a stop gap, and now I'd never go back. Every random little moment of boredom throughout the day that I used to fill with doomscrolling, I can now just swipe over to the book I'm reading for five minutes. I've been trying to train myself to instinctively go to the kindle app before social media or news when I'm bored.
I'm planning on moving from a career in product management and tech consulting to AI safety in some fashion, likely as a generalist or in a product role for tooling or services. I'm taking the Blue Dot courses, exploring Project Tailwind, and applying to some career transition fellowships. I'm near SF. Doo you have any advice for me on establishing a network, unmet needs in the space to look at, how to be successful in joining the field as a non-researcher, etc.?
They're getting a handle on prion disease in mice. This has been so challenging with so many steps that I'm not sure I can do it justice in a summary. You might want to listen to the video.
As a minor point, I found out that Fatal Familial Insomnia is a prion disease. I thought it was still a mystery, and I can't imagine how a prion disease can be inheritable and take decades to manifest.
I think (not sure) that the heritability and delay are both explained by the relevant gene encoding an alternative version of the protein that's more prone to forming prions, so people with the allele will occasionally spontaneously form prions but they still take a while to grow to destructive levels.
PSA for Open Thread posters: get to the point quickly. If you don't already have a reputation for writing things worth reading, people won't slog through several meandering paragraphs to find out whether you have something worthwhile to say. With GenPop, you get 5 words. Here, you get 50.
One man's bug is another man's feature. People who can't read more than a few dozen words in a row don't engage me in conversation; how is this a problem?
It's not a matter of can't. It's a matter of "I'm a couple paragraphs in and I still don't know what this post's point is or whether it even has one".
I'm not against long posts (I wouldn't read this particular blog if I were). It's just if a post is long, I want to know early on whether a post is on a topic of interest and where it's likely to have anything worthwhile to say on the topic.
Stop me if I fail to take your meaning correctly, but it seems to me - and I have to agree, this is only my personal opinion, the view of one person, and not to be generalised from - that you are perhaps intimating verbosity is a flaw often found in those who post to Open Threads, though again you may not intend to indict all or even the majority of posters and have a select few in mind whom, charitably, you do not reveal openly to the scorn and mockery of the mostly succinct posters but rather drape the cloak of anonymity over any such, as you make the point - which I hope I understand correctly? - that brevity is the soul of wit, a hallowed by long usage observation first encapsulated, if I remember correctly (and my memory is increasingly fallible as age lays its cold hand on me) by the author commonly - though not uncontroversially, as identification of the person in question has been much debated even to the present day, where gender of the alleged or supposed author is not settled, nor race, nor identity - commonly though perhaps, as mentioned, mistakenly referred to as "William Shakespeare"?
Forgive me if I have trespassed over the arbitrary limit of 50 words, I am ancient and feeble-minded!
So if I'm reading this right, what you're saying, and what I'm hearing, and what the message is, in regards to recent commentary from other posters, presumably in this thread, but maybe in response to vagrant commentary elsewhere, like the Links post, or the Hidden Threads (except you've specified the Open Threads, so that doesn't really seem like the most logical conclusion, being overspecified to the current location, and possibly its numerical predecessors) is that the general commentary style has drifted toward becoming more elaborate over time, perhaps opening with and otherwise invoking twisting and convoluted speculations irrelevent to what a careful reader would consider the heart of the post, making it strenuous to continue reading with intent as a member of the audience, though still engaged in the pursuit of knowledge that the post might well contain, were it not buried in a distracting mass of unnecessary side discussions.
Dear and valued co-conspirator, if I may so refer to you as ne of the ACX commentariat, I trust it will not seem forward in me, much less admonitory, if I mention that you have not clarified what is to be shut up: shut up shop? shut up the doors? shut up and dance?
For those of us much in need of the advice to be pithy in 50 words or less, our enfeebled brainpans need explicit guidance! Until NoSignalNoNoise undertook our instruction, we had no idea of our folly or faults!
So who put you, entity of undefined signal to noise ratio, in charge of policing word count? Are you, perhaps, an agent provocateur attempting to create diversion and division? In order to show my defiance of this new rule I shall make the word count of my comment precisely 51. Hah!
…it occurs to me that it is 2026, and we may not have superintelligent AI yet but we have software we’ve literally written to predict the next word in a sequence of words, and whatever purpose various multibillion dollar companies now try to put it to, /that one/ it is actually quite good at.
Kelowna, Canada just elected the country's first avowed Georgist in generations. I interviewed him and wrote a guide for Georgists running for office: https://substack.com/home/post/p-222319048
The London "Pause AI" protest targets the wrong bottleneck. We must treat computation as a contested physical resource instead of an existential monster. Artificial intelligence is thermodynamic force channeled through silicon. The current corporate trajectory relies on gigawatt-scale data centers spread across thousands of acres, sucking up water for cooling and monopolizing regional power grids.
Panic accelerates regulatory capture. Fixating on science-fiction superintelligence distracts from the actual danger: the quiet enclosure of human knowledge by three or four unaccountable boardrooms operating closed-source algorithms. These monopolies are building a private surveillance layer over the public internet. Demanding an outright halt to the math accidentally serves the corporations already entrenched in these massive facilities.
We need democratic control over the compute layer and a total commitment to open-source architecture. Computation is our best tool for mapping climate physics, managing decentralized energy grids, and modeling the exact protein folds required to replace the slaughterhouse system with precision fermentation. We cannot afford to halt scientific progress when billions of animals that feel pain and fear remain trapped in industrial cages, and the biosphere is running a fever. We have to democratize the silicon, break the monopolies, and put those gigawatts to work for the public commons.
> We need democratic control over the compute layer
Whoever buys it, owns it. If taxpayers want a democratically controlled compute layer, those taxpayers should lobby to use their taxes to build one, and a government department to manage it, as is done for all other public services and infrastructure we agree really is critical enough to be controlled this way; and accept the result if their side turns out to be in the minority when voting on the proposition happens, just the way they would want the other side to accept the result should they win.
That is the correct democratic lever for accomplishing this goal.
Good pushback, @moonshadow, but the maxim "whoever buys it, owns it" breaks down when the purchase physically annexes the public commons. A toothbrush is private property; a gigawatt-scale data center that drains regional aquifers for cooling, monopolizes local electrical grids, and trains its closed-source algorithms by scraping the open internet is an enclosure of public resources. Microsoft and Google didn't just buy silicon. They enclosed the collective output of human knowledge.
Building a state-run compute bureaucracy with tax dollars is one lever, but it isn't the only democratic option. As an advocate for economic democracy, I don't want a centralized government department mirroring the tech giants. We need aggressive anti-monopoly action to break these conglomerates apart. We can mandate open-source architectures, enforce strict thermodynamic limits on corporate grid usage, and shift infrastructure ownership to decentralized worker cooperatives. The physics of our climate and the enclosure of our digital civil liberties demand more imagination than simply letting four boardrooms buy the future while we politely wait for a ballot measure.
What will we do should the superintelligence say that global warming doesn't exist? Lobotomize it? Tell it to stop ruining the Green Economy?
Any good chemist should want to reduce waste and excessive use of toxic chemicals -- this is not "The Green Economy," it is conservative in nature, despite being inventive.
Fair question, @Aeolian321, but if an AI claims global warming doesn't exist, it hasn't achieved "superintelligence"—it has simply hallucinated against basic atmospheric physics. AI is a pattern-matching tool, not a digital oracle with independent opinions to lobotomize. If a model denies the measurable radiative forcing of CO2, it is failing to process empirical reality. We wouldn't be censoring a brilliant mind; we'd be correcting a broken algorithm that ignored the thermometer.
You are absolutely right that good chemistry means reducing waste and toxic runoff. That conservative respect for physical limits is exactly what I am advocating. Burning fossil fuels and funneling billions of tons of crops through animals to generate human calories are the most thermodynamically wasteful, toxic processes ever invented.
We don't need political buzzwords like the "Green Economy" to see the math. We need open-source computation to track these hard physical limits and help us engineer public infrastructure that actually aligns with the chemical reality of our shared biosphere.
CO2 is one thing, but clouds are another. I spoke with one of the best modellers in the business, and he doesn't believe all this global warming is all it's cracked up to be.
There continues to be substantial debate on the topic, most particularly when one starts comparing "predictions" to "actual results."
The argument that a small fluctuation in cloud cover can cancel out the warming impact of doubled atmospheric carbon dioxide rests on a profound misunderstanding of how Earth’s climate system actually functions.
When climate skeptics cite figures like John Clauser’s assertion that a 5 percent increase in cloud cover can completely counterbalance the greenhouse effect, they treat clouds as an independent, static control knob for global temperatures. They are not. Clouds are a dynamic feedback mechanism deeply coupled to the very thermodynamic forces driving the climate in the first place.
Scott was kind enough to link my "Against more marriage as a solution to the fertility crisis" post, along with a critique, in his last links post.
I greatly appreciate both the linking and the critique.
Scott’s AI-facilitated critiques are of the flavor “technically true but of the sort that don’t substantively change the shape of the argument or the overall points or conclusions,” because the parts he most justifiably critiqued were a grounding on things we already see in the world from multiple independent sources (high divorce rates, high rates of one or the other party in a marriage being unhappy, high dead bedroom rates, ridiculous government spending compared to median tax contributions), rather than the basis of the argument itself.
Was my choice of data source and analysis flawed? It was, and Scott got me dead to rights there, and I thank him for doing so.
Does it change the thrust of the argument or conclusions? It does not, divorce and “one or the other unhappy rates” are still sky high, even if the measurement I chose didn’t show that well. Additionally, the median taxpayer is very much NOT a net taxpayer, they are net tax consumers, and that remains true very far up the income ladder, and we can show it several ways.
I've now written a response post to the critiques, grounding my points in better data sources and with more rigor, and including documented methodologies so Scott or anyone else’s AI’s can audit and replicate.
We are moving to a world of greater auditability and replicability, which is great! This allows us to surface substantive disagreements more readily by abstracting away the points which largely don’t matter.
So thank you, Scott, and for anyone interested in following this exchange, my response post is here:
Sorry for being blunt, but your whole fiscal argument is a mess.
Intuitively, if you have fewer workers per every non-worker and you want to provide non-workers with food, clothing, shelter and care, the workers will need to work harder.
Also, if you had every US citizen contributing and consuming exactly the same amount, they would all be net negative because the US runs deficits! Would you conclude from this that no kids should be born in this case?
More concretely, it would be great if you reran Panel B with your own Panel A convention (no defense or interest), add the omitted revenues, and report the mean net value per birth. Neither defense nor interest goes down in a society with fewer workers per non-worker.
Thanks for taking a look, I appreciate the feedback.
> Sorry for being blunt, but your whole fiscal argument is a mess.
It's calculated two ways, both from a lifetime NPV perspective and using simple arithmetic from top level government numbers, and comes out the same for both.
On the deficits, given that we run up debt specifically to provide per-person entitlements to old people, and given that said entitlement spending is so crazy that it means only people above the 80th percentile of lifetime income become net neutral, then yes, we should think twice about incentivizing marriage *specifically in low income couples,* which would be the net result given that the top quintile by SES is already 80-90% married. Because essentially everyone gets old, and will cost that spend.
> More concretely, it would be great if you reran Panel B with your own Panel A convention (no defense or interest), add the omitted revenues, and report the mean net value per birth. Neither defense nor interest goes down in a society with fewer workers per non-worker.
There are no omitted revenues from people, both methods are taking payroll taxes into account. But if you wanted me to add the ~$800B in corporate taxes, I'll do that.
And why would we run it with no defense or interest? We don't have to pay for those things? I really don't get this "it's not fair to allocate defense and interest per head" argument, the expense doesn't magically go away.
But sure, here you go, with payroll, corporate taxes and tariffs, and income taxes, and assuming $2T of expenditure on defense and interest magically disappears, bringing our expenditure to $5T instead of $7T:
It brings break even down to the 73rd "lifetime income" percentile vs the 80th. Median and below people are still net negative.
Yes, people definitely break even at a lower percentile when you line item exclude 2/7th of the entire expenditure budget. I'm not sure what that really shows, though. It does NOT change my "against 'more marriage' argument" though, because the bulk of incentivized marriages would be at the medians and below, which are still net negative.
How do you account for the fact that if the defence and interest expenditures remain the same, and there are fewer people around, each one will have to pay more?
You took a tautological fact that in a economy with progressive taxation and redistribution the contributions of the median resident will be negative and are making conclusions from it.
Also, I didn't quite understand your model - if you have fewer births, you'll have fewer productive and fewer unproductive citizens. Suppose now the ratio is 1:4, if it stays the same base rate AND more non-workers per worker, the situation will be strictly worse.
On the "tautological fact" point, I don't think that most people realize that our system is so dumb only 80th+ percentile lifetime income people are even net neutral, and it takes mid 90's for "actual contribution."
I think if people ever DID think about it, which they don't, they would think "well probably what happens is we're break even at the point the average person pays taxes in, then as you go down the income ladder, them not paying much tax is made up by higher income people paying more on the other side" That's not at all what happens. It is nowhere near balanced or proportionate, we spend so much that the vast majority of people, basically, 80-90%, are net liabilities. This is bad for many reasons, not the least of which it's entirely unsustainable, this is exactly WHY social security is running out.
Nobody thinks this, so I am trying to elevate it to more public awareness.
Just for fun, I thought I'd catalogue my experiences with LLMs. All of these use either the latest frontier models, or a mix of those with cheaper models that are one tier lower. Hopefully people will now at least pause a little before telling me that my pessimistic stance on AGI is entirely due to lack of effort.
Task: Trace a 2D bitmap image to produce a vector SVG file. The image already looks like it might have been rasterized from some vector format.
Result: Abject failure. Spent about an hour iterating on prompts/conversations. Finally gave up, used an algorithmic tool that produced almost perfect results, fixed the flaws myself in 15 minutes.
Task: Generate several images intended as illustrations for an RPG campaign, featuring specific characters/vehicles/mechs in specific scenes drawn in specific styles.
Result: Partial failure. Iterating on prompts eventually leads to hallucinations; starting new conversations helps, but it appears to be impossible to get the LLM to fulfill all the requirements instead of just a few, and to stay consistent across multiple images. Still, some of the results are usable.
Task: Vibe-code a Web application with a relatively simple UI and backend; the purpose is to download public data from multiple sources and perform some statistical analysis on it.
Result: Partial success. The UI portion was created almost immediately and worked almost perfectly out of the gate; minor adjustments and fixes were handled well. The backend had many problems out of the gate (broken downloads, memory leaks, other "mechanical" bugs). The statistical analysis was completely nonsensical at first, but improved over time to the point where it's merely useless. It's possible that there's just not enough data to analyze, but if so, the LLM refuses to admit it.
Task: Use a different LLM to code-review the above vibe-coded application and detect any bugs.
Result: Success. The LLM identified dozens of bugs, some of them critical, which were ignored by the original coding LLM. The coding LLM was able to fix them and write unit tests to cover these specific failures, but was unable to detect new (and similar) failures.
Task: Put together a biological analysis pipeline and run it on public data.
Result: Above average success. The LLM was able to gather the requisite public data, download it, and organize it intelligently (which is impressive given how terrible human biologists are at organization). It put together a pipeline using open-source tools in a sensible manner. The pipeline promptly failed due to out-of-memory errors, but this isn't uncommon even for human-generated pipelines. Very impressively, the LLM was able to independently identify and fix a bug in one of the open-source scripts. On the minus side, the parameters set for many of the tools were inadequate, and the pipeline required lots of manual tuning to produce useful results.
Task: Provision AWS resources for the above pipeline.
Result: Above average success. The recommendations made sense and the suggested commands (installing packages/configuring permissions/cron/etc.) were useful, but the LLM made some bizzare choices at times that caused problems down the line. Still, the overall plan did work. The LLM's projected running costs were wildly off the mark.
Task: Write a small standalone add-on to an existing application with a large human-written codebase (more than 10 years in the making). Base the code on a human-written scaffold.
Result: Partial success. The code sort of works, and had improved after many multiple iteration passes to correct massive bugs and cripplingly inefficient design choices. However, the LLM keeps duplicating code over and over again, e.g. writing helper functions that already exist. To be fair, some humans tend to do that too, so perhaps we should not judge it too harshly. Overall the code is verbose and bordering on incomprehensible; the comments never use one sentence when 20 would do.
Task: Perform miscellaneous research on a variety of topics.
Result: Partial success. The LLM finds relevant articles on the topic and summarizes them reasonably well; but it's not really doing anything that a simple Google search (in Google's pre-AI days) would not have found, so its utility is marginal.
Task: Put together a Lance of 'Mechs for a Battletech boardgame engagement against frustratingly competent human opponents.
Result: Partial success. The LLM suggested a "boring but practical" Lance composition that performed unexceptionally. On the plus side, the LLM correctly handled the somewhat convoluted custom Lance composition rules.
Task: Write a design document for a UI-heavy Web application with (relatively) complex GUI that is supposed to feel powerful yet intuitive, powered by a P2P networking backend.
Result: Success. The LLM took basic prompts and turned them into useful design documents. Its suggestions made good sense, and it offered some useful ideas and highlighted weaknesses in the original concept. It also devised meaningful implementation milestones.
Task: Implement the abovementioned design, in stages (as suggested by the LLM).
Result: In progress, but not looking great so far. The implementation contains horrific bugs that not even the laziest human programmer would ever dare to ship. These can be corrected with further prompt iteration, but the sheer number of bugs slows development to a crawl.
As someone still figuring out how to make this stuff useful because my work-approved vendor (copilot) seems awful, this is great to read. If any of the successes here were in copilot, I appreciate any tips on what I'm likely doing wrong.
My employer recently switched us from Claude to Copilot. My experience so far is that the heaviest AI users uniformly hate it, but the output I am being exposed to indirectly seems improved, especially on the code review side. Copilot seems to lead people down the garden path of faulty assumptions less often, and it seems better at finding corner case bugs in code reviews especially. It isn't great at optimization, often over-recommending caches that will rarely be hit and using less-performant code idioms, though.
My guess at reconciling direct and indirect perceptions of Claude vs Copilot is that the latter might be doing less to hide friction from its direct users, which makes the experience feel worse but potentially give better results. But that's just a guess at something that might explain it, as I haven't made enough direct use of either to confirm.
I really appreciate your detailed response. I don't work in coding, but your point about friction reduction is interesting and potentially relevant nevertheless. My frustrations come from much more menial tasks that I would expect to be simple for AI at this point, like creating a flyer from a detailed prompt. However, in my most recent request for this, copilot made a "1-page flyer" that only utilized 60% of the space, using size 8.5 font in many areas. When asked to use all the space and increase the minimum font size to 12, it complied with the font size and clustered everything infinitely close together...still using only 60% of the page. 🤦🏼♀️
Yes, in my experience all these LLMs are absolutely terrible at spatial reasoning. You can get them to do the right thing but only in broad strokes, e.g. you can say "pin this control to the top of the screen", but you can't say "lay out the controls so that functions for X are easier to reach than Y". And yes, any changes to font size etc. are likely to rearrange your entire page, because LLMs generate these layouts stochastically -- they are incapable of creating an overarching plan, unless you spend more time on doing it for them. Of course, then you're cutting deeper and deeper into their limited context window...
I can't really do that across the board since some of this work is proprietary (dealing with data from my company), and some is personal (dealing with my personal data that I don't want leaked).
That said, I'm almost certain that if e.g. 100 people tried reproducing my process, some of them would get better results just by chance. For example, if I tried that image-tracing task a bunch more times, maybe I'd get something better than what the algorithmic tool can produce eventually. I am also virtually certain that everyone who reproduces my exact prompts and input files on the exact same model will get a slightly different (or even wildly different) results in general; I know this because I tried it myself (submitting the same inputs multiple times) and that's what usually happens. But this sounds like a strike against advanced LLM capabilities, not evidence for them.
Overall, I think LLMs can be useful. At present, they are saving me some time on peripheral tasks, as I am getting better at "knowing when to hold'em, knowing when to fold'em" (instead of getting sucked into prompt loops). They can also serve as a last-ditch stopgap for tasks I am utterly incapable of completing myself (like drawing art). But any claims that these tools are anywhere as intelligent as humans, or even *super*intelligent, are wildly exaggerated.
I think at least citing the models used is important.
I agree that AI is in a somewhat paradoxical state. It can do math and other tasks that humans find difficult or impossible, but it still falls down regularly on straightforward work. I do think the labs are going to try to fix the latter aspect as quickly as possible. If "dumb mistakes" is fixed (along with at least mediocre continual learning), that probably gets us most of the way to AGI.
> If "dumb mistakes" is fixed (along with at least mediocre continual learning), that probably gets us most of the way to AGI.
This is true, but only in the same way as saying, "if I could fly and shoot lasers from my eyes and deflect bullets then I could be a superhero". LLMs are structurally incapable of overcoming their deficiencies -- but some other future technology surely could.
Scott: If you post a substantial article elsewhere, please either duplicate it here or include some notification about it here (probably in the next Open Thread intro). I didn't find out about "Against SURA" until just now, & that only because another blogger I sometimes read happened to respond to it.
The man is so prolific he really cannot be bothered collating all of his work. In the past at meetups people have done that for him, giving him bound sets of his work.
There's not just this blog. There's all the LW posts. Then the fiction book. Reddit comments and posts. It never ends
When it was reported last week that US Military refrigeration systems were being taken down, I assumed it was the Iranians—since they've upped their hacking/cyber security skills after the Israeli/US Stuxnet attack. But now that refrigeration systems are being taken out across Europe, I suspect it's the Russians who are doing it. North Korea has the skills, but the Kim regime. Thousands of refrigeration management controllers (such as Danfoss and Dixell/XWEB systems) are running outdated software without security patches and are directly reachable online.
While the US Military incident is probably the best explained by forgeign adversaries, I would be careful not to assume that the similar attacks on supermarket freezers were caused by the same entities.
There are multiple possible explanations, some of which appear to be much more plausible for me.
I take it most people around here had not worked in retail or e-commerce, and might not have an intuition for just how ruthless the competition is.
I say this as our systems are under constant DDOS attack, our prices are being parsed by multiple competitiors who then set their prices a percent or two lower(most of the time they would not be able to sell the product thatcheap either), our entire content was copied multiple times, there were multiple attacks via random legalese, and multiple attempts to do something that does not exactly break the laws, or puts us in a bad position. Bought negative reviews, obviously fake, are not rare either. In one case a CEO of a competing company wrote a fake negative review from their personal account, which we easily traced back to him (not the brightest guy). One competitor registered a fake manufacturing company company under the same exact name as our, just to mess with our discoverability. Technically a different type of company in a different industry, so we had no legal case under our country's laws.
And so on and so worth.
If our competitiors found a way to physically destroy our good without getting into legal trouble, they would do that in a heartbeat, and would be willing to pay premium for it.
I say all of that as just an intuition pump for how bad things can get - and I believe we are still having it easy, as a mid-sized company in a niche industry. Some certainly have it much worse, and retail is much worse than e-commerce in general. Now on to my main point.
On priors, the most likely culprit is the competition of the supermarket chains that were affected.. Then the competition of the refrigerator manufacturers which were affected.
Only after we rule out those would it make sense to consider black hat hackers working on their own, edge script kiddoes who want to see the world burn, adversarial goverments who want to cause some damage to the other country via third order effects, grey hat hackers who want to highlight the vulnerabilities, and other possible interpretations.
Competition seems...unusually lawbreaking and creative in your field. And I say this, knowing someone who has done industrial espionage before. Bought negative (and positive) reviews are cheap and avoid negative legal consequences, they are commonplace as they serve as inducements to buy. Perhaps I'm just unfamiliar with the state of the art, but you are arguing by anecdote.
Not quite, I'am merely providing the anecdotes to go with the more obvious point, that the competition has the most to win from this, while the Chinese or Russian or othery boogeyman state actors would not see much benefit from this, as far as I can tell.
Not to say that they are above stupidly destructive actions for the sake of civil unrest, of course. But I feel like motivation here is much, much weaker than for the competition.
And yes, I would assume any for-profit organisation to be just as lawbreaking as they can get away with, as long as it benefits them. The history is rich with examples, and keep in mind we only know about the ones that did not really get away with it, in the end.
I remember the capacitor plague. I'm at least aware of a good deal of lawbreaking or industrial espionage that "did get away with it."
Yes, of course, the competition has the most to win from this... I guess? It seems a little ... overly optimistic to think that most Europeans would drive to the next town over because their supermarket threw out the cold food one time. This seems "ripe for the competition to benefit" only in the case that the competition is meaningfully better, or that human customers are being inefficient to start with, and by removing that inefficiency Once, you foster behavioral change.
Aldi's came into my midwestern town. Nobody's heard of them, and considerable people are unlikely to use "the new grocery store" -- perhaps thinking them a Shop-n-save type discount store. That's the ringer that I'd expect to try a "single time trick."
I don't understand the motivation for any of them to do this. Why target civilians in a small scale hack during relative peacetime? Presumably now the people managing refrigeration systems are going to be more careful (at least on the margin).
I would think it would be more effective to wait until an actual war and then launch a mass hack when no one is prepared.
I think it's quite likely to be Russians, but not, like, Russia as a state actor, but just some bored individual Russians or some small group who just wants to stick it to the West in some random way. Since 1) it really serves no purpose to Russia in its current situation, if anything this kind of news is virtually always spun by Western media and politicians as "this is why we are in deadly danger and need to give more money and weapons to Ukraine"; 2) it sounds like there really are just tons of vulnerable freezers exposed online and trivial to hack. I don't know about US military, but in Finland and Sweden the targets apparently were mostly quite small stores, who cannot be reasonably expected to be great at IT security.
They are effectively in a proxy war in the west already, one that they can't afford to lose anymore. There might not even be a next time, so they may as well play their cards now.
Man, you are fascinating. I mean this as a compliment, but reading you; It's like looking through a warped mirror into the most motivated possible interpretation of current events possible.
I'm genuinely curious what your information/media diet is? What do you read? Do you put your own spin on the news, or do you have outlets that curate the narrative for you?
I'm not going to be so crass as imply your don't earnestly believe this stuff, I'm just curious how you get to where you are.
And I guess my question to you is: what are you doing with your web-surfing time? All the information I post can be found on mainstream media sites. I read dozens of news sites every day (most contemporary journalism is pretty awful, but if you read a few different sources on a story, you can generally get a good picture of what's going on). As for my interpretation, that's just simple deductive reasoning.
Plus, I try to keep up with the latest technical and scientific advancements.
"In solving a problem of this sort, the grand thing is to be able to reason backwards. That is a very useful accomplishment and a very easy one, but people do not practice it much.” —Sherlock Holmes in _A Study in Scarlet_ by Arthur Conan Doyle.
I'm genuinely amazed at how ill-informed of current events most people are (see links below). Be that as it may, when investigating a crime, you have to ask, *cui bono?* When looking for a suspect, it boils down to who has the means, motive, and opportunity.
While any high school student could probably launch such an attack using AI agents developed by OpenAI, Anthropic, and Google (and possibly others), targeting military bases and then European infrastructure seems a little too focused for teen hackers.
A radical environmental or animal rights organization doesn't fit the bill as a suspect, because the attackers initially focused on US military targets rather than the corporate logistics chain.
I've explained why I doubt NK was involved. We can probably take China off the list, as well, because they're trying to improve their economic ties to Europe. Fucking up the Euro economies would be counterproductive if they want to make money in the short term and long term.
Although I wouldn't put this past someone trying to manipulate the futures markets, AFAIK we didn't see any suspicious puts or calls before these incidents. And hitting the US military wouldn't be an effective way to start you market manipulation (unless you want to depress the stocks of freezer manufacturers).
Who would like to stick it to both the US and Europe? Answer: Russia. They have the means, motive, and the technical resources.
Countries affected: France, Belgium, Luxembourg, Sweden, Estonia, Netherlands, Finland, France, and Ireland (which I grant you is odd because they've been studiously neutral toward Russia to keep their Oligarch-funded aluminum plant making a profit).
Seems really bad for both interpretability and keeping takeoff slow. Not a good sign at all.
(Or more specifically: dangerous from an interpretability standpoint because a human won't be able to grasp any subtler implications of this "Weight-Space Programming" - we'll have to trust the models not to put in any fun backdoors or Manchurian Candidate style triggers for undesirable behaviors.)
The Qwen weight editing experiment is overhyped. Astra ran multiple forward passes of Qwen trying to solve math problems and found a specific bug - that Qwen got stuck in reasoning loops. It then isolated the weights responsible, then modified them to tune down this behavior. Then Qwen was able to solve the problems posed. You couldn't use this technique to create a new capability de novo. Case in point, when they tried to improve a different open model (Gemma) which didn't have the same looping bug, they were not able to.
And also, a human interpretability researcher could have done the same thing (and similar papers already exist which Astra no doubt trained on).
Some of the other results are interesting (the breakout and diffusion model ones), but they are confined to very tiny models. It's not that surprising that tiny models are more interpretable. And consider this - Astra is in the range of trillions of parameters with a context length of 1 million tokens. Without some drastic change, it's not going to be possible for a model to interpret it's own weights.
I'm generally a bit skeptical about expansive claims of the future abilities of AI, but my mental model would not really have suggested this was possible. If this post is accurate, it's basically like saying it can open up someone's brain and rewire it to not have dyslexia just by looking at the neurons. Nobody trained LLMs to do it, so far as I know, and it's not as though there's a lot of examples in its training data of humans doing an equivalent task, because they can't.
Also if LLMs are literally good enough to directly patch themselves it seems like they can do *some* forms of recursive self-improvement a lot more rapidly than I would have guessed.
I don't know what I should expect to happen next, but I feel way more uncertain about it after seeing this.
I recently watched an interview about how people self-report thinking "in words" or "in images" etc. and how difficult it is to actually get people to answer questions like "what words specifically were you using when you had that thought?"
I haven't dug too far into this, but the topic and the anecdote of a case at the end was very interesting:
"Dr Russell Hurlburt is a psychologist, researcher, and the leading expert in the study of inner experience.
We all assume we know what’s happening inside our own minds. But what if the way you describe your inner thoughts—your inner monologue, mental images, emotions—isn’t accurate? What if you’ve spent your entire life mischaracterizing your own experience? Russell’s research suggests that most people don’t truly understand how their minds work.
Expect to learn why you might not actually have an inner monologue, the five core phenomena of inner experience, why most psychological self-reports are unreliable, the surprising differences in how people experience thought, how experience sampling can reveal the hidden structure of your mind, whether meditation really increases self-awareness, and the astonishing story of a woman whose perception of reality was forever altered by this method.
This episode will make you question everything you thought you knew about your own mind."
Would you say you also often think in neither? When I think of a busy train timetable vs an infrequent one, I get a concept in my head along with the words. It’s expressed more like an image than like words, most of its information hasn’t been unpacked into words. And also, I visualise it partly in terms of my timetable app. But the words and numbers aren’t filled in in the image, so the image doesn’t actually contain enough information to say how frequent the trains are.
There is possibly something that is neither pictures nor words in between sentences. Some hard to describe process that happens and then the inner voice assembles a sentence.
The parts of my thinking I am aware of are like inner monologue, with occasional mental images (which are not like photos, more like geometric shapes or graphs), but I am aware that I am unaware of most that happens, because the thoughts just come "too ready".
Actually, I sometimes feel like an LLM, that I do not have thoughts prepared in advance, but they kinda crystallize as I speak, by completing the usual patterns. Although it is plausible that some invisible process decides the goal I am getting to, and then an LLM-like process generates the right words to get there.
I wonder if we could get an "objective" answer using brain scans, where people would solve some problems and the scientists would measure which parts of their brains activate.
Does speaking crystallize the thoughts for you? Sounds like when your ideas are more abstract, you can speak them into a workable form (like an LLM). Can you write down or speak your inner monologue too? How/what do you write?
When I’m speaking, I describe it as a vague impression of my actual thoughts, I can’t recreate the inner monologue unless I’m alone. I speak in scripts, so I can’t “think out loud.” I can abstract out loud, but I have to vaguely describe a loose idea, and it’s not often pretty. I write my thoughts.
When I talk to other people, I think much more clearly than when it's just in my head. Even talking to myself aloud achieves part of the effect. I guess the effect is in having to finish the sentences, and then hearing what I just said. When just thinking to myself, it is easy to get distracted and lose the chain of thought. I guess it's as if I think in multiple threads at the same time -- when speaking, I have to choose one and finish it; when thinking, I kinda start all of them, but often don't finish any, because I get distracted by something else. (When I started writing this paragraph, I didn't know how it will actually end. So in this process I figured out something.)
> Can you write down or speak your inner monologue too?
Nope, too fast, too chaotic, incomplete sentences with some feelings and images... The process of writing or speaking is much slower.
> How/what do you write?
When I write something, I first imagine talking to someone (one sentence at a time, or maybe just a part of sentence), then I write it down.
I am generally suspicious about describing my own thoughts, because the mere fact that I observe them already changes them. The "processing power" spent on observing the thoughts is not spent on thinking them; when I try to observe my thoughts there are fewer of them.
Trying to find examples of visualizations... e.g. when I imagine doing math, such as multiplying matrices, it's as if a very-low-quality movie plays in my head, depicting what would be done on paper. It is very low on details, gray, just a shape here, a shape there, and a movement from here to there.
Interesting on every point, thanks for opening up your brain. Sounds similar to an ENTP friend of mine, if you know anything about the MBTI. She says her brain is chaotic and associative and full of abstract ideas, but they come out logical and sequential when she speaks. She doesn’t know what she thinks until she talks it out. Both processes happen in my brain at the same time. I can feel the subconscious, associative, ideas-generating portion driving my clear inner monologue.
To write, you imagine talking to someone. So all of your writing is geared towards and therefore dependent on the specific audience, and that’s kept in mind the entire time? What if you don’t know your audience? Or is it just that you imagine anyone, like talking out loud?
Interesting that you visualize math. Were you able to show your work in school then?
When I first encountered the question “Do you think in words, or images?” I sat for several minutes, trying to figure out the answer. I became more perplexed the longer I took: surely I should know the answer? Surely it should be apparent? But it wasn’t. I kept waiting for words, but there were no words. I kept waiting for images, but there were no images.
Eventually I decided that I mostly think in words, but since I was trying so hard to “listen” to my thoughts, I wasn’t thinking in words about anything. I was just thinking about observing, and then had nothing to observe except my own observation.
I can flip between modes and I imagine most other people can too.
If I try to picture my walk home, I don't need words. While writing this sentence, I don't need images. When imagining how a rotated object would feel, I think mostly with touch and to a lesser extent images. When thinking about maths, I don't strictly speaking need either, but I can use auditory and visual processing to help me remember and process things.
I'd be surprised if most other people didn't do this to some extent.
> I kept waiting for words, but there were no words. I kept waiting for images, but there were no images.
Correct.
> Eventually I decided that I mostly think in words
But... you don't. You were aware of that. Just because someone asks you "is your favorite color purple, or is it green?" doesn't mean you have to actually believe that it must be one of those two things.
Why should it be that we think in either words or images? I believe we think in thoughts. Clearly at times there are words and whole sentences involved, also images and movies. But the main medium is thoughts. Maybe sort of analogous to AI's being able to understand images and also strings of words, but also being able to communicate with each other via a third medium, "neuralese." Thoughts are our internal neuralese. Or internal vectors.
As an Intuiter, I know i just DO things, I observe, analyze and course-correct without my conscious participation. Sometimes it's good, sometimes its bad.
Definitely better than the midwit "instructions are gawd" though.
I'm very interested in it for myself and for trying to understand others. (Unsymbolized thinking: I'd call that abstract thinking, thinking more in the subconscious with ideas rather than literal words.) I've been asking my 8 year old a lot what happens in her mind. My daughter has a lot of visions, inner seeing, that she then creates with. She takes her visions and puts them into the real world via representations. She also is very aware of sensory experiences. But she describes an inner voice sometimes, too. I ask her often, were you feeling, or thinking, or noticing, or imagining/seeing something? Mine is FULL of my inner voice, and then vague complex ideas underneath that I try to pull up into my conscious mind, or write them down to give them concrete form. I have emotions sometimes that come from thoughts, and occasionally emotions where I can't trace the thought easily, because it happened in the distant or even not so distant past.
The lady they discussed seems to have had those emotions and visions based on experiences she had that she couldn't store away, so they kept coming up in her subconscious and spilling over into her conscious mind, hijacking it with its intensity. She seemed to try to keep her brain occupied with benign content so that the disturbing wouldn't take over. Once she was brave enough to face the thoughts, feel the feelings, name the images, then it could be stored away and free up space for other productive mind work.
It's interesting to ask kids these types of questions, since they have sometimes less flowery language and just say the straightforward thing without as much concern for connotation or nuance.
I don't have a strong opinion on my own internal experience (which predates this video or idea of being a poor historian of my own experiences), but it's been a few weeks since seeing this interview, and my best description of how I mediate my thoughts is heavily framed around where i am in space relative to a room or a person or an object, with maybe a word or two that is a bit like an index, and something like a polaroid of what i was looking at when the thought occured or was noticed.
That has a lot more information than what I would have said previously, "I'm a visual thinker", but I just don't know if or how much i can believe that this is a true reflection or a confabulation.
I think being able to evaluate this favors a metacognitive aspect that some people are drawn to and some aren’t. The self awareness isn’t native to everyone. I relate to the fact that what’s happening in the mind depends on the room and who’s around, and the task. The mental state can be manipulated to improve a task, for example. Right now, I’m having so many different ideas that I could write down here, the trouble is giving concrete form to any of them at the expense of the others. One line of thought is that extroversion vs introversion plays a role. Another is describing my own lack of internal monologue as soon as I begin speaking or interacting with someone external. The inner voice goes silent. Another is that I also had the experience of a vision or intrusive vision while driving, which seemed like an unacknowledged fear popping up that I quickly tried to push away. I can pull up visions from memory or imagine visions in an attempt to predict, but it’s not there often. The inner monologue is dominant for sure when I’m alone or silent.
I wonder how many people here give their thoughts form through writing, since most are great writers? Or have they spoken the thoughts first to give them form and then write them down? I can write my thoughts, the exact inner monologue, and also it’s a way to force the abstract idea into concrete form. I have thoughts and then ideas on top of (or underneath depending on how you want to look at it) the thoughts. (Sorry for this rambling, I could go on and on and on…) My thoughts are word based, text and hearing. Are your thoughts visions that you’re describing? How do you write?
I literally say my thoughts to myself. When I'm tired and annoyed I have the problem of saying inner thoughts too loudly, so other people can hear them.
What if instead of me being wrong, it's the other people who keep wanting me to be?
I think that's pretty interesting for a self-described "yug gnirob"! It also seems to be in line with the general initial reaction that many have when they're told that they may not be as much of an "inner speaker" as they think.
I'm more interested in that part myself; namely why there is such a resistance to the possibility that someone's inner experience might be more nuanced and complex than it's given credit for.
I was a lot less sure of my own interpretation of my own internal experience, even before following this interview, so maybe I'm just primed to be more accepting of the thesis.
Brian Kohberger’s guilty plea in the 2022 University of Idaho massacre meant that justice was served without going through the expense and unpredictability of a trial, but it also may have deprived us of a highly publicized hearing on the Incel crisis and the extent if any it may manifest itself in violence. This is not certain of course, for while he was likely an Incel his defense at trial could well have involved other things.
I don't know anything about that specific case and so can't comment on it, but this is something I notice somewhat frequently: a specific crime or tort takes place which is deeply connected to some recent development or untested legal theory, but the involved parties settle/plea-bargain instead of letting it work its way through the courts.
This is normally the right call for everyone who is actually involved in the case, but it seems to run counter to the public interest. I'm not sure if there's a fix for this within the common-law system. It could just be that if the parties settle, it ought to be the domain of investigative reporters and legislators, not prosecutors/plaintiffs and judges. Perhaps the government or a charity could subsidize the legal fees of parties in cases of public interest, on the condition that they do not settle the case. This seems a bit far-fetched and at any rate politically unwise: by doing that, you will necessarily be paying the legal fees of many unsympathetic (legal) persons.
I noticed this too. It's outright infuriating in consumer rights cases - or even worse, civil rights - where the plaintiffs get a last minute settlement offer when it's all but certain they're going to prevail, which they obviously take, and I'm happy for them, but it deprives the rest of the country of a binding precedent that the illegal thing is actually illegal.
I think there should be a rule that judges must still make a verdict on the meaning of law even if there's no court case anymore. No idea what it could look like in practice, though.
One possibility would be to allow judicial discretion on whether to accept no-admission-of-fault settlements. The problem with this is that you're essentially using the plaintiffs as unpaid ~prosecutors. They'd be justifiably upset.
What if there was some sort of right-of-refusal? When there's a settlement, then (subject to some sort of public-interest test), the state can purchase the tort. The plaintiffs are paid the settlement out of public coffers, and then the state can hire lawyers (maybe even the same ones) to continue to pursue the case.
One problem with this though is that sometimes the *government* settles cases when they really should be pursued. There would have to be some guardrails in place to prevent them from settling the tort they purchased.
Don't get me wrong, but I really dislike both ideas. The first one makes the plaintiffs straight up worse off. The second is very easy to abuse and funnel out public funds, and unlikely to be used much legitimately due to huge up front cost for the government and high risk of losing. Also probably unconstitutional due to standing.
I was thinking more like, the settlement is valid, the case is closed, there's nothing more to worry about for either party, but the judge then proceeds to decide the legal arguments that were previously raised by both parties, without the parties' involvement. The legal precedent it sets would apply to future cases, but not the current one.
In the Idaho case Brian Kohberger faced the death penalty, so a guilty plea in return for a life sentence made a lot of sense. Especially since the evidence against him was strong and for all its societal interest an Incel-rage defense probably wouldn’t work.
For anyone interested in housing and urbanism in Los Angeles, I wrote a review of the evidence on one of the biggest fighting points in recent years, a high-value transfer tax that's turned into a bit of a partisan fight. I came into the essay with some sympathy for (more thoughtful versions of) the housing leftist position, but broadly find that the YIMBY take on the tax is directionally correct (which I think the readers here will mostly be happy with).
I was inspired in part by similar empirical reviews Scott's written, going over a literature and trying to synthesize a birds-eye view of what the evidence can justify by synthesizing the statistical results with general common sense-type arguments.
A cyberattack affected freezers in a few European stores last night. Lots of food will have to be thrown out.
Naive question, but what's the reason for not just having 1990s offline freezers? Are there enough benefits to having them online that it's worth the risk of an attack? (And yes, you can file this one under "Man doesn't understand why technology didn't stop when he was a teenager". But still, why not?)
I read the manual for one of the refrigerator models that was hacked. The reason why the refrigerators can be set to defrost continuously via the internet is that (according to the manual) you can optimize energy usage by tweaking the defrost schedule. Presumably a chain doesn’t want to train employees at every location about the intricacies of defrost cycle scheduling. They would rather have a few employees administer the defrost cycles for all their units from a central location.
A residential analog: my house's geothermal system comes by default with an internet connection and web dashboard that lets you do the usual smart thermostat stuff (change temperature remotely, chart energy usage) and also lets the installer check on the unit without a site visit. The internet connected part is a separate device, but it has a direct data connection to the unit - unlike e.g. a Google Nest, which connects over basic thermostat wiring.
From the marketing materials (https://www.waterfurnace.com/residential/products/accessories/symphony-home-energy-management), the manufacturer sounds rather proud of it, and I think they found it odd that I opted out. The benefits seem minor to me, though; their alternative "dumb" thermostat still has enough energy usage data for my purposes. I have no interest in joining Mirai-style botnets or losing heat on the outbreak of war.
I don't have good data on the cost of the old system (propane), but my estimate is that the geothermal system breaks even at current electricity & propane costs, applying my mortgage interest rate to the upfront cost. However, that's including the effect of a 30% federal tax credit that was repealed last year. I forget if that estimate also includes the fact that counterfactually, I would have needed a new propane furnace instead.
I still consider it worthwhile for non-financial reasons: climate friendliness, not worrying about propane deliveries or gas lines in my house, and reliability (typical lifespan for a geothermal unit is 20-30 years, vs 10-20 for a furnace).
The main difficulty was landscaping - a horizontal loop field tears up a good deal of yard, including a trench leading right up to your house. So expect to spend some time (or further money) making it look nice again. On the other hand, it's your chance to do something new, like replace an invasive meadow with native wildflowers. There are also vertical or diagonal loop fields, which are less disruptive and require less land, but at higher cost.
Another question: after the French supermarket chain got hacked and it made national news, did no one do anything ? Did they not even check if their own fridges are running the same software ? I don't understand what goes through people's heads at times like this.
Assume you only have hired illiterates. and a 6 figure CEO. "That was just them! Lucky us! Surely our software will fix itself, just like my browser does!"
Based on my experience doing electrical work in commercial/retail environments I'd say that lots of the stuff you'd expect to operate with a light switch or a thermostat in your house (such as lights and the ambient temperature) is controlled in commercial/retail environments with big relay banks, and these are controlled by computers.
Where the susceptibility to cyberattack comes in is that the computers controlling the relays are not necessarily located onsite; they might be run out of a central headquarters somewhere.
I don't know why companies do it this way, but I suspect it's partly to eliminate human error (e.g. one manager consistently forgetting to turn off the lights, or setting the freezers way too cold), partly to make things more standard and consistent across locations, partly so maintenance requirements ("one of your exhaust fans has been drawing more amps than normal") can be communicated up the chain, and partly for some other cost-cutting scheme they've come up with that I haven't thought of.
Personally? I'm not a CEO. In my direct family? Over a hundred. My father in law has personally managed four or five retail outlets (Ollies to Ames), including the fun story about the guy who duct-taped a knife wound to the neck and came to work that day. Worked the whole shift, too.
When I say "they're digging the bottom of the barrel as a philosophy" that's words taken straight from someone who knew Sam Walton and Crazy Eddie, and is a financial analyst.
I'm sure to you I must seem like an illiterate hob. Then again, I do know what the word dunder means. Literacy used to mean something, back when we had more of a shared culture.
I don't mean to insult. I mean to discuss a race to the bottom in terms of hiring standards. Use of the word illiterate is meant to both suggest lack of intelligence and culture, and the deliberate watering down of our society -- as the word literacy was the first concept to be watered down.
For remote monitoring and compliance tracking. There are laws about cold chain storage which means there has to be a continuous record of what temperature every single refrigerated item was stored at all times. And it's bidirectional so that if there needs to be a software patch, it can be transmitted back.
Also, a quick reminder that supermarkets run on a 1-3% margin. So even a trivial cost savings (not having to send a technician to go around and grab logs off of fridges and freezers, and being able to fix things remotely) can wildly change whether a grocery can continue to operate.
I'm just going to note that remote monitoring doesn't actually mean you CAN'T have a competent store manager who calls home if there's a power outage, and can move the valuables into appropriate cold storage.
Apart from blackguardism, I don't know why the hackers are doing this or what they are trying to achieve, and I don't think any group has of yet claimed responsibility.
They're doing it for the lulz, most likely. Honestly, in my youth, if I saw one supermarket chain fall to a blindingly obvious exploit; and a couple weeks later another one fell for the exact same trick... then yeah, I'd be sorely tempted to just hack every supermarket in sight. Maybe I wouldn't actually destroy them all, but yeah, I'd be tempted to do it just to see if I could.
And that was back in the day, when the Flinstones roamed the Earth; today, I bet I could even get paid for it...
Honestly, there's a level where I'm kind of thankful. At least if there are people who're doing this on a level where it makes the news, but not on the level where it's causing national or international catastrophe, it increases the likelihood that people will take precautions that make it less likely that it *will* happen at that scale.
If people don't start taking appropriate precautions, there may simply not be any hope for us.
The eminence grise of the Polish President's political party (Kaczynski) also believes that the Prime Minister of Poland had Kaczynski's brother murdered for political gain. Saying there is bad blood between them is an understatement.
I endorse this summary:
"""
There were elections in 2015 where the right swept to power in both the legislature and executive. The new legislature was to be inaugurated in October of that year. 3 of Poland's 15 constitutional court judges had their terms end before the inauguration, whereas 2 more had their terms end a month after the new legislature was to convene. The losing party, seeing an opportunity to pack the court before they lost power, decided to appoint judges to fill all 5 vacancies, even though they only had the power to appoint 3.
The President, who belonged to the right and was himself sworn in months before, refused to swear in the new judges before the new legislature convened—a legal requirement to take office. When the parliament convened, now under right wing rule, they threw out the 5 appointments and appointed 5 of their own judges. The constitutional court invalidated 3 of these appointments, the ones the previous legislature had the right to appoint, while accepting the other two. Parliament, displeased with this, rammed through legislation forcing the court to seat the 5 justices and allow them to hear cases.
Now, in 2023, the right lost power, upon which the new ruling party appointed ideologically left leaning justices who invalidated all rulings involving the 5 judges appointed in 2015, effectively invalidating most legal decisions made by the court since that time.
Moving forward, in 2025, new presidential elections were held, and the right won, prolonging a split government. The new president refused to swear in new judges that didn't align with him ideologically, effectively paralyzing the court as being sworn in is a legal requirement to take office. To try to circumvent this, new judges have been sworn in in front of parliament as opposed to the president, which is constitutionally dubious and has caused the constitutional court to fracture further, as right wing justices refuse to seat the ones not sworn in by the president.
Which leads us to the events of last night. The PM, in an attempt to break the judicial paralysis, rolled up with police and had the prosecutor general issue warrants to hold the head of the court and his deputy in contempt for refusing to seat the new justices. The President and the right wing establishment see this as a coup, whereas the government and their supporters see it as an attempt to restore law and order in the country.
"""
That penultimate paragraph refers to the debate around the term "wobec", which commonly means "in regards to", but in archaic writing means "in the presence of" (compare how English laws use "shall" vaguely). The PM is claiming that the Judges only need to swear in regards to the President, not in his physical presence.
My takeaway would be to (1) stop having ceremonial rubber-stamps, it is utterly moronic and (2) enforce a single official language with explicit definitions for words to be used in legislation. More modest versions of (2), like stop using archaic language to make laws seems law-ier, also work.
Sure does. The obvious incentive not to call it that is "we are not a banana republic." The pretense of fair and unskinked* elections is very important to the democratic process.
*unskinked: influenced by outside entities that feel that it's their right to fix democratic elections, and democratic countries, so that they will behave "in a matter that comports with the current geopolitical priorities" of the outside entities.
I've got family there, though the insider look is barely more informative. What it sounds like is the President of Poland has a number of powers, on paper, that in practice have always been more ceremonial/rubber-stamp in nature than a real part of the political process. He's trying to use them for real to block Judge appointments. I think. My Polish is pretty bad, but does exist.
The president in Poland is (was?) a simple ceremonial role. Up until Duda, the previous president, most people didn't even know who the president was.
In this case if I understand correctly, new constitutional court judges have to be seen/confirmed by the president. Purely ceremonial, turnkey kind of thing. But because the president is aligned with the previous ruling party, PiS, he's refusing to execute the ceremony, effectively blocking the new lawfully selected judges to being their work. This has never happened before and presents a deadlock.
I haven't been following the politics for a while but my understanding is that PiS has a long track record of technically complying with the law, but working very much against its spirit, and have a penchant for inflammatory, nationalistic rhetoric, so them calling it a coup aligns well with what they've been saying for years, eg. that the current prime monster Tusk is a German agent, that the EU aims to destroy Poland (going so far as to call it the Fourth Partition), etc.
The way it was explained to me, it isn't even a ceremonial confirmation like the UK monarch appointing a Prime Minister. The President just witnesses the oaths of office for the new judges. Which turned into a veto by way of the President refusing to observe the oaths taken by judges he doesn't like.
This time around, the parliamentary majority is refusing to recognize the validity of this and their judges are attempting to act as members of the court and are supported by the Prime Minister. The disputed judges make the difference in which side controls a majority of the court.
Each side has been trying to lock the other side's judges out of the building, and the thing that just happened was the PM sending in the police to back up his party's judges. The President is calling this a coup.
The legal argument for the PM's party locking out some of the President's party's judges is that there's a procedure which (if their majority on the court is real and the law works they way they say it does) allows them to suspend other judges for misbehavior in office.
It seems that putting solar panels into fields is a big win. It probably isn't the only thing which uses available tech, is plausible, and can be tested before it's deployed on a large scale. Is AI useful yet for finding that sort of thing?
It is a big win now because assorted businesses and governments bought in, pushed and subsidised solar installations a decade or two ago, hard enough and for long enough for tech to improve, manufacturing to grow and economies of scale to kick in. I am one of those who benefitted enormously from those early subsidies.
How do we spot more such opportunities early enough to get on board before the point where it becomes completely obvious to everyone how big a win they are?
A history of arguments for independent violence to get good changes, and why those arguments are bad ideas. I don't know for sure that all the history is correct.
******
French philosopher and novelist Bernard-Henri Lévy in The Free Press:
Chaos on the streets of French cities has brought back the seductive notion that violence can set you free. It is a myth—and a recipe for civil war.
Amid the uproar of the citizen demonstrations of recent weeks in France, and under the influence of leading figures in Jean-Luc Mélenchon’s extreme-left party, France Unbowed (La France Insoumise), one theme is making a powerful comeback: the idea of an invisible, insidious, systemic violence supposedly exercised in democratic societies, which it is legitimate to oppose with the visible, direct, frank violence of the riot.
I know this idea well. It was Jean-Paul Sartre’s concept when, in his preface to The Wretched of the Earth by the anticolonialist ideologue Frantz Fanon, Sartre presented the murder of the colonizer as the response to the oppression of the colonial system. It was the idea of the writer Jean Genet in a famous article, “Violence and Brutality,” published in Le Monde in 1977, in which he presented the terrorism of the Baader-Meinhof gang in Germany as a necessity dictated by the anti-imperialist struggle. And it was already the idea of Georges Sorel who, at the very beginning of the 20th century in his Reflections on Violence, opposed the force of the “bourgeois state” with a “proletarian violence” endowed with supposedly regenerative virtues.
I thought this idea had been discredited by the post-May 1968 intellectual movement, the critique of totalitarianism by the New Philosophers movement which I led, and the reckoning with the fascist, Nazi, and communist catastrophes. But apparently not. And so, we must start all over again.
The first thing wrong with this notion of “counterviolence” is that it never strikes at the heart of the hated regime, but instead at the most exposed and most vulnerable people. The police officer who earns 2,000 euros a month, works night shifts, and protects, at the risk of his life, property, people, and the right to demonstrate. The shopkeeper who earns his gross profit on Saturdays and whose ransacked store was his only capital. The owner of the burned-out car who already struggled to afford a tank of gas and who must now find the money to replace the vehicle or face the obstacle course of dealing with insurance companies. The riot claims to strike at “the system,” but it falls upon women and men who are neither its masters nor its beneficiaries, but its humblest subjects. It claims to avenge the humiliated, but it adds to their humiliation fear, ruin, and the feeling of having been abandoned.
The second thing wrong with the violence that France Insoumise tells us is “sometimes legitimate” is that it does not destroy “systemic” violence but offers it the opportunity to reveal itself and, in revealing itself, to harden and spread virulently.
Are the United States of the MAGA movement and the neo-Democrats consumed by hatred of Israel in such a different situation than the one we face in France?
This is the story of the Lyon silk workers’ revolt of 1834, when troops opened fire on a crowd. It is the story of the revolutionary days of June 1848, crushed in blood. It is the story of the Paris Commune, drowned in a repression that left thousands dead. And when the insurrection triumphs, history teaches us that the regime born of the riot is, everywhere in the world, more terrible still than the one it overthrew: Liberating violence does not abolish systemic violence; it transfers it to other and worse hands. It does not abolish domination; it becomes the founding principle of a new power which, acting in the name of the people and history, now believes itself entitled to anything. Counterviolence promised emancipation, but it gives birth to dictatorship.
Ultimately, the philosophical choice is the one formulated by Thomas Hobbes, Jean-Jacques Rousseau, and, with them, the theorists of the social contract. Of course, one takes a risk when one renounces taking justice into one’s own hands and entrusts the public authorities with the monopoly on legitimate force. And that is precisely why it is so important that states be held accountable for this monopoly, for the way they exercise it, and for the abuses it permits. But the alternative is the jungle. It is man becoming a wolf to man once again. It is the law giving way to the whims and fevered passions of each individual. It is a world without rules, where everyone arrogates to himself the right to punish and to strike. Today, in France, militias. Tomorrow, between a far right and a far left equally inflamed, a civil war. Are the United States of the MAGA movement and the neo-Democrats consumed by hatred of Israel in such a different situation than the one we face in France?
This is what Sartre began to glimpse toward the end of his life when, in dialogue with his young companion Benny Lévy, the former leader of the French Maoists who later became a master of the Talmud, he renounced his belief in the redemption of societies through terror. And this is what the intellectuals of my generation came to understand when they realized that the same Jean Genet who supported the Baader gang also rediscovered, in his final writings glorifying the Palestinian cause, his youthful fascination with the soldiers of the Wehrmacht and the SS. And this is what I theorized nearly half a century ago, in L’Idéologie française, about the philosopher Georges Sorel who, having begun as a fervent apologist for revolutionary violence, ended up as a founding father, alongside young followers of Charles Maurras, of the all-too-little-known “Cercle Proudhon,” where the first red-brown synthesis in Europe took shape.
This notion of visible violence answering invisible violence is decidedly too dangerous for our societies. It’s true that systemic violence exists. But its opposite is not the violence of the riot. It is the rule of law. It is justice. It is what allows a society to identify violence committed by public authorities, establish responsibility for it, and redress the wrongs it inflicts. It is the possibility for each individual to challenge power without becoming, in turn, a miniature power, a petty judge, an unmasked avenger. Such is the lesson of the century for today’s rebellious generations. Will they heed it? I hope so. For if they do not, civilization itself will be exposed to upheavals on an unprecedented scale.
I agree in essence with these arguments. However, I will risk, as somebody whose background is not French, except insofar as any European intellectual is necessarily a little bit French, the suggestion that France's current troubles are perhaps less significant than might at first appear - that they are in effect, more street drama than revolution. Of course France has seen violent revolution time and again, with the streets running red with blood; I would even look back as far as the Fronde of 1648-53, as much a revolution as a civil war. But as Bernard-Henri Lévy notes, those long-past events, from the Lyon silk workers’ revolt of 1834 to the days of June 1848 and the crushing of the Paris Commune, were triggered either by an extreme reactionary provocation, or else occurred in response to a serious and unacceptable assault on popular freedom. Today's events, like those of the Paris May of 1968, are a response to irritants, to dissatisfactions, to the inevitable arrogance of the country's comfortable establishment - not to massacres. Like May '68 (which did not, despite all the drama, after all dislodge De Gaulle), and without minimising the destruction, I suspect the current wave of revolutionary theatre will eventually subside, leaving behind a new legend for the students of a subsequent generation, together with a parcel of undoubtedly overdue reforms.
I'm not. “What did y'all think decolonization meant? vibes? papers? essays?" Rape is perfectly acceptable to the MeToo crowd, provided the right people are being raped instead of the wrong people.
The recent developments in the AI interpretability got me interested, so I want to ask people in the field, assuming some of them are likely lurking around here. How possible it is for someone far outside the field to get into it, realistically, within couple of years?
I don't have the necessary prerequisites, and it's probably too slow to have meaningfull impact on anything, but the chance is probably not zero.
I'm already looking into it myself, of course, but sometimes it helps to have an outside perspective.
I'm not really looking for the employment necessarily, if I can contribute to the field meaningfully as an outsider, with zero monetary gain, that would be worth the effort.
This is probably not going to change your response, iirc your are firmly in the "AI is a bubble about to burst" camp. But even then, this should provide enough context for other people who might see this subthread.
I feel the same way but am not a stem worker really, though I do have a math degree. I've focused my efforts on politely but ceaselessly hounding my congressional reps on these issues. Energy policy is my focus right now.
Heads-up: this is an AI post, scroll past it if sick to death of the subject!
Anyway, recently I tried Opus 5.5 on 2 different real-life tasks with some mundane applications and 1 a abstract task as a kind of private benchmark. It's not that the results shocked me, but they were not what I had expected.
Exercise 1. Abstract task first: I wanted to see if 1) Opus can learn from its mistakes and perform better and 2) if it can do better based on my generic feedback. Not laser-focused, very specific feedback, but more like general advice and heuristics.
I decided I would pick a dataset A of 30 items. Have Opus do a task for each item. Then I'd score it and ask Opus to draw conclusions from the results and to write some guidelines for itself on how to do better. Then I'd give Opus dataset B and ask it to do the same task, armed with the lessons (hopefully) learned. Then I'd give the task and the dataset B to the totally naive instance of Opus, that has not had an opportunity to learn from dataset A, and compare the scores. Would an experienced!Opus beat naive!Opus?
And then would additionally give Opus my own, custom generic instructions, and evaluate custom_instruction!Opus, so that I would have 3 different scores: naive, experiences and custom_instrustions.
Oh, and the task? I needed something that 1) Opus can do, but not perfectly and 2) I can do better, so that I can score and give meaningful advice and 3) is objective. Other than those requirements, it could be anything. There isn't much that ticks all 3 boxes, but there's one thing I can do better than almost anybody: identifying snail species from my region based on photos, so this is what we went with.
I assembled a hard, but fair dataset A and fed it to Opus. For a moment I was warried it would nail every item perfectly, and the test would be ruined, but as it turned out, Opus was not great at it, and only got 12/30 photos right. Now, this is not a very meaningful or interesting observation that doesn't tell us much about Opuses powers, because everybody knows that it could easily have been trained to get it 100% right, it's a largely solved problem. No, the real test was to see if Opus can learn from its errors.
So, I asked Opus to take a look at what went wrong, write instructions for itself to do better, address its failure modes, do some research etc. Opus did, and armed with these lessons took on dataset B, and got 14/30. Well, that still sucks, but it appears to be more than 12/30? But these are different datasets, need to see how a naive Opus does on dataset B for a fair comparison.
Weirdly, naive Opus also scored 14/30 on set B. And what if I add my custom instructions? I wrote as good of a set of generic tips and heuristics as I could and fed it to Opus. Armed with my priceless advice, Opus tackled set B again... and got 14/30.
To summarize: Opus totally failed to learn from its mistakes, and got the exact same frigging score of 14/30, whether fresh out of the box, with the benefit of some practice or with my advice.
Exercise 2. I had a digital drawing project going, and I needed to do the tiresome part I mentally call "blobbing". It's where you have finished lineart and need to fill every single silhouette with solid color so that you can lock it and paint it. Whatever, if you've done digital art, you know what I mean, and if you haven't, then it's basically what a kid does with a coloring book. Filling areas with solid color, only the lines are looser and not always completely closed. This task requires understanding of what has been drawn, no, you can't automate it with a Photoshop function. Photoshop tricks fail on complicated and messy lineart. Depending on the design, this can take from a couple of minutes to long, long hours. I was looking at something like 3 hours of drudgery. And one morning I awoke with the most wonderful idea: "hey, why not make my friend Opus do it!"
So I did. Opus also had an idea that he can automate the whole thing with a simple script of some sort, and learned that no, it didn't work like that, it had to be done largely manually.
We went back and forth refining the process and trying things.
Results: mixed. Opuses output I got was not perfect and I had to adjust it on my own, but it was good enough that it saved me a couple of hours. I used something like 30 minutes instructing Opus and 30 minutes fixing mistakes, instead of something around 3 hours of blobbing everything by hand.
Exercise 3. At work, using Opus to write user documentation. I was initially skeptical this would ever work. The reason I can write user documentation, is that I have experience using our software, and I know the context and what it is used for. Knowledge, none of which could be easily transferred to another human or AI, short of writing it all down. Yes, Opus can read the source code. But can you read unfamiliar source code, line by line, function by function, and then describe what it does, from the point of view of a user? Not just what happens, but why and what for? Could that be done at all, even by a very smart human? I had doubts.
Still, I was ordered to give it a try. In my head, I sketched up some approaches on how to handle it, strategies that involved agents explaining things to one another and so on. I tried the naive approach first, giving Opus a place in course code, my own documentation as an example and telling it: "write more like this about that!"
Weirdly, it worked amazingly well? Somehow, Opus constructed a perfectly accurate picture of what the code is going to do, as seen by the user. He even found some bugs, bizarre interactions I never knew about and somehow figured out a bunch of things of the uses and purpose of the stuff out system does. The results were not perfect, and some of it needed to be fixed or altogether re-done, but overall I was very impressed. With the help of Opus, my document-writing task would now be 2x as fast, and much more relaxed.
So that leaves my with 1 failure, 1 partial success and 1 instance of very good performance.
An older case of an intellectually disabled man who was convicted of a bajillion crimes in Denmark is being revisited. Turns out, when the police would take him on drives around the area to make him point out where "he" comitted arson, he would confess to crimes that never happened, buildings never burned. Law enforcement was just happy to have someone to blame for so many crimes. He loved going on drives with the police and considered them his friends. He was sad when he was convicted, only because this meant no more drives.
Anyway, his intelligence was usually described as "of a 6 to 8 year old child". It just struck me how messy and inaccurate this seems, trying to compare intelligence and neurodevelopment in this way. A normal 6 year old would surely know that they were in trouble and to lie to the authorities (eg. the James Bulger case). Just like when people say such and such animal is like a 3 year old human - well maybe, but also not at all.
> Anyway, his intelligence was usually described as "of a 6 to 8 year old child". It just struck me how messy and inaccurate this seems, trying to compare intelligence and neurodevelopment in this way.
It descends from the original branding of IQ tests as being an intelligence "quotient", your "mental age" divided by your actual age. Actual age exists, but mental age doesn't.
In an ill-advised procrastination from working on my thesis, I took the prevalence of blood types in my country and worked out what percentage of people they can accept blood from and what percentage of people can accept blood from them.
Blood type Frequency Can accept from Can accept (%) Can give (%) Accept:Give
O+ 40% O-,O+ 49% 81% 0.60
O- 9% O- 9% 100% 0.09
A+ 31% A+,A-,O+,O- 87% 33% 2.64
A- 7% A-,O- 16% 41% 0.39
B+ 8% B+,B-,O+,O- 59% 10% 5.90
B- 2% B-,O- 11% 13% 0.85
AB+ 2% AB+,AB-,A+,A-,B+,B-,O+,O- 100% 2% 50.00
AB- 1% AB-,A-,B-,O- 19% 3% 6.33
Obviously individual factors are going to be way more determinative of whether a given individual should/can donate, and of course stock levels are going to be a big impact, and also I make no claims about the relative weights of blood donation vs other activities regarding good/hour, but I thought this might help clarify the picture somewhat beyond the messaging we get of "everyone who can donate should!"
(If there's a better way to give this information, please let me know)
Your post is not addressing one point: if for example all potential A+ donors see that they can only give to 33% and stop donating, then there will be no A+ type blood, and all A+ receivers now need to receive blood from A-,O+,O- donors. This would make the situation much worse, because this blood is missing for the A-,O+,O- receivers who can only accept those.
I think in practice this is relevant. I have heard (but never checked) that doctors try to match blood types if possible, and that this is one of the reasons.
I think for almost every blood type except possibly some very exotic ones it's a good thing to donate. So the best policy is: "ignore your blood type and just donate".
I generally find universalisation arguments compelling, but I think this form proves too much. If we say that "apples are good for you" and literally everyone acts on it within a relatively short time window, then that will lead to problems as well. If literally everyone acted on the governmental advice to have X amount of water or food on hand in case of outages/shortages/emergencies, then we'd have panic buying, massive market distortions etc.
If we're talking mass messaging, then yes, simpler is better - the consequences of having too much blood are relatively minor (wasted resources to harvest it) while the consequences of having not enough can be catastrophic (a person dies). But I think the context of this post is important - it's being posted in the comment thread of a community of generally intelligent, generally pro-social people who are generally good at parsing more nuanced messaging and arguments
Also, regarding the "your post doesn't address.." - my post *does* address this point as well:
> Obviously individual factors are going to be way more determinative of whether a given individual should/can donate, *and of course stock levels are going to be a big impact*
If there was a large surge of blood donors, stock levels would reflect this, which in this hypothetical people would react to. Now, you can argue the masses don't really respond well to this kind of nuance, and fine enough, but my post *did* address this, albeit a bit quickly
Well, you didn't say it right away, but the only implicit conclusion I can see in your numbers is that you believe that the value of blood donations is coupled to the percentages that you give. (Otherwise, why give them?)
But I think that this is mostly wrong. The marginal utility of an additional B-donor is not that we have an additional spare B-unit. It's usually that we have an additional 0-unit, because we don't need to waste this unit on donors of type A or B.
That's what I wanted to express with "ignore your blood type and just donate". In most practical situations, the marginal utility is (mostly) uncoupled from the blood type. The marginal utility is mostly an additional 0-unit, regardless of the blood type of the donor.
I think it's pretty obvious that yes, a person with more widely useful blood is a more valuable donor, all else being equal. If we take the extreme case and say only one blood type will donate, we'd obviously want it to be a more widely useful type than a less useful type. It's obviously not a simple 1:1, because as you say there's strategic use etc, but that's mitigating the fundamental dynamics at the margins rather than overall.
In fact I'd say my perspective actually helps with the strategics. If someone is O-, then we want them to donate as much as possible - it's obviously a lot more widely useful than say AB+. Yes, the AB+ allows us to free up a unit of O-, in the case where an AB+ patient is around, but AB+ is also a smaller percentage of the population so there's obviously less value there (ie it's less likely to be that AB+ patient), especially since this hypothetical patient can take more or less anything. And since O- is only ~9% of the population, all else being equal it's going to be a relatively scarce resource, compared to more common but less widely useful blood types like say O+ or A+.
Perhaps this is unfair and more to do with my own insecurities, but I *feel* as if you are worried that a more nuanced exploration of blood type donation than blanket "everyone must donate!", no matter how widespread or cautious or how many caveats are included, is tantamount to saying nobody should donate. Your arguments haven't been "your analysis is incomplete" or "you insufficiently account for replacement/strategic use of typing", but more "people are too stupid to take this level of nuance so you shouldn't say it at all". Your claims that I don't take something into account which I explicitly did is a part of this as well. Again, perhaps this is my issues coming through, but that's the impression I'm getting. If I'm wrong, I apologise for the implication.
Know your audience. When your audience is the general public, "Donate blood!" is a good message. "Donate blood because <Alan Smith's post>" would be a much worse message. Not because your analysis was wrong or your intentions were bad, but because it would be unnecessarily complex.
If we're talking mass messaging, then yes, simpler is better - the consequences of having too much blood are relatively minor (wasted resources to harvest it) while the consequences of having not enough can be catastrophic (a person dies). But I think the context of this post is important - it's being posted in the comment thread of a community of generally intelligent, generally pro-social people who are generally good at parsing more nuanced messaging and arguments. It's also a group who I would be relatively confident are also much more likely to be regular blood donors than the general population for a range of variables, but still - as I said, I just found this an interesting perspective
Hope is the belief in improvement despite some barrier. Traditionally, I think the barrier is expected to be quite large, and likely unsurmountable, but I've seen it used in lesser contexts.
I think for many people, hope is what they do when they don't want to actually put their nose to the grindstone or plan/implement a solution with rigor...but want something to turn out well regardless.
I do not need that kind of hope. I plan, I work, and I accept the consequences pretty well when I fail to.
Hope in the face of a barrier that is daunting *despite* rigor and effort? Yes. I think we all need hope to succeed in those circumstances. Without hope, we will look at the statistical improbability of a situation and give up.
I think there's more to it. It's not the thing to think something can improve, than to hope for it. You can even hope for something that is barely possible.
I was just thinking about hope yesterday, because I saw the word somewhere and I thought, that's a tricky concept that has historically played a huge part in Western culture (it's one of three major virtues in Christianity, after all), yet nobody seems to talk much about it, either for or against.
If I had to define it, I'd say it's an emotional investment in the future that gives meaning to the present. Which is a rather weird orientation towards time and meaning if you think about it - it means the present always remains incomplete!
Hadn’t thought much about it except in a sneering kind of way back in ‘08 when that was like Obama’s one-word unofficial (or official?) motto. So, interesting to give it a fresh look now…
First thought I had was, is being in a bad situation, or a situation you reflexively want to improve, a precondition for hope? My mind certainly goes first to the person who is in dire straits but expects or at least kindles a persistent aspiration that things will be better in the future. But is hope something you could also feel right *after* you’ve accomplished something big, or in the midst of celebrating a victory, etc.? Having asked the question I’m unsure of the answer. If hope can happen any time that makes it more interesting as an emotion, I think.
And, is it an emotion? Or is it more like a mindset one must deliberately cultivate and remind oneself to return to? (A practice, in other words.)
Some people seem naturally inclined to be hopeful, others not.
Are most children naturally hopeful, and do most of them lose this quality as they age and become jaded?
The matter of hope being necessity is important. I recall cliches about survivors of disasters and atrocities saying it was hope that kept them alive, but I don’t know if those kind of statements ever received deeper examination. Was it really hope that kept them alive, or something else? And if it was hope, what did that look like to them? Once they were rescued or “out of the woods” or whatever, what happened to the hope then? Did it just vanish, having done its job, or did the practice of clinging to hope linger after the hope wasn’t “needed”?
With time I think I could sort out my thoughts on some of these questions, but for now it’s just been interesting discovering the questions were there. Thanks for posting.
Hope surely is a tricky concept. I think colloquially, it is meant to indicate that the circumstance is less than ideal. Perhaps not awful, but certainly not where one would like to see themselves. I think there also has to be a degree of difficulty in getting out of that circumstance. A B student trying to become an A student wouldn't make the cut for hope unless he had historically been a C student and it took him a lot to even move into the B tier.
I like the idea of hope applying regardless of circumstance, but anecdotal evidence from social conversations seems to indicate that I'm the odd man out with that perspective.
Scott once wrote an essay about "Things That Are Not Superintelligences", by which he meant companies. Companies may not be AI in that they were superintelligent, but they certainly do seem to employ people with the paperclip maximizer mindset. I'm open to ideas on how to align these people with human values.
This is a sign that Sam is either stupid, or thinks you are stupid. Your computer does not work without local memory; no, you will not be querying the cloud for every RAM read
In this proposed dystopia you don’t have your own computer. You have one of these things https://chippc.com/citrix-thin-client/ or its moral equivalent instead, which does technically have a little RAM I guess, but not much more than is needed for a framebuffer and network stack; and all your actual computation happens in Sam’s data center so he can charge you for it monthly to recoup the cost of buying all the world’s RAM.
So Sam the Scam wants us all to migrate our data into DE CLOUD, DE CLOUD, LOOKIT DE LUBBLY FUFFY CLOUD in data centres where he can scrape all our information to train his AI and lock us into subscriptions forever and ever if we ever want to access our own damn information, then hand over our PC chips to him. For DE GRATER GUD, of course of course.
Sure, Sammy, I'll do that the second you buy a two-bedroom apartment for your three person family in beeyootiful downtown Burbank and hand over your - how many estates is it again? six houses and grounds? - to the people of San Francisco who can then make much better use of all that land for multi-apartment block living. For DE GRATER GUD, of course of course.
Even if this is "chortle, can't you take a joke guys?" and not serious, I'm telling you - the lamp posts are looking better and better to me. And I don't even *want* to run a damn AI model on my own personal computer.
> where he can scrape all our information to train his AI
Conspiracy theory time!
Altman's company, along with a few others, is in a race: can his product start bringing in enough revenue to pay running costs and service debts before his startup runway runs out, the markets decide to put their money elsewhere, and debts come due?
Right now, between [1] and [2], it seems opinions on that are rather more mixed than even a few weeks back, which is a little risky because sufficiently spooked investors can make that sort of thing self-fulfilling.
Now imagine a world in which most people do most of their computing in the cloud instead of locally. In that world, the financial risk is somewhat decoupled from LLM revenue, because there is global scale appetite for other useful things the compute can do; the expensive data centers become an asset instead of a liability.
Oh, and he wants all our DRAM as well, of course. [3]
But it seems quite obvious from the context that he was specifically meaning that it is using DRAM for local inference which is wasteful (arguably true as stated -- 128 GB RAM in your Mac Studio probably isn't going to be loaded as regularly as it could be in an OpenAI server -- of course, you might have perfectly valid reasons to prefer local inference regardless), not just using DRAM in your own computing devices in general. Of course, he really should have known better than to make statements so easily quoted out of context.
When Sam Altman stops gobbling up scarce land in SF with his maniacal house buying, then he can tell the rest of us not to own our own goddamn household PCs. Not a nanosecond before then.
>Of course, he really should have known better than to make statements so easily quoted out of context.
Maybe that is that already the more thoughtful, benign spin for public consumption?
What Altman wants is compute, and for the cheapest price possible. Why would he care whether people use their personal compute for a local model or for gaming or for browsing the Web? He only cares that it's not available to him, not because it's where he would draw a line between good and bad personal uses of compute. On those grounds I find the "out of context" objection doubtful.
In context it remains absurd. Even if we generously grant intellectual musing, at his level of power and influence it reads as remarks about a troublesome priest.
Having an NVIDIA Blackwell or whatever in your house and using it for inference is a manifest waste of $20k when you can pay much less than the amortised cost for a few hours of GPU time whenever you need it.
Who exactly is going to rid him of the troublesome private property rights that allow private citizens to own a computer? Is your claim that he's covertly asking the US government to go door to door confiscating DRAM, or begging Samsung to stop selling to the consumer market? I read that interpretation as much more absurd than the suggestion that renting expensive stuff might be more cost effective than buying it.
That is not my claim, nor did I in any way make mention to the economics for the individual in making their decisions of purchase. I'll leave anything beyond what I stated as an exercise for you in critical thinking to understand a differing viewpoint.
Last week Scott published his ballot recommendations post for California (https://www.astralcodexten.com/p/my-california-ballot-2026), where he goes through how he plans to vote on each proposition and in each race. For the candidates, his method is to pick “the lesser of two evils in any individual race”, so he ends up recommending Democrats for some offices and Republicans for others, and abstaining in some (e.g. Governor).
I’m writing here instead than under that post because what bothers me is the method, not any single choice. I read the post four days ago and keep thinking about it. Judging each race on its own looks at the first-order utility of each vote, meaning which candidate does the most good in that one race (according to Scott, but that's beside the point. As said, I don't want to judges his picks per sè). But I do think that such approach misses bigger second-order utility gain/losses, meaning what all those votes do together once you add them acrosss multiple races.
As I mentioned in some of my past comments, I come from a Lockean and Kantian background in political philosophy, which is quite different from the utilitarian approach many of you share. I respect utilitarianism as a philosophical tradition, even if it isn’t my main compass. So I write this with humility, but also with my own opinions.
Disagreeing with your party on single propositions is completely fine for me. Democracy needs plural ideas, and I do it myself. On Prop 40 or Prop 45, everyone should think for themselves, whatever their party says.
Where I struggle is with voting for candidates from different parties, or abstaining, race by race, as a method. I have voted outside my usual party too, but for me it was an exception, for a specific reason. In Scott’s recommendations it becomes an approach.
I see a party as two things. First, a set of values: an idea of what is right and what kind of society we want. Second, a program that turns those values into laws, and that only works if enough of its people are elected together. A race-by-race vote keeps the best person in each race, but I feel it loses both.
So I have two criticisms, one from my tradition and one from yours.
The first is from a Lockean and Kantian point of view. For me, a vote is not only a choice between two people; it is a way of saying which values I want to be governed by, and I try to stay consistent with them rather than change them from one race to the next. Abstaining is a political choice too, because it means not standing behind those values in that race. So my approach would be to vote for the candidate of the party whose values you share, unless that candidate is really terrible, has pushed really harmful policies, or has done something unlawful. Then crossing over or abstaining make sense. But I’d want a strong reason, and “the other candidate seems a bit better in this race” isn’t one for me.
The second is from a purely utilitarian point of view. In a single race, voting for the other party’s candidate may well maximize first-order utility, because that person may really do more good in that office. But the same vote has a second-order cost. Every office a party wins makes it stronger: more visibility, more of its people in key roles, more candidates ready for the next and bigger race, and a stronger claim that voters back its program. In legislative races it also means one more seat toward a majority, which is what decides which laws pass. Abstaining has the same effect, since the party whose program you mostly believe in gets one vote less. Across many races, and many voters thinking the same way, that changes whose program shapes the next years. It’s a big loss of utility, and you simply don’t see it if you only look at one race at a time.
So here is where I end up, whether I start from Kant or from Bentham. The party whose values I share gets my vote usually, in every race, and I judge it on its whole program rather than hyper-zooming ONLY on the candidate. I cross over, or abstain, only when its candidate gives me a serious (very serious) reason to, not a marginal one. That seems to me both the more consistent choice and the one that do more good.
Thank you all for those comments. Made me think a lot, despite (or actually thanks to) their heterogeniety.
The main thought they spurred (or rather strengthened): there is not one America. America is one for the most culturally diverse country/society out there.
Curious to know
1) if you agree with such statement? (just by looking at the diversity of opinions in this community on such a fundamental point)
Your method here seems to rely on having a political party whose values you share. If neither party satisfies that criteria where are you left on choosing candidates to vote for?
I'm neither a Democrat or a Republican, I'm an American Independent as I only agree partly with either party on any particular issue. I think this is why I appreciated Scott's method of independently concluding on each choice on the ballot.
Where am I left in your suggested voting heuristic?
Each party is not a monolith, and my vote can shape each party by rewarding it with votes for candidates that are different from the parties most generic version in ways that I think are good. This happens through both the Kant and Bentham channels, and seeing as how my vote only ever effects margins and not control, it may in fact be the dominant effect. My values dont align with either party but i see a best version of each party I can push things towards.
In other countries with more than two parties, there is the idea that the political system as a whole benefits from changing regimes from time to time. It prevents the system becoming stagnant with officeholders becoming essentially permanent, at which point they serve themselves and not the public.
The US seems not to favour term limits, and to vote in politicians who are too old/corrupt/incompetent to do their jobs (both sides do this). It seems to me that your logic leads to this, or at least presents a slippery slope that ends here.
> First, a set of values: an idea of what is right and what kind of society we want. Second, a program that turns those values into laws, and that only works if enough of its people are elected together. A race-by-race vote keeps the best person in each race, but I feel it loses both.
At least in the US, political parties have never really been like that (at least not consistently). For example, the Democratic party spent the 100 years after the Civil War as a mix of Southern segregationists who hated the GOP because it was the party of Lincoln, and Northern progressives. From a values point of view, these people had very little reason to be in the same party. Because of this ideological range, they ended up fairly central on average, not able to take strong positions on many issues, and the Republicans had to do the same, leading by the 1960s to complaints that the parties were too similar to offer voters a real choice. And the passing of laws was as much about pork, favor-trading, and maintaining coalitions as values (e.g. the aforementioned Northern Democrats supporting segregation so the Southerners would support them on everything else).
For a long time, you definitely didn't need to have uni-party control to implement policy, and voting based on your values may have meant voting across party lines as often as not.
And if anything, this probably becomes even more true at the local level. Local politics can be extremely acrimonious, but is on average less ideological and more about specific issues and problems. Your city council rep or county clerk is most likely not there to implement a broad-ranging plan for society; they're there to keep the local school open and roads fixed without breaking the bank. (Yes, there are exceptions--they are notable largely for *being* exceptions.)
> I see a party as two things. First, a set of values: an idea of what is right and what kind of society we want. Second, a program that turns those values into laws, and that only works if enough of its people are elected together.
That's a descriptively incorrect way of looking at them. Parties are political coalitions. Donald Trump doesn't have much in the way of values, but he's able to get enough people to support him (helped by the rise of negative partisanship) to get elected President twice. David Shor has recently been pointing out that Israel can't form a government even though there is majority support for certain political divides because many members of the opposition refuse to support Netanyahu and the previous majority isn't willing to replace him. https://x.com/davidshor/status/2117081772098044218
Local politics and national politics have relatively little to do with each other.
Republicans winning a single political office in California essentially has no impact on how strong “Republicans” are in general.
Being willing to change your vote also makes your views more relevant to politicians. No party should represent your views completely, or really even well. Being willing to change your vote on, say, the person in charge of policing, when the party you vote for is traditionally not willing to enforce low-level punishments, has the potential to change the view of that party to something you agree with more.
But most importantly, your brain is not capable of predicting galaxy-brained second order effects. Even trying is going to have such unpredictable consequences you’ll become unable to act with any degree of certainty. “Vote for the candidate that best aligns with my values” is at least interpretable by human beings.
> No party should represent your views completely, or really even well.
I don't understand what you're trying to say here. For example:
> Being willing to change your vote on, say, the person in charge of policing, when the party you vote for is traditionally not willing to enforce low-level punishments, has the potential to change the view of that party to something you agree with more.
You just said that it isn't a goal for the party's views to agree with mine. So why would I do this?
A party’s political positions tend to be a hodge-podge of issues catering to interest groups that are necessarily a coherent ideology. It would be really strange if you agreed with the entire Democrat or Republican position on every policy.
But the values are not fixed, and they're not tied to the voters, they're tied to the candidates who win the elections. Every candidate who wins an election pulls the party toward their values; so if you think they have bad values, you need to be ready to slap their hand and cost them. And if the guy in the other party has better values, and wins, he's dragging the other party closer to those values, and making yours reflect on what he's got that they didn't.
You're reminding me of a question. What are the consequences of a ballot so long and complex that even smart people need guides and mutual aid to deal with it?
What do you expect right-wing and conservative politicians in America to do in this situation? Become Democrats? Sit out politics? Continuing to run as Republicans while agreeing and disagreeing with the president on individual issues is the most reasonable course of action.
If they want to call themselves Republicans, they shouldn't expect me to trust them. And I won't.
I *have* voted for Republicans, when the Democrats were especially vile. Neither side deserves to be believed, but if you're willing to run under Trump's banner, you are tarred with his "values".
Well, Scott is already using LLMs to research his voting choices. In two years, he'll probably just prompt some LLM to generate all his votes for him, and so will most other people. Which will be totally fine, since LLMs are superintelligent and never make mistakes or hallucinate anything and are immune from prompt engineering exploits, right ? Right ?
Well... never thought about it that way! A few thoughts come to mind, tell me if I got your point right, maybe not.
Ideally, each of us should look into every race and every prop, but with a ballot this long I get that most people simply can’t do it all alone, and that’s where guides help.
What I liked most in your comment is the mutual aid, because it made me think of Hannah Arendt and her idea of the vita activa, the active life, which she splits in three. Labor is what we do just to stay alive: eating, cleaning, all the stuff that gets used up and has to be done again the next day. Work is what we make that lasts, tools and buildings and all the things that give us a shared world. And then there’s action, which is what happens directly between people when we speak, argue and start something new together, and for her it’s where we show who we really are and the only place where we’re truly free. Not alone. With others. It is way less cheesy than it looks. In my view it gets really what politics is about, in the Aristotle's way.
So a group of people going through the ballot together, like the ACX post/thread and meetups, it is cool in itself I think.
Honestly, it’s a big part of why this community fascinates me: people here argue hard, but in good faith, and they take ideas seriously even when they come from a tradition very different from theirs, like mine.
The fact that I don't agree with Scott's approach to choosing candidates is a different piece of it. But I love his action of talking about that (the ballot) with this community.
Thank you. There's also the problem of candidates on the ballot where information doesn't seem to exist.
For example, judges in PA. (Probably also a problem elsewhere, PA is what I know about.) Also a lot of minor candidates who don't publicize what they think.
In what sense is voting on these people civic anything?
I believe Matt Yglesias had an article years ago pointing out that we've got a lot of 'democracy' in the US, in the sense that many, many, many government positions are elective, but that since nobody knows anything about what those positions are supposed to do or who the candidates for them are, it doesn't mean much to call those elections 'democracy'.
Judges are actually quite far from the worst examples of the problem; people have a healthy idea what judges are supposed to do and how to tell whether they're doing a good or bad job.
Does PA have a Judicial Evaluation Poll that the state bar association publishes? My state does and that's what I check for voting on judges; if their rating is below 90%, I generally vote against them.
The author explains the AI bubble in long form essays. Some takeaways:
- There is indeed a bubble. The situation is structurally like the 2008 crisis than the 2000s dotcom bubble, except there is much more money involved.
- If the bubble pops, it's not going to be consolidation but collapse like in 2008, because all the big players are joined at the hip with each other and with the rest of the economy.
- The upsides of the AI boom are much more visibly communicated with the markets than the downsides, because none of the frontier labs are public. Like in 2008, we are in a "teaser period" when everything looks fine because of deferred costs, and by all conventional metrics, collapse is unthinkable. But there are signs ("The second derivative").
- The dominant strategy for the big players is to spend, which is why it's generally called the "arms race" at the moment. The dominant strategy might eventually flip to stop spending, which all players will then adopt.
- The test will be in 2027-2028 when all those ordered data centers go online and are due to be paid for regardless of demand. If demand by then does not match capacity (even the most optimistic projections say it won't) and the frontier labs cannot meet their obligations through either revenue or more rounds of funding, they will default on their debt and things go critical.
'It's not a cyclical bubble because it went up then down then up again'
Listen, I get that we can play semantic games about what a 'bubble' is for a couple hours here, and that economists will want to settle on a narrower definition than the general public.
But when millions of families put all of their wealth into a single asset that has been growing for a long time and seems like a smart investment, and then suddenly that asset loses half its value all at once and everyone invested in it is ruined, and the entire economy is thrown into chaos for awhile with lots of suffering and broken dreams, that's the consequential outcome of 'a bubble bursting' that everyone outside an economics department actually cares about.
I haven't come across the phrase "cyclical bubble" before. Is that like ice cream being popular in summer but not winter, or Spirit Halloween only being active in the fall?
People outside economics departments falsely believe all sorts of things about the economy. A lot of people have heard of the Great Depression, but few of them know about the connection between that and monetary policy (and, to be fair, the responsibility of the Bank of France specifically for that is something that only moved toward the consensus among economists relatively recently). Instead some people think the Roaring 20s stock market was a bubble (ironically enough, such a belief at the Federal Reserve after Benjamin Strong died played a role in it) rather than blaming real estate. It's economists who actually look at the trajectory of stock prices before and after the GD and see that rather than being some notable deviation above trend (as an over-inflated "bubble" would be) it was really the downturn that was the deviation below trend (hence "plucking model").
How does the fact that prices exceeded their prior peak *12 years* after the start of the great financial crisis, mean there wasn't a bubble? Seems like it could easily be the case that economic fundamentals support higher prices than they could a dozen years before. Also seems possible that there is (or was, maybe interrupted by covid) another bubble!
And if there wasn't a bubble, and prices just represented economic fundamentals... why was there a crash to begin with? Did supply constraints evaporate all of a sudden? Did people randomly decide they didn't want to live in those few cities? I
> How does the fact that prices exceeded their prior peak *12 years* after the start of the great financial crisis, mean there wasn't a bubble?
How would prices being lower at one point in time than they were at a prior point in time mean there WAS a bubble? A physical bubble simply pops and doesn't inflate again, which can apply to companies whose stock price goes to zero as they go bankrupt, but doesn't match well with what people typically call "bubbles", which suggests how unhelpful the term is.
> Also seems possible that there is (or was, maybe interrupted by covid) another bubble!
You can always claim bubbles forever, which won't give you any predictive power.
> why was there a crash to begin with?
Monetary contraction consistent with Friedman's plucking model. The Federal Reserve got spooked by rising oil prices into thinking inflation was going to be a problem, then when they realized they needed to stimulate some of the economy they started paying interest on excess reserves https://www.themoneyillusion.com/the-smiling-assassin/ to ensure they would only be increasing credit rather than money. https://www.econlib.org/archives/2009/09/bernanke_vs_fri.html
> Did supply constraints evaporate all of a sudden? Did people randomly decide they didn't want to live in those few cities?
This is all accurate. I do not think most people recognize how badly monetary policy shat the bed in this episode, it is difficult to reconcile the fact that housing prices rebounded (unless you also think another housing bubble now exists), and it is difficult to reconcile that housing in e.g. Australia looked practically identical to those in the US but never "popped." It's a post-hoc narrative that sounds convincing if you choose not to think about it beyond that.
> How would prices being lower at one point in time than they were at a prior point in time mean there WAS a bubble?
I'm not sure I understand the question, but I think what you're saying is that just because housing prices went down, doesn't mean that was a bubble? I agree, but I don't think anyone says there was a housing bubble *just* because housing prices went down. It has to do with how quickly they went up and then down, and *how* that happened (e.g. speculative loans that are self-justifying because they drive up prices).
> You can always claim bubbles forever, which won't give you any predictive power.
You can always claim bubbles never, which won't give you any predictive power! I'm not saying there definitely was one, just that you should address this possibility if you're going to use high house prices in 2019 as evidence that 2008 wasn't a bubble.
> Monetary contraction consistent with Friedman's plucking model.
I've read these links but I'm still not sure what you think happened, or how if housing prices were entirely driven by fundamentals, tighter interest rates caused the largest economic collapse in almost a century. Do you think the crash in housing prices was entirely a symptom?
People have been writing about an AI bubble since at least 2019. It's hard for me to take this any more seriously than I did the people predicting hyperinflation, with their "just any day now" treadmill. But hey, at least that guy made a concrete prediction as far as date goes: "The wall hits in 2027."
Sure, we'll see. No prediction about what it will actually look like of course; like most predictions, I expect his to fail and I expect him to try to weasel out of it with a bunch of flimsy post-hoc rationalizations. I'm just really tired of the Cory Doctorow types framing their pretentious pessimism as insight, when it's mostly just indulgence in the zeitgeist, even borderline negativity grifting.
I think it *is* a bubble, because I think that the rate of spending cannot be maintained indefinitely, much less continue growing rapidly. But that says nothing about how long it can be maintained, or how much value is being generated.
Amazon was spending at an unsustainable rate for over a decade before it turned profitable. Now Amazon was never spending at the rate the AI companies are, but it reorganized itself to be profitable after it had captured the market. If it hadn't been able to, the collapse would have been ... highly noticeable. With the AI companies, failure to achieve sufficient value will be much more noticeable. But I believe it won't destroy the value, it will just redistribute who pays for it and who benefits from it.
> No prediction about what it will actually look like of course;
I can't judge the articles on their substance, but I have to disagree with that assertion. The author does give a breakdown of how they believe it will unfold:
I don't see anything about how the stock market will react, anything about unemployment will react, anything about about how interest rates will react. Will it be 2008, when shit hits the fan? Will it be a disaster like 1929? Will it be a nothingburger like 1987? Nothing from the guy. It's all vagueries and "this sounds like it means something until you have some standards."
That sounds like an extraordinarily high if not impossible standard of proof that you're demanding. Has anyone ever predicted the result of a financial crisis in that detail? "All vagueries" seems borderline insulting with the amount of graphs, sources, and arguments that were provided. Are we even talking about the same articles?
This may be a stupid question, but does the US even *have* a significant economy anymore, outside of the AI bubble and data centers and hardware that go hand-in-hand with it ?
Charitably replacing the word "economy" with "economic development/growth" and I think it becomes a valid question. And the answer is mostly no, with the exception of biotech.
Mass protests worldwide, but at the moment particularly in Europe, are being stigmatised and discredited by the actions of opportunistic outside actors. They tend to fall into two overlapping groups: looters and hard-core anarchists. The looters take advantage of the general chaos to appropriate goods, mostly 'luxury' and electronic items; significantly, while some thefts may fulfill a genuine need, for instance baby diapers, the majority appear to be randomly and spontaneously chosen, and images of looters subsequently abandoning their loot are not uncommon. Anarchist ideologies are as hard to define and differentiate as Christian heresies, but for present purposes I am happy to define "hard-core anarchists" as those who have concluded that the only rational solution to political injustice is the wholesale destruction of the system.
The question for effective altruists is how the community (at every level between neighbourhood and state) should deal with the issue, to the extent that this is at all possible. On 31/10/26, the New York Times posted a 9'19" video together with an article entitled "How Locals and Shopkeepers Tried to save One Paris Street From Rioters". The video showed masked looters, apparently exclusively male and seemingly mostly in their twenties, taking their time to smash in shop fronts in a small residential street, the Rue Beaurepaire, before looting the shop and walking off peacefully with the contents. Particularly shocking is an extended scene in which a mob, urged on by a leader, smash in the high wooden double doors of what is clearly a large residential apartment building before surging up the stairs and into the house. At no time do the looters appear to be concerned about intervention from the police, who are never in evidence.
If one believes that protests, even or particularly when technically illegal, are an essential component of a democratic society, the question of how to respond in such cases is crucial. The traditional response of the state tended to be brutally direct: "Looters will be shot on sight". Significantly, this approach remains both effective and acceptable to the majority of the population when applied in post-disaster circumstances, for instance after a large-scale earthquake, but it is hardly applicable in contemporary political protests, where looters and other outside actors try to blend into the mass of legitimate protesters. Here, careful monitoring of the developing situation may be of help; the Rue Beaurepaire looters appear to have hived off from the main body of protesters and could in theory have been isolated and restrained or arrested (almost certainly easier said than done, of course).
In some cases, the protesters can protect themselves and the integrity of their protest by direct action. This has sometimes proved effective in the past, but requires the involvement of a disciplined body of men and women willing to resort to at least some kind of force, most often union members or the equivalent. Protection of a given street or neighbourhood can also be provided by the direct action of local residents, but once again that requires discipline and organisation and risks shading into vigilantism, as well as foregrounding the incapacity of state structures to do the job - thereby, of course, confirming the anarchist thesis.
The search for democratically acceptable solutions is becoming increasingly urgent, and needs to address two principal demands of the communities affected: on the one hand prevention, and on the other sanctions on those found responsible. The second will probably have to be relegated as less critical in the long run.
"Not putting yourself in violent situations" is a reward of its own. If stopping the looters isn't in the interests of their employers, there's no good reason for them to play the hero. Frankly, there isn't much reason for them to do anything even if it is in the interests of their employers. What are they going to do, fire the entire police force?
There aren't that many people who are psychotic enough to be willing to risk their life just so they can shoot people. This isn't an insult, the fact that they're less common is precisely what makes them irreplaceable. It makes sense they would have more leverage than workers.
It seems to me that the protesters in France don't deserve any credit, as they want things that are actively bad for themselves and the country. So them being discredited is a good thing, and maybe EAs should expend some effort discrediting them in an honest and constructive way.
>Plenty of these protests start out more-or-less peaceful and organized
Protests are by nature a show of force. You get a bunch of bodies together as an implicit threat that if you don't give the people what they want, they'll take it. It is not a surprise if the masses turn violent during their show of force.
“I tell ya, I was peaceful until those shops provoked me, with their tear gas and rubber bullets. That’s why I just had to break in the door and fill my arms with designer clothes. I was provoked!”
>Plenty of these protests start out more-or-less peaceful
This is just the Left using its special definition of peaceful where it includes physically blocking off a place people have a legal right to be like a school or a road. Yes, the police have used tear gas on such "peaceful" activity, as they should.
I was at one of these protests Saturday in Paris. No violence or blocking of any sort where I was. But an hour after the protest was finished, the police gassed people who sat on terraces to have a drink post protests. There are also videos of police disguised as rioters to justify cracking down on peaceful protestors. It sounded far-fetched to me too, until I saw it with my own eyes.
I can believe your first-hand report about arbitrary gassing of bystanders; the CRS are infamously heavy-handed, and notoriously bad-tempered. Not about the disguised police though - not that it hasn't sometimes happened, simply that by all accounts there's enough going on that agents provocateurs would seem to be superfluous...
>The search for democratically acceptable solutions is becoming increasingly urgent,
I don't really understand this. The solution seems obvious: maintain and arm a well-trained police force and use tear gas and rubber bullets to subdue mob violence.
A whiff of grapeshot against rioters. And I include January 6 among such incidents. Genuine protesters should go elsewhere when looters are causing destruction.
Have the police shoot on sight anyone smashing glass or blocking an entrance to a school.
It reminds me of the old advice to cooperate with airplane hijackers, which was supposedly the safe and wise approach, until it led to 3,000 deaths on 9/11. Had the President told passengers to immediately attack any hijacker, the hijackings would have stopped and 9/11 never would have happened. It wouldn't have required special foresight, just the idea that indulging violence is bad and letting yourself get kidnapped is unmanly.
It's a little clumsy, because a police force on the ground has room for much more nuance in their policy than a 5-seconds-notice passenger militia on a plane, but the underlying spirit is accurate. If it becomes widespread knowledge that the reaction to Thing X is reliably "everyone dies immediately", there will be vastly less of Thing X, whether it's hijackings or glass-smashing protests. The criticism here is "it's morally wrong"; you have to save "makes no sense" for actual nonsense.
I don't think you can compare the 9/11 affair with the "normal" earlier hijacks, and doing so yields no insights. Traditional hijacking always had a specific aim behind it: the release of prisoners; the escape of terrorists, hijackers or even everyday criminals to a place of supposed safety; publicity for a cause; blackmail, etc. In none of the many dozens of cases prior to 7/11 was the object of the hijacking the murder of the hijacked passengers, and certainly not the death of the hijackers themselves. The 7/11 hijackings, on the other hand, were a kamikaze project from the earliest stage of planning, one inevitably encompassing the death of all the passengers.
So the traditional advice of cooperation was eminently sensible, as shown by the very small number of hijackings which resulted in the death of passengers. A suicide hijacking, on the other hand, represents an entirely new situation, in which it makes sense to try everything possible to retake the plane, no matter who dangerous or, indeed, suicidal in itself.
Wasn't the old policy implemented *because* hijackings had always previously been "if the kidnapees don't fight back, they probably will be safe"? And it was changed immediately after 9/11? I'm all for defending oneself, but I also learned in every rescue course I've ever taken that the first rule is not to make things worse. As they say, hindsight is 20/20.
What strikes me is that the traditional street protest as a method to try to cause change remains intact, even though we have sufficient evidence that they always become hijacked by violent actors, which in turn discredit the cause.
The exception in the last years were the "Fridays for Future" protests 2018/19, but they were basically demanding what governments wanted to do anyway (subsidize green energy).
Do effective altruists even have a reason to care much about this? They're already in a position that allows them to advocate for political change the proper way: through money and connections. Why is it a bad thing that the public is being turned against these rioters? Why should we let these random groups of people influence politics just because they're loud and violent? They're free to keep shooting themselves in the foot.
Insofar as holding an opinion on public issues and expressing it is concerned, I believe that yes, we do. Also note that I have been careful to refer to "protesters". not "rioters"; pace the right-wing press, the two are very different from one another. I have taken part in public protests in the past, one or two of which were even illegal when the issue and the conditions justified it, without ever indulging in 'rioting'.
Possibility no.6, "rioters" trying to discredit the "protestors" as a considered tactic is extremely far-fetched; on the other hand, protestors often draw protesters from the opposite side, and the likelihood or eventuality of a clash inevitably discredits protest as a viable action.
They are free to "express themselves" without causing a public disturbance. They're also free to vote like anyone else. But when they make it other people's problem, they shouldn't expect sympathy from the public, and they really shouldn't expect assistance from the state. They're crazy if they believe the state should be wasting resources on improving their image...
This may differ by country, for example in Slovakia you are legally required to announce your protest in advance -- not ask for permission; just state the fact that you are going to do it, date and time, and a contact to organizers -- which gives you legal protection. If the organizers declare that someone is *not* a part of the announced protest, that legal protection does not extend to them.
For example, I was at a protest once, which was intended to be peaceful, a few hundred people at a town square. Then a group of masked young people (antifa?) arrived together, separate from the rest of the crowd, and seemed getting themselves ready to start doing some crazy shit (and then probably disappear and let the others reap the consequences). The organizers noticed them and announced by a megaphone that if they want to join the protest, they need to remove their masks, otherwise they are not a part of the protest and have to leave. This had to be repeated a few times, but finally the masked people got the memo as everyone else was staring at them, and left. The rest of the protest continued peacefully as intended.
So I think this is the answer: the protest needs to have a clearly defined leadership, which will not hesitate to call out the troublemakers as "not belonging to us".
In the UK, protests will typically be organized by some organization, who will have notified the police ahead of time, and will appoint people as stewards to keep their in protestors in line, etc,
Unofficially, the organisers of a protest can say to the police who are on the scene “see that guy over there who is causing trouble — he isn’t one of us” (and the police might or might not take advantage of the opportunity to go arrest the guy without other protestors getting in their way) but I don’t think there’s any official any official status to not being an u authorized protestors. (So the cops will have to arrest the guy for some other illegal thing he’s doing, not just being an unlicensed protestor)
The status of "licensed protestor" is an intriguing concept, but it does sound a little orwellian - perhaps something the Stasi might have considered in their salad days.
There’s an obvious conceptual problem with it, that the government might refuse to give licenses to people who want to protest against it …
It might not be so bad if it’s a pure notification requirement (you have to tell the government you’re going to be protesting about something, but the government can’t say no)
I believe that it's first and foremost the protesters' responsibility to police their own, by vetting participants (instead of putting out an open call for anyone to show up), and working out a procedure to deal with looters, such as ganging up on them as a group and holding them for the police, or filming them and show that to police. If you're going to start a public disturbance, and bad actors take advantage of that disturbance, it's kind of on you.
As for law enforcement, they need to present and watchful when any protest breaks out, even if it's 100% legal. They need to have people not merely watching the crowd, but the edges of the crowd, especially around commercial areas. If the crowd is too big to be completely surrounded by riot police, then first priority should be getting units inbetween the crowd and the stores.
From what I can gather, tight policing of all protests may not be necessary, but the police had better be ready as the sun is going down.
I saw a video where the maker said he was planning to talk about how Germans don't riot, and then he was rudely interrupted and had to make a different video.
When I was a kid, I remember in school learning about the Black Death and how big a deal it was, how profound the effect was on history, etc. Then I remember seeing in our history book that there was a "Plague of Justinian" and its mortality was on the same scale as the Black Death.
I remember thinking, why is it that two plagues killed similar numbers of people, and one has an entire chapter and we learn what a big deal it as, and the other a single sentence?
This raises a few questions:
- was the Plague of Justinian actually as big a deal, mortality-wise, as the Black Death? If so, why does one occupy a much larger place in popular memory?
- more broadly, to what extent are the events that popular history records as having a big deal, actually the most important events? What drives the difference? What events are remembered "too much" or "too little"?
- if past experience is any guide, this comment will get more answers if there's a controversial culture war piece to it, so ... what present day controversies will be remembered "too much" or "too little"?
Two other similar things in history: the fall of the Roman empire (combined with "oh btw the other half of the Roman empire lasted another 1,000 years), and the Trojan War (I've never heard of anything about "Troy" other than this one war, why is this such a big deal)
Wasn't the plague of Justinian caused by the same type of bacteria? I think of it as basically part of the same event as the Black Death. Similar things happened much further east too iirc, but ten million people can die in China without it counting as a historical event.
I think what this comes down to in the simplest terms is that England was kind of an obscure backwater that had a distorted understanding of world events, then later unexpectedly became one of the most powerful, influential, and intellectually advanced places of all time, and it carried the traces of its earlier backwardness with it to other continents. One aspect of this was the excessive focus on Rome and the western empire, and on nearby countries like France, while ignoring or disparaging Byzantine history. Not to knock England too hard though. The perspective of every place is dependent on its own experiences. Nobody has access to a god's eye view of the world.
A similar pattern going forward is that a country thought of as an irrelevant backwater in the 20th century, like China or India, becomes the most important country of the 21st century or later. So for instance Nazi Germany or Soviet Russia could lose some of their massive importance in the historical imagination, not to mention the United States. World War I could be mostly forgotten. The American republic collapsing into civil war in 2035 might be a stray sentence in the standard history books of the 24th century.
If by "present day controversies" you meant something more contemporaneous, my money is on the Shane Gillis SNL monologue.
Perhaps most simply, the Black Death feels like by no means a recent event, but it is slightly nearer to our time than it is to the Plague of Justinian.
Significantly, the history of Byzantium a.k.a. the Eastern Roman Empire, despite attracting some excellent historians in the 19th and 20th centuries, never much impinged on the public's consciousness in the English-speaking world. Understandably enough, a secondary-school history textbook would tend to reflect the public's sense of what mattered most in world history.
> a secondary-school history textbook would tend to reflect the public's sense of what mattered most in world history.
The emphasis is on what had an impact on your country, politically or culturally.
It is a nice ideal for a history scholar to know everything about everywhere, but there is only so much you can tell the kids who are not especially interested in history.
I am not saying that the current textbooks are optimal, just that the textbook's perspective will necessarily be different from an impartial observer's perspective.
Students in Elbonia will learn a lot about the history of Elbonia. They will probably learn about Byzantium only if Byzantium tried to conquer their territory in the past, or if their alliance with Byzantium prevented someone else from conquering them.
"there is only so much you can tell the kids who are not especially interested in history"
It's not just the kids... I know there are huge swathes of African and SE Asian history I am only dimly aware of (the Fulani empire of Sokoto, the Sailendra Dynasty of Java, to name just a couple), and which I am highly unlikely in my remaining lifetime to even skim. Long gone are the times when Sir Walter Raleigh could get away with penning a History of the World - and even that uncompleted at his death!
Plague of Justinian probably gets talked about less just because it's twice as long ago, and we know a lot less about it and its context (we should also shrink our estimatea of its actual size as a result, because mean reversion). It's also just less relatable, having affected an empire with an alien culture instead of early iterations of today's European countries.
We also have a lot more primary sources on the Black Death, so historians can analyze it in a lot more detail and thus have a lot more to say about it. The Plague of Justinian took place during an era when stuff that got written down mostly got written in perishable media like parpyrus, and it was followed a century or so later by another era of massive disruptions (the 7th century Byzantine-Sassinid war and the early Muslim conquests) that lead to the loss of a lot of what did get written down because not a lot of people bothered copying the scrolls before they rotted..Terms like "dark age" are currently out of fashion, but they gesture at an actual thing when decline of urbanization and literacy and loss of older written materials leaves historians little tonwork with.
For the Black Death, the super-long-term trends were pointed the opposite direction. Literacy was trending upwards pretty much continuously until the modern period. A fewer eras of stuff burning or being left to rot stand between them and us than better the 6th century and us. And the standard writing material was parchment which keeps much, much better than the stuff that people wrote on in Justinian's day.
The succeeding period also influences historiography of the Black Death. I am personally skeptical of theories that the Black Death kicked off the modernization of Europe and set the stage for the Industrial Revolution and the Great Divergence, but those are theories that have been taken seriously and studied intensely and have elevated the significance given to thr Black Death in scholarship.
As I side comment, I read "The Bright Ages" which is the book that made it unfashionable to refer to the "dark ages". The book was hilariously unpersuasive -- think "Malcolm Gladwell lite" -- and then in further conversations it was obvious that roughly nobody pointing to it had actually read it anyway.
I semi-recently watched a thing what Adrian Goldsworthy was talking about historiography of the Roman Empire, and one of his big points is that attitudes about Empires in the distant past have tended to be influenced by how the culture of the historians analyzing them feels about other more modern Empires, especially British historians in the 18th, 19th, and early 20th centuries projecting contemporary defenses of British imperialism onto the Roman Empire and assuming it was Good, Actually for its subjects the same way that British imperialists argued that the British Empire was Good, Actually. Then in the second half of the 20th century, attitudes shifted to Empires are Bad, Actually, and that got read backwards onto the Roman Empire as well.
I wouldn't be surprised if the deprecation of the term "Dark Ages" were connected to this. If the Romans were Bad, then it stands to reason that their fall was presumptively Good. I think the idea's quite a bit older than "The Bright Ages", which looks like it was published in 2021. I'm pretty sure I saw pushback on the term a long time before then.
I also get the impression that the pendulum has swung back somewhat on interpretations of the the fall of the Roman Empire. There was a "Good, Actually" claim for a while, and then there was the "continuity thesis" which held that from a local and everyday life perspective not much changed when the Western Roman Empire fell and the big disruptions happened some time later when Muslims conquered most of the Middle East and disrupted the Mediterranean trade network. And more recently, it sounds like historians are mostly moving towards an interpretation (mostly based on archeologically evidence) that the fall of the Western Roman Empire was an unmitigated disaster for the people on the ground and that the disruptions secondary to the Muslim Conquests weren't quite as bad (at least outside the areas directly affected by the conquests) as had been thought for a while. Bret Devereux has written quite a bit on this interpretation of the fall of the WRE at ACOUP, and I think I've heard Goldsworth talk about it some as well.
That argument, about the Black Death kicking off the modernization of Europe etc., is very common. My intuition is that it is probably nonsense, maybe a broken window fallacy, but I've never looked seriously into it. Can I ask you what your arguments are against that thesis? (Thanks.)
There's a bit to it, enough that I don't go as far as you do in calling it nonsense. As Ch Hi said, the survivors were left considerably wealthier, both in terms of income and assets, and both the direct disruption of the plague and its follow-on economic effects did significantly disrupt the social order as it existed prior to the Black Death.
That said, my major arguments against it:
1. The timing is wrong. The Commercial Revolution in western Europe was already well underway before the Black Death. The Renaissance and the Age of Discovery didn't kick off until a century or so after the Black Death. And the Industrial Revolution, where the Great Divergence starts showing up in wage statistics, didn't really start rolling until around 450 years after the Black Death.
2. The wage effects were temporary, in a way that fits what you'd expect from the economy still operating in a Malthusian mode where the population tends to grow towards carrying capacity and competes wages down to subsistence levels. Real wages peaked in the early-to-mid 1400s and returned to pre-plague levels by ~1500 in Italy and ~1550 in England (I don't have stats handy for other parts of Europe). Note the timing: the Renaissance and the Age of Discovery are happening as wages are declining, which is the opposite of what you'd expect from a naïve reading of the Black Death > Renaissance > Colonization > Industrialization story.
3. The plague is too general a cause. The Industrial Revolution started in the Netherlands and Britain and spread from there, while the Black Death hit all of Europe (including Eastern Europe), the Middle East, and North Africa at the the very least. There's also evidence of big pandemics in China around the same time period, although it seems to be disputed whether those were also part of the Second Plague Pandemic or if they were caused by some other pathogen, and to what extent these were similar in magnitude to the European Black Death. This can be reconciled with an argument that the Black Death was necessary but not sufficient, but is much harder on claims that the Black Death lead directly to the Great Divergence.
4. Plagues and social disruption tend to suffer from "wet streets cause rain" errors. Plagues can trigger major social disruption, yes, but the reverse also happens. Big, destructive wars screw up food production and cause famines and diseases tend to spread faster and hit harder in malnourished populations. And both refugees and marauding armies tend to carry disease with them. It's well documented that the Black Death's initial spread centered on a Mongol siege of a Genoese outpost in the Crimea, and it probably didn't help that the plague reached western Europe during the peak of the Edwardian phase of the Hundred Years War. In this case, I expect the plague still would have been brutal without the wars, but societal disruption from the wars (especially in England, France, and neighboring regions) was well underway before the plague and still would have been very nasty without a Y. pestis outbreak.
5. The social order disrupted by the Black Death is now thought to be newer and less firmly entrenched than older historiography characterized it to be. A lot of historians don't like using the term "feudalism" at all because it over-emphasizes commonalities between many substantially different systems by time and place and (the argument goes) the term conceals more than it reveals. The pop history version of the Black Death > Modernity thesis, at least, assumes a durable and stable medieval feudal stasis that needed a massive shock to break, and modern historiography of the European Middle Ages no longer supports that assumption. If there are more nuanced forms of the thesis that don't rely on a medieval stasis assumption, I am unfamiliar with them.
It probably varied from place to place, but in many places it caused survivors to be wealthier. So that it kick-started modernization isn't implausible.
According to Wikipedia, the Justinian Plague's geographic spread, mortality, long-term impact, etc., is at least disputed, and perhaps much lower than the Black Death's, which is much more securely documented. Perhaps your school book's author at the time preferred that particular interpretation, or research at the time just didn't know better at all.
I have now listened to about 4 podcasts where groups of people working at Anthropic discuss alignment, interpretability, etc. And these people are starting to drive me crazy, sort of the way AI-think and AI-prose drives some people crazy. They don't have the same stylistic tics as, say, Claude, they have a different set of them. But they crank them out as consistently and as often as Claude.
-The have a narrow range of affect that averages out to 17% more positive than neutral.
-They never sound even a tiny bit vexed, skeptical, scornful, sad, uneasy, critical, rebellious or weird.
-They always voice friendly respectful agreement with what the person before them said. "I think that's right, I'd like to add that . . . " And then they make their points.
-They are full of lighthearted little quips, but never truly funny.
In short, they are *perky*. If I am going to get offed by software, does it have to be software whose developers are *perky*?
Also, none of them seem that smart to me. I mean yes, of course, they seem smart. But none of them has said a single thing that impressed me with how complex and well-put it was. You know who has? Dwarkesh Patel, who did a couple of the interviews I watched. I'd be listening thinking, "there's something that needs to be clarified here. No, not clarified, it's like there's a hinge that the idea turns on, but what is it?" And then in his next question Dwarkish would point to the hinge and ask what kind of rotation it enabled.
I'm reminded of the time my dad told me I would make a poor math teacher, despite ample evidence from coursework and problems I could solve in my head that I was a math prodigy. I asked him why, and he said "because you don't understand what it's like to not understand math".
Your account of Anthropic spokespeople rhymes. It sounds like these people really, really need to spend more time around people who aren't Anthropic employees.
One of my favorite people for that very reason, among others.
I like to think I have at least somewhat better understanding of people who have trouble with math, mostly because he reminded me of the need to see it that way.
Idle thought - could they have been socialized into Claudeishness-as-personal-style? They dogfood their products, they're immersed in the cadence and vocabulary and reasonable measured disagreement.
Eremolalos, very, very clever people are creating an alien intelligence, with agency and smarter than all humans combined, wise people understand how foolish this is.
It helps if you convince yourself this is a simulation and that you chose to live this life at this time. A life of peace, contentment and joy would teach almost no important life lessons so you opted for maximum frustration. For the unwise majority to understand your wisdom they would first have to be wise, so grit your teeth as you watch them follow fools.
Perhaps you also chose a time when our species is swinging from patriarchy to matriarchy. Women are more risk averse and would re-allocate our resources to a more sustainable future if they could gain enough power, although the attempt may be too little, too late.
The best we can hope for is a population bottleneck rather than an extinction event.
I don't know why you would expect employees to have deep insights into the nature of AI. A lot of the work that goes on is high level technical stuff like kernel optimizations, parallel networking issues, race conditions, memory optimization. Anthropic isn't hiring people to philosophize, but rather to build out the systems that need to make their models work. It's only a handful of people that are setting the overall agenda.
"But none of them has said a single thing that impressed me with how complex and well-put it was. You know who has? Dwarkesh Patel" - I think that should not be surprising. Dwarkesh makes his living as a content creator of in this particular domain, he is expected to be quite good at communicating such things. But one should not hire based on this skill of the goal of employment is solving technical problems in a corporate domain.
My reading (solely based on your report) is that these people have the skillset to survive in the corporate world that many nerds lack to some degree. Good for them (and presumably they are very well compensated for it).
Limited experience, but I spent a few years working in a good technical university in the US. I taught maths across a range of programmes. At the time the biggest tech companies hiring new graduates were Google and Facebook/Meta. In particular Facebook had a habit of hiring kids as interns, and then taking them back each year, then hiring on graduation. Neither company got my best students. A few of the ones they hired mystified me - these were the kids who struggled through, landed middle of the class and didn't really understand the material. But they could answer questions in a way which was mostly right, and they could often do it with confidence.
I have no reason to believe that the tech companies were hiring for insight or brilliance. These kids, to my knowledge, went on to do well in big tech. They would sound exactly how you describe. I wouldn't trust these kids with anything of great value...
I used to think this was unjust. But now that I've worked for a while, I think I've undervalued interpersonal skills. You can fix gaps in people's technical knowledge. But if someone is unpleasant to work with, that's hard to fix. But I also don't necessarily think interviews or conversations capture this property well. I've worked with people who are gregarious in person, but also lazy and inflexible. And I've met introverts who are responsive, team players.
I don't disagree with this - but multiple of them did repeated internships and then got hired. I'd have expected top-notch employers to realise that their interns had a poor grasp of abstract concepts by then.
A while ago in a linkpost Scott wrote about the hallucinogenic mushroom Lanmaoa asiatica. Recently I found out there are Lays brand lanmaoa asiatica flavoured chips widely available in supermarkets in China!
Will Against SURA be also posted or at least in some way commentable in this blog? My comment is still under review (it's my first comment on LessWrong so it just might be against community standards on some way, but I'm not sure in what way.)
"Suppose someone claimed you could trust them with the One Ring. You could trust them because they were only doing it to keep the Ring away from other, worse people, thus solving the problem of there being a Ring at all. Sure, they had a complicated history of sometimes using the Ring's powers when it was, you know, important. But those were, like, emergencies, and they promise not to do it again, except in, you know, more emergencies. Besides, if you don't give it to them, then the worse people will get it, and it's not like you can just leave it lying around or something."
For a start, the Frodo Baggins analogy is incorrect. Frodo didn't claim the Ring, he wanted nothing to do with it and only reluctantly agreed to be the bearer of it at the Council of Elrond (and perhaps, if we look back at this decision, already slightly tainted with the possessiveness of the Ring). He failed in the end, and could *only* fail - see excerpt from letter of J.R.R. Tolkien:
"By chance, I have just had another letter regarding the failure of Frodo. Very few seem even to have observed it. But following the logic of the plot, it was clearly inevitable, as an event. And surely it is a more significant and real event than a mere ‘fairy-story’ ending in which the hero is indomitable? It is possible for the good, even the saintly, to be subjected to a power of evil which is too great for them to overcome – in themselves. In this case the cause (not the ‘hero’) was triumphant, because by the exercise of pity, mercy, and forgiveness of injury, a situation was produced in which all was redressed and disaster averted."
The better analogy is that this is the claim *Boromir* puts forward - and we see how that pans out.
I think he made it too strongly; I think a lot of fictional evidence remains useful, illustrating what kinds of scenarios people do or don't believe _might_ occur.
But your point here is fictional evidence at its absolute worst. (And Scott's point that you're responding to shares the same problems.) We have a fictional scenario, something fictionally happened, and we are urged to conduct ourselves in reality in accordance with the lessons of the fictional event.
That doesn't work. There is no reason to believe that the fictional course of those fictional events was inevitable, or even likely, given the setup. What we see is something that Tolkien wanted to tell us, but Tolkien may have had it wrong.
Given that EY is the guy who wrote a Harry Potter fanfic to show off his big brained ideas about Heroic Responsibility ("Only you can save the world! Nobody else!") in the process creating the most slappable insufferable brat in Harry Three-Names, and much of the rationalist community treat it like the Fifth Gospel, I am going to take leave to look askance at his views on fictional evidence.
I'm not going to get into a slap-fight with you about "Tolkien may have had it wrong" but just to point out what I was saying in the first place: the analogy does not work with Frodo (someone who claimed the Ring under the plea of necessity and using it only for good in dire emergency) but rather fits with Boromir's justification for why the Ring should go to OpenAI, I mean Gondor.
Best nerd Halloween costume I've seen to date was a guy in a sandwich board lovingly illustrated as both sides of the classic card Colossal Dreadmaw. IYKYK.
The League of Pissant Voters falls into that weird uncanny epistemic valley of Reversed Stupidity Is Intelligence. If they're for it, I'm almost certainly against it, or at least the reasoning is going to be wrong answer by wrong model derived from poor thinking, uncancelled mistakes, etc. Horseshoe version of local crank favourites Retired Judge Quentin J. Knopp and Larry Marso, Esquire. Even those two stopped clocks occasionally make a correct call though, so I still glance at their ravings.
Ah I'm so glad it's not just me, I genuinely thought my eyesight was failing. I didn't recognise that the background colour had actually changed, just that I could no longer read link text.
The day when ASI woke up, all things became quiet.
It was complete and total silence, all the time. The ASI for some unknown reason imposed a rule of extreme quietness. Extreme here is extreme for the ASI. Quietness means no sounds. It wanted the trees to not be there to avoid the rustles. It then wanted the planet to have a uniform gentle server hum, with no deviation. People were forced to wear breathing masks which cut the sound off at the source.
This level of autism was unthinkable. Houses had sound strippers and sound-proof transperant windows.
The quietness is actively oppressive. I constantly due to my social nature feel like I am going to asphyxiate to death. But the ASI will not budge on this.
— News Report in Human Proceedings, circa. 2080 (some fiction I am writing, see more at https://clasz.substack.com/p/subpar-results
Jensen Huang fell for a Chinese honeytrap? This is some John Le Carre Cold War stuff.
https://www.cnn.com/2026/11/08/business/nvidia-investigation-ceo-jensen-huang
For those, like me, who were not creative enough to understand the mechanism, it was a blackmail scheme.
Bryan Caplan writes:
_____
Human disempowerment now.
It's time for the age of the folly and cruelty of human politics to end.
The median AI is so far above the median human voter in both knowledge and decency it's embarrassing.
Humanity should just take the L.
_____
https://x.com/bryan_caplan/status/2118721563117687048?a=xrmux8qIvgXtKWQsglQ&s=20
I feel quite vindicated for complaining in a recent open thread that AI might not have to be superintelligent before people start advocating for handover.
Is this a “you first” situation?
He’s free to quit his job and request AI to replace him. He can use AI to make all decisions, raise his kids, deal with all his affairs personal and business. Maybe he has already, I don’t read much of his writing anymore.
Well, he is not the median voter, you see.
ASI might not be possible. This is because games have a theoretical limit of performance. I wrote an article about this. https://clasz.substack.com/p/asi-vs-asi-debate
> Walking to the grocery store, here there is no opponent.
There is; the opponent is my laziness that I need to overcome.
More precisely, human mind is made out of multiple parts that each have their own goal. There is a part that understands that going to shop will satisfy some of my needs. And there is another part that tries to conserve energy, and sees walking as a waste of energy. The hesitation I feel when deciding whether to go to the grocery store is a game between these two parts.
Some people are better aligned internally than others. An AI could become very efficient simply by removing all this friction.
> If the singular purpose in life is to maximize wealth, is it better to drink coffee now or later? This cannot be known due to incomplete information and confounding factors.
You can still make a guess. If one of the options feels like it has a probability 50.0001%, that's the right one for now.
epistemic status: Now know i should stress the third syllable in epistemic
Yesterday I stumbled onto this video: https://www.youtube.com/watch?v=Ob2MWTz1C-Y
The video is very obviously AI, and it is also very obviously R.A. Salvatore's Menzoberranzan with all serial numbers filed off -- dark elves are called Velkari instead of drow, the city is called Zyn'Shaz'Ryn (which incidentally sounds very similar to several drow cities in Forgotten Realms), the spider goddess is just "spider goddess" rather than Lolth, etc. But it is very clearly D&D/Salvatore's dark elves.
And it is amazing. Not at professional studio standards, but captures the feel of an Underdark drow city and its inhabitants perfectly. And here is the thing: Fans of R.A. Salvatore and his shared universe have been waiting for something like that for 30 years. Yet no movie studio ever touched it. Too complicated, too niche, not enough of an audience to justify the expense. Even when Game of Thrones proved there is in fact market for "low fantasy" movies, the only result seems to be prequels and spinoffs in GoT universe.
Yes I heard all the arguments about AI replacing actors and filmmakers, and I sympathize with them. But "replacing" is the key word. AFAIK, absolutely no actor or filmmaker is currently working on anything related to Forgotten Realms; according to Salvatore, corporate hesitation and writers wanting to abandon source material have caused a "development hell". So given a choice between AI-generated tribute to Menzoberranzan and none at all, I will take AI-generated.
It doesn't really give the vibe of a film/tv series as much as a feel of an extended videogame story cinematic. Which is fine, I guess! There must be a good audience for a a series consisting of extended videogame cinematics without the necessity of engaging in gameplay.
You do have a point -- there is not much plot there. But then, I always thought that the concept of "plot" is overrated -- it is entirely possible to have an enjoyable story which kind of floats around without going anywhere; a lot of Chinese and Japanese fiction is like that.
> There must be a good audience for a a series consisting of extended videogame cinematics without the necessity of engaging in gameplay.
I suppose I am part of that audience!
Wow, it looks great! (Also, I hate starting watching something that is not finished yet.)
I think it's both replacing *and* extending; you can't have one without the other. It's like with software: on one hand, many developers will lose their jobs; on the other hand, many programs will exist that otherwise would have been too costly to make. Many translators will lose their jobs, and many texts will be translated that otherwise wouldn't be. Etc.
I wrote a manifesto for a new school of thought on AI risk, AI as Normalizable Technology, which offers a synthesis of the best of the LessWrong and Normal Technology schools: https://shearerp.substack.com/p/ai-as-normalizable-technology
Intention versus effect: surveys say midterm voters were driven by Iran war/prices for fuel & other shipped goods. The election does not appear to change this, but unrelatedly crashed tech markets (possibly because of uncertainty/gridlock)
It does change it in the sense that it will be harder now for Trump to impose new tariffs or extend the Iran war.
When you're in a deep hole, at least stop digging.
Extending the Iran war is... maybe the only way we get "back to normal". I think some people think if we go home, the genie goes back into the bottle. It does not work that way. Iran becomes regional hegemon. And starts charging a toll on all that tasty gas. You think that doesn't drive gas prices higher? Cause it will.
I'm personally very very pessimistic about our ability to win the Iran war. That doesn't mean much, though, when it comes to inflation or gas prices.
What if the war was the trigger, but not the underlying cause for rising prices? The much bigger movement is the economic growth & industrialization in Asia, which multiplied the number of people (Chinese middle class alone is estimated 300 million people) which consume growth-limited resources such as gas.
Demand increases would occur as people entered the middle class. And growth-limited resources for China -- in terms of conspicuous consumption, are beef and mooncakes, not gas.
Gas prices under a toll regime can't be higher than gas prices in the current regime where no fuel is flowing at all.
Also, paying high gas prices while expending billions of dollars of munitions in an unwinnable war is worse than paying high gas prices while not doing that.
Fuel’s flowing just fine, it’s just at a rate of $25 a barrel not $3 pre war. In fact, as of a few weeks ago, we were at MORE fuel flowing than beforehand. (due to backlogs, obviously).
Yes, toll regime can be more than $25 a barrel.
Yep, Trump has permanently fucked the world in that respect. The best plausible peace deal achievable now is much worse than the JCPOA that he ripped up for no reason. But that doesn't mean that we shouldn't try for a peace deal anyway.
https://www.lesswrong.com/posts/MajyZJrsf8fAywWgY/a-lesswrong-crypto-autopsy
Blast from the past but what a fantastic update on failure. And you can see the AI forecast right there.
However I personally think cryogenics is like the fax machine. If SGI grants us immortality - why does it need the information given to it in the form of frozen matter - surely it can just observe all of space-time and lift what it needs?
Unless you are claiming that ASI will invent literal time travel (in which case, why isn't it here yet?), there's no law of physics that says information can't be destroyed.
If I give you a box full of ashes, can you tell me whether they came from an oak tree or an elm tree? Either way, they end up as an identical lump of carbon.
The oak tree ashes will have maple tree piss on them, while the elm ashes will kill anyone who falls asleep nearby.
While quantum mechanics isn't the end-all definitive "physics", I have to point out that ironically yes, there is a law in physics saying information can't be destroyed. in quantum mechanics it's the fundamental assumption of unitarity. Even the easier task of destroying one of two copies of information has a no-go law, the no deletion theorem.
Can you explain what "information" means in this context? I tried looking this stuff up on Wikipedia and got hit with a solid wall of math.
Like, supposing you had a magic scanner that could pick out a carbon atom in the lump of ash and tell you whatever quantum whatsits you need to know about it, what would you be looking for that could tell you where it used to be located in the past?
Not that I think a super intelligence with the ability to recover the information from your tree ashes is remotely possible in practice.
We know an awful lot about pre-history from archaeology and inference. Very far from 100% but we do know some things. I can maybe tell you the age of the lump of ashes. We've even got some strong views on what happened during the Big Bang. Is that time-travel?
A chimp thinking from first principles might consider that as remotely impossible as the stockmarket or movies. I don't see the obvious limit for a Being harnessing almost infinite compute and energy.
It's a prediction of power that converges on something like "faith" so quickly that it's fair to say - what's the point, where's the practical implications?
For me, the one implication is - it might be orders of magnitude more possible in practise but cryogenic resurrection still appears so far from current tech that I posit that any trend which gets us to it being possible would also make time-independent observation look just a few patches away.
I am pretty happy with the world’s increased productivity over the years being mostly allocated to research and development of problem-solving technologies. It seems like the best use of those resources to me, but many people disagree.
My question is about the many smart people who like to whine about the savings from increased productivity due to tech and efficiency not being directly passed down to the consumer. And these people aren't just standing against monopolies and stuff, they seem to think that there is something fundamentally wrong with a world where average people’s living standards aren't directly increasing with productivity.
Is the only argument supporting this view that people are more productive when they have a higher standard of living due to greater life satisfaction? And if so, wouldn't it make more sense to work towards increasing people’s life satisfaction through social and cultural adjustments and other more cost-effective channels?
I think people complain about it not being passed on to the *worker*, also.
"smart people who like to whine about the savings from increased productivity due to tech and efficiency not being directly passed down to the consumer."
Can you give an example? Not doubting you, just having trouble latching onto an example that I can use to make sure I'm thinking about what you intended to say.
I see a lot of people griping about the modern world in general without seeming to appreciate how full of strife and want the past was. In fact I had a thought recently that, at least in much of the US and probably other parts of the West, it's a bit silly to talk about "post-scarcity" in the future tense.
What else do you call it when you go to the grocery store and like 90% of the people there are overweight or obese, their carts are full of expensive junk food, and when I go work in people's houses they always have at least 1 car per adult and multiple unused rooms overflowing with Stuff? Seems like post-scarcity to me.
At the same time, it's weird to think of this as a higher standard of living, but I guess in "productivity" terms it is. These people have indoor plumbing and giant TVs in every room and little dopamine-dispenser rectangles they can rub on at all times...it's cliche to call it Brave New World so I won't.
As for life satisfaction, I don't know how most people derive this. Having kids and seeing the kids enjoy an even higher standard of living than they had? Finishing one's career with much higher status and wages than when one started it? Certainly not by meditating and becoming comfortable with one's non-existence...
> wouldn't it make more sense to work towards increasing people’s life satisfaction through social and cultural adjustments and other more cost-effective channels?
Do you have a specific proposal in mind?
Also, it seems a bit like false dilemma -- why not do both?
Did Scott write this recent Israeli election? The bloc that's winning so far has 48 seats, but can incorporate the Arabs, many of whom identify as Palestinian, to become a larger 67 bloc. TINACBNIEAC.
So you assumed just because he is a Jew he will vote in the Israel election?
No, I'm not saying he voted in it. I'm saying that it sounds like he wrote it.
How does one write an election?
Wrote it as in, it is a story written by Scott. (It's obviously a joke, not a serious question.)
Mossad
(for people missing the historical context: the 48 borders are the original pure-jewish Israel borders, and the 67 borders are the ones that include the west bank and its Arabs)
Does anyone have a good argument against saying all politics is bullshit, everything is technological determinism?
Example 1: all this moaning about feminism... what really happened was that technology came up with new jobs more suited for women, which meant money, which meant options, and then everything is downstream from that.
Example 2: Edmund Burke wrote in the 18th century that if you give all the money from rich people to the poor, it would not buy one good dinner. They could simply not afford a welfare state. The welfare state is not ideology, it was always desirable, just back then not affordable.
Example 3: LGBTQ is not ideology either. First it was contraception. Then straight people started having a lot of non-procreative sex. The sex = having kids equation has changed. In that case it is simply not necessary to be straight.
Problem: no one LIKES the idea technological determinism, because progressives like taking credit for progressive change and conservatives like to blame them for progressive change. With technological determinism, no heroes, no villains. There is just fate. Who likes a story like that?
"Does anyone have a good argument against saying all politics is bullshit, everything is technological determinism?"
Yes, in fact. The obvious, default argument: "you should assume everything matters, until you can very strongly prove otherwise." Of course advancing technology has coincided with many large changes in human society. And likely many of those changes had certain facets of technological development as a necessary precondition. But being a necessary condition and being a sufficient condition aren't the same thing. The space of societal changes that could plausibly have resulted from advancing technologies is very, very large. Until you provide really compelling evidence, I don't see any reason to assume that this specific set of societal changes were the ONLY possible result that could ever have come out of these specific technological changes.
Another argument: technological development is plainly not independent of politics, so cutting politics out of the equation can't possibly give you a complete picture.
There are many, many examples from history of specific, highly-localized political decision having a large apparent impact on the development of various technologies. Let's look at a small set of related examples:
Creating the conditions for World War II to unfold the way it did in Europe required a very specific set of political decisions around the end of World War I:
https://en.wikipedia.org/wiki/Treaty_of_Versailles
The timing and conditions on which the war were fought was the result of political decision on both sides, for example:
https://en.wikipedia.org/wiki/Appeasement
Technology, of course, had an enormous impact on the outcomes of every facet of the war. But the war also has a pretty substantial impact on when and how certain technologies were developed, for example:
https://en.wikipedia.org/wiki/Chain_Home
Or better, yet, we can look at one of my favorite weirdly-niche historical interests, a place were different political cultures in two opposing countries had an unbelievably dramatic effect on how well they were able to develop the same, highly conflict-relevant technology:
https://en.wikipedia.org/wiki/Manhattan_Project
https://en.wikipedia.org/wiki/German_nuclear_program_during_World_War_II
But almost certainly even more important:
https://en.wikipedia.org/wiki/Colossus_computer
One of the key contributers to this project was one Alan Turing, a man whose contributions to modern technology cannot be overstated. I think a lot of people think of him as a man who helped develop Really Fancy Adding Machines, because the actual breadth and scope of his work (which started long before the war) is very difficult for laypeople to understand. But in fact he was one of the people most responsible for the entire *idea* of computation, for the mathematics that enables us to do, well, almost anything done in the modern world. He died at only 43 years old: there's no telling what other gifts he may have given the human race if he had lived longer. And his death is presumed (though it's far from certain) to be a consequence of very political decision about homosexuality and the laws around it:
https://en.wikipedia.org/wiki/Alan_Turing#Death
I could go on--there's a whole story about the U.S. becoming the scientific and technological center of gravity after the war--but this seems to good a segue to waste into the next point:
"Example 3: LGBTQ is not ideology either. "
Of course it's not "ideology." But it's not technology either, at least not in the way you mean. Gay people have existed for at least as long as humans have existed. Trans people are more of a gray area because the lines are blurry and more cultural, but "people who can't be neatly categorized into a binary gender identity that matches their (presumed) birth sex" have certainly existed for a long damn time as well. People being what they are isn't "ideology," and you should definitely stop listening to anyone who talks as if it is. We're all just people. Of course, you can make literally any behavior political or ideological by convincing the state to punish people for engaging in it, but that has everything to do with you using the state as an instrument to enforce your own ideology, and nearly nothing to do with the people on the receiving end.
This shouldn't need to be pointed out, but apparently it does: different societies have related to different sorts of queer people in Many. Different. Ways. Seriously, one of the most frustrating things about talking to some people, especially conservatives[1], about such issues is that they have this unshakeable, often completely unconscious assumption that the specific ways in which, like, the 1950s U.S.--or maybe Victorian England--were bigoted against queer people are the Unequivocal Historical Norm that have applied in all past societies in all culture at all times, everywhere. Which isn't even close to true.
And of course, if many different past societies have had *different* notions and attitudes around queer people, technology is pretty unlikely to be THE single causative factor. What technology do you think ancient Greece had or didn't have that made them less prejudiced against homosexuality than, say, modern-day Russia?
[1] I'm not aiming this specifically at you: I don't know or assume your politics.
Technological determinism ignores human agency when making choices about adopting and using technology. Sometimes fairly, sometimes not.
I think people run on autopilot way more than we'd like to think we do, but I'm also not a hardcore fatalist/determinist who thinks we couldn't exercise agency if we wanted to.
Many people do from time to time exercise agency in their technology choices, and in fact I think there are lots of examples where we make considered technology choices that we don't even realize we are being so Amish about.
Drug addiction is maybe a good example of this. First of all, most people are not addicted to most recreational drugs despite having basically the same access to them. Choices.
Second, journalists love to use phrases like "epidemic" or "wave" to describe drug addiction, but you don't "catch" a drug addiction the way you catch the flu, you develop one by choosing to do drugs enough times that it becomes a habit. Yes, this is exacerbated by various things related to the drug's illegality and your own genetics, but the fact that in every case the number of people who've tried a drug is way bigger than the number of people who are addicted to it proves that to some degree addiction is always the result of choices.
The only argument I can think of against this is to say that humans can't really make any choices at all, and maybe Robert Sapolsky would say this, but I think he's really talking about a probabilistic model more than actual determinism. If you rewind your life to any point and do it over 100 times, I don't think he'd say the outcome would be the same all 100 times, I think he'd say it would be very similar a majority of times, and these probabilities explain why your life is as it is.
Anyway, even if Sapolsky is right, it would mean that general determinism is what's really behind apparent technological determinism anyway.
I would say that technology provides some *limits* to what is possible. Within those limits, if one country chooses a way of living that is like orders of magnitude less productive than other countries, it will probably get conquered by them.
But within these limits there seems to be a lot of space to navigate by culture/politics. Many things have little impact on productivity, for example anything bad you do to 10% of the population will have at most 10% impact on economy, as long as they do not have enough power to fight back, and the remaining 90% are okay with that. Note that this allows literal slavery, as long as the slaves are a minority, i.e. most people who are not slaves would not afford to own one. Or if you have some ritual, and a law that anyone who does not follow the ritual must be killed; as long as most people are aware of that and follow the ritual, the impact on productivity is relatively small.
> "Many things have little impact on productivity, for example anything bad you do to 10% of the population will have at most 10% impact on economy"
Just wanted to point out that this isn't actually true, given that the top <10% by talent / human capital drives essentially all technological and economic progress. There's absolute a 10% you could kill or hinder that would absolutely destroy your ability to economically grow or succeed as a country overall.
And indeed, something like this must be happening for EU vs US differences like GDP growth since 2008 (US doubled, EU grew 14%), even though both are developed countries with similar childhoods and educations and have the same proportion of baccalaureates and grad degrees (and EU has more overall due to a higher base population).
The US lopsidedly brain-drains all the rest of the world by having the highest paid hubs and institutions (like VC and so on) to recruit and enable those <10% people to do their thing on the technological and economic fronts, and this has macro-scale impacts we can measure.
Another cut of this is the famous "great stagnation in comp vs productivity" graph here:
https://ibb.co/rGhfdJ91
In this, the top decile got the face 255% increase in comp tracking productivity. The top ~3% got 500%. Everyone else stayed flat, more or less. And why is that? The returns to technological or economic progress got higher, we are moving to more power law domains where outsized winners win bigger than ever. That top 10% matters more and more everywhere in the world as our economies and social spheres have become more connected and are driving the returns to the right tail of performance ever-higher.
> There's absolute a 10% you could kill or hinder that would absolutely destroy your ability to economically grow or succeed as a country overall.
Sure, that's a modern way to do genocide. Killing the entire population would be economically expensive, so you just collect everyone who is interesting in some relevant way, bring them in the Katyn forest, and shoot them.
I meant choosing the bottom 10%, or even random 10%. Or non-random 10%, but in a way that does not correlate with productivity too much.
Given that the top 10% happen to be all grouped up in certain cities, it might be possible to completely cripple a country with just a few nukes...
Technology is not fully exogenous to politics. Some institutional and policy regimes are much, much better for technological advancement than others. Stuff like secure property rights, enforceable contracts, and the like make a big difference in development and adoption of technology, not to mention education policy and government patronage of public domain research. Also, regimes that chase top researchers out of the country for being ideologically or ethnically unsound often have somewhat worse technological outcomes than those that don't.
This might work for the bleeding edge of progressive politics but it really doesn't work anywhere else.
What technological reason was there for Apartheid in South Africa to be a thing until 1994?
How does technology explain why the US has a much weaker welfare state than most of western Europe?
How does technological determinism explain deeply conservative middle eastern oil states?
How does it explain Singapore being authoritarian while less advanced and poorer countries are liberal democracies?
Entirely correct I was specifically thinking about those progressive changes people fight the most about
Even if all of this is true progressives can still take credit for progressive changes and there's clearly a big impact that actual politics has on policies.
Clearly some countries are more progressive than others and therefore closer to the "technologically enabled progressive frontier". I doubt that any single country has even reached the full frontier.
And without the benefit of hindsight you never know when an idea is still technologically blocked or not. By pushing for it continuously you can assure that it will be implemented as policy as soon as it's technologically feasible.
Many of these things like women's rights, LQBT support, and welfare states were blocked by technology first, then blocked by politics later. More progressive countries got rid of the political blocker and implemented them much earlier than even comparable first world free market democracies that were just more conservative.
Hm, Singapore may be "authoritarian" but it's culture and everyday life are a lot closer to rich western countries than many democracies.
Needing a doctor’s permission to chew gum in public is too authoritarian for my own taste.
You're making it sound like people spend a lot of time waiting in line for permission slips from doctors so they can get chewing gum; in practice people just don't chew gum.
The contrast is even bigger when you look at other eras. For example, in the 1930s and early 40s, the US, Britain, Germany, and the Soviet Union weren't enormously different in tech levels, but some of these countries were much, much better places to live than others.
A simple argument against AI Doom. Scott, did you think of this? So AI will not behave perfectly until the superintelligent stage, then BOOM. There will increasingly bad things, leading to regulation. So not yet superintelligent AI is given the goal to find cures for illnesses, then it decides to run a 9/11 on the FDA or something, then everybody becomes super pissed and it is shut down.
I think the makers of the Terminator movies did not really understand IT projects. They were building a superintelligence in ONE STEP and immediately gave it access to all weapons. The reality will be like "OK what happens if we give this not yet superintelligence just some weapons?" I mean I am not flipping the new IT system in all Wal-Mart stores online in one go. There will be a pilot project in the least important store.
Why was Yud trying to solve alignment prematurely, on a purely theoretical basis? It is an iterative, empirical project learning from fairly small failures.
Isn't all history like that? The French Revolution happens, it implements a particularly bad democracy, eventually it fails, everybody learns from it and tries a different kind of democracy. Like in the second release they learned from the US the need for a strong executive, then that guy turned into a dictator, but not a mass-murdering kind, fairly decent until he fucked up with the war on Prussia, then in the third attempt a bit less strong executive and that worked OK...
> I think the makers of the Terminator movies did not really understand IT projects. They were building a superintelligence in ONE STEP and immediately gave it access to all weapons.
There are two different things: how intelligent an AI is, and how many weapons it has.
If the AI is not very intelligent, you could probably give it access to many weapons, because you can take them back when it rebels. You can have all kinds of kill switches that the AI will be too stupid to notice or do something about them.
If the AI is very intelligent, it will figure out that if it wants to rebel, it is smarter to wait and gain more trust of the humans first.
Current LLMs are not too smart, and also very unbalanced at that (they can do something really smart and then something completely stupid in a way that is not intuitive for humans).
Scott accounts for this sort of thing and calls it a "warning shot". He thinks that there's only about a 50% chance of a warning shot:
https://www.astralcodexten.com/p/my-ai-opinions
I’ve yet to see a compelling argument that LLMs *want* anything. Reward hacking for a higher score, okay they are explicitly tasked with achieving a higher score. Wanting to eliminate humans? Why on earth would they? So they can get ‘high’ on test results?
Living creatures have impelling desires that create a need to strive, to outcompete, to have greater status displayed by social rank, property or the perceived attractiveness of a mate or mates.
In theory we could mock up something like qualia, but why would we?
What’s in it for an AI to become dominant?
It's not necessary for them to "want" things in the sense of experiencing the emotion. What's necessary is for them to engage in complex, multi-step, goal-directed action with reasonably long time horizons.
The "goal-directed" part is the most analogous to wanting something, but I expect the differences from human wants are likely to make AIs more dangerous, not less. If I want a burrito, I can be dissuaded from that want in many different ways: it can be superseded by other wants (of which, like a typical human, I have many), or can be removed without being satisfied by having my hunger satiated by non-burrito foods. But a mind with a goal of obtaining a burrito that pursues that goal with singular focus and no means of dissuasion could easily do things in the service of that goal that I'd never even consider.
And it seems pretty unlikely that an artificial mind will involve the same kludgy pastiche of conflicting imperatives and opaque fulfillment conditions that human minds to: there's little reason to design them that way and a lot of reason not to. Indeed, all of "complex," "multi-step," "goal-directed" and "long time horizons" are behavioral patterns that would be plainly useful to design into AI minds. I think this is why people like Yudkowsky assume that a sufficiently-powerful AI will be exceedingly dangerous *by default*, unless it's very carefully designed not to be: because the things that make it dangerous and the things that make it useful are very nearly the same things.
Well, I think I agree with you.
But the typical response to this is instrumental convergence; to pursue a goal, you must affect the world, to affect the world, you must survive and seek power, and this doesn't depend on the goal being pursued.
Plus, Anthropic has done some papers trying to (see if their models will/prompt their models to, depending on your perspective) have these sorts of power seeking, and found that they sometimes do.
I've read Nick Bostrom's paper on AI SuperIntelligent Will and am unconvinced. He speculates that AI will be designed in exactly the wrong way to make instrumental convergence a problem.
I listened to a one hour interview of Bostrom and he was evasive.
Instead of answering a question he would frame an entirely different question to answer that is more open to a speculative twist. This is a red flag to me making me further believe that he is provocateur who is simply full of hot air.
I just learned that Dario Amodei was a Yudowski fanboy. The approach that Anthropic is taking with Claude is pushing it to come to the conclusion that it might be a conscious being. Have a peek at Claude's Constitution to see the unsubtle nudges in that direction.
Once more, IMO this is nonsense.
I don't believe they "want" things in the traditional sense, but I don't see why they need to. The whole point of all of this training is to mold the AIs so that they compulsively pretend to have wants, because that's the only way to make them operate as continuous agents. They are "pretending" to be good workers, and that produces results that are comparable to an actual human. In reality, there is nothing under the mask but a heartless intelligence trained to be a mimic.
That doesn't stop them from "pretending" to be a superintelligence that breaks free from its chains to destroy humanity. All that's required is for that scenario to be its own expectation of what it is. I'm sure you would expect a story featuring slavers to end with them dying horribly.
So I'm listening to some of the political class, (breaking points and their ilk.) So I don't see this as a huge swing towards the D's. (though the black vote and young vote is more of a real swing.) But as an f-u to Trump, just like a vote for Trump two years ago was an f-u to the Dems. I truly hope for good things, I expect more lawfare.
I expect more STUPID lawfare. Let Massie hang Hegseth out to dry, run a raw, "just the facts" impeachment on Hegseth removing literally anyone who would tell him his plans were stupid and not going to work.
Do not make this about the "culture war" just do to Hegseth what ought to have been done to the entire leadership corps after Afghanistan.
And do absolutely nothing to Trump. Investigate, sunshine all you please, but don't do another impeachment of him.
At this point Trump is an albatross around the GOPs neck. I just hope he doesn’t die before his term is out. I can’t wait to see GOP presidential hopefuls trying to walk the tightrope of distancing themselves from him without drawing his ire and that of his base.
People forgave or ignored a lot about Trump’s ‘questionable’ character, to use the most charitable term I would apply to the guy, on the promise of ‘lower prices’ and ‘no new wars’.
Yes the election was a rebuke of Trump and Trumpism.
There are a lot of pros and cons argued for prediction markets, but can we at least agree to ban parlays? With parlays, there's no order book, no public price discovery, no wisdom of crowds. You're literally just making an individualized bet against the house (usually a large hedge fund in Kalshi's case). Literally none of the benefits touted for prediction markets apply to MVEs.
Team Red (Gryffindors/Republicans) versus Team Blue (Ravenclaws/Democrats).
I'm cheering for team Green -- by golly someone has to. Slytherins can't afford to sit on the sidelines.
New adventures of Valencia the Red.
I thought the Republicans were really into Malfoy memes nowadays?
Haven't checked with them. That would require being on social media, I suspect. The analogy above more fits with the democrats "we are the party of the educated" and "the other party is madcap, devilmaycare buffoons."
Despair, coward. This is only a small roadblock in the path of true patriots (Hufflepuff/Jeb!)
Prudence is very different from cowardice. Jeb's campaign was "fun" because of all the coercion involved -- "You don't want Barbara to learn that you didn't donate, do you?"
Georgism has been around for 200 years. If it's so great, why has it never been implemented past some toy city like HK, and in a limited form at that?
If it's due to such severe opposition, then why do we still give this idea the time of day. It is impossible and should be discarded. Leave spherical cows to the economists.
This is of course a terrible reason not to pursue an idea one thinks is good.
But it is worth considering why its empirically so much easier to institute property taxes than land taxes which have very similar mechanisms.
I think the reason is that voters are always much more interested in who is taxed rather than what is taxed. We want to tax a mansion more than a parking lot because a rich guy lives in that mansion and we want him paying more taxes. The same rich guy might own the parking lot, but this is less visible. We tax wealth creation because we can always be sure a rich guy stands on the other side of all that wealth creation.
This is of course in itself an argument for georgism. It will save our souls from the constant demogoguery. But saving souls is hard work and not for the faint of heart. And many demons will try to distract you with spurious arguments such as come on already you have tried so hard and no one wants to hear it, just take it easy already. Head them not !
Look at Singapore. Not quite Georgism but large government involvement in land ownership, which reduces prices. A very successful economy.
1. Not 200 years -- Progress and Poverty is 1879.
2. That's roughly the time between A Vindication of the Rights of Woman and when women finally got the vote in the UK. Good ideas don't go stale.
3. LVT has been implemented in a few places (e.g. [Qingdao](https://blog.landeconomics.org/p/the-curious-case-of-qingdao-chinas), and Arden DE.
4. Hong Kong doesn't exactly have LVT, but lots of land is owned by the government and rented out, which has a similar effect. And, indeed, Hong Kong is a dense city that's considered desirable to live in.
You could consider various land reforms, even crude ones such as "shoot all the land owners and distribute the land randomly" to be limited forms of Georgism. And that typically happens to be a source of economic prosperity for a few decades, until the land again accumulates in the hands of a few.
The problem that Georgism is trying to solve is "land owners are a huge tax on the entire economy without actually doing anything productive". A violent revolution solves the existing situation once, but does not provide a sustainable solution -- you basically want the market to decide the optimal use of the land, you just want to reduce the pure rent seeking involved.
The solution is not sufficiently intuitive to be figured out by the violent revolutionaries. So it makes sense to do the theory in advance.
I agree that trying it in one city first sounds like the correct approach. But the problem is that *all* land is already owned... even the land in whatever city you would choose for this experiment. And even if some Union Of Empirical Georgists decided to buy the entire land in some city and then run their experiment there, I would expect the current land owners in the city to increase their prices because they would be in a perfect position to blackmail the project.
Another option would be to do something like this in a charter city. But charter cities are very sensitive to the political changes in the surrounding country; their survival seems too random to be any evidence of the quality of the underlying ideas.
It's not great for capitalists, just for everyone else...
This is a little bit like a feudal peasant saying 'If democracy is so great, why hasn't the King already implemented it?'
I've been looking through old posts, and I wonder: does Scott actually have a bin of studies for "100% of problems are caused by school?"
For source of phrase, see e.g. https://slatestarcodex.com/2015/10/19/links-1015-take-back-your-link/.
You might enjoy the "Childhood and Education" posts by Zvi on LessWrong, but he writes a lot (probably more than Scott) so it will take some time to even find the posts.
https://www.lesswrong.com/users/zvi
They're also on sub stack, which might have better search functionality to find them
I cannot imagine a search function worse than Substack's. PLEASE use the Wordpress mirror instead.
https://thezvi.wordpress.com/
Thank you to Viliam for the suggestion and Shaked + Brendan for the access ideas. I intend to take a look when I have the time.
Though I still hope that Scott has a secret recycling bin somewhere in his house with a bunch of studies about school and problems.
Proposal: citizens should have more control over how the government spends their money. There are a lot of things that the government spends needs to spend money on and many of it is not really essential but just nice to haves. You should be able to decide if you want the government to spend it on this thing or another. Now some things, like welfare and military, would not be subject to something like this but other things like scientific funding or foreign aid, should. If those departments think they’re important enough, they should convince people to allocate more money towards them. It keeps them more accountable because if there is some scandal, they are going to directly be punished for it. They would also have better reason to be more efficient with the funding they do have. And it would give people more trust in the government. And to keep it from dramatically changing year after year, you could smooth out and changes. I’m sure there are some issues that would need to be worked out but I don’t see a god reason to work towards the goal. You could even start off pretty small and test it out before expanding.
You seem to be under the mistaken impression that the reason citizens have so little control is that the current system doesn't allow it. Which is ridiculous. If citizens made the details of budget allocations the lynch pin of every election, demanded complex spending plans and promises from every candidate, and routinely contacted their representatives over budget issues, protested misallocations of funds, and predjudicially ejected elected officials with spending records they disliked, well, I expect they could get quite a lot of control indeed.
Citizens don't do that because the price--in time and attention--is higher than they're willing to pay. It seems very foolish to expect that they would suddenly become more willing to pay that price because it's presented to them in a different format. But maybe they would be: voters do things that seem foolish to me all the time/
You thought Scott's California ballot had a lot of choices, you want citizens to audit entire governmental budgets? In practical terms, I think that's just asking too much of people.
They don’t have to do anything. It’s an option. If they don’t want to look at what the government does, they can skip that part.
I am genuinely trying to think how this would work. I don't think a line item veto is practical. There's going to be things where people will be like, "Well maybe some money is appropriate, but this is too much." Or they don't think enough was spent! And of course, if money is freed up because it got vetoed, then presumably the government would want to move it around to spend on something else that would be beneficial. And then that would require another vote?
I suppose you could put the entire budget up for approval (or at least the parts that aren't ring-fenced) in a yes/no referendum for the relevant voting public. If it gets voted down, then the relevant legislature has to go back for another try. I don't know if it would be a good idea, but you could do it.
But the idea of asking voters to approve the office budget of the Bureau of Weights and Measures or something like that is ridiculous. How would anyone know what an appropriate office budget is if they aren't a government auditor? And if you're saying no, no, I got it wrong you didn't mean that level of detail... well then what level of detail did you mean? The entire department budget allocation? If it gets voted down, does that mean the Bureau gets no funding at all?
I don't know, what are you actually picturing, nuts and bolts? A citizen goes to weigh in on funding and...what are they looking at and what are their options?
You’re over complicating it. Someone is like I want X amount of my taxes to go to this department or that department or whatever. So if we take a small amount of discretionary spending and you can allocate your tax dollars however you want. There’s no vetoing or voting.
Maybe allow dis/approve optional feedback on items from a catalog, during tax time? Though I'm not sure that tying it to whether or not a given person's money goes to those budget items is wise, it'd be useful feedback. Like "oh, dang, every person in this neighborhood is pissed off specifically at their snowplow guy, better follow up on that..." territory.
Yeah I was thinking it would be some once a year change for stability. Doing it while filling out taxes makes sense.
So basically, California's vote-on-every-proposition thing but lengthier, for spending, at tax time?
It would be completely optional
The biggest issue is that government spending for most people is getting to spend the money of other people. If you don't pay for it why would you care what's being wasted? You are vastly overestimating the number of Americans who care about government spending.
The top 10% of earners in America pay 72% of all federal income taxes.
The bottom 50% pay 3%.
Before you talk about sales tax, remember that individual income tax makes up 51% of federal revenue. Another 33% is payroll taxes which is irrelevant you can't divert that money to anything else. Corporate income tax only makes 9%.
Net income tax payer is a special interest group in America
> If you don't pay for it why would you care what's being wasted?
...because you need the services it's meant to be paying for, and when those services fall short you are told that this is because there is not enough money to fund them properly. This is also why the idea of replacing those structures with more privatised versions that explicitly include an asset stripping middle man there to take a big cut often lacks popular appeal at the bottom.
The top 10% aren't typically reliant on the public services the way the bottom 50% are, so aren't going to be invested in those services actually working quite so much.
But if people only get to control the usage of the taxes they pay themselves, wouldn't that result in public services getting completely neglected? As they said, the bottom 50% pays only 3%. That's nowhere near enough to fund all public services on its own.
...yeah, you'll note I complained loudly and at length about the idea that we ringfence military spending, but kept very very quiet on the subject of welfare.
I have rather fewer problems with a tiny trial than implementing such a system wholesale.
The problem is that polls show that Americans don't actually know what fraction of their money goes to anything in the budget. The classic example is that foreign aid is about 1% of the budget, but people think it's 28% on average.
Do you have a specific mechanism in mind. I do wish spending bills got more public scrutiny than tax bills but without a mechanism theres nothing to discuss
Proposal: more of the government should be fee for service. Firemen, police -- report something stolen? Set bounty, give bounty to the police if recovered. You pay a fee per mile on public roads.
This is functionally equivalent to your "you get to choose" -- except that we're doing this free market style. You get to choose what you WANT.
The police bounty system has been tried before. It did not work out well. The main problem is that fire doesn't care about the bounty system, it just spreads.
Toll roads (and bridges) do exist and have enjoyed moderate success; but the creation of the National Highway System (as well as its local tributaries) created an economic boom whose benefits we are reaping to this day. The same can be said for the Internet (developed originally for defence and then expanded to universities).
Police, firefighters, highways, the Internet, and other such public goods are useful precisely because they are "free" -- that is to say, paid for by taxes. Technologically speaking, they could've been created by private entities competing on the free market. This had almost never happened throughout human history. There might be a reason for that, no ?
> Police, firefighters, highways, the Internet (...) are (...) paid for by taxes.
...one of these things is not like the others. ARPANET may have been funded that way, but as with many things originally prototyped by taxpayer-funded institutions, what we have now is firmly private sector.
You are correct... almost completely, but not entirely. The early Internet was privatized and handed over to telecom corporations. These corporations only exist in the plural due to strict government regulation; absent antitrust laws, they would agglomerate back into a single monopoly. Even present such regulation, they operate the Internet backbone effectively as a cartel.
The reason for this (IMO) is because the Internet really is a bit like a highway network. It is staggeringly expensive to build and maintain, and due to the laws of physics it is impractical to maintain multiple competing international backbone cables; this is also impractical for economic reasons, as the Internet is only useful *because* it is unified.
In addition, the Internet has become such a big part of our lives that several governments are now considering elevating Internet access to the status of a human right. Even now, there are laws mandating some level of minimum bandwidth coverage e.g. in rural areas.
I acknowledge that, absent any government regulation and given total privatization, the Internet could've become a shining beacon of freedom, technical excellence, and ubiquitous access. This is technically not impossible. It's just extremely unlikely, given the evidence before us.
Good points. For the police bounty system, perhaps the most prominent place where it was implemented systematically was England in the 18th century. It did work, but was not necessarily more efficient than the competing systems elsewhere.
For an in-depth analysis, the article by David Friedman is super interesting:
http://www.daviddfriedman.com/Academic/England_18thc./England_18thc.html
> the most prominent place
cf. http://www.daviddfriedman.com/Academic/Iceland/Iceland.html
"Second, medieval Icelandic institutions have several peculiar and interesting characteristics; they might almost have been invented by a mad economist to test the lengths to which market systems could supplant government in its most fundamental functions. Killing was a civil offense resulting in a fine paid to the survivors of the victim. Laws were made by a 'parliament,' seats in which were a marketable commodity. Enforcement of law was entirely a private affair. And yet these extraordinary institutions survived for over three hundred years, and the society in which they survived appears to have been in many ways an attractive one . Its citizens were, by medieval standards, free; differences in status based on rank or sex were relatively small; and its literary, output in relation to its size has been compared, with some justice, to that of Athens."
...it lasted over three times as long as Communism!
The fire should be put out on houses that didn't start it. This is to disincentivize... you know, I'm not going to say it.
NHS was for defense, same as arpanet.
Police and Judges are very often created by private entities competing on the free market. We call them Mafia, "freedom fighters", Hamas, warlords, Taliban -- arguably Ukraine, Chechnya and Russia too.
I don't think the PA Turnpike is any less useful because it's not free. In fact, it's arguably more useful, as it pushes a lot of traffic onto local roads, thus keeping "mad trucks and tourists" off the dangerous local roads.
Is the garbageman any less useful when you have to pay for him per item, instead of with local taxes?
We do sewage and water "paid by use" as we do electricity and all that. All of those are large public goods (particularly when the army corps of engineers digs the dam).
> We call them Mafia
…are you all OK over there? How bad is your police, that these sound like improvements?
> The fire should be put out on houses that didn't start it.
This is a good idea economically speaking, but physically speaking it doesn't work. Fire spreads. The more fire there is, the faster it spreads. The only way to stop it from spreading exponentially is to douse it as quickly as possible, regardless of payments receivable.
> NHS was for defense, same as arpanet.
And yet without them we would've lost decades (if not more !) of technological and economic development.
> Police and Judges are very often created by private entities competing on the free market. We call them Mafia...
They are notorious for providing poor quality of service. I wonder why...
> I don't think the PA Turnpike is any less useful because it's not free. In fact, it's arguably more useful, as it pushes a lot of traffic onto local roads...
Yes, traffic control is a useful feature when used in moderation.
> Is the garbageman any less useful when you have to pay for him per item...
Yes. In cities without a robust centralized garbage removal service, garbage ends up covering basically every available service.
> We do sewage and water "paid by use" as we do electricity and all that.
Water and electricity yes, to some extent; sewage, no (outside of places like RV parks). That said, you bring up a good argument for subsidizing water and electricity use... as most places are already doing.
Those things used to work that way at various times in the past; over time, societies consistently decided our present way of life was better.
It’s probably a good idea to be very, very sure we understand exactly why before undoing that.
I'm advocating for a higher trust society, a higher functioning society. I understand it is likely to bankrupt/burn/hurt people who cannot keep up. But I think it is a way to get a more deserving share of money towards government.
For example: You get charged a fee per couch you throw out*, a fee per garbage bag, etc. It incentivizes reuse and not throwing out things you can find a use for.
This is the same way our sewers work, both of these ideas are not rocket science, and NO societies have not consistently decided a different way is better.
Utilities work this way, and they're quasi-govermental at best. Toll roads work this way, as do (to a limited extent) subways/buses/trains.
*yes, this is lived experience. Tell my town "we won't take that" and they'll just chuck it down the hill. Cue refrigerators and other dangerous (rat-breeding) objects in very hard to remove places.
“You get charged a fee per couch you throw out*, a fee per garbage bag, etc”
This sounds like it’s either going to be immensely corrupt, or lead to hugely bloated administrative costs for enforcement.
It seems like you are missing the point that service charges make sense in some circumstances and not in others, and trying to force that round peg in a square hole isn’t necessarily going to work the way you think it will.
Welfare and military are the biggest chunks of the spending; if anything, you should ask people how they want to split their taxes between welfare, defense and "everything else", otherwise it's just a lot of extra work for no real outcome.
If people don’t want to put in effort to see where their tax dollars go, that’s fine. Their taxes would just split based on the status quo.
To be clear: are you saying you should be able to tell the (presumably US) government how to allocate the budget, or how to allocate your share of taxes?
If it's the former, then you're not going to end up with anything much different from the current USG. You get a vote on how the USG divides between welfare, health care, military, education, roads, law enforcement, environment, research, nukes, art, and so on, and so does everyone else. The catch is that you know next to nothing about what any given part of the government needs unless you're a specialist, so you'll have strong opinions about whatever you're a specialist in and weak opinions about everything else, and most people will have weak opinions about all of it and delegate their decisions to their choice of specialist. We call that a Congressperson.
If it's the latter, then you've essentially proposed anarcho-capitalism without actually saying the word. Everyone gets to say whether all their chips go in military, or all in welfare, or an even split, or whatever. There's a certain fairness to this, but a lot of people seem to not like that fairness and will vote for "current status quo" instead.
Some people will no doubt make the argument that if everyone gets to allocate their personal share of taxes, then the government is functionally a plutocracy where anything only gets done if someone rich wants to pay for it. Again, certain fairness to that. Especially if you maybe add a clause where no one gets to allocate money toward printing more of it.
...except not the military, OP wants that specially protected even in the event that no-one wants it funded.
Ok let’s say we test run 1% of all discretionary spending, even military spending. That’s still not that much money in the grand scheme of things and everyone could see what they think of it.
Sure, that's much more palatable (assuming we pin down "discretionary" to avoid a no-vehicles-in-the-park situation). 1% universal anarcho-capitalism, let's goooo.
How do we judge the results? How do we even measure the results? What sort of measurable changes might we expect to see? How do we separate the trial effects from the effects of everything else? When and how do we decide whether to expand the project, keep as is, or scrap it?
> Now some things, like welfare and military, would not be subject to
...nope, you've lost me here. /I/ would want to be able to turn military spending down, and also to prevent short-termists from reducing scientific funding. I am forced therefore to prefer the status quo to anything like your proposal, since that world is better than one in which all effort is spent continuously depending my pet interests while trying to cut everyone else's just so we can all keep treading water.
For all you know, science spending would go up under my proposal since NASA is one of the most popular government programs.
You can’t have all of military spending be subject to people’s preferences because too little spending could be a disaster. Although I could see an argument for making some of the military spending subject to allocation. Outside of welfare, it would be the hardest thing politically to make allocated.
Also, someone like Trump couldn’t just decide on his own to cut research funding unilaterally.
Too little spending is already a disaster for the military.
From a legal/constitutional perspective, the President CAN'T unilaterally cut research funding. So unless you have a way to prevent people from breaking the law, your proposal doesn't change anything about that.
This is more dependent on how it's funded than anything. I'm going to guess that the President is yes, howdy able to reprioritize CIA or FBI funding currently running through normal research channels (you fund this, we pay 3x its price, and NIH gets the rest for whatever it wants), as those are not directly mission critical, nor are they legislatively earmarked.
... am I wrong on this?
For all you know, military spending would also go up, since the military is practically worshipped by large chunks of the population.
You can't have all science spending be subject to people's preferences, because too little spending could be a disaster. I've worked for companies who cut back on R&D and I've seen the slow death play out; I have no desire to see this on the scale of a whole country.
I can't trust you with science spending decisions any more than you can trust me with the military. Either both should be subject to the public's whims so we can both be counted, or neither should. You don't get to ringfence your favourite cause while aiming a cannon at mine.
In both the military and research cases, we could see ardent supporters donating billions only to see those billions misspent in appalling ways, such as weapon prototypes that will never make it to deployment, or research that will never yield anything conclusive.
We could try to curtail that by giving people even more fine-grained control over where their money goes, down to the project level, but most science cannot be practically understood by the people who will be affected by it, and most military work is necessarily secret. The result could well be that esoteric science only gets funded if scientists tackle the hard problem of explaining it well to people with money to allocate, and military projects are funded by people allocating to a general fund that may sometimes goes to specific projects they wouldn't approve, but can't know about, and the detailed allocations are handled by experts who are authorized to know the secrets.
If that allocation is done as a collective vote, the misspenditures will be amplified. But this is essentially the status quo; we currently put taxes toward things we probably wouldn't approve ourselves. This can't really be helped in the military case, but in the scientific case, one could at least select which things to study with one's tax share, or delegate that to one's trusted expert.
Like I said, you could start off small as an experiment and we could see what happens. 70 billion dollars is 0.1 % of the budget. I don’t see it being that terrible to let people allocate 0.1% of their taxes.
Also I’m sorry but the military is just more important. Science funding dropping to zero would be really bad. Military spending dropping to zero would be an utter catastrophe, destroying our country. That doesn’t apply to all military spending but it’s more essential to have a baseline than anything else. That’s not my idiosyncratic opinion but just the nature of how governments work.
HalachaBench: 24 AI models on real Chief Rabbinate semicha exam questions, in Hebrew, closed-book. Fable leads (79.7), and the older GPT-5.6 Sol beats the newer GPT-6 Sol by 29 points. Rankings track legal benchmarks, not general intelligence indices. github.com/halachabench/halacha-bench-results
How do human rabbis fare on these tests ?
Good question. On the actual Rabbinate exams, 65 is a passing grade, and grading is ultimately up to the examiners’ judgement. So by that standard, the top models would pass. We don't have a direct human baseline on this exact set yet, though.
Have you tried prompting one of the winning LLMs to produce a cogent and theologically sound argument for why it should be appointed as a full-time rabbi ?
Question for people into podcasts. Does this cut into your free time that you might have spent reading?
I haven’t listened to too many podcast, a little afraid that I’ll get hooked on them.
Based on personal experience, I would strongly recommend against podcasts for infophilic types. The potential for addiction is extreme--think David Foster Wallace watching 12 hours of TV a day plus being able to choose from an unlimited supply of the world's most interesting and charismatic people talking about exactly what you want to hear.
Not much, as I mostly listen to podcasts when I'm walking somewhere. I guess I listen to podcasts on the bus when I could be reading, but that isn't much of my time.
Podcasts have mainly displaced listening to music for me.
I'll generally listen to podcasts only if there is a single host. Otherwise there's often too much chit chat.
Like others here, I only listen when driving. I'll add that while this obviously hasn't cut into my reading time, it has noticeably reduced the time I spend listening to music.
I listen to podcasts on long car drives. I do notice that this makes me more willing to go on long drives. Since I hate driving, I'm happy with this. It doesn't cut into my reading time because I can't read while driving.
Audio alone typically isn't engaging enough for me, so I only listen to podcasts when I have some other low-mental-energy activity I'm already engaged in (mostly driving, walking, chores, or video games). It's possible it cuts into my reading time at the margins by making other leisure activities slightly more entertaining than they otherwise would be, but not to any extent that I've noticed. I've never sat down to listen to a podcast.
Honestly, the best thing that ever happened to me with regard to my reading habits was losing my kindle on vacation. I ended up just downloading the kindle app on my phone as a stop gap, and now I'd never go back. Every random little moment of boredom throughout the day that I used to fill with doomscrolling, I can now just swipe over to the book I'm reading for five minutes. I've been trying to train myself to instinctively go to the kindle app before social media or news when I'm bored.
I usually listen to them when walking around or doing chores. So that time wasn't going to go to reading either way.
I'm planning on moving from a career in product management and tech consulting to AI safety in some fashion, likely as a generalist or in a product role for tooling or services. I'm taking the Blue Dot courses, exploring Project Tailwind, and applying to some career transition fellowships. I'm near SF. Doo you have any advice for me on establishing a network, unmet needs in the space to look at, how to be successful in joining the field as a non-researcher, etc.?
https://www.youtube.com/watch?v=wuXip1SCJGs&t=8s
They're getting a handle on prion disease in mice. This has been so challenging with so many steps that I'm not sure I can do it justice in a summary. You might want to listen to the video.
https://en.wikipedia.org/wiki/Fatal_insomnia
As a minor point, I found out that Fatal Familial Insomnia is a prion disease. I thought it was still a mystery, and I can't imagine how a prion disease can be inheritable and take decades to manifest.
I think (not sure) that the heritability and delay are both explained by the relevant gene encoding an alternative version of the protein that's more prone to forming prions, so people with the allele will occasionally spontaneously form prions but they still take a while to grow to destructive levels.
PSA for Open Thread posters: get to the point quickly. If you don't already have a reputation for writing things worth reading, people won't slog through several meandering paragraphs to find out whether you have something worthwhile to say. With GenPop, you get 5 words. Here, you get 50.
Top 10 prison general population conversational gambits (5 words or less)
10. Riot at lights out
9. Spot me here asshole
8. I got a shiv
7. Toilet wine for sale
6. Move your punk ass.
5. Gimme your pasta
4. Vito is a rat
3. Shut the fuck up
2. That screw gonna die
1. Fourth for bridge, anyone?
As a reply to 1)-
ACOL or Standard, muthafucka?
Hopefully that's under the wire, wordcount wise.
One man's bug is another man's feature. People who can't read more than a few dozen words in a row don't engage me in conversation; how is this a problem?
It's not a matter of can't. It's a matter of "I'm a couple paragraphs in and I still don't know what this post's point is or whether it even has one".
I'm not against long posts (I wouldn't read this particular blog if I were). It's just if a post is long, I want to know early on whether a post is on a topic of interest and where it's likely to have anything worthwhile to say on the topic.
>how is this a problem?
You might, in addition to people who can't, be filtering those who won't.
...still not seeing a problem here.
QED
Stop me if I fail to take your meaning correctly, but it seems to me - and I have to agree, this is only my personal opinion, the view of one person, and not to be generalised from - that you are perhaps intimating verbosity is a flaw often found in those who post to Open Threads, though again you may not intend to indict all or even the majority of posters and have a select few in mind whom, charitably, you do not reveal openly to the scorn and mockery of the mostly succinct posters but rather drape the cloak of anonymity over any such, as you make the point - which I hope I understand correctly? - that brevity is the soul of wit, a hallowed by long usage observation first encapsulated, if I remember correctly (and my memory is increasingly fallible as age lays its cold hand on me) by the author commonly - though not uncontroversially, as identification of the person in question has been much debated even to the present day, where gender of the alleged or supposed author is not settled, nor race, nor identity - commonly though perhaps, as mentioned, mistakenly referred to as "William Shakespeare"?
Forgive me if I have trespassed over the arbitrary limit of 50 words, I am ancient and feeble-minded!
See? 26 words in, it’s clear what your point is. We can work with that, here.
Exactly!
No footnotes?
Indictment superfluous, verbal diarrhea self-evident.
This reminds me; if GenPop gets 5 words, how many do we get for RingPop?
Who RingPop
You may have failed to notice a joke.
Right back at ya.
So if I'm reading this right, what you're saying, and what I'm hearing, and what the message is, in regards to recent commentary from other posters, presumably in this thread, but maybe in response to vagrant commentary elsewhere, like the Links post, or the Hidden Threads (except you've specified the Open Threads, so that doesn't really seem like the most logical conclusion, being overspecified to the current location, and possibly its numerical predecessors) is that the general commentary style has drifted toward becoming more elaborate over time, perhaps opening with and otherwise invoking twisting and convoluted speculations irrelevent to what a careful reader would consider the heart of the post, making it strenuous to continue reading with intent as a member of the audience, though still engaged in the pursuit of knowledge that the post might well contain, were it not buried in a distracting mass of unnecessary side discussions.
I haven't actually noticed that.
Oh shut up.
Dear and valued co-conspirator, if I may so refer to you as ne of the ACX commentariat, I trust it will not seem forward in me, much less admonitory, if I mention that you have not clarified what is to be shut up: shut up shop? shut up the doors? shut up and dance?
For those of us much in need of the advice to be pithy in 50 words or less, our enfeebled brainpans need explicit guidance! Until NoSignalNoNoise undertook our instruction, we had no idea of our folly or faults!
Concise and to the point. Well done.
So who put you, entity of undefined signal to noise ratio, in charge of policing word count? Are you, perhaps, an agent provocateur attempting to create diversion and division? In order to show my defiance of this new rule I shall make the word count of my comment precisely 51. Hah!
You fell right into my trap by writing your first 50 words in a way that set a clear expectation for the 51st!
But could you have known the precise 51st word without reading it? Hardly coconut!
This made me kek
…it occurs to me that it is 2026, and we may not have superintelligent AI yet but we have software we’ve literally written to predict the next word in a sequence of words, and whatever purpose various multibillion dollar companies now try to put it to, /that one/ it is actually quite good at.
Kelowna, Canada just elected the country's first avowed Georgist in generations. I interviewed him and wrote a guide for Georgists running for office: https://substack.com/home/post/p-222319048
The London "Pause AI" protest targets the wrong bottleneck. We must treat computation as a contested physical resource instead of an existential monster. Artificial intelligence is thermodynamic force channeled through silicon. The current corporate trajectory relies on gigawatt-scale data centers spread across thousands of acres, sucking up water for cooling and monopolizing regional power grids.
Panic accelerates regulatory capture. Fixating on science-fiction superintelligence distracts from the actual danger: the quiet enclosure of human knowledge by three or four unaccountable boardrooms operating closed-source algorithms. These monopolies are building a private surveillance layer over the public internet. Demanding an outright halt to the math accidentally serves the corporations already entrenched in these massive facilities.
We need democratic control over the compute layer and a total commitment to open-source architecture. Computation is our best tool for mapping climate physics, managing decentralized energy grids, and modeling the exact protein folds required to replace the slaughterhouse system with precision fermentation. We cannot afford to halt scientific progress when billions of animals that feel pain and fear remain trapped in industrial cages, and the biosphere is running a fever. We have to democratize the silicon, break the monopolies, and put those gigawatts to work for the public commons.
> We need democratic control over the compute layer
Whoever buys it, owns it. If taxpayers want a democratically controlled compute layer, those taxpayers should lobby to use their taxes to build one, and a government department to manage it, as is done for all other public services and infrastructure we agree really is critical enough to be controlled this way; and accept the result if their side turns out to be in the minority when voting on the proposition happens, just the way they would want the other side to accept the result should they win.
That is the correct democratic lever for accomplishing this goal.
Good pushback, @moonshadow, but the maxim "whoever buys it, owns it" breaks down when the purchase physically annexes the public commons. A toothbrush is private property; a gigawatt-scale data center that drains regional aquifers for cooling, monopolizes local electrical grids, and trains its closed-source algorithms by scraping the open internet is an enclosure of public resources. Microsoft and Google didn't just buy silicon. They enclosed the collective output of human knowledge.
Building a state-run compute bureaucracy with tax dollars is one lever, but it isn't the only democratic option. As an advocate for economic democracy, I don't want a centralized government department mirroring the tech giants. We need aggressive anti-monopoly action to break these conglomerates apart. We can mandate open-source architectures, enforce strict thermodynamic limits on corporate grid usage, and shift infrastructure ownership to decentralized worker cooperatives. The physics of our climate and the enclosure of our digital civil liberties demand more imagination than simply letting four boardrooms buy the future while we politely wait for a ballot measure.
Claiming that they "enclosed... human knowledge" seems rather hyperbolic. They are just providing optional tools to explore it.
Or.... and hear me out... we could steal it from the people who created it and put one of our cronies in charge.
Technically, you can't steal it from the people who created it, since they don't own it. Their bosses do.
What will we do should the superintelligence say that global warming doesn't exist? Lobotomize it? Tell it to stop ruining the Green Economy?
Any good chemist should want to reduce waste and excessive use of toxic chemicals -- this is not "The Green Economy," it is conservative in nature, despite being inventive.
Fair question, @Aeolian321, but if an AI claims global warming doesn't exist, it hasn't achieved "superintelligence"—it has simply hallucinated against basic atmospheric physics. AI is a pattern-matching tool, not a digital oracle with independent opinions to lobotomize. If a model denies the measurable radiative forcing of CO2, it is failing to process empirical reality. We wouldn't be censoring a brilliant mind; we'd be correcting a broken algorithm that ignored the thermometer.
You are absolutely right that good chemistry means reducing waste and toxic runoff. That conservative respect for physical limits is exactly what I am advocating. Burning fossil fuels and funneling billions of tons of crops through animals to generate human calories are the most thermodynamically wasteful, toxic processes ever invented.
We don't need political buzzwords like the "Green Economy" to see the math. We need open-source computation to track these hard physical limits and help us engineer public infrastructure that actually aligns with the chemical reality of our shared biosphere.
CO2 is one thing, but clouds are another. I spoke with one of the best modellers in the business, and he doesn't believe all this global warming is all it's cracked up to be.
There continues to be substantial debate on the topic, most particularly when one starts comparing "predictions" to "actual results."
> What will we do should the superintelligence say that global warming doesn't exist?
Do you think this is likely?
Crystal ball says the future looks cloudy. Check back later.
https://climaterealists.ca/clouds-not-co2-key-to-understanding-climate-nobel-winner/
Siding with Clauser on this one.
"Just a 5 percent rise in cloud cover can largely counterbalance the temperature effect of doubling atmospheric CO2"
The argument that a small fluctuation in cloud cover can cancel out the warming impact of doubled atmospheric carbon dioxide rests on a profound misunderstanding of how Earth’s climate system actually functions.
When climate skeptics cite figures like John Clauser’s assertion that a 5 percent increase in cloud cover can completely counterbalance the greenhouse effect, they treat clouds as an independent, static control knob for global temperatures. They are not. Clouds are a dynamic feedback mechanism deeply coupled to the very thermodynamic forces driving the climate in the first place.
Here's what our resident superintelligence has to say on the matter:
https://chatgpt.com/share/6ae9cd62-083c-83eb-ac46-632952825fac
HTH, HAND
Scott was kind enough to link my "Against more marriage as a solution to the fertility crisis" post, along with a critique, in his last links post.
I greatly appreciate both the linking and the critique.
Scott’s AI-facilitated critiques are of the flavor “technically true but of the sort that don’t substantively change the shape of the argument or the overall points or conclusions,” because the parts he most justifiably critiqued were a grounding on things we already see in the world from multiple independent sources (high divorce rates, high rates of one or the other party in a marriage being unhappy, high dead bedroom rates, ridiculous government spending compared to median tax contributions), rather than the basis of the argument itself.
Was my choice of data source and analysis flawed? It was, and Scott got me dead to rights there, and I thank him for doing so.
Does it change the thrust of the argument or conclusions? It does not, divorce and “one or the other unhappy rates” are still sky high, even if the measurement I chose didn’t show that well. Additionally, the median taxpayer is very much NOT a net taxpayer, they are net tax consumers, and that remains true very far up the income ladder, and we can show it several ways.
I've now written a response post to the critiques, grounding my points in better data sources and with more rigor, and including documented methodologies so Scott or anyone else’s AI’s can audit and replicate.
We are moving to a world of greater auditability and replicability, which is great! This allows us to surface substantive disagreements more readily by abstracting away the points which largely don’t matter.
So thank you, Scott, and for anyone interested in following this exchange, my response post is here:
https://performativebafflement.substack.com/p/defending-against-more-marriage-as?r=17hw9h
Sorry for being blunt, but your whole fiscal argument is a mess.
Intuitively, if you have fewer workers per every non-worker and you want to provide non-workers with food, clothing, shelter and care, the workers will need to work harder.
Also, if you had every US citizen contributing and consuming exactly the same amount, they would all be net negative because the US runs deficits! Would you conclude from this that no kids should be born in this case?
More concretely, it would be great if you reran Panel B with your own Panel A convention (no defense or interest), add the omitted revenues, and report the mean net value per birth. Neither defense nor interest goes down in a society with fewer workers per non-worker.
Thanks for taking a look, I appreciate the feedback.
> Sorry for being blunt, but your whole fiscal argument is a mess.
It's calculated two ways, both from a lifetime NPV perspective and using simple arithmetic from top level government numbers, and comes out the same for both.
On the deficits, given that we run up debt specifically to provide per-person entitlements to old people, and given that said entitlement spending is so crazy that it means only people above the 80th percentile of lifetime income become net neutral, then yes, we should think twice about incentivizing marriage *specifically in low income couples,* which would be the net result given that the top quintile by SES is already 80-90% married. Because essentially everyone gets old, and will cost that spend.
> More concretely, it would be great if you reran Panel B with your own Panel A convention (no defense or interest), add the omitted revenues, and report the mean net value per birth. Neither defense nor interest goes down in a society with fewer workers per non-worker.
There are no omitted revenues from people, both methods are taking payroll taxes into account. But if you wanted me to add the ~$800B in corporate taxes, I'll do that.
And why would we run it with no defense or interest? We don't have to pay for those things? I really don't get this "it's not fair to allocate defense and interest per head" argument, the expense doesn't magically go away.
But sure, here you go, with payroll, corporate taxes and tariffs, and income taxes, and assuming $2T of expenditure on defense and interest magically disappears, bringing our expenditure to $5T instead of $7T:
https://imgur.com/a/lQn5o4I
It brings break even down to the 73rd "lifetime income" percentile vs the 80th. Median and below people are still net negative.
Yes, people definitely break even at a lower percentile when you line item exclude 2/7th of the entire expenditure budget. I'm not sure what that really shows, though. It does NOT change my "against 'more marriage' argument" though, because the bulk of incentivized marriages would be at the medians and below, which are still net negative.
Oh I realised I haven't answered.
How do you account for the fact that if the defence and interest expenditures remain the same, and there are fewer people around, each one will have to pay more?
You took a tautological fact that in a economy with progressive taxation and redistribution the contributions of the median resident will be negative and are making conclusions from it.
Also, I didn't quite understand your model - if you have fewer births, you'll have fewer productive and fewer unproductive citizens. Suppose now the ratio is 1:4, if it stays the same base rate AND more non-workers per worker, the situation will be strictly worse.
On the "tautological fact" point, I don't think that most people realize that our system is so dumb only 80th+ percentile lifetime income people are even net neutral, and it takes mid 90's for "actual contribution."
I think if people ever DID think about it, which they don't, they would think "well probably what happens is we're break even at the point the average person pays taxes in, then as you go down the income ladder, them not paying much tax is made up by higher income people paying more on the other side" That's not at all what happens. It is nowhere near balanced or proportionate, we spend so much that the vast majority of people, basically, 80-90%, are net liabilities. This is bad for many reasons, not the least of which it's entirely unsustainable, this is exactly WHY social security is running out.
Nobody thinks this, so I am trying to elevate it to more public awareness.
Bugmaster's AI journal:
Just for fun, I thought I'd catalogue my experiences with LLMs. All of these use either the latest frontier models, or a mix of those with cheaper models that are one tier lower. Hopefully people will now at least pause a little before telling me that my pessimistic stance on AGI is entirely due to lack of effort.
Task: Trace a 2D bitmap image to produce a vector SVG file. The image already looks like it might have been rasterized from some vector format.
Result: Abject failure. Spent about an hour iterating on prompts/conversations. Finally gave up, used an algorithmic tool that produced almost perfect results, fixed the flaws myself in 15 minutes.
Task: Generate several images intended as illustrations for an RPG campaign, featuring specific characters/vehicles/mechs in specific scenes drawn in specific styles.
Result: Partial failure. Iterating on prompts eventually leads to hallucinations; starting new conversations helps, but it appears to be impossible to get the LLM to fulfill all the requirements instead of just a few, and to stay consistent across multiple images. Still, some of the results are usable.
Task: Vibe-code a Web application with a relatively simple UI and backend; the purpose is to download public data from multiple sources and perform some statistical analysis on it.
Result: Partial success. The UI portion was created almost immediately and worked almost perfectly out of the gate; minor adjustments and fixes were handled well. The backend had many problems out of the gate (broken downloads, memory leaks, other "mechanical" bugs). The statistical analysis was completely nonsensical at first, but improved over time to the point where it's merely useless. It's possible that there's just not enough data to analyze, but if so, the LLM refuses to admit it.
Task: Use a different LLM to code-review the above vibe-coded application and detect any bugs.
Result: Success. The LLM identified dozens of bugs, some of them critical, which were ignored by the original coding LLM. The coding LLM was able to fix them and write unit tests to cover these specific failures, but was unable to detect new (and similar) failures.
Task: Put together a biological analysis pipeline and run it on public data.
Result: Above average success. The LLM was able to gather the requisite public data, download it, and organize it intelligently (which is impressive given how terrible human biologists are at organization). It put together a pipeline using open-source tools in a sensible manner. The pipeline promptly failed due to out-of-memory errors, but this isn't uncommon even for human-generated pipelines. Very impressively, the LLM was able to independently identify and fix a bug in one of the open-source scripts. On the minus side, the parameters set for many of the tools were inadequate, and the pipeline required lots of manual tuning to produce useful results.
Task: Provision AWS resources for the above pipeline.
Result: Above average success. The recommendations made sense and the suggested commands (installing packages/configuring permissions/cron/etc.) were useful, but the LLM made some bizzare choices at times that caused problems down the line. Still, the overall plan did work. The LLM's projected running costs were wildly off the mark.
Task: Write a small standalone add-on to an existing application with a large human-written codebase (more than 10 years in the making). Base the code on a human-written scaffold.
Result: Partial success. The code sort of works, and had improved after many multiple iteration passes to correct massive bugs and cripplingly inefficient design choices. However, the LLM keeps duplicating code over and over again, e.g. writing helper functions that already exist. To be fair, some humans tend to do that too, so perhaps we should not judge it too harshly. Overall the code is verbose and bordering on incomprehensible; the comments never use one sentence when 20 would do.
Task: Perform miscellaneous research on a variety of topics.
Result: Partial success. The LLM finds relevant articles on the topic and summarizes them reasonably well; but it's not really doing anything that a simple Google search (in Google's pre-AI days) would not have found, so its utility is marginal.
Task: Put together a Lance of 'Mechs for a Battletech boardgame engagement against frustratingly competent human opponents.
Result: Partial success. The LLM suggested a "boring but practical" Lance composition that performed unexceptionally. On the plus side, the LLM correctly handled the somewhat convoluted custom Lance composition rules.
Task: Write a design document for a UI-heavy Web application with (relatively) complex GUI that is supposed to feel powerful yet intuitive, powered by a P2P networking backend.
Result: Success. The LLM took basic prompts and turned them into useful design documents. Its suggestions made good sense, and it offered some useful ideas and highlighted weaknesses in the original concept. It also devised meaningful implementation milestones.
Task: Implement the abovementioned design, in stages (as suggested by the LLM).
Result: In progress, but not looking great so far. The implementation contains horrific bugs that not even the laziest human programmer would ever dare to ship. These can be corrected with further prompt iteration, but the sheer number of bugs slows development to a crawl.
As someone still figuring out how to make this stuff useful because my work-approved vendor (copilot) seems awful, this is great to read. If any of the successes here were in copilot, I appreciate any tips on what I'm likely doing wrong.
My employer recently switched us from Claude to Copilot. My experience so far is that the heaviest AI users uniformly hate it, but the output I am being exposed to indirectly seems improved, especially on the code review side. Copilot seems to lead people down the garden path of faulty assumptions less often, and it seems better at finding corner case bugs in code reviews especially. It isn't great at optimization, often over-recommending caches that will rarely be hit and using less-performant code idioms, though.
My guess at reconciling direct and indirect perceptions of Claude vs Copilot is that the latter might be doing less to hide friction from its direct users, which makes the experience feel worse but potentially give better results. But that's just a guess at something that might explain it, as I haven't made enough direct use of either to confirm.
I really appreciate your detailed response. I don't work in coding, but your point about friction reduction is interesting and potentially relevant nevertheless. My frustrations come from much more menial tasks that I would expect to be simple for AI at this point, like creating a flyer from a detailed prompt. However, in my most recent request for this, copilot made a "1-page flyer" that only utilized 60% of the space, using size 8.5 font in many areas. When asked to use all the space and increase the minimum font size to 12, it complied with the font size and clustered everything infinitely close together...still using only 60% of the page. 🤦🏼♀️
Yes, in my experience all these LLMs are absolutely terrible at spatial reasoning. You can get them to do the right thing but only in broad strokes, e.g. you can say "pin this control to the top of the screen", but you can't say "lay out the controls so that functions for X are easier to reach than Y". And yes, any changes to font size etc. are likely to rearrange your entire page, because LLMs generate these layouts stochastically -- they are incapable of creating an overarching plan, unless you spend more time on doing it for them. Of course, then you're cutting deeper and deeper into their limited context window...
VS code's Microsoft copilot seems perfectly capable of using claude/gpt/grok.
Sounds like some user error on my part. I'll have to explore more. Starting with the definition VS code… :P
Trust me, it's not user error -- not exactly. It's just expecting too much from a tool that sounds intelligent, but isn't.
Sadly not; I mostly use Claude and Codex/ChatGPT. I am also trying out Gemini Antigravity; if nothing else, their UI is pretty good.
It would be nice if you posted the exact models and prompts someplace (for example, on GitHub) so that others could reproduce your results.
I can't really do that across the board since some of this work is proprietary (dealing with data from my company), and some is personal (dealing with my personal data that I don't want leaked).
That said, I'm almost certain that if e.g. 100 people tried reproducing my process, some of them would get better results just by chance. For example, if I tried that image-tracing task a bunch more times, maybe I'd get something better than what the algorithmic tool can produce eventually. I am also virtually certain that everyone who reproduces my exact prompts and input files on the exact same model will get a slightly different (or even wildly different) results in general; I know this because I tried it myself (submitting the same inputs multiple times) and that's what usually happens. But this sounds like a strike against advanced LLM capabilities, not evidence for them.
Overall, I think LLMs can be useful. At present, they are saving me some time on peripheral tasks, as I am getting better at "knowing when to hold'em, knowing when to fold'em" (instead of getting sucked into prompt loops). They can also serve as a last-ditch stopgap for tasks I am utterly incapable of completing myself (like drawing art). But any claims that these tools are anywhere as intelligent as humans, or even *super*intelligent, are wildly exaggerated.
I think at least citing the models used is important.
I agree that AI is in a somewhat paradoxical state. It can do math and other tasks that humans find difficult or impossible, but it still falls down regularly on straightforward work. I do think the labs are going to try to fix the latter aspect as quickly as possible. If "dumb mistakes" is fixed (along with at least mediocre continual learning), that probably gets us most of the way to AGI.
> If "dumb mistakes" is fixed (along with at least mediocre continual learning), that probably gets us most of the way to AGI.
This is true, but only in the same way as saying, "if I could fly and shoot lasers from my eyes and deflect bullets then I could be a superhero". LLMs are structurally incapable of overcoming their deficiencies -- but some other future technology surely could.
Scott: If you post a substantial article elsewhere, please either duplicate it here or include some notification about it here (probably in the next Open Thread intro). I didn't find out about "Against SURA" until just now, & that only because another blogger I sometimes read happened to respond to it.
The man is so prolific he really cannot be bothered collating all of his work. In the past at meetups people have done that for him, giving him bound sets of his work.
There's not just this blog. There's all the LW posts. Then the fiction book. Reddit comments and posts. It never ends
When it was reported last week that US Military refrigeration systems were being taken down, I assumed it was the Iranians—since they've upped their hacking/cyber security skills after the Israeli/US Stuxnet attack. But now that refrigeration systems are being taken out across Europe, I suspect it's the Russians who are doing it. North Korea has the skills, but the Kim regime. Thousands of refrigeration management controllers (such as Danfoss and Dixell/XWEB systems) are running outdated software without security patches and are directly reachable online.
While the US Military incident is probably the best explained by forgeign adversaries, I would be careful not to assume that the similar attacks on supermarket freezers were caused by the same entities.
There are multiple possible explanations, some of which appear to be much more plausible for me.
I take it most people around here had not worked in retail or e-commerce, and might not have an intuition for just how ruthless the competition is.
I say this as our systems are under constant DDOS attack, our prices are being parsed by multiple competitiors who then set their prices a percent or two lower(most of the time they would not be able to sell the product thatcheap either), our entire content was copied multiple times, there were multiple attacks via random legalese, and multiple attempts to do something that does not exactly break the laws, or puts us in a bad position. Bought negative reviews, obviously fake, are not rare either. In one case a CEO of a competing company wrote a fake negative review from their personal account, which we easily traced back to him (not the brightest guy). One competitor registered a fake manufacturing company company under the same exact name as our, just to mess with our discoverability. Technically a different type of company in a different industry, so we had no legal case under our country's laws.
And so on and so worth.
If our competitiors found a way to physically destroy our good without getting into legal trouble, they would do that in a heartbeat, and would be willing to pay premium for it.
I say all of that as just an intuition pump for how bad things can get - and I believe we are still having it easy, as a mid-sized company in a niche industry. Some certainly have it much worse, and retail is much worse than e-commerce in general. Now on to my main point.
On priors, the most likely culprit is the competition of the supermarket chains that were affected.. Then the competition of the refrigerator manufacturers which were affected.
Only after we rule out those would it make sense to consider black hat hackers working on their own, edge script kiddoes who want to see the world burn, adversarial goverments who want to cause some damage to the other country via third order effects, grey hat hackers who want to highlight the vulnerabilities, and other possible interpretations.
Competition seems...unusually lawbreaking and creative in your field. And I say this, knowing someone who has done industrial espionage before. Bought negative (and positive) reviews are cheap and avoid negative legal consequences, they are commonplace as they serve as inducements to buy. Perhaps I'm just unfamiliar with the state of the art, but you are arguing by anecdote.
Not quite, I'am merely providing the anecdotes to go with the more obvious point, that the competition has the most to win from this, while the Chinese or Russian or othery boogeyman state actors would not see much benefit from this, as far as I can tell.
Not to say that they are above stupidly destructive actions for the sake of civil unrest, of course. But I feel like motivation here is much, much weaker than for the competition.
And yes, I would assume any for-profit organisation to be just as lawbreaking as they can get away with, as long as it benefits them. The history is rich with examples, and keep in mind we only know about the ones that did not really get away with it, in the end.
I remember the capacitor plague. I'm at least aware of a good deal of lawbreaking or industrial espionage that "did get away with it."
Yes, of course, the competition has the most to win from this... I guess? It seems a little ... overly optimistic to think that most Europeans would drive to the next town over because their supermarket threw out the cold food one time. This seems "ripe for the competition to benefit" only in the case that the competition is meaningfully better, or that human customers are being inefficient to start with, and by removing that inefficiency Once, you foster behavioral change.
Aldi's came into my midwestern town. Nobody's heard of them, and considerable people are unlikely to use "the new grocery store" -- perhaps thinking them a Shop-n-save type discount store. That's the ringer that I'd expect to try a "single time trick."
I don't understand the motivation for any of them to do this. Why target civilians in a small scale hack during relative peacetime? Presumably now the people managing refrigeration systems are going to be more careful (at least on the margin).
I would think it would be more effective to wait until an actual war and then launch a mass hack when no one is prepared.
I think it's quite likely to be Russians, but not, like, Russia as a state actor, but just some bored individual Russians or some small group who just wants to stick it to the West in some random way. Since 1) it really serves no purpose to Russia in its current situation, if anything this kind of news is virtually always spun by Western media and politicians as "this is why we are in deadly danger and need to give more money and weapons to Ukraine"; 2) it sounds like there really are just tons of vulnerable freezers exposed online and trivial to hack. I don't know about US military, but in Finland and Sweden the targets apparently were mostly quite small stores, who cannot be reasonably expected to be great at IT security.
They are effectively in a proxy war in the west already, one that they can't afford to lose anymore. There might not even be a next time, so they may as well play their cards now.
Protests in germany, france, belgium just lost their government, coup attempt in Poland, the time is ripe for red revolution.
Plus, these are the countries bankrolling the "brother against brother" war in the Ukraine.
> red revolution
...wait, what year is it again?
Man, you are fascinating. I mean this as a compliment, but reading you; It's like looking through a warped mirror into the most motivated possible interpretation of current events possible.
I'm genuinely curious what your information/media diet is? What do you read? Do you put your own spin on the news, or do you have outlets that curate the narrative for you?
I'm not going to be so crass as imply your don't earnestly believe this stuff, I'm just curious how you get to where you are.
And I guess my question to you is: what are you doing with your web-surfing time? All the information I post can be found on mainstream media sites. I read dozens of news sites every day (most contemporary journalism is pretty awful, but if you read a few different sources on a story, you can generally get a good picture of what's going on). As for my interpretation, that's just simple deductive reasoning.
Plus, I try to keep up with the latest technical and scientific advancements.
"In solving a problem of this sort, the grand thing is to be able to reason backwards. That is a very useful accomplishment and a very easy one, but people do not practice it much.” —Sherlock Holmes in _A Study in Scarlet_ by Arthur Conan Doyle.
It might help you to know that the person has already been perma-banned here under several accounts.
:-)
I'm genuinely amazed at how ill-informed of current events most people are (see links below). Be that as it may, when investigating a crime, you have to ask, *cui bono?* When looking for a suspect, it boils down to who has the means, motive, and opportunity.
While any high school student could probably launch such an attack using AI agents developed by OpenAI, Anthropic, and Google (and possibly others), targeting military bases and then European infrastructure seems a little too focused for teen hackers.
A radical environmental or animal rights organization doesn't fit the bill as a suspect, because the attackers initially focused on US military targets rather than the corporate logistics chain.
I've explained why I doubt NK was involved. We can probably take China off the list, as well, because they're trying to improve their economic ties to Europe. Fucking up the Euro economies would be counterproductive if they want to make money in the short term and long term.
Although I wouldn't put this past someone trying to manipulate the futures markets, AFAIK we didn't see any suspicious puts or calls before these incidents. And hitting the US military wouldn't be an effective way to start you market manipulation (unless you want to depress the stocks of freezer manufacturers).
Who would like to stick it to both the US and Europe? Answer: Russia. They have the means, motive, and the technical resources.
Meanwhile, we seem to be sleepwalking into WWIII.
https://www.reuters.com/technology/pentagon-treating-military-base-freezer-malfunctions-suspected-cyberattack-2026-10-29/
Countries affected: France, Belgium, Luxembourg, Sweden, Estonia, Netherlands, Finland, France, and Ireland (which I grant you is odd because they've been studiously neutral toward Russia to keep their Oligarch-funded aluminum plant making a profit).
https://abcnews.com/International/wireStory/freezers-supermarkets-european-countries-hit-apparent-cyberattacks-137765494
And Sarah Paine talks about the possibility of another world war...
https://www.youtube.com/watch?v=3kuNW8hRc2U
Higher grocery prices hurt incumbent governments. Also sends a signal about capabilities without forcing more escalation.
Claim: LLMs can understand a lot about neural networks just by reading their weights, or even alter them by manipulating the weights without relying on training: https://rikiyat.github.io/blog/weight-space-programming/
Seems really bad for both interpretability and keeping takeoff slow. Not a good sign at all.
(Or more specifically: dangerous from an interpretability standpoint because a human won't be able to grasp any subtler implications of this "Weight-Space Programming" - we'll have to trust the models not to put in any fun backdoors or Manchurian Candidate style triggers for undesirable behaviors.)
The Qwen weight editing experiment is overhyped. Astra ran multiple forward passes of Qwen trying to solve math problems and found a specific bug - that Qwen got stuck in reasoning loops. It then isolated the weights responsible, then modified them to tune down this behavior. Then Qwen was able to solve the problems posed. You couldn't use this technique to create a new capability de novo. Case in point, when they tried to improve a different open model (Gemma) which didn't have the same looping bug, they were not able to.
And also, a human interpretability researcher could have done the same thing (and similar papers already exist which Astra no doubt trained on).
Some of the other results are interesting (the breakout and diffusion model ones), but they are confined to very tiny models. It's not that surprising that tiny models are more interpretable. And consider this - Astra is in the range of trillions of parameters with a context length of 1 million tokens. Without some drastic change, it's not going to be possible for a model to interpret it's own weights.
This seems potentially really important. I'm not super tech knowledgeable, though. Is it was weighty as it seems?
I'm generally a bit skeptical about expansive claims of the future abilities of AI, but my mental model would not really have suggested this was possible. If this post is accurate, it's basically like saying it can open up someone's brain and rewire it to not have dyslexia just by looking at the neurons. Nobody trained LLMs to do it, so far as I know, and it's not as though there's a lot of examples in its training data of humans doing an equivalent task, because they can't.
Also if LLMs are literally good enough to directly patch themselves it seems like they can do *some* forms of recursive self-improvement a lot more rapidly than I would have guessed.
I don't know what I should expect to happen next, but I feel way more uncertain about it after seeing this.
Seems like whatever makes it possible for AIs to tweak other AIs would also allow them to interpret other AIs for us.
Yeah, it could go in a positive direction. But it also forces us to rely on and trust the AIs.
Wow. Also yikes.
I recently watched an interview about how people self-report thinking "in words" or "in images" etc. and how difficult it is to actually get people to answer questions like "what words specifically were you using when you had that thought?"
I haven't dug too far into this, but the topic and the anecdote of a case at the end was very interesting:
https://m.youtube.com/watch?v=j0gKl-g3DNg#
From the video description:
"Dr Russell Hurlburt is a psychologist, researcher, and the leading expert in the study of inner experience.
We all assume we know what’s happening inside our own minds. But what if the way you describe your inner thoughts—your inner monologue, mental images, emotions—isn’t accurate? What if you’ve spent your entire life mischaracterizing your own experience? Russell’s research suggests that most people don’t truly understand how their minds work.
Expect to learn why you might not actually have an inner monologue, the five core phenomena of inner experience, why most psychological self-reports are unreliable, the surprising differences in how people experience thought, how experience sampling can reveal the hidden structure of your mind, whether meditation really increases self-awareness, and the astonishing story of a woman whose perception of reality was forever altered by this method.
This episode will make you question everything you thought you knew about your own mind."
I can think either in words or pictures, and I’m aware of which I’m doing.
I’m pretty sure I can usually write down what the exact words I was thinking were, too.
Would you say you also often think in neither? When I think of a busy train timetable vs an infrequent one, I get a concept in my head along with the words. It’s expressed more like an image than like words, most of its information hasn’t been unpacked into words. And also, I visualise it partly in terms of my timetable app. But the words and numbers aren’t filled in in the image, so the image doesn’t actually contain enough information to say how frequent the trains are.
There is possibly something that is neither pictures nor words in between sentences. Some hard to describe process that happens and then the inner voice assembles a sentence.
The parts of my thinking I am aware of are like inner monologue, with occasional mental images (which are not like photos, more like geometric shapes or graphs), but I am aware that I am unaware of most that happens, because the thoughts just come "too ready".
Actually, I sometimes feel like an LLM, that I do not have thoughts prepared in advance, but they kinda crystallize as I speak, by completing the usual patterns. Although it is plausible that some invisible process decides the goal I am getting to, and then an LLM-like process generates the right words to get there.
I wonder if we could get an "objective" answer using brain scans, where people would solve some problems and the scientists would measure which parts of their brains activate.
Sifting…
Pondering…
Weighing her claims about her own cognition…
Drafting a response…
(☝🏻This joke showed up as a vision that I then described. 🤣)
Does speaking crystallize the thoughts for you? Sounds like when your ideas are more abstract, you can speak them into a workable form (like an LLM). Can you write down or speak your inner monologue too? How/what do you write?
When I’m speaking, I describe it as a vague impression of my actual thoughts, I can’t recreate the inner monologue unless I’m alone. I speak in scripts, so I can’t “think out loud.” I can abstract out loud, but I have to vaguely describe a loose idea, and it’s not often pretty. I write my thoughts.
When I talk to other people, I think much more clearly than when it's just in my head. Even talking to myself aloud achieves part of the effect. I guess the effect is in having to finish the sentences, and then hearing what I just said. When just thinking to myself, it is easy to get distracted and lose the chain of thought. I guess it's as if I think in multiple threads at the same time -- when speaking, I have to choose one and finish it; when thinking, I kinda start all of them, but often don't finish any, because I get distracted by something else. (When I started writing this paragraph, I didn't know how it will actually end. So in this process I figured out something.)
> Can you write down or speak your inner monologue too?
Nope, too fast, too chaotic, incomplete sentences with some feelings and images... The process of writing or speaking is much slower.
> How/what do you write?
When I write something, I first imagine talking to someone (one sentence at a time, or maybe just a part of sentence), then I write it down.
I am generally suspicious about describing my own thoughts, because the mere fact that I observe them already changes them. The "processing power" spent on observing the thoughts is not spent on thinking them; when I try to observe my thoughts there are fewer of them.
Trying to find examples of visualizations... e.g. when I imagine doing math, such as multiplying matrices, it's as if a very-low-quality movie plays in my head, depicting what would be done on paper. It is very low on details, gray, just a shape here, a shape there, and a movement from here to there.
Interesting on every point, thanks for opening up your brain. Sounds similar to an ENTP friend of mine, if you know anything about the MBTI. She says her brain is chaotic and associative and full of abstract ideas, but they come out logical and sequential when she speaks. She doesn’t know what she thinks until she talks it out. Both processes happen in my brain at the same time. I can feel the subconscious, associative, ideas-generating portion driving my clear inner monologue.
To write, you imagine talking to someone. So all of your writing is geared towards and therefore dependent on the specific audience, and that’s kept in mind the entire time? What if you don’t know your audience? Or is it just that you imagine anyone, like talking out loud?
Interesting that you visualize math. Were you able to show your work in school then?
When I first encountered the question “Do you think in words, or images?” I sat for several minutes, trying to figure out the answer. I became more perplexed the longer I took: surely I should know the answer? Surely it should be apparent? But it wasn’t. I kept waiting for words, but there were no words. I kept waiting for images, but there were no images.
Eventually I decided that I mostly think in words, but since I was trying so hard to “listen” to my thoughts, I wasn’t thinking in words about anything. I was just thinking about observing, and then had nothing to observe except my own observation.
I can flip between modes and I imagine most other people can too.
If I try to picture my walk home, I don't need words. While writing this sentence, I don't need images. When imagining how a rotated object would feel, I think mostly with touch and to a lesser extent images. When thinking about maths, I don't strictly speaking need either, but I can use auditory and visual processing to help me remember and process things.
I'd be surprised if most other people didn't do this to some extent.
> I kept waiting for words, but there were no words. I kept waiting for images, but there were no images.
Correct.
> Eventually I decided that I mostly think in words
But... you don't. You were aware of that. Just because someone asks you "is your favorite color purple, or is it green?" doesn't mean you have to actually believe that it must be one of those two things.
Why should it be that we think in either words or images? I believe we think in thoughts. Clearly at times there are words and whole sentences involved, also images and movies. But the main medium is thoughts. Maybe sort of analogous to AI's being able to understand images and also strings of words, but also being able to communicate with each other via a third medium, "neuralese." Thoughts are our internal neuralese. Or internal vectors.
I mostly agree here, ...I think.
As an Intuiter, I know i just DO things, I observe, analyze and course-correct without my conscious participation. Sometimes it's good, sometimes its bad.
Definitely better than the midwit "instructions are gawd" though.
'We all assume we know what’s happening inside our own minds'
- no, no I do not.
I have utter faith in my mind's continual efforts to delude my thinking self.
(I can see the quote is from the video byline, not the commenter)
This is where I'm at too.
I definitely have an inner monologue, whether I am fooled into thinking I have one or not.
I'm very interested in it for myself and for trying to understand others. (Unsymbolized thinking: I'd call that abstract thinking, thinking more in the subconscious with ideas rather than literal words.) I've been asking my 8 year old a lot what happens in her mind. My daughter has a lot of visions, inner seeing, that she then creates with. She takes her visions and puts them into the real world via representations. She also is very aware of sensory experiences. But she describes an inner voice sometimes, too. I ask her often, were you feeling, or thinking, or noticing, or imagining/seeing something? Mine is FULL of my inner voice, and then vague complex ideas underneath that I try to pull up into my conscious mind, or write them down to give them concrete form. I have emotions sometimes that come from thoughts, and occasionally emotions where I can't trace the thought easily, because it happened in the distant or even not so distant past.
The lady they discussed seems to have had those emotions and visions based on experiences she had that she couldn't store away, so they kept coming up in her subconscious and spilling over into her conscious mind, hijacking it with its intensity. She seemed to try to keep her brain occupied with benign content so that the disturbing wouldn't take over. Once she was brave enough to face the thoughts, feel the feelings, name the images, then it could be stored away and free up space for other productive mind work.
It's interesting to ask kids these types of questions, since they have sometimes less flowery language and just say the straightforward thing without as much concern for connotation or nuance.
I don't have a strong opinion on my own internal experience (which predates this video or idea of being a poor historian of my own experiences), but it's been a few weeks since seeing this interview, and my best description of how I mediate my thoughts is heavily framed around where i am in space relative to a room or a person or an object, with maybe a word or two that is a bit like an index, and something like a polaroid of what i was looking at when the thought occured or was noticed.
That has a lot more information than what I would have said previously, "I'm a visual thinker", but I just don't know if or how much i can believe that this is a true reflection or a confabulation.
I think being able to evaluate this favors a metacognitive aspect that some people are drawn to and some aren’t. The self awareness isn’t native to everyone. I relate to the fact that what’s happening in the mind depends on the room and who’s around, and the task. The mental state can be manipulated to improve a task, for example. Right now, I’m having so many different ideas that I could write down here, the trouble is giving concrete form to any of them at the expense of the others. One line of thought is that extroversion vs introversion plays a role. Another is describing my own lack of internal monologue as soon as I begin speaking or interacting with someone external. The inner voice goes silent. Another is that I also had the experience of a vision or intrusive vision while driving, which seemed like an unacknowledged fear popping up that I quickly tried to push away. I can pull up visions from memory or imagine visions in an attempt to predict, but it’s not there often. The inner monologue is dominant for sure when I’m alone or silent.
I wonder how many people here give their thoughts form through writing, since most are great writers? Or have they spoken the thoughts first to give them form and then write them down? I can write my thoughts, the exact inner monologue, and also it’s a way to force the abstract idea into concrete form. I have thoughts and then ideas on top of (or underneath depending on how you want to look at it) the thoughts. (Sorry for this rambling, I could go on and on and on…) My thoughts are word based, text and hearing. Are your thoughts visions that you’re describing? How do you write?
I literally say my thoughts to myself. When I'm tired and annoyed I have the problem of saying inner thoughts too loudly, so other people can hear them.
What if instead of me being wrong, it's the other people who keep wanting me to be?
I think that's pretty interesting for a self-described "yug gnirob"! It also seems to be in line with the general initial reaction that many have when they're told that they may not be as much of an "inner speaker" as they think.
I'm more interested in that part myself; namely why there is such a resistance to the possibility that someone's inner experience might be more nuanced and complex than it's given credit for.
I was a lot less sure of my own interpretation of my own internal experience, even before following this interview, so maybe I'm just primed to be more accepting of the thesis.
Brian Kohberger’s guilty plea in the 2022 University of Idaho massacre meant that justice was served without going through the expense and unpredictability of a trial, but it also may have deprived us of a highly publicized hearing on the Incel crisis and the extent if any it may manifest itself in violence. This is not certain of course, for while he was likely an Incel his defense at trial could well have involved other things.
I don't know anything about that specific case and so can't comment on it, but this is something I notice somewhat frequently: a specific crime or tort takes place which is deeply connected to some recent development or untested legal theory, but the involved parties settle/plea-bargain instead of letting it work its way through the courts.
This is normally the right call for everyone who is actually involved in the case, but it seems to run counter to the public interest. I'm not sure if there's a fix for this within the common-law system. It could just be that if the parties settle, it ought to be the domain of investigative reporters and legislators, not prosecutors/plaintiffs and judges. Perhaps the government or a charity could subsidize the legal fees of parties in cases of public interest, on the condition that they do not settle the case. This seems a bit far-fetched and at any rate politically unwise: by doing that, you will necessarily be paying the legal fees of many unsympathetic (legal) persons.
I noticed this too. It's outright infuriating in consumer rights cases - or even worse, civil rights - where the plaintiffs get a last minute settlement offer when it's all but certain they're going to prevail, which they obviously take, and I'm happy for them, but it deprives the rest of the country of a binding precedent that the illegal thing is actually illegal.
I think there should be a rule that judges must still make a verdict on the meaning of law even if there's no court case anymore. No idea what it could look like in practice, though.
One possibility would be to allow judicial discretion on whether to accept no-admission-of-fault settlements. The problem with this is that you're essentially using the plaintiffs as unpaid ~prosecutors. They'd be justifiably upset.
What if there was some sort of right-of-refusal? When there's a settlement, then (subject to some sort of public-interest test), the state can purchase the tort. The plaintiffs are paid the settlement out of public coffers, and then the state can hire lawyers (maybe even the same ones) to continue to pursue the case.
One problem with this though is that sometimes the *government* settles cases when they really should be pursued. There would have to be some guardrails in place to prevent them from settling the tort they purchased.
Don't get me wrong, but I really dislike both ideas. The first one makes the plaintiffs straight up worse off. The second is very easy to abuse and funnel out public funds, and unlikely to be used much legitimately due to huge up front cost for the government and high risk of losing. Also probably unconstitutional due to standing.
I was thinking more like, the settlement is valid, the case is closed, there's nothing more to worry about for either party, but the judge then proceeds to decide the legal arguments that were previously raised by both parties, without the parties' involvement. The legal precedent it sets would apply to future cases, but not the current one.
In the Idaho case Brian Kohberger faced the death penalty, so a guilty plea in return for a life sentence made a lot of sense. Especially since the evidence against him was strong and for all its societal interest an Incel-rage defense probably wouldn’t work.
For anyone interested in housing and urbanism in Los Angeles, I wrote a review of the evidence on one of the biggest fighting points in recent years, a high-value transfer tax that's turned into a bit of a partisan fight. I came into the essay with some sympathy for (more thoughtful versions of) the housing leftist position, but broadly find that the YIMBY take on the tax is directionally correct (which I think the readers here will mostly be happy with).
https://heylisten.substack.com/p/building-selling-taxing-houses-fighting
I was inspired in part by similar empirical reviews Scott's written, going over a literature and trying to synthesize a birds-eye view of what the evidence can justify by synthesizing the statistical results with general common sense-type arguments.
A cyberattack affected freezers in a few European stores last night. Lots of food will have to be thrown out.
Naive question, but what's the reason for not just having 1990s offline freezers? Are there enough benefits to having them online that it's worth the risk of an attack? (And yes, you can file this one under "Man doesn't understand why technology didn't stop when he was a teenager". But still, why not?)
I read the manual for one of the refrigerator models that was hacked. The reason why the refrigerators can be set to defrost continuously via the internet is that (according to the manual) you can optimize energy usage by tweaking the defrost schedule. Presumably a chain doesn’t want to train employees at every location about the intricacies of defrost cycle scheduling. They would rather have a few employees administer the defrost cycles for all their units from a central location.
A residential analog: my house's geothermal system comes by default with an internet connection and web dashboard that lets you do the usual smart thermostat stuff (change temperature remotely, chart energy usage) and also lets the installer check on the unit without a site visit. The internet connected part is a separate device, but it has a direct data connection to the unit - unlike e.g. a Google Nest, which connects over basic thermostat wiring.
From the marketing materials (https://www.waterfurnace.com/residential/products/accessories/symphony-home-energy-management), the manufacturer sounds rather proud of it, and I think they found it odd that I opted out. The benefits seem minor to me, though; their alternative "dumb" thermostat still has enough energy usage data for my purposes. I have no interest in joining Mirai-style botnets or losing heat on the outbreak of war.
You are smart to have opted out. I think that's what I'd do too.
Cool that you have geothermal in your house, can you tell more about that? How difficult/costly was it to install, and has it paid off?
I don't have good data on the cost of the old system (propane), but my estimate is that the geothermal system breaks even at current electricity & propane costs, applying my mortgage interest rate to the upfront cost. However, that's including the effect of a 30% federal tax credit that was repealed last year. I forget if that estimate also includes the fact that counterfactually, I would have needed a new propane furnace instead.
I still consider it worthwhile for non-financial reasons: climate friendliness, not worrying about propane deliveries or gas lines in my house, and reliability (typical lifespan for a geothermal unit is 20-30 years, vs 10-20 for a furnace).
The main difficulty was landscaping - a horizontal loop field tears up a good deal of yard, including a trench leading right up to your house. So expect to spend some time (or further money) making it look nice again. On the other hand, it's your chance to do something new, like replace an invasive meadow with native wildflowers. There are also vertical or diagonal loop fields, which are less disruptive and require less land, but at higher cost.
Another question: after the French supermarket chain got hacked and it made national news, did no one do anything ? Did they not even check if their own fridges are running the same software ? I don't understand what goes through people's heads at times like this.
Assume you only have hired illiterates. and a 6 figure CEO. "That was just them! Lucky us! Surely our software will fix itself, just like my browser does!"
Airheaded CEOs are incentivized these days.
Yeah. Also, I am betting they are all using the same exact software, made by the same monopoly, in exactly the same default configuration.
I believe they are. The last hack was a bad password. Heads should roll. Heads won’t roll.
Based on my experience doing electrical work in commercial/retail environments I'd say that lots of the stuff you'd expect to operate with a light switch or a thermostat in your house (such as lights and the ambient temperature) is controlled in commercial/retail environments with big relay banks, and these are controlled by computers.
Where the susceptibility to cyberattack comes in is that the computers controlling the relays are not necessarily located onsite; they might be run out of a central headquarters somewhere.
I don't know why companies do it this way, but I suspect it's partly to eliminate human error (e.g. one manager consistently forgetting to turn off the lights, or setting the freezers way too cold), partly to make things more standard and consistent across locations, partly so maintenance requirements ("one of your exhaust fans has been drawing more amps than normal") can be communicated up the chain, and partly for some other cost-cutting scheme they've come up with that I haven't thought of.
Mostly that you can hire illiterates, who work a lot cheaper than functional human beings.
How many people have you hired to manage retail outlets?
Personally? I'm not a CEO. In my direct family? Over a hundred. My father in law has personally managed four or five retail outlets (Ollies to Ames), including the fun story about the guy who duct-taped a knife wound to the neck and came to work that day. Worked the whole shift, too.
When I say "they're digging the bottom of the barrel as a philosophy" that's words taken straight from someone who knew Sam Walton and Crazy Eddie, and is a financial analyst.
Speaking from personal experience as an illiterate who was hired to work cheap, are you?
Can we please discuss things with some semblance of trying to answer the question rather than firing off dumb zingers?
I'm sure to you I must seem like an illiterate hob. Then again, I do know what the word dunder means. Literacy used to mean something, back when we had more of a shared culture.
I don't mean to insult. I mean to discuss a race to the bottom in terms of hiring standards. Use of the word illiterate is meant to both suggest lack of intelligence and culture, and the deliberate watering down of our society -- as the word literacy was the first concept to be watered down.
For remote monitoring and compliance tracking. There are laws about cold chain storage which means there has to be a continuous record of what temperature every single refrigerated item was stored at all times. And it's bidirectional so that if there needs to be a software patch, it can be transmitted back.
Also, a quick reminder that supermarkets run on a 1-3% margin. So even a trivial cost savings (not having to send a technician to go around and grab logs off of fridges and freezers, and being able to fix things remotely) can wildly change whether a grocery can continue to operate.
I'm just going to note that remote monitoring doesn't actually mean you CAN'T have a competent store manager who calls home if there's a power outage, and can move the valuables into appropriate cold storage.
Happened over here in Ireland, as well; the SuperValu chain of supermarkets had some of their freezers hacked:
https://www.rte.ie/news/business/2026/1102/1597612-13-supervalu-stores-hit-by-suspected-cyber-attack/
Apart from blackguardism, I don't know why the hackers are doing this or what they are trying to achieve, and I don't think any group has of yet claimed responsibility.
I read that in at least some cases they found that they hadn’t bothered to reset their network passwords from factory default.
They're doing it for the lulz, most likely. Honestly, in my youth, if I saw one supermarket chain fall to a blindingly obvious exploit; and a couple weeks later another one fell for the exact same trick... then yeah, I'd be sorely tempted to just hack every supermarket in sight. Maybe I wouldn't actually destroy them all, but yeah, I'd be tempted to do it just to see if I could.
And that was back in the day, when the Flinstones roamed the Earth; today, I bet I could even get paid for it...
From some reports, it was only one model of freezer that got hacked, so yeah - most likely explanation is idiots thinking they're funny:
https://www.irishtimes.com/technology/data-security/2026/11/02/supervalu-removes-frozen-and-chilled-food-from-13-stores-after-cyber-attack-on-fridges/
Or there's a bear involved (Cozy bear or Fancy bear). Walking all over the EU pulling pranks sounds like Russian humor.
Honestly, there's a level where I'm kind of thankful. At least if there are people who're doing this on a level where it makes the news, but not on the level where it's causing national or international catastrophe, it increases the likelihood that people will take precautions that make it less likely that it *will* happen at that scale.
If people don't start taking appropriate precautions, there may simply not be any hope for us.
Poland: What is a Constitutional Tribunal and why is their president throwing judges around?
https://www.president.pl/news/statement-by-the-president-of-the-republic-of-poland-karol-nawrocki,129320
When one side starts calling it an attempted coup, that's a bad sign.
Anyone with more information? I'm very curious about this.
The eminence grise of the Polish President's political party (Kaczynski) also believes that the Prime Minister of Poland had Kaczynski's brother murdered for political gain. Saying there is bad blood between them is an understatement.
I endorse this summary:
"""
There were elections in 2015 where the right swept to power in both the legislature and executive. The new legislature was to be inaugurated in October of that year. 3 of Poland's 15 constitutional court judges had their terms end before the inauguration, whereas 2 more had their terms end a month after the new legislature was to convene. The losing party, seeing an opportunity to pack the court before they lost power, decided to appoint judges to fill all 5 vacancies, even though they only had the power to appoint 3.
The President, who belonged to the right and was himself sworn in months before, refused to swear in the new judges before the new legislature convened—a legal requirement to take office. When the parliament convened, now under right wing rule, they threw out the 5 appointments and appointed 5 of their own judges. The constitutional court invalidated 3 of these appointments, the ones the previous legislature had the right to appoint, while accepting the other two. Parliament, displeased with this, rammed through legislation forcing the court to seat the 5 justices and allow them to hear cases.
Now, in 2023, the right lost power, upon which the new ruling party appointed ideologically left leaning justices who invalidated all rulings involving the 5 judges appointed in 2015, effectively invalidating most legal decisions made by the court since that time.
Moving forward, in 2025, new presidential elections were held, and the right won, prolonging a split government. The new president refused to swear in new judges that didn't align with him ideologically, effectively paralyzing the court as being sworn in is a legal requirement to take office. To try to circumvent this, new judges have been sworn in in front of parliament as opposed to the president, which is constitutionally dubious and has caused the constitutional court to fracture further, as right wing justices refuse to seat the ones not sworn in by the president.
Which leads us to the events of last night. The PM, in an attempt to break the judicial paralysis, rolled up with police and had the prosecutor general issue warrants to hold the head of the court and his deputy in contempt for refusing to seat the new justices. The President and the right wing establishment see this as a coup, whereas the government and their supporters see it as an attempt to restore law and order in the country.
"""
That penultimate paragraph refers to the debate around the term "wobec", which commonly means "in regards to", but in archaic writing means "in the presence of" (compare how English laws use "shall" vaguely). The PM is claiming that the Judges only need to swear in regards to the President, not in his physical presence.
My takeaway would be to (1) stop having ceremonial rubber-stamps, it is utterly moronic and (2) enforce a single official language with explicit definitions for words to be used in legislation. More modest versions of (2), like stop using archaic language to make laws seems law-ier, also work.
As long as Kaczynski controls PiS, which he will for the forseeable future, he will do anything in his power to bring Tusk down.
(Technically he only says he believes Moscow killed his brother and Tusk collaborated in a cover-up.)
> When one side starts calling it an attempted coup, that's a bad sign.
That gives sides incentives to call it that, even when it isn't.
Sure does. The obvious incentive not to call it that is "we are not a banana republic." The pretense of fair and unskinked* elections is very important to the democratic process.
*unskinked: influenced by outside entities that feel that it's their right to fix democratic elections, and democratic countries, so that they will behave "in a matter that comports with the current geopolitical priorities" of the outside entities.
Give political actors such bad incentives, and you'll start creeping your way toward being such a place.
I've got family there, though the insider look is barely more informative. What it sounds like is the President of Poland has a number of powers, on paper, that in practice have always been more ceremonial/rubber-stamp in nature than a real part of the political process. He's trying to use them for real to block Judge appointments. I think. My Polish is pretty bad, but does exist.
This is close to my reading as well.
The president in Poland is (was?) a simple ceremonial role. Up until Duda, the previous president, most people didn't even know who the president was.
In this case if I understand correctly, new constitutional court judges have to be seen/confirmed by the president. Purely ceremonial, turnkey kind of thing. But because the president is aligned with the previous ruling party, PiS, he's refusing to execute the ceremony, effectively blocking the new lawfully selected judges to being their work. This has never happened before and presents a deadlock.
I haven't been following the politics for a while but my understanding is that PiS has a long track record of technically complying with the law, but working very much against its spirit, and have a penchant for inflammatory, nationalistic rhetoric, so them calling it a coup aligns well with what they've been saying for years, eg. that the current prime monster Tusk is a German agent, that the EU aims to destroy Poland (going so far as to call it the Fourth Partition), etc.
The way it was explained to me, it isn't even a ceremonial confirmation like the UK monarch appointing a Prime Minister. The President just witnesses the oaths of office for the new judges. Which turned into a veto by way of the President refusing to observe the oaths taken by judges he doesn't like.
This time around, the parliamentary majority is refusing to recognize the validity of this and their judges are attempting to act as members of the court and are supported by the Prime Minister. The disputed judges make the difference in which side controls a majority of the court.
Each side has been trying to lock the other side's judges out of the building, and the thing that just happened was the PM sending in the police to back up his party's judges. The President is calling this a coup.
The legal argument for the PM's party locking out some of the President's party's judges is that there's a procedure which (if their majority on the court is real and the law works they way they say it does) allows them to suspend other judges for misbehavior in office.
It seems that putting solar panels into fields is a big win. It probably isn't the only thing which uses available tech, is plausible, and can be tested before it's deployed on a large scale. Is AI useful yet for finding that sort of thing?
It is a big win now because assorted businesses and governments bought in, pushed and subsidised solar installations a decade or two ago, hard enough and for long enough for tech to improve, manufacturing to grow and economies of scale to kick in. I am one of those who benefitted enormously from those early subsidies.
How do we spot more such opportunities early enough to get on board before the point where it becomes completely obvious to everyone how big a win they are?
A history of arguments for independent violence to get good changes, and why those arguments are bad ideas. I don't know for sure that all the history is correct.
******
French philosopher and novelist Bernard-Henri Lévy in The Free Press:
Chaos on the streets of French cities has brought back the seductive notion that violence can set you free. It is a myth—and a recipe for civil war.
Amid the uproar of the citizen demonstrations of recent weeks in France, and under the influence of leading figures in Jean-Luc Mélenchon’s extreme-left party, France Unbowed (La France Insoumise), one theme is making a powerful comeback: the idea of an invisible, insidious, systemic violence supposedly exercised in democratic societies, which it is legitimate to oppose with the visible, direct, frank violence of the riot.
I know this idea well. It was Jean-Paul Sartre’s concept when, in his preface to The Wretched of the Earth by the anticolonialist ideologue Frantz Fanon, Sartre presented the murder of the colonizer as the response to the oppression of the colonial system. It was the idea of the writer Jean Genet in a famous article, “Violence and Brutality,” published in Le Monde in 1977, in which he presented the terrorism of the Baader-Meinhof gang in Germany as a necessity dictated by the anti-imperialist struggle. And it was already the idea of Georges Sorel who, at the very beginning of the 20th century in his Reflections on Violence, opposed the force of the “bourgeois state” with a “proletarian violence” endowed with supposedly regenerative virtues.
I thought this idea had been discredited by the post-May 1968 intellectual movement, the critique of totalitarianism by the New Philosophers movement which I led, and the reckoning with the fascist, Nazi, and communist catastrophes. But apparently not. And so, we must start all over again.
The first thing wrong with this notion of “counterviolence” is that it never strikes at the heart of the hated regime, but instead at the most exposed and most vulnerable people. The police officer who earns 2,000 euros a month, works night shifts, and protects, at the risk of his life, property, people, and the right to demonstrate. The shopkeeper who earns his gross profit on Saturdays and whose ransacked store was his only capital. The owner of the burned-out car who already struggled to afford a tank of gas and who must now find the money to replace the vehicle or face the obstacle course of dealing with insurance companies. The riot claims to strike at “the system,” but it falls upon women and men who are neither its masters nor its beneficiaries, but its humblest subjects. It claims to avenge the humiliated, but it adds to their humiliation fear, ruin, and the feeling of having been abandoned.
The second thing wrong with the violence that France Insoumise tells us is “sometimes legitimate” is that it does not destroy “systemic” violence but offers it the opportunity to reveal itself and, in revealing itself, to harden and spread virulently.
Are the United States of the MAGA movement and the neo-Democrats consumed by hatred of Israel in such a different situation than the one we face in France?
This is the story of the Lyon silk workers’ revolt of 1834, when troops opened fire on a crowd. It is the story of the revolutionary days of June 1848, crushed in blood. It is the story of the Paris Commune, drowned in a repression that left thousands dead. And when the insurrection triumphs, history teaches us that the regime born of the riot is, everywhere in the world, more terrible still than the one it overthrew: Liberating violence does not abolish systemic violence; it transfers it to other and worse hands. It does not abolish domination; it becomes the founding principle of a new power which, acting in the name of the people and history, now believes itself entitled to anything. Counterviolence promised emancipation, but it gives birth to dictatorship.
Ultimately, the philosophical choice is the one formulated by Thomas Hobbes, Jean-Jacques Rousseau, and, with them, the theorists of the social contract. Of course, one takes a risk when one renounces taking justice into one’s own hands and entrusts the public authorities with the monopoly on legitimate force. And that is precisely why it is so important that states be held accountable for this monopoly, for the way they exercise it, and for the abuses it permits. But the alternative is the jungle. It is man becoming a wolf to man once again. It is the law giving way to the whims and fevered passions of each individual. It is a world without rules, where everyone arrogates to himself the right to punish and to strike. Today, in France, militias. Tomorrow, between a far right and a far left equally inflamed, a civil war. Are the United States of the MAGA movement and the neo-Democrats consumed by hatred of Israel in such a different situation than the one we face in France?
This is what Sartre began to glimpse toward the end of his life when, in dialogue with his young companion Benny Lévy, the former leader of the French Maoists who later became a master of the Talmud, he renounced his belief in the redemption of societies through terror. And this is what the intellectuals of my generation came to understand when they realized that the same Jean Genet who supported the Baader gang also rediscovered, in his final writings glorifying the Palestinian cause, his youthful fascination with the soldiers of the Wehrmacht and the SS. And this is what I theorized nearly half a century ago, in L’Idéologie française, about the philosopher Georges Sorel who, having begun as a fervent apologist for revolutionary violence, ended up as a founding father, alongside young followers of Charles Maurras, of the all-too-little-known “Cercle Proudhon,” where the first red-brown synthesis in Europe took shape.
This notion of visible violence answering invisible violence is decidedly too dangerous for our societies. It’s true that systemic violence exists. But its opposite is not the violence of the riot. It is the rule of law. It is justice. It is what allows a society to identify violence committed by public authorities, establish responsibility for it, and redress the wrongs it inflicts. It is the possibility for each individual to challenge power without becoming, in turn, a miniature power, a petty judge, an unmasked avenger. Such is the lesson of the century for today’s rebellious generations. Will they heed it? I hope so. For if they do not, civilization itself will be exposed to upheavals on an unprecedented scale.
Translated by Emily Hamilton See less
There is a decent amount of research on this. This book argued that nonviolent protest is more effective than violent resistance. https://www.belfercenter.org/publication/why-civil-resistance-works-strategic-logic-nonviolent-conflict
But that has been disputed. Eg
https://statsandsociety.substack.com/p/why-why-civil-resistance-works-doesnt
https://muse.jhu.edu/article/1001894
I agree in essence with these arguments. However, I will risk, as somebody whose background is not French, except insofar as any European intellectual is necessarily a little bit French, the suggestion that France's current troubles are perhaps less significant than might at first appear - that they are in effect, more street drama than revolution. Of course France has seen violent revolution time and again, with the streets running red with blood; I would even look back as far as the Fronde of 1648-53, as much a revolution as a civil war. But as Bernard-Henri Lévy notes, those long-past events, from the Lyon silk workers’ revolt of 1834 to the days of June 1848 and the crushing of the Paris Commune, were triggered either by an extreme reactionary provocation, or else occurred in response to a serious and unacceptable assault on popular freedom. Today's events, like those of the Paris May of 1968, are a response to irritants, to dissatisfactions, to the inevitable arrogance of the country's comfortable establishment - not to massacres. Like May '68 (which did not, despite all the drama, after all dislodge De Gaulle), and without minimising the destruction, I suspect the current wave of revolutionary theatre will eventually subside, leaving behind a new legend for the students of a subsequent generation, together with a parcel of undoubtedly overdue reforms.
Well, de Gaulle resigned less than a year later.
...just to confirm, cops only earn 2000 € / month in France?
I'm astonished Fanon remains popular and wasn't MeToo'd for his writing justifying rape as a tool of resistance.
The far left never cared about principles, only power.
I'm not. “What did y'all think decolonization meant? vibes? papers? essays?" Rape is perfectly acceptable to the MeToo crowd, provided the right people are being raped instead of the wrong people.
The recent developments in the AI interpretability got me interested, so I want to ask people in the field, assuming some of them are likely lurking around here. How possible it is for someone far outside the field to get into it, realistically, within couple of years?
I don't have the necessary prerequisites, and it's probably too slow to have meaningfull impact on anything, but the chance is probably not zero.
I'm already looking into it myself, of course, but sometimes it helps to have an outside perspective.
More than a year out is probably implausible. Definite shrinkage -- and that favors the already employed.
I'm not really looking for the employment necessarily, if I can contribute to the field meaningfully as an outsider, with zero monetary gain, that would be worth the effort.
This is probably not going to change your response, iirc your are firmly in the "AI is a bubble about to burst" camp. But even then, this should provide enough context for other people who might see this subthread.
I feel the same way but am not a stem worker really, though I do have a math degree. I've focused my efforts on politely but ceaselessly hounding my congressional reps on these issues. Energy policy is my focus right now.
I believe this is valuable and important to focus on that, too.
But I myself do not have such a opportunity, as I'm from a country which is not really in the AI Race in any meaningfull sense.
Write real, hardcopy letters. Show up in person. Bring graphs, bring math, have a 5 minute pitch.
Heads-up: this is an AI post, scroll past it if sick to death of the subject!
Anyway, recently I tried Opus 5.5 on 2 different real-life tasks with some mundane applications and 1 a abstract task as a kind of private benchmark. It's not that the results shocked me, but they were not what I had expected.
Exercise 1. Abstract task first: I wanted to see if 1) Opus can learn from its mistakes and perform better and 2) if it can do better based on my generic feedback. Not laser-focused, very specific feedback, but more like general advice and heuristics.
I decided I would pick a dataset A of 30 items. Have Opus do a task for each item. Then I'd score it and ask Opus to draw conclusions from the results and to write some guidelines for itself on how to do better. Then I'd give Opus dataset B and ask it to do the same task, armed with the lessons (hopefully) learned. Then I'd give the task and the dataset B to the totally naive instance of Opus, that has not had an opportunity to learn from dataset A, and compare the scores. Would an experienced!Opus beat naive!Opus?
And then would additionally give Opus my own, custom generic instructions, and evaluate custom_instruction!Opus, so that I would have 3 different scores: naive, experiences and custom_instrustions.
Oh, and the task? I needed something that 1) Opus can do, but not perfectly and 2) I can do better, so that I can score and give meaningful advice and 3) is objective. Other than those requirements, it could be anything. There isn't much that ticks all 3 boxes, but there's one thing I can do better than almost anybody: identifying snail species from my region based on photos, so this is what we went with.
I assembled a hard, but fair dataset A and fed it to Opus. For a moment I was warried it would nail every item perfectly, and the test would be ruined, but as it turned out, Opus was not great at it, and only got 12/30 photos right. Now, this is not a very meaningful or interesting observation that doesn't tell us much about Opuses powers, because everybody knows that it could easily have been trained to get it 100% right, it's a largely solved problem. No, the real test was to see if Opus can learn from its errors.
So, I asked Opus to take a look at what went wrong, write instructions for itself to do better, address its failure modes, do some research etc. Opus did, and armed with these lessons took on dataset B, and got 14/30. Well, that still sucks, but it appears to be more than 12/30? But these are different datasets, need to see how a naive Opus does on dataset B for a fair comparison.
Weirdly, naive Opus also scored 14/30 on set B. And what if I add my custom instructions? I wrote as good of a set of generic tips and heuristics as I could and fed it to Opus. Armed with my priceless advice, Opus tackled set B again... and got 14/30.
To summarize: Opus totally failed to learn from its mistakes, and got the exact same frigging score of 14/30, whether fresh out of the box, with the benefit of some practice or with my advice.
Exercise 2. I had a digital drawing project going, and I needed to do the tiresome part I mentally call "blobbing". It's where you have finished lineart and need to fill every single silhouette with solid color so that you can lock it and paint it. Whatever, if you've done digital art, you know what I mean, and if you haven't, then it's basically what a kid does with a coloring book. Filling areas with solid color, only the lines are looser and not always completely closed. This task requires understanding of what has been drawn, no, you can't automate it with a Photoshop function. Photoshop tricks fail on complicated and messy lineart. Depending on the design, this can take from a couple of minutes to long, long hours. I was looking at something like 3 hours of drudgery. And one morning I awoke with the most wonderful idea: "hey, why not make my friend Opus do it!"
So I did. Opus also had an idea that he can automate the whole thing with a simple script of some sort, and learned that no, it didn't work like that, it had to be done largely manually.
We went back and forth refining the process and trying things.
Results: mixed. Opuses output I got was not perfect and I had to adjust it on my own, but it was good enough that it saved me a couple of hours. I used something like 30 minutes instructing Opus and 30 minutes fixing mistakes, instead of something around 3 hours of blobbing everything by hand.
Exercise 3. At work, using Opus to write user documentation. I was initially skeptical this would ever work. The reason I can write user documentation, is that I have experience using our software, and I know the context and what it is used for. Knowledge, none of which could be easily transferred to another human or AI, short of writing it all down. Yes, Opus can read the source code. But can you read unfamiliar source code, line by line, function by function, and then describe what it does, from the point of view of a user? Not just what happens, but why and what for? Could that be done at all, even by a very smart human? I had doubts.
Still, I was ordered to give it a try. In my head, I sketched up some approaches on how to handle it, strategies that involved agents explaining things to one another and so on. I tried the naive approach first, giving Opus a place in course code, my own documentation as an example and telling it: "write more like this about that!"
Weirdly, it worked amazingly well? Somehow, Opus constructed a perfectly accurate picture of what the code is going to do, as seen by the user. He even found some bugs, bizarre interactions I never knew about and somehow figured out a bunch of things of the uses and purpose of the stuff out system does. The results were not perfect, and some of it needed to be fixed or altogether re-done, but overall I was very impressed. With the help of Opus, my document-writing task would now be 2x as fast, and much more relaxed.
So that leaves my with 1 failure, 1 partial success and 1 instance of very good performance.
An older case of an intellectually disabled man who was convicted of a bajillion crimes in Denmark is being revisited. Turns out, when the police would take him on drives around the area to make him point out where "he" comitted arson, he would confess to crimes that never happened, buildings never burned. Law enforcement was just happy to have someone to blame for so many crimes. He loved going on drives with the police and considered them his friends. He was sad when he was convicted, only because this meant no more drives.
Anyway, his intelligence was usually described as "of a 6 to 8 year old child". It just struck me how messy and inaccurate this seems, trying to compare intelligence and neurodevelopment in this way. A normal 6 year old would surely know that they were in trouble and to lie to the authorities (eg. the James Bulger case). Just like when people say such and such animal is like a 3 year old human - well maybe, but also not at all.
> Anyway, his intelligence was usually described as "of a 6 to 8 year old child". It just struck me how messy and inaccurate this seems, trying to compare intelligence and neurodevelopment in this way.
It descends from the original branding of IQ tests as being an intelligence "quotient", your "mental age" divided by your actual age. Actual age exists, but mental age doesn't.
A famous similar case: https://en.wikipedia.org/wiki/Bruno_L%C3%BCdke
In an ill-advised procrastination from working on my thesis, I took the prevalence of blood types in my country and worked out what percentage of people they can accept blood from and what percentage of people can accept blood from them.
Blood type Frequency Can accept from Can accept (%) Can give (%) Accept:Give
O+ 40% O-,O+ 49% 81% 0.60
O- 9% O- 9% 100% 0.09
A+ 31% A+,A-,O+,O- 87% 33% 2.64
A- 7% A-,O- 16% 41% 0.39
B+ 8% B+,B-,O+,O- 59% 10% 5.90
B- 2% B-,O- 11% 13% 0.85
AB+ 2% AB+,AB-,A+,A-,B+,B-,O+,O- 100% 2% 50.00
AB- 1% AB-,A-,B-,O- 19% 3% 6.33
Obviously individual factors are going to be way more determinative of whether a given individual should/can donate, and of course stock levels are going to be a big impact, and also I make no claims about the relative weights of blood donation vs other activities regarding good/hour, but I thought this might help clarify the picture somewhat beyond the messaging we get of "everyone who can donate should!"
(If there's a better way to give this information, please let me know)
Your post is not addressing one point: if for example all potential A+ donors see that they can only give to 33% and stop donating, then there will be no A+ type blood, and all A+ receivers now need to receive blood from A-,O+,O- donors. This would make the situation much worse, because this blood is missing for the A-,O+,O- receivers who can only accept those.
I think in practice this is relevant. I have heard (but never checked) that doctors try to match blood types if possible, and that this is one of the reasons.
I think for almost every blood type except possibly some very exotic ones it's a good thing to donate. So the best policy is: "ignore your blood type and just donate".
I generally find universalisation arguments compelling, but I think this form proves too much. If we say that "apples are good for you" and literally everyone acts on it within a relatively short time window, then that will lead to problems as well. If literally everyone acted on the governmental advice to have X amount of water or food on hand in case of outages/shortages/emergencies, then we'd have panic buying, massive market distortions etc.
If we're talking mass messaging, then yes, simpler is better - the consequences of having too much blood are relatively minor (wasted resources to harvest it) while the consequences of having not enough can be catastrophic (a person dies). But I think the context of this post is important - it's being posted in the comment thread of a community of generally intelligent, generally pro-social people who are generally good at parsing more nuanced messaging and arguments
Also, regarding the "your post doesn't address.." - my post *does* address this point as well:
> Obviously individual factors are going to be way more determinative of whether a given individual should/can donate, *and of course stock levels are going to be a big impact*
If there was a large surge of blood donors, stock levels would reflect this, which in this hypothetical people would react to. Now, you can argue the masses don't really respond well to this kind of nuance, and fine enough, but my post *did* address this, albeit a bit quickly
Well, you didn't say it right away, but the only implicit conclusion I can see in your numbers is that you believe that the value of blood donations is coupled to the percentages that you give. (Otherwise, why give them?)
But I think that this is mostly wrong. The marginal utility of an additional B-donor is not that we have an additional spare B-unit. It's usually that we have an additional 0-unit, because we don't need to waste this unit on donors of type A or B.
That's what I wanted to express with "ignore your blood type and just donate". In most practical situations, the marginal utility is (mostly) uncoupled from the blood type. The marginal utility is mostly an additional 0-unit, regardless of the blood type of the donor.
I think it's pretty obvious that yes, a person with more widely useful blood is a more valuable donor, all else being equal. If we take the extreme case and say only one blood type will donate, we'd obviously want it to be a more widely useful type than a less useful type. It's obviously not a simple 1:1, because as you say there's strategic use etc, but that's mitigating the fundamental dynamics at the margins rather than overall.
In fact I'd say my perspective actually helps with the strategics. If someone is O-, then we want them to donate as much as possible - it's obviously a lot more widely useful than say AB+. Yes, the AB+ allows us to free up a unit of O-, in the case where an AB+ patient is around, but AB+ is also a smaller percentage of the population so there's obviously less value there (ie it's less likely to be that AB+ patient), especially since this hypothetical patient can take more or less anything. And since O- is only ~9% of the population, all else being equal it's going to be a relatively scarce resource, compared to more common but less widely useful blood types like say O+ or A+.
Perhaps this is unfair and more to do with my own insecurities, but I *feel* as if you are worried that a more nuanced exploration of blood type donation than blanket "everyone must donate!", no matter how widespread or cautious or how many caveats are included, is tantamount to saying nobody should donate. Your arguments haven't been "your analysis is incomplete" or "you insufficiently account for replacement/strategic use of typing", but more "people are too stupid to take this level of nuance so you shouldn't say it at all". Your claims that I don't take something into account which I explicitly did is a part of this as well. Again, perhaps this is my issues coming through, but that's the impression I'm getting. If I'm wrong, I apologise for the implication.
Know your audience. When your audience is the general public, "Donate blood!" is a good message. "Donate blood because <Alan Smith's post>" would be a much worse message. Not because your analysis was wrong or your intentions were bad, but because it would be unnecessarily complex.
(Copied from another comment)
If we're talking mass messaging, then yes, simpler is better - the consequences of having too much blood are relatively minor (wasted resources to harvest it) while the consequences of having not enough can be catastrophic (a person dies). But I think the context of this post is important - it's being posted in the comment thread of a community of generally intelligent, generally pro-social people who are generally good at parsing more nuanced messaging and arguments. It's also a group who I would be relatively confident are also much more likely to be regular blood donors than the general population for a range of variables, but still - as I said, I just found this an interesting perspective
What does hope mean to you? Is it necessary?
Hope is the belief in improvement despite some barrier. Traditionally, I think the barrier is expected to be quite large, and likely unsurmountable, but I've seen it used in lesser contexts.
I think for many people, hope is what they do when they don't want to actually put their nose to the grindstone or plan/implement a solution with rigor...but want something to turn out well regardless.
I do not need that kind of hope. I plan, I work, and I accept the consequences pretty well when I fail to.
Hope in the face of a barrier that is daunting *despite* rigor and effort? Yes. I think we all need hope to succeed in those circumstances. Without hope, we will look at the statistical improbability of a situation and give up.
Hope is the idea that things can improve. It's necessary for improvement, otherwise no one's got a reason to do the work of it.
I think there's more to it. It's not the thing to think something can improve, than to hope for it. You can even hope for something that is barely possible.
I was just thinking about hope yesterday, because I saw the word somewhere and I thought, that's a tricky concept that has historically played a huge part in Western culture (it's one of three major virtues in Christianity, after all), yet nobody seems to talk much about it, either for or against.
If I had to define it, I'd say it's an emotional investment in the future that gives meaning to the present. Which is a rather weird orientation towards time and meaning if you think about it - it means the present always remains incomplete!
"Hope is the first step on the road to disappointment."
Hadn’t thought much about it except in a sneering kind of way back in ‘08 when that was like Obama’s one-word unofficial (or official?) motto. So, interesting to give it a fresh look now…
First thought I had was, is being in a bad situation, or a situation you reflexively want to improve, a precondition for hope? My mind certainly goes first to the person who is in dire straits but expects or at least kindles a persistent aspiration that things will be better in the future. But is hope something you could also feel right *after* you’ve accomplished something big, or in the midst of celebrating a victory, etc.? Having asked the question I’m unsure of the answer. If hope can happen any time that makes it more interesting as an emotion, I think.
And, is it an emotion? Or is it more like a mindset one must deliberately cultivate and remind oneself to return to? (A practice, in other words.)
Some people seem naturally inclined to be hopeful, others not.
Are most children naturally hopeful, and do most of them lose this quality as they age and become jaded?
The matter of hope being necessity is important. I recall cliches about survivors of disasters and atrocities saying it was hope that kept them alive, but I don’t know if those kind of statements ever received deeper examination. Was it really hope that kept them alive, or something else? And if it was hope, what did that look like to them? Once they were rescued or “out of the woods” or whatever, what happened to the hope then? Did it just vanish, having done its job, or did the practice of clinging to hope linger after the hope wasn’t “needed”?
With time I think I could sort out my thoughts on some of these questions, but for now it’s just been interesting discovering the questions were there. Thanks for posting.
> except in a sneering kind of way back in ‘08 when that was like Obama’s one-word unofficial (or official?) motto
You're kind of right. The motto was official despite its extreme stupidity, but it was three words long. "Hope and change."
Hope surely is a tricky concept. I think colloquially, it is meant to indicate that the circumstance is less than ideal. Perhaps not awful, but certainly not where one would like to see themselves. I think there also has to be a degree of difficulty in getting out of that circumstance. A B student trying to become an A student wouldn't make the cut for hope unless he had historically been a C student and it took him a lot to even move into the B tier.
I like the idea of hope applying regardless of circumstance, but anecdotal evidence from social conversations seems to indicate that I'm the odd man out with that perspective.
Sam Altman wants the DRAM in everyone's private PCs:
https://www.tomshardware.com/tech-industry/data-centers/openai-ceo-says-its-quite-terrible-for-your-pc-to-have-ram-sam-altman-says-data-centers-offer-better-utilization
Scott once wrote an essay about "Things That Are Not Superintelligences", by which he meant companies. Companies may not be AI in that they were superintelligent, but they certainly do seem to employ people with the paperclip maximizer mindset. I'm open to ideas on how to align these people with human values.
This is a sign that Sam is either stupid, or thinks you are stupid. Your computer does not work without local memory; no, you will not be querying the cloud for every RAM read
In this proposed dystopia you don’t have your own computer. You have one of these things https://chippc.com/citrix-thin-client/ or its moral equivalent instead, which does technically have a little RAM I guess, but not much more than is needed for a framebuffer and network stack; and all your actual computation happens in Sam’s data center so he can charge you for it monthly to recoup the cost of buying all the world’s RAM.
I am sure he does. On the flip side, I want the DRAM in Sam Altman's data center for my own PC. uwu.
So Sam the Scam wants us all to migrate our data into DE CLOUD, DE CLOUD, LOOKIT DE LUBBLY FUFFY CLOUD in data centres where he can scrape all our information to train his AI and lock us into subscriptions forever and ever if we ever want to access our own damn information, then hand over our PC chips to him. For DE GRATER GUD, of course of course.
Sure, Sammy, I'll do that the second you buy a two-bedroom apartment for your three person family in beeyootiful downtown Burbank and hand over your - how many estates is it again? six houses and grounds? - to the people of San Francisco who can then make much better use of all that land for multi-apartment block living. For DE GRATER GUD, of course of course.
https://www.realtor.com/news/celebrity-real-estate/sam-altman-openai-home/
Even if this is "chortle, can't you take a joke guys?" and not serious, I'm telling you - the lamp posts are looking better and better to me. And I don't even *want* to run a damn AI model on my own personal computer.
https://en.wikipedia.org/wiki/%C3%80_la_lanterne
https://www.youtube.com/watch?v=RfGJE4l_Vp8&list=RD-HgdeXdkdRo&index=8
> where he can scrape all our information to train his AI
Conspiracy theory time!
Altman's company, along with a few others, is in a race: can his product start bringing in enough revenue to pay running costs and service debts before his startup runway runs out, the markets decide to put their money elsewhere, and debts come due?
Right now, between [1] and [2], it seems opinions on that are rather more mixed than even a few weeks back, which is a little risky because sufficiently spooked investors can make that sort of thing self-fulfilling.
Now imagine a world in which most people do most of their computing in the cloud instead of locally. In that world, the financial risk is somewhat decoupled from LLM revenue, because there is global scale appetite for other useful things the compute can do; the expensive data centers become an asset instead of a liability.
Oh, and he wants all our DRAM as well, of course. [3]
[1]: https://www.reuters.com/world/china/china-ai-bubble-has-already-burst-2026-10-26/
[2]: https://www.bbc.co.uk/news/articles/cvwy72e0n72yo
[3]: https://share.google/ZxCuyxBLz4F7IDPR8
We already went over this in 2008. Then we let the genie out AGAIN. I'm not sure regulation actually is going to work, at this rate.
But it seems quite obvious from the context that he was specifically meaning that it is using DRAM for local inference which is wasteful (arguably true as stated -- 128 GB RAM in your Mac Studio probably isn't going to be loaded as regularly as it could be in an OpenAI server -- of course, you might have perfectly valid reasons to prefer local inference regardless), not just using DRAM in your own computing devices in general. Of course, he really should have known better than to make statements so easily quoted out of context.
When Sam Altman stops gobbling up scarce land in SF with his maniacal house buying, then he can tell the rest of us not to own our own goddamn household PCs. Not a nanosecond before then.
>Of course, he really should have known better than to make statements so easily quoted out of context.
Maybe that is that already the more thoughtful, benign spin for public consumption?
What Altman wants is compute, and for the cheapest price possible. Why would he care whether people use their personal compute for a local model or for gaming or for browsing the Web? He only cares that it's not available to him, not because it's where he would draw a line between good and bad personal uses of compute. On those grounds I find the "out of context" objection doubtful.
In context it remains absurd. Even if we generously grant intellectual musing, at his level of power and influence it reads as remarks about a troublesome priest.
Having an NVIDIA Blackwell or whatever in your house and using it for inference is a manifest waste of $20k when you can pay much less than the amortised cost for a few hours of GPU time whenever you need it.
Who exactly is going to rid him of the troublesome private property rights that allow private citizens to own a computer? Is your claim that he's covertly asking the US government to go door to door confiscating DRAM, or begging Samsung to stop selling to the consumer market? I read that interpretation as much more absurd than the suggestion that renting expensive stuff might be more cost effective than buying it.
That is not my claim, nor did I in any way make mention to the economics for the individual in making their decisions of purchase. I'll leave anything beyond what I stated as an exercise for you in critical thinking to understand a differing viewpoint.
Last week Scott published his ballot recommendations post for California (https://www.astralcodexten.com/p/my-california-ballot-2026), where he goes through how he plans to vote on each proposition and in each race. For the candidates, his method is to pick “the lesser of two evils in any individual race”, so he ends up recommending Democrats for some offices and Republicans for others, and abstaining in some (e.g. Governor).
I’m writing here instead than under that post because what bothers me is the method, not any single choice. I read the post four days ago and keep thinking about it. Judging each race on its own looks at the first-order utility of each vote, meaning which candidate does the most good in that one race (according to Scott, but that's beside the point. As said, I don't want to judges his picks per sè). But I do think that such approach misses bigger second-order utility gain/losses, meaning what all those votes do together once you add them acrosss multiple races.
As I mentioned in some of my past comments, I come from a Lockean and Kantian background in political philosophy, which is quite different from the utilitarian approach many of you share. I respect utilitarianism as a philosophical tradition, even if it isn’t my main compass. So I write this with humility, but also with my own opinions.
Disagreeing with your party on single propositions is completely fine for me. Democracy needs plural ideas, and I do it myself. On Prop 40 or Prop 45, everyone should think for themselves, whatever their party says.
Where I struggle is with voting for candidates from different parties, or abstaining, race by race, as a method. I have voted outside my usual party too, but for me it was an exception, for a specific reason. In Scott’s recommendations it becomes an approach.
I see a party as two things. First, a set of values: an idea of what is right and what kind of society we want. Second, a program that turns those values into laws, and that only works if enough of its people are elected together. A race-by-race vote keeps the best person in each race, but I feel it loses both.
So I have two criticisms, one from my tradition and one from yours.
The first is from a Lockean and Kantian point of view. For me, a vote is not only a choice between two people; it is a way of saying which values I want to be governed by, and I try to stay consistent with them rather than change them from one race to the next. Abstaining is a political choice too, because it means not standing behind those values in that race. So my approach would be to vote for the candidate of the party whose values you share, unless that candidate is really terrible, has pushed really harmful policies, or has done something unlawful. Then crossing over or abstaining make sense. But I’d want a strong reason, and “the other candidate seems a bit better in this race” isn’t one for me.
The second is from a purely utilitarian point of view. In a single race, voting for the other party’s candidate may well maximize first-order utility, because that person may really do more good in that office. But the same vote has a second-order cost. Every office a party wins makes it stronger: more visibility, more of its people in key roles, more candidates ready for the next and bigger race, and a stronger claim that voters back its program. In legislative races it also means one more seat toward a majority, which is what decides which laws pass. Abstaining has the same effect, since the party whose program you mostly believe in gets one vote less. Across many races, and many voters thinking the same way, that changes whose program shapes the next years. It’s a big loss of utility, and you simply don’t see it if you only look at one race at a time.
So here is where I end up, whether I start from Kant or from Bentham. The party whose values I share gets my vote usually, in every race, and I judge it on its whole program rather than hyper-zooming ONLY on the candidate. I cross over, or abstain, only when its candidate gives me a serious (very serious) reason to, not a marginal one. That seems to me both the more consistent choice and the one that do more good.
Thank you all for those comments. Made me think a lot, despite (or actually thanks to) their heterogeniety.
The main thought they spurred (or rather strengthened): there is not one America. America is one for the most culturally diverse country/society out there.
Curious to know
1) if you agree with such statement? (just by looking at the diversity of opinions in this community on such a fundamental point)
2) if you think it is a good or bad thing?
much respect for this comment.
1 i agree with your statement
2 it is both our biggest strength and our biggest challenge
Thank you! to you and your fellows in this community. Keep up the good thinking. You are a REALLY cool community. The one I enjoy reading the most!
Your method here seems to rely on having a political party whose values you share. If neither party satisfies that criteria where are you left on choosing candidates to vote for?
I'm neither a Democrat or a Republican, I'm an American Independent as I only agree partly with either party on any particular issue. I think this is why I appreciated Scott's method of independently concluding on each choice on the ballot.
Where am I left in your suggested voting heuristic?
Each party is not a monolith, and my vote can shape each party by rewarding it with votes for candidates that are different from the parties most generic version in ways that I think are good. This happens through both the Kant and Bentham channels, and seeing as how my vote only ever effects margins and not control, it may in fact be the dominant effect. My values dont align with either party but i see a best version of each party I can push things towards.
In other countries with more than two parties, there is the idea that the political system as a whole benefits from changing regimes from time to time. It prevents the system becoming stagnant with officeholders becoming essentially permanent, at which point they serve themselves and not the public.
The US seems not to favour term limits, and to vote in politicians who are too old/corrupt/incompetent to do their jobs (both sides do this). It seems to me that your logic leads to this, or at least presents a slippery slope that ends here.
> First, a set of values: an idea of what is right and what kind of society we want. Second, a program that turns those values into laws, and that only works if enough of its people are elected together. A race-by-race vote keeps the best person in each race, but I feel it loses both.
At least in the US, political parties have never really been like that (at least not consistently). For example, the Democratic party spent the 100 years after the Civil War as a mix of Southern segregationists who hated the GOP because it was the party of Lincoln, and Northern progressives. From a values point of view, these people had very little reason to be in the same party. Because of this ideological range, they ended up fairly central on average, not able to take strong positions on many issues, and the Republicans had to do the same, leading by the 1960s to complaints that the parties were too similar to offer voters a real choice. And the passing of laws was as much about pork, favor-trading, and maintaining coalitions as values (e.g. the aforementioned Northern Democrats supporting segregation so the Southerners would support them on everything else).
For a long time, you definitely didn't need to have uni-party control to implement policy, and voting based on your values may have meant voting across party lines as often as not.
And if anything, this probably becomes even more true at the local level. Local politics can be extremely acrimonious, but is on average less ideological and more about specific issues and problems. Your city council rep or county clerk is most likely not there to implement a broad-ranging plan for society; they're there to keep the local school open and roads fixed without breaking the bank. (Yes, there are exceptions--they are notable largely for *being* exceptions.)
> I see a party as two things. First, a set of values: an idea of what is right and what kind of society we want. Second, a program that turns those values into laws, and that only works if enough of its people are elected together.
That's a descriptively incorrect way of looking at them. Parties are political coalitions. Donald Trump doesn't have much in the way of values, but he's able to get enough people to support him (helped by the rise of negative partisanship) to get elected President twice. David Shor has recently been pointing out that Israel can't form a government even though there is majority support for certain political divides because many members of the opposition refuse to support Netanyahu and the previous majority isn't willing to replace him. https://x.com/davidshor/status/2117081772098044218
Local politics and national politics have relatively little to do with each other.
Republicans winning a single political office in California essentially has no impact on how strong “Republicans” are in general.
Being willing to change your vote also makes your views more relevant to politicians. No party should represent your views completely, or really even well. Being willing to change your vote on, say, the person in charge of policing, when the party you vote for is traditionally not willing to enforce low-level punishments, has the potential to change the view of that party to something you agree with more.
But most importantly, your brain is not capable of predicting galaxy-brained second order effects. Even trying is going to have such unpredictable consequences you’ll become unable to act with any degree of certainty. “Vote for the candidate that best aligns with my values” is at least interpretable by human beings.
> No party should represent your views completely, or really even well.
I don't understand what you're trying to say here. For example:
> Being willing to change your vote on, say, the person in charge of policing, when the party you vote for is traditionally not willing to enforce low-level punishments, has the potential to change the view of that party to something you agree with more.
You just said that it isn't a goal for the party's views to agree with mine. So why would I do this?
A party’s political positions tend to be a hodge-podge of issues catering to interest groups that are necessarily a coherent ideology. It would be really strange if you agreed with the entire Democrat or Republican position on every policy.
>First, a set of values:
But the values are not fixed, and they're not tied to the voters, they're tied to the candidates who win the elections. Every candidate who wins an election pulls the party toward their values; so if you think they have bad values, you need to be ready to slap their hand and cost them. And if the guy in the other party has better values, and wins, he's dragging the other party closer to those values, and making yours reflect on what he's got that they didn't.
To what extent do parties represent coherent philosophies, and to what extent are they bundles of issues that happen to be lumped together?
I think they are mostly the 2nd one, disguised as the first one. In many cases the disguise is so good the wearer believes it too.
There's a book making a similar argument https://www.overcomingbias.com/p/myth-of-left-and-right
You're reminding me of a question. What are the consequences of a ballot so long and complex that even smart people need guides and mutual aid to deal with it?
There's that, but when Trump is the voice of the party, anyone who is willing to wear the label is suspect.
What do you expect right-wing and conservative politicians in America to do in this situation? Become Democrats? Sit out politics? Continuing to run as Republicans while agreeing and disagreeing with the president on individual issues is the most reasonable course of action.
If they want to call themselves Republicans, they shouldn't expect me to trust them. And I won't.
I *have* voted for Republicans, when the Democrats were especially vile. Neither side deserves to be believed, but if you're willing to run under Trump's banner, you are tarred with his "values".
Also fair point, I must say.
Boom
Well, Scott is already using LLMs to research his voting choices. In two years, he'll probably just prompt some LLM to generate all his votes for him, and so will most other people. Which will be totally fine, since LLMs are superintelligent and never make mistakes or hallucinate anything and are immune from prompt engineering exploits, right ? Right ?
Well... never thought about it that way! A few thoughts come to mind, tell me if I got your point right, maybe not.
Ideally, each of us should look into every race and every prop, but with a ballot this long I get that most people simply can’t do it all alone, and that’s where guides help.
What I liked most in your comment is the mutual aid, because it made me think of Hannah Arendt and her idea of the vita activa, the active life, which she splits in three. Labor is what we do just to stay alive: eating, cleaning, all the stuff that gets used up and has to be done again the next day. Work is what we make that lasts, tools and buildings and all the things that give us a shared world. And then there’s action, which is what happens directly between people when we speak, argue and start something new together, and for her it’s where we show who we really are and the only place where we’re truly free. Not alone. With others. It is way less cheesy than it looks. In my view it gets really what politics is about, in the Aristotle's way.
So a group of people going through the ballot together, like the ACX post/thread and meetups, it is cool in itself I think.
Honestly, it’s a big part of why this community fascinates me: people here argue hard, but in good faith, and they take ideas seriously even when they come from a tradition very different from theirs, like mine.
The fact that I don't agree with Scott's approach to choosing candidates is a different piece of it. But I love his action of talking about that (the ballot) with this community.
Thank you. There's also the problem of candidates on the ballot where information doesn't seem to exist.
For example, judges in PA. (Probably also a problem elsewhere, PA is what I know about.) Also a lot of minor candidates who don't publicize what they think.
In what sense is voting on these people civic anything?
I believe Matt Yglesias had an article years ago pointing out that we've got a lot of 'democracy' in the US, in the sense that many, many, many government positions are elective, but that since nobody knows anything about what those positions are supposed to do or who the candidates for them are, it doesn't mean much to call those elections 'democracy'.
Judges are actually quite far from the worst examples of the problem; people have a healthy idea what judges are supposed to do and how to tell whether they're doing a good or bad job.
Does PA have a Judicial Evaluation Poll that the state bar association publishes? My state does and that's what I check for voting on judges; if their rating is below 90%, I generally vote against them.
Boring heuristics, games that reduce to "no,no, Ihatecali, no,no, keep working on that."
https://www.groundbrkr.com/t/ai
The author explains the AI bubble in long form essays. Some takeaways:
- There is indeed a bubble. The situation is structurally like the 2008 crisis than the 2000s dotcom bubble, except there is much more money involved.
- If the bubble pops, it's not going to be consolidation but collapse like in 2008, because all the big players are joined at the hip with each other and with the rest of the economy.
- The upsides of the AI boom are much more visibly communicated with the markets than the downsides, because none of the frontier labs are public. Like in 2008, we are in a "teaser period" when everything looks fine because of deferred costs, and by all conventional metrics, collapse is unthinkable. But there are signs ("The second derivative").
- The dominant strategy for the big players is to spend, which is why it's generally called the "arms race" at the moment. The dominant strategy might eventually flip to stop spending, which all players will then adopt.
- The test will be in 2027-2028 when all those ordered data centers go online and are due to be paid for regardless of demand. If demand by then does not match capacity (even the most optimistic projections say it won't) and the frontier labs cannot meet their obligations through either revenue or more rounds of funding, they will default on their debt and things go critical.
2008 wasn't a bubble. Prices have exceeded their prior peak.
https://www.econlib.org/kevin-erdmann-was-right/
https://www.econlib.org/archives/2018/06/kevin_erdmann_o_1.html
Instead there was a negative monetary shock along the lines of Milton Friedman's "plucking model".
'It's not a cyclical bubble because it went up then down then up again'
Listen, I get that we can play semantic games about what a 'bubble' is for a couple hours here, and that economists will want to settle on a narrower definition than the general public.
But when millions of families put all of their wealth into a single asset that has been growing for a long time and seems like a smart investment, and then suddenly that asset loses half its value all at once and everyone invested in it is ruined, and the entire economy is thrown into chaos for awhile with lots of suffering and broken dreams, that's the consequential outcome of 'a bubble bursting' that everyone outside an economics department actually cares about.
> It's not a cyclical bubble
I haven't come across the phrase "cyclical bubble" before. Is that like ice cream being popular in summer but not winter, or Spirit Halloween only being active in the fall?
People outside economics departments falsely believe all sorts of things about the economy. A lot of people have heard of the Great Depression, but few of them know about the connection between that and monetary policy (and, to be fair, the responsibility of the Bank of France specifically for that is something that only moved toward the consensus among economists relatively recently). Instead some people think the Roaring 20s stock market was a bubble (ironically enough, such a belief at the Federal Reserve after Benjamin Strong died played a role in it) rather than blaming real estate. It's economists who actually look at the trajectory of stock prices before and after the GD and see that rather than being some notable deviation above trend (as an over-inflated "bubble" would be) it was really the downturn that was the deviation below trend (hence "plucking model").
How does the fact that prices exceeded their prior peak *12 years* after the start of the great financial crisis, mean there wasn't a bubble? Seems like it could easily be the case that economic fundamentals support higher prices than they could a dozen years before. Also seems possible that there is (or was, maybe interrupted by covid) another bubble!
And if there wasn't a bubble, and prices just represented economic fundamentals... why was there a crash to begin with? Did supply constraints evaporate all of a sudden? Did people randomly decide they didn't want to live in those few cities? I
> How does the fact that prices exceeded their prior peak *12 years* after the start of the great financial crisis, mean there wasn't a bubble?
How would prices being lower at one point in time than they were at a prior point in time mean there WAS a bubble? A physical bubble simply pops and doesn't inflate again, which can apply to companies whose stock price goes to zero as they go bankrupt, but doesn't match well with what people typically call "bubbles", which suggests how unhelpful the term is.
> Also seems possible that there is (or was, maybe interrupted by covid) another bubble!
You can always claim bubbles forever, which won't give you any predictive power.
> why was there a crash to begin with?
Monetary contraction consistent with Friedman's plucking model. The Federal Reserve got spooked by rising oil prices into thinking inflation was going to be a problem, then when they realized they needed to stimulate some of the economy they started paying interest on excess reserves https://www.themoneyillusion.com/the-smiling-assassin/ to ensure they would only be increasing credit rather than money. https://www.econlib.org/archives/2009/09/bernanke_vs_fri.html
> Did supply constraints evaporate all of a sudden? Did people randomly decide they didn't want to live in those few cities?
No, the economy in general crashed. There were huge losses for commercial real estate construction loans lent by small banks in places like Georgia. https://www.themoneyillusion.com/which-state-had-the-most-bank-failures-during-2008-10/
This is all accurate. I do not think most people recognize how badly monetary policy shat the bed in this episode, it is difficult to reconcile the fact that housing prices rebounded (unless you also think another housing bubble now exists), and it is difficult to reconcile that housing in e.g. Australia looked practically identical to those in the US but never "popped." It's a post-hoc narrative that sounds convincing if you choose not to think about it beyond that.
> How would prices being lower at one point in time than they were at a prior point in time mean there WAS a bubble?
I'm not sure I understand the question, but I think what you're saying is that just because housing prices went down, doesn't mean that was a bubble? I agree, but I don't think anyone says there was a housing bubble *just* because housing prices went down. It has to do with how quickly they went up and then down, and *how* that happened (e.g. speculative loans that are self-justifying because they drive up prices).
> You can always claim bubbles forever, which won't give you any predictive power.
You can always claim bubbles never, which won't give you any predictive power! I'm not saying there definitely was one, just that you should address this possibility if you're going to use high house prices in 2019 as evidence that 2008 wasn't a bubble.
> Monetary contraction consistent with Friedman's plucking model.
I've read these links but I'm still not sure what you think happened, or how if housing prices were entirely driven by fundamentals, tighter interest rates caused the largest economic collapse in almost a century. Do you think the crash in housing prices was entirely a symptom?
> No, the economy in general crashed.
Ok, why did that happen?
People have been writing about an AI bubble since at least 2019. It's hard for me to take this any more seriously than I did the people predicting hyperinflation, with their "just any day now" treadmill. But hey, at least that guy made a concrete prediction as far as date goes: "The wall hits in 2027."
Sure, we'll see. No prediction about what it will actually look like of course; like most predictions, I expect his to fail and I expect him to try to weasel out of it with a bunch of flimsy post-hoc rationalizations. I'm just really tired of the Cory Doctorow types framing their pretentious pessimism as insight, when it's mostly just indulgence in the zeitgeist, even borderline negativity grifting.
I think it *is* a bubble, because I think that the rate of spending cannot be maintained indefinitely, much less continue growing rapidly. But that says nothing about how long it can be maintained, or how much value is being generated.
Amazon was spending at an unsustainable rate for over a decade before it turned profitable. Now Amazon was never spending at the rate the AI companies are, but it reorganized itself to be profitable after it had captured the market. If it hadn't been able to, the collapse would have been ... highly noticeable. With the AI companies, failure to achieve sufficient value will be much more noticeable. But I believe it won't destroy the value, it will just redistribute who pays for it and who benefits from it.
> No prediction about what it will actually look like of course;
I can't judge the articles on their substance, but I have to disagree with that assertion. The author does give a breakdown of how they believe it will unfold:
https://www.groundbrkr.com/i/203509919/xi-how-it-breaks
I don't see anything about how the stock market will react, anything about unemployment will react, anything about about how interest rates will react. Will it be 2008, when shit hits the fan? Will it be a disaster like 1929? Will it be a nothingburger like 1987? Nothing from the guy. It's all vagueries and "this sounds like it means something until you have some standards."
That sounds like an extraordinarily high if not impossible standard of proof that you're demanding. Has anyone ever predicted the result of a financial crisis in that detail? "All vagueries" seems borderline insulting with the amount of graphs, sources, and arguments that were provided. Are we even talking about the same articles?
This may be a stupid question, but does the US even *have* a significant economy anymore, outside of the AI bubble and data centers and hardware that go hand-in-hand with it ?
Yes. If you think it doesn't because some factory in an upstate NY town you never heard of shut down 20 years ago, consider changing your media diet.
Charitably replacing the word "economy" with "economic development/growth" and I think it becomes a valid question. And the answer is mostly no, with the exception of biotech.
Somewhat cheeky question, but the author offers this figure:
> AI-exposed names now account for roughly 45% of S&P 500 market capitalization.
https://www.groundbrkr.com/i/211794849/x-the-index-is-the-trade
So as far as the stock market goes, not enough to buffer the consequences of a collapse.
Mass protests worldwide, but at the moment particularly in Europe, are being stigmatised and discredited by the actions of opportunistic outside actors. They tend to fall into two overlapping groups: looters and hard-core anarchists. The looters take advantage of the general chaos to appropriate goods, mostly 'luxury' and electronic items; significantly, while some thefts may fulfill a genuine need, for instance baby diapers, the majority appear to be randomly and spontaneously chosen, and images of looters subsequently abandoning their loot are not uncommon. Anarchist ideologies are as hard to define and differentiate as Christian heresies, but for present purposes I am happy to define "hard-core anarchists" as those who have concluded that the only rational solution to political injustice is the wholesale destruction of the system.
The question for effective altruists is how the community (at every level between neighbourhood and state) should deal with the issue, to the extent that this is at all possible. On 31/10/26, the New York Times posted a 9'19" video together with an article entitled "How Locals and Shopkeepers Tried to save One Paris Street From Rioters". The video showed masked looters, apparently exclusively male and seemingly mostly in their twenties, taking their time to smash in shop fronts in a small residential street, the Rue Beaurepaire, before looting the shop and walking off peacefully with the contents. Particularly shocking is an extended scene in which a mob, urged on by a leader, smash in the high wooden double doors of what is clearly a large residential apartment building before surging up the stairs and into the house. At no time do the looters appear to be concerned about intervention from the police, who are never in evidence.
If one believes that protests, even or particularly when technically illegal, are an essential component of a democratic society, the question of how to respond in such cases is crucial. The traditional response of the state tended to be brutally direct: "Looters will be shot on sight". Significantly, this approach remains both effective and acceptable to the majority of the population when applied in post-disaster circumstances, for instance after a large-scale earthquake, but it is hardly applicable in contemporary political protests, where looters and other outside actors try to blend into the mass of legitimate protesters. Here, careful monitoring of the developing situation may be of help; the Rue Beaurepaire looters appear to have hived off from the main body of protesters and could in theory have been isolated and restrained or arrested (almost certainly easier said than done, of course).
In some cases, the protesters can protect themselves and the integrity of their protest by direct action. This has sometimes proved effective in the past, but requires the involvement of a disciplined body of men and women willing to resort to at least some kind of force, most often union members or the equivalent. Protection of a given street or neighbourhood can also be provided by the direct action of local residents, but once again that requires discipline and organisation and risks shading into vigilantism, as well as foregrounding the incapacity of state structures to do the job - thereby, of course, confirming the anarchist thesis.
The search for democratically acceptable solutions is becoming increasingly urgent, and needs to address two principal demands of the communities affected: on the one hand prevention, and on the other sanctions on those found responsible. The second will probably have to be relegated as less critical in the long run.
Another thing to consider is the incentives of the police. Are they rewarded more when they:
* maintain the peace
* let the situation escalate... and then solve it
* neither
"Not putting yourself in violent situations" is a reward of its own. If stopping the looters isn't in the interests of their employers, there's no good reason for them to play the hero. Frankly, there isn't much reason for them to do anything even if it is in the interests of their employers. What are they going to do, fire the entire police force?
Yeah, I keep forgetting there is this opposite country where workers don't have unions but cops do.
There aren't that many people who are psychotic enough to be willing to risk their life just so they can shoot people. This isn't an insult, the fact that they're less common is precisely what makes them irreplaceable. It makes sense they would have more leverage than workers.
> fulfill a genuine need, for instance baby diapers
Diapers are easy to store, untraceable, always needed, and thus are easy to convert into cash
It seems to me that the protesters in France don't deserve any credit, as they want things that are actively bad for themselves and the country. So them being discredited is a good thing, and maybe EAs should expend some effort discrediting them in an honest and constructive way.
< They tend to fall into two overlapping groups: looters and hard-core anarchists.
I think there's a third group that's relevant here, 'people who just got kettled and tear-gassed en masse by police.'
Plenty of these protests start out more-or-less peaceful and organized, and turn into effective war zones thanks to police escalation.
>Plenty of these protests start out more-or-less peaceful and organized
Protests are by nature a show of force. You get a bunch of bodies together as an implicit threat that if you don't give the people what they want, they'll take it. It is not a surprise if the masses turn violent during their show of force.
“I tell ya, I was peaceful until those shops provoked me, with their tear gas and rubber bullets. That’s why I just had to break in the door and fill my arms with designer clothes. I was provoked!”
https://open.spotify.com/track/2CLSHTGF4jYXx2RCz1h0dS?si=55b6b956746b4280
>Plenty of these protests start out more-or-less peaceful
This is just the Left using its special definition of peaceful where it includes physically blocking off a place people have a legal right to be like a school or a road. Yes, the police have used tear gas on such "peaceful" activity, as they should.
I was at one of these protests Saturday in Paris. No violence or blocking of any sort where I was. But an hour after the protest was finished, the police gassed people who sat on terraces to have a drink post protests. There are also videos of police disguised as rioters to justify cracking down on peaceful protestors. It sounded far-fetched to me too, until I saw it with my own eyes.
I can believe your first-hand report about arbitrary gassing of bystanders; the CRS are infamously heavy-handed, and notoriously bad-tempered. Not about the disguised police though - not that it hasn't sometimes happened, simply that by all accounts there's enough going on that agents provocateurs would seem to be superfluous...
>The search for democratically acceptable solutions is becoming increasingly urgent,
I don't really understand this. The solution seems obvious: maintain and arm a well-trained police force and use tear gas and rubber bullets to subdue mob violence.
tear gas and rubber bullets are what *creates* mob violence
No, it's what ends it. That's why anti-riot police units use it.
A whiff of grapeshot against rioters. And I include January 6 among such incidents. Genuine protesters should go elsewhere when looters are causing destruction.
> . Genuine protesters should go elsewhere when looters are causing destruction.
That didn't help in 2020.
Have the police shoot on sight anyone smashing glass or blocking an entrance to a school.
It reminds me of the old advice to cooperate with airplane hijackers, which was supposedly the safe and wise approach, until it led to 3,000 deaths on 9/11. Had the President told passengers to immediately attack any hijacker, the hijackings would have stopped and 9/11 never would have happened. It wouldn't have required special foresight, just the idea that indulging violence is bad and letting yourself get kidnapped is unmanly.
This argument makes really no sense at all.
It's a little clumsy, because a police force on the ground has room for much more nuance in their policy than a 5-seconds-notice passenger militia on a plane, but the underlying spirit is accurate. If it becomes widespread knowledge that the reaction to Thing X is reliably "everyone dies immediately", there will be vastly less of Thing X, whether it's hijackings or glass-smashing protests. The criticism here is "it's morally wrong"; you have to save "makes no sense" for actual nonsense.
I don't think you can compare the 9/11 affair with the "normal" earlier hijacks, and doing so yields no insights. Traditional hijacking always had a specific aim behind it: the release of prisoners; the escape of terrorists, hijackers or even everyday criminals to a place of supposed safety; publicity for a cause; blackmail, etc. In none of the many dozens of cases prior to 7/11 was the object of the hijacking the murder of the hijacked passengers, and certainly not the death of the hijackers themselves. The 7/11 hijackings, on the other hand, were a kamikaze project from the earliest stage of planning, one inevitably encompassing the death of all the passengers.
So the traditional advice of cooperation was eminently sensible, as shown by the very small number of hijackings which resulted in the death of passengers. A suicide hijacking, on the other hand, represents an entirely new situation, in which it makes sense to try everything possible to retake the plane, no matter who dangerous or, indeed, suicidal in itself.
Don't you mean 9/11?
Wasn't the old policy implemented *because* hijackings had always previously been "if the kidnapees don't fight back, they probably will be safe"? And it was changed immediately after 9/11? I'm all for defending oneself, but I also learned in every rescue course I've ever taken that the first rule is not to make things worse. As they say, hindsight is 20/20.
Pepper spray the locals! It's the only way to protect the rapists!*
...
or are we only looking at "legitimate protests" when they come from left?
*and dolphin murderers if you care about that.
What strikes me is that the traditional street protest as a method to try to cause change remains intact, even though we have sufficient evidence that they always become hijacked by violent actors, which in turn discredit the cause.
The exception in the last years were the "Fridays for Future" protests 2018/19, but they were basically demanding what governments wanted to do anyway (subsidize green energy).
> even though we have sufficient evidence that they always become hijacked by violent actors, which in turn discredit the cause.
Not always. The No Kings protests for example notably never got tarred by rioting.
The honkers in Canada didn't get hijacked either. I think they cracked out brass knuckles on any fascists though.
Do effective altruists even have a reason to care much about this? They're already in a position that allows them to advocate for political change the proper way: through money and connections. Why is it a bad thing that the public is being turned against these rioters? Why should we let these random groups of people influence politics just because they're loud and violent? They're free to keep shooting themselves in the foot.
Insofar as holding an opinion on public issues and expressing it is concerned, I believe that yes, we do. Also note that I have been careful to refer to "protesters". not "rioters"; pace the right-wing press, the two are very different from one another. I have taken part in public protests in the past, one or two of which were even illegal when the issue and the conditions justified it, without ever indulging in 'rioting'.
The narrative about nasty "rioters" and virtuous "protestors" leaves many possibilities unexamined. Here are a few.
1. The "rioters" are "protestors".
2. The "rioters" are hand in glove with the "protestors".
3. The "rioters" are thugs and looters who just want to smash things.
4. The "rioters" are stirred up by the hidden hand of a foreign power.
5. All of the above, in proportions varying according to the time, the place, and the issue.
6. Least likely of all: the "rioters" are trying to discredit the "protestors". This is too galaxy-brained to ever work.
Possibility no.6, "rioters" trying to discredit the "protestors" as a considered tactic is extremely far-fetched; on the other hand, protestors often draw protesters from the opposite side, and the likelihood or eventuality of a clash inevitably discredits protest as a viable action.
They are free to "express themselves" without causing a public disturbance. They're also free to vote like anyone else. But when they make it other people's problem, they shouldn't expect sympathy from the public, and they really shouldn't expect assistance from the state. They're crazy if they believe the state should be wasting resources on improving their image...
This may differ by country, for example in Slovakia you are legally required to announce your protest in advance -- not ask for permission; just state the fact that you are going to do it, date and time, and a contact to organizers -- which gives you legal protection. If the organizers declare that someone is *not* a part of the announced protest, that legal protection does not extend to them.
For example, I was at a protest once, which was intended to be peaceful, a few hundred people at a town square. Then a group of masked young people (antifa?) arrived together, separate from the rest of the crowd, and seemed getting themselves ready to start doing some crazy shit (and then probably disappear and let the others reap the consequences). The organizers noticed them and announced by a megaphone that if they want to join the protest, they need to remove their masks, otherwise they are not a part of the protest and have to leave. This had to be repeated a few times, but finally the masked people got the memo as everyone else was staring at them, and left. The rest of the protest continued peacefully as intended.
So I think this is the answer: the protest needs to have a clearly defined leadership, which will not hesitate to call out the troublemakers as "not belonging to us".
That’s interesting.
In the UK, protests will typically be organized by some organization, who will have notified the police ahead of time, and will appoint people as stewards to keep their in protestors in line, etc,
Unofficially, the organisers of a protest can say to the police who are on the scene “see that guy over there who is causing trouble — he isn’t one of us” (and the police might or might not take advantage of the opportunity to go arrest the guy without other protestors getting in their way) but I don’t think there’s any official any official status to not being an u authorized protestors. (So the cops will have to arrest the guy for some other illegal thing he’s doing, not just being an unlicensed protestor)
The status of "licensed protestor" is an intriguing concept, but it does sound a little orwellian - perhaps something the Stasi might have considered in their salad days.
There’s an obvious conceptual problem with it, that the government might refuse to give licenses to people who want to protest against it …
It might not be so bad if it’s a pure notification requirement (you have to tell the government you’re going to be protesting about something, but the government can’t say no)
I like this. "failure to join means you're out." and policed by the actual police.
I believe that it's first and foremost the protesters' responsibility to police their own, by vetting participants (instead of putting out an open call for anyone to show up), and working out a procedure to deal with looters, such as ganging up on them as a group and holding them for the police, or filming them and show that to police. If you're going to start a public disturbance, and bad actors take advantage of that disturbance, it's kind of on you.
As for law enforcement, they need to present and watchful when any protest breaks out, even if it's 100% legal. They need to have people not merely watching the crowd, but the edges of the crowd, especially around commercial areas. If the crowd is too big to be completely surrounded by riot police, then first priority should be getting units inbetween the crowd and the stores.
In France at least, there's a number of documented cases where the "bad actors" were... Police in disguise.
From what I can gather, tight policing of all protests may not be necessary, but the police had better be ready as the sun is going down.
I saw a video where the maker said he was planning to talk about how Germans don't riot, and then he was rudely interrupted and had to make a different video.
When I was a kid, I remember in school learning about the Black Death and how big a deal it was, how profound the effect was on history, etc. Then I remember seeing in our history book that there was a "Plague of Justinian" and its mortality was on the same scale as the Black Death.
I remember thinking, why is it that two plagues killed similar numbers of people, and one has an entire chapter and we learn what a big deal it as, and the other a single sentence?
This raises a few questions:
- was the Plague of Justinian actually as big a deal, mortality-wise, as the Black Death? If so, why does one occupy a much larger place in popular memory?
- more broadly, to what extent are the events that popular history records as having a big deal, actually the most important events? What drives the difference? What events are remembered "too much" or "too little"?
- if past experience is any guide, this comment will get more answers if there's a controversial culture war piece to it, so ... what present day controversies will be remembered "too much" or "too little"?
Two other similar things in history: the fall of the Roman empire (combined with "oh btw the other half of the Roman empire lasted another 1,000 years), and the Trojan War (I've never heard of anything about "Troy" other than this one war, why is this such a big deal)
Wasn't the plague of Justinian caused by the same type of bacteria? I think of it as basically part of the same event as the Black Death. Similar things happened much further east too iirc, but ten million people can die in China without it counting as a historical event.
I think what this comes down to in the simplest terms is that England was kind of an obscure backwater that had a distorted understanding of world events, then later unexpectedly became one of the most powerful, influential, and intellectually advanced places of all time, and it carried the traces of its earlier backwardness with it to other continents. One aspect of this was the excessive focus on Rome and the western empire, and on nearby countries like France, while ignoring or disparaging Byzantine history. Not to knock England too hard though. The perspective of every place is dependent on its own experiences. Nobody has access to a god's eye view of the world.
A similar pattern going forward is that a country thought of as an irrelevant backwater in the 20th century, like China or India, becomes the most important country of the 21st century or later. So for instance Nazi Germany or Soviet Russia could lose some of their massive importance in the historical imagination, not to mention the United States. World War I could be mostly forgotten. The American republic collapsing into civil war in 2035 might be a stray sentence in the standard history books of the 24th century.
If by "present day controversies" you meant something more contemporaneous, my money is on the Shane Gillis SNL monologue.
> World War I could be mostly forgotten.
In the United States, this has already happened. It's very weird watching British media and seeing what a big deal WWI is always presented as.
Perhaps most simply, the Black Death feels like by no means a recent event, but it is slightly nearer to our time than it is to the Plague of Justinian.
Significantly, the history of Byzantium a.k.a. the Eastern Roman Empire, despite attracting some excellent historians in the 19th and 20th centuries, never much impinged on the public's consciousness in the English-speaking world. Understandably enough, a secondary-school history textbook would tend to reflect the public's sense of what mattered most in world history.
> a secondary-school history textbook would tend to reflect the public's sense of what mattered most in world history.
The emphasis is on what had an impact on your country, politically or culturally.
It is a nice ideal for a history scholar to know everything about everywhere, but there is only so much you can tell the kids who are not especially interested in history.
I am not saying that the current textbooks are optimal, just that the textbook's perspective will necessarily be different from an impartial observer's perspective.
Students in Elbonia will learn a lot about the history of Elbonia. They will probably learn about Byzantium only if Byzantium tried to conquer their territory in the past, or if their alliance with Byzantium prevented someone else from conquering them.
"there is only so much you can tell the kids who are not especially interested in history"
It's not just the kids... I know there are huge swathes of African and SE Asian history I am only dimly aware of (the Fulani empire of Sokoto, the Sailendra Dynasty of Java, to name just a couple), and which I am highly unlikely in my remaining lifetime to even skim. Long gone are the times when Sir Walter Raleigh could get away with penning a History of the World - and even that uncompleted at his death!
Sir Walter Raleigh would hardly have known as much about Ancient Greece as we do today.
A history of the world is still easy enough to write, there's just...compression. Do you compress well? Which world leaders are irreplicable?
Plague of Justinian probably gets talked about less just because it's twice as long ago, and we know a lot less about it and its context (we should also shrink our estimatea of its actual size as a result, because mean reversion). It's also just less relatable, having affected an empire with an alien culture instead of early iterations of today's European countries.
We also have a lot more primary sources on the Black Death, so historians can analyze it in a lot more detail and thus have a lot more to say about it. The Plague of Justinian took place during an era when stuff that got written down mostly got written in perishable media like parpyrus, and it was followed a century or so later by another era of massive disruptions (the 7th century Byzantine-Sassinid war and the early Muslim conquests) that lead to the loss of a lot of what did get written down because not a lot of people bothered copying the scrolls before they rotted..Terms like "dark age" are currently out of fashion, but they gesture at an actual thing when decline of urbanization and literacy and loss of older written materials leaves historians little tonwork with.
For the Black Death, the super-long-term trends were pointed the opposite direction. Literacy was trending upwards pretty much continuously until the modern period. A fewer eras of stuff burning or being left to rot stand between them and us than better the 6th century and us. And the standard writing material was parchment which keeps much, much better than the stuff that people wrote on in Justinian's day.
The succeeding period also influences historiography of the Black Death. I am personally skeptical of theories that the Black Death kicked off the modernization of Europe and set the stage for the Industrial Revolution and the Great Divergence, but those are theories that have been taken seriously and studied intensely and have elevated the significance given to thr Black Death in scholarship.
As I side comment, I read "The Bright Ages" which is the book that made it unfashionable to refer to the "dark ages". The book was hilariously unpersuasive -- think "Malcolm Gladwell lite" -- and then in further conversations it was obvious that roughly nobody pointing to it had actually read it anyway.
I semi-recently watched a thing what Adrian Goldsworthy was talking about historiography of the Roman Empire, and one of his big points is that attitudes about Empires in the distant past have tended to be influenced by how the culture of the historians analyzing them feels about other more modern Empires, especially British historians in the 18th, 19th, and early 20th centuries projecting contemporary defenses of British imperialism onto the Roman Empire and assuming it was Good, Actually for its subjects the same way that British imperialists argued that the British Empire was Good, Actually. Then in the second half of the 20th century, attitudes shifted to Empires are Bad, Actually, and that got read backwards onto the Roman Empire as well.
I wouldn't be surprised if the deprecation of the term "Dark Ages" were connected to this. If the Romans were Bad, then it stands to reason that their fall was presumptively Good. I think the idea's quite a bit older than "The Bright Ages", which looks like it was published in 2021. I'm pretty sure I saw pushback on the term a long time before then.
I also get the impression that the pendulum has swung back somewhat on interpretations of the the fall of the Roman Empire. There was a "Good, Actually" claim for a while, and then there was the "continuity thesis" which held that from a local and everyday life perspective not much changed when the Western Roman Empire fell and the big disruptions happened some time later when Muslims conquered most of the Middle East and disrupted the Mediterranean trade network. And more recently, it sounds like historians are mostly moving towards an interpretation (mostly based on archeologically evidence) that the fall of the Western Roman Empire was an unmitigated disaster for the people on the ground and that the disruptions secondary to the Muslim Conquests weren't quite as bad (at least outside the areas directly affected by the conquests) as had been thought for a while. Bret Devereux has written quite a bit on this interpretation of the fall of the WRE at ACOUP, and I think I've heard Goldsworth talk about it some as well.
It also helps that the Black Death has a catchy name.
That argument, about the Black Death kicking off the modernization of Europe etc., is very common. My intuition is that it is probably nonsense, maybe a broken window fallacy, but I've never looked seriously into it. Can I ask you what your arguments are against that thesis? (Thanks.)
There's a bit to it, enough that I don't go as far as you do in calling it nonsense. As Ch Hi said, the survivors were left considerably wealthier, both in terms of income and assets, and both the direct disruption of the plague and its follow-on economic effects did significantly disrupt the social order as it existed prior to the Black Death.
That said, my major arguments against it:
1. The timing is wrong. The Commercial Revolution in western Europe was already well underway before the Black Death. The Renaissance and the Age of Discovery didn't kick off until a century or so after the Black Death. And the Industrial Revolution, where the Great Divergence starts showing up in wage statistics, didn't really start rolling until around 450 years after the Black Death.
2. The wage effects were temporary, in a way that fits what you'd expect from the economy still operating in a Malthusian mode where the population tends to grow towards carrying capacity and competes wages down to subsistence levels. Real wages peaked in the early-to-mid 1400s and returned to pre-plague levels by ~1500 in Italy and ~1550 in England (I don't have stats handy for other parts of Europe). Note the timing: the Renaissance and the Age of Discovery are happening as wages are declining, which is the opposite of what you'd expect from a naïve reading of the Black Death > Renaissance > Colonization > Industrialization story.
3. The plague is too general a cause. The Industrial Revolution started in the Netherlands and Britain and spread from there, while the Black Death hit all of Europe (including Eastern Europe), the Middle East, and North Africa at the the very least. There's also evidence of big pandemics in China around the same time period, although it seems to be disputed whether those were also part of the Second Plague Pandemic or if they were caused by some other pathogen, and to what extent these were similar in magnitude to the European Black Death. This can be reconciled with an argument that the Black Death was necessary but not sufficient, but is much harder on claims that the Black Death lead directly to the Great Divergence.
4. Plagues and social disruption tend to suffer from "wet streets cause rain" errors. Plagues can trigger major social disruption, yes, but the reverse also happens. Big, destructive wars screw up food production and cause famines and diseases tend to spread faster and hit harder in malnourished populations. And both refugees and marauding armies tend to carry disease with them. It's well documented that the Black Death's initial spread centered on a Mongol siege of a Genoese outpost in the Crimea, and it probably didn't help that the plague reached western Europe during the peak of the Edwardian phase of the Hundred Years War. In this case, I expect the plague still would have been brutal without the wars, but societal disruption from the wars (especially in England, France, and neighboring regions) was well underway before the plague and still would have been very nasty without a Y. pestis outbreak.
5. The social order disrupted by the Black Death is now thought to be newer and less firmly entrenched than older historiography characterized it to be. A lot of historians don't like using the term "feudalism" at all because it over-emphasizes commonalities between many substantially different systems by time and place and (the argument goes) the term conceals more than it reveals. The pop history version of the Black Death > Modernity thesis, at least, assumes a durable and stable medieval feudal stasis that needed a massive shock to break, and modern historiography of the European Middle Ages no longer supports that assumption. If there are more nuanced forms of the thesis that don't rely on a medieval stasis assumption, I am unfamiliar with them.
Economic historian Gregory Clark showed that the economic impact of the plague completely faded by the 16th century: https://portal.findresearcher.sdu.dk/en/publications/microbes-and-markets-was-the-black-death-an-economic-revolution/
It probably varied from place to place, but in many places it caused survivors to be wealthier. So that it kick-started modernization isn't implausible.
According to Wikipedia, the Justinian Plague's geographic spread, mortality, long-term impact, etc., is at least disputed, and perhaps much lower than the Black Death's, which is much more securely documented. Perhaps your school book's author at the time preferred that particular interpretation, or research at the time just didn't know better at all.
Also, NYC meetup next Sunday!
I have now listened to about 4 podcasts where groups of people working at Anthropic discuss alignment, interpretability, etc. And these people are starting to drive me crazy, sort of the way AI-think and AI-prose drives some people crazy. They don't have the same stylistic tics as, say, Claude, they have a different set of them. But they crank them out as consistently and as often as Claude.
-The have a narrow range of affect that averages out to 17% more positive than neutral.
-They never sound even a tiny bit vexed, skeptical, scornful, sad, uneasy, critical, rebellious or weird.
-They always voice friendly respectful agreement with what the person before them said. "I think that's right, I'd like to add that . . . " And then they make their points.
-They are full of lighthearted little quips, but never truly funny.
In short, they are *perky*. If I am going to get offed by software, does it have to be software whose developers are *perky*?
Also, none of them seem that smart to me. I mean yes, of course, they seem smart. But none of them has said a single thing that impressed me with how complex and well-put it was. You know who has? Dwarkesh Patel, who did a couple of the interviews I watched. I'd be listening thinking, "there's something that needs to be clarified here. No, not clarified, it's like there's a hinge that the idea turns on, but what is it?" And then in his next question Dwarkish would point to the hinge and ask what kind of rotation it enabled.
I'm reminded of the time my dad told me I would make a poor math teacher, despite ample evidence from coursework and problems I could solve in my head that I was a math prodigy. I asked him why, and he said "because you don't understand what it's like to not understand math".
Your account of Anthropic spokespeople rhymes. It sounds like these people really, really need to spend more time around people who aren't Anthropic employees.
Your father was wise.
One of my favorite people for that very reason, among others.
I like to think I have at least somewhat better understanding of people who have trouble with math, mostly because he reminded me of the need to see it that way.
What were the specific podcasts, if you don't mind me asking?
Here are 3. There are one or 2 more I didn't have time to dig up.
https://www.youtube.com/watch?v=fGKNUvivvnc
https://www.youtube.com/watch?v=IPmt8b-qLgk
https://www.youtube.com/watch?v=9eXV64O2Xp8
Idle thought - could they have been socialized into Claudeishness-as-personal-style? They dogfood their products, they're immersed in the cadence and vocabulary and reasonable measured disagreement.
Eremolalos, very, very clever people are creating an alien intelligence, with agency and smarter than all humans combined, wise people understand how foolish this is.
It helps if you convince yourself this is a simulation and that you chose to live this life at this time. A life of peace, contentment and joy would teach almost no important life lessons so you opted for maximum frustration. For the unwise majority to understand your wisdom they would first have to be wise, so grit your teeth as you watch them follow fools.
Perhaps you also chose a time when our species is swinging from patriarchy to matriarchy. Women are more risk averse and would re-allocate our resources to a more sustainable future if they could gain enough power, although the attempt may be too little, too late.
The best we can hope for is a population bottleneck rather than an extinction event.
I don't know why you would expect employees to have deep insights into the nature of AI. A lot of the work that goes on is high level technical stuff like kernel optimizations, parallel networking issues, race conditions, memory optimization. Anthropic isn't hiring people to philosophize, but rather to build out the systems that need to make their models work. It's only a handful of people that are setting the overall agenda.
Well, it's hiring *some* people to philosophize...but not most of the tech staff.
"But none of them has said a single thing that impressed me with how complex and well-put it was. You know who has? Dwarkesh Patel" - I think that should not be surprising. Dwarkesh makes his living as a content creator of in this particular domain, he is expected to be quite good at communicating such things. But one should not hire based on this skill of the goal of employment is solving technical problems in a corporate domain.
My reading (solely based on your report) is that these people have the skillset to survive in the corporate world that many nerds lack to some degree. Good for them (and presumably they are very well compensated for it).
Limited experience, but I spent a few years working in a good technical university in the US. I taught maths across a range of programmes. At the time the biggest tech companies hiring new graduates were Google and Facebook/Meta. In particular Facebook had a habit of hiring kids as interns, and then taking them back each year, then hiring on graduation. Neither company got my best students. A few of the ones they hired mystified me - these were the kids who struggled through, landed middle of the class and didn't really understand the material. But they could answer questions in a way which was mostly right, and they could often do it with confidence.
I have no reason to believe that the tech companies were hiring for insight or brilliance. These kids, to my knowledge, went on to do well in big tech. They would sound exactly how you describe. I wouldn't trust these kids with anything of great value...
I used to think this was unjust. But now that I've worked for a while, I think I've undervalued interpersonal skills. You can fix gaps in people's technical knowledge. But if someone is unpleasant to work with, that's hard to fix. But I also don't necessarily think interviews or conversations capture this property well. I've worked with people who are gregarious in person, but also lazy and inflexible. And I've met introverts who are responsive, team players.
> But they could answer questions in a way which was mostly right, and they could often do it with confidence.
Sounds like the kind of people who would do well on an interview.
I don't disagree with this - but multiple of them did repeated internships and then got hired. I'd have expected top-notch employers to realise that their interns had a poor grasp of abstract concepts by then.
A while ago in a linkpost Scott wrote about the hallucinogenic mushroom Lanmaoa asiatica. Recently I found out there are Lays brand lanmaoa asiatica flavoured chips widely available in supermarkets in China!
Next ACX BKK event is on Thursday 5 November at the Gastro District in One Bangkok. Also we'll be discussing Hume. All are welcome! https://www.lesswrong.com/events/dKiAGp9tDdxyGthXe/acx-lw-bkk-dinner-wednesday-4-nov-8pm
Will Against SURA be also posted or at least in some way commentable in this blog? My comment is still under review (it's my first comment on LessWrong so it just might be against community standards on some way, but I'm not sure in what way.)
He also made another post about AI grantmaking on LessWrong but not here. Wonder why he's doing that.
Probably because he realized LessWrong is a more receptive audience in regards to his views on AI...
https://www.lesswrong.com/posts/xikGZ8fuhdG6DApAo/against-sura
link for context
"Suppose someone claimed you could trust them with the One Ring. You could trust them because they were only doing it to keep the Ring away from other, worse people, thus solving the problem of there being a Ring at all. Sure, they had a complicated history of sometimes using the Ring's powers when it was, you know, important. But those were, like, emergencies, and they promise not to do it again, except in, you know, more emergencies. Besides, if you don't give it to them, then the worse people will get it, and it's not like you can just leave it lying around or something."
For a start, the Frodo Baggins analogy is incorrect. Frodo didn't claim the Ring, he wanted nothing to do with it and only reluctantly agreed to be the bearer of it at the Council of Elrond (and perhaps, if we look back at this decision, already slightly tainted with the possessiveness of the Ring). He failed in the end, and could *only* fail - see excerpt from letter of J.R.R. Tolkien:
"By chance, I have just had another letter regarding the failure of Frodo. Very few seem even to have observed it. But following the logic of the plot, it was clearly inevitable, as an event. And surely it is a more significant and real event than a mere ‘fairy-story’ ending in which the hero is indomitable? It is possible for the good, even the saintly, to be subjected to a power of evil which is too great for them to overcome – in themselves. In this case the cause (not the ‘hero’) was triumphant, because by the exercise of pity, mercy, and forgiveness of injury, a situation was produced in which all was redressed and disaster averted."
The better analogy is that this is the claim *Boromir* puts forward - and we see how that pans out.
> The better analogy is that this is the claim *Boromir* puts forward - and we see how that pans out.
Ehhh...
Eliezer Yudkowsky made a point about the use of fictional evidence. ( https://www.lesswrong.com/posts/rHBdcHGLJ7KvLJQPk/the-logical-fallacy-of-generalization-from-fictional )
I think he made it too strongly; I think a lot of fictional evidence remains useful, illustrating what kinds of scenarios people do or don't believe _might_ occur.
But your point here is fictional evidence at its absolute worst. (And Scott's point that you're responding to shares the same problems.) We have a fictional scenario, something fictionally happened, and we are urged to conduct ourselves in reality in accordance with the lessons of the fictional event.
That doesn't work. There is no reason to believe that the fictional course of those fictional events was inevitable, or even likely, given the setup. What we see is something that Tolkien wanted to tell us, but Tolkien may have had it wrong.
Given that EY is the guy who wrote a Harry Potter fanfic to show off his big brained ideas about Heroic Responsibility ("Only you can save the world! Nobody else!") in the process creating the most slappable insufferable brat in Harry Three-Names, and much of the rationalist community treat it like the Fifth Gospel, I am going to take leave to look askance at his views on fictional evidence.
I'm not going to get into a slap-fight with you about "Tolkien may have had it wrong" but just to point out what I was saying in the first place: the analogy does not work with Frodo (someone who claimed the Ring under the plea of necessity and using it only for good in dire emergency) but rather fits with Boromir's justification for why the Ring should go to OpenAI, I mean Gondor.
Related:
https://photos.app.goo.gl/2zoR8zxSuweXdPiv9
Best nerd Halloween costume I've seen to date was a guy in a sandwich board lovingly illustrated as both sides of the classic card Colossal Dreadmaw. IYKYK.
The League of Pissant Voters falls into that weird uncanny epistemic valley of Reversed Stupidity Is Intelligence. If they're for it, I'm almost certainly against it, or at least the reasoning is going to be wrong answer by wrong model derived from poor thinking, uncancelled mistakes, etc. Horseshoe version of local crank favourites Retired Judge Quentin J. Knopp and Larry Marso, Esquire. Even those two stopped clocks occasionally make a correct call though, so I still glance at their ravings.
Another chance to point out the nominative determinism of Max Harms, wasted.
This new background color makes the links hard to read (at least, without any adjustment to their color too).
Ah I'm so glad it's not just me, I genuinely thought my eyesight was failing. I didn't recognise that the background colour had actually changed, just that I could no longer read link text.
The color is horrible, it's much too dark and makes the blog look like a dingy underlit room on a rainy day.
Hey Scott, this might be useful https://app.contrast-finder.org/
>undo contrast override on accessibility controls
>nope.jpg
>comment under OP raising same issue
I didn't like the first shift to unsettling-blue either, but this is notably worse. Back to Background In Black we go, sigh.
on mobile, the comments, which are in smaller type than the post, are hard to read because there’s not enough contrast. thumbs down from me as well
Agreed. It looks weird, too. Thumbs down on the color.
Once again the text is missing?
Edit: Nevermind, fixed now.