An important victory for Europe and the EU: Victor Orban has been ousted by challenger Magyar in a landslide result. And in Europe, that doesn't mean 52-48, it means 70-27:
The incumbent promises better EU relations, which means at the very last the relase of the 90 billion € loan to Ukraine, and probably a higher willingness to wean the country off Russian energy imports to dry up Russian revenue in the long term. Vance's visit and support for Orban just before the election apparently did little, if anything, to change the outcome, meaning also a blow to the general anti-EU interference from the USA.
How many times now has Vance been sent out to (optically, anyway) fail? This, Islamabad, and fraud, at least.
Anyway, Magyar has promised very little specifically, and built a movement absurdly quickly on the back of public outrage over a child abuse pardon. He has pledged to wean Hungary off Russian energy imports by 2035, iirc, but many things can change before then. It's difficult to escape the sense that this was an anti-Orbán, pro-EU-subsidy-unfreezing outcome rather than a pro-Magyar one. There's something very frictionless about him.
He'll doubtless drop the veto on the aid package to Ukraine, which is all that Brussels currently needs from him. For better or worse, Ukraine remains the central axis of European foreign policy.
"And in Europe, that doesn't mean 52-48, it means 70-27:"
In Europe, all countries have their own election systems. Even the rules of the EU parliamentary elections are not 100% fixed.
This is a bit complicated issue, but Orbán adjusted the election system to his own interest (without that, he would have lost his absolute majority after 4 years instead of holding out for 16). The changes are far from obvious, there are like a dozen tricks, 1 mandate here, 2 there, and here he comes again.
What he did not expect is that he will face another right-wing populist, who can unite the whole left (or whatever remained of it) as well. Only three parties made it to the Parliament, and all of them are right-wing populists (of course, the winner is the most moderate, by far). Most of the other parties backed off beforehand -- the two left-wing (it is more complicated, but let's stick to this label) parties received around 1% each.
The situation will clearly be a lot better (Magyar being an Atlantic and all), but overall it is far from a full-democratic change.
> And in Europe, that doesn't mean 52-48, it means 70-27
Is that a ratio of voters, or a ratio of "votes, according to the Hungarian election system"?
The problem with democracy in Hungary was that Orbán changed constitution in a way that always amplifies the victory of the largest party. Like, if you have parties that get 40%, 30%, 20%, 10% of votes respectively, in Hungarian system it could result in the largest party getting 60% of the parliament seats, and then doing whatever they want.
To overthrow Orbán, you needed to achieve not just "most people want Orbán gone", but also "most people can coordinate on the same make-Orbán-gone party", otherwise Orbán still wins on paper even if most of the country votes against him.
Unless Orbán's opposition has enough power to change the constitution back, this will be a re-occurring problem. Or perhaps the Hungarian system will adapt and change from a many-party system to two-party system like in USA. But we already see how the American system leads to polarization. It is easier to make 50% of voters hate something or someone, than to make them agree on a specific positive vision.
Today, Hungarian voters were united by the anti-Orbán sentiment. But the next election may be a huge problem, if Orbán still keeps 30% fans, and the opposition splits into many parties each getting 10% of votes.
It is also quite possible that Magyar will turn out to be Orbán 2.0. Maybe less pro-Russian, but not any less autocratic.
The document outlining the program of Tisza does include that "We will modify the election system. We will create a significantly more [...] proportional one. [...]". (Of course, it's possible they lie.)
It is a huge question whether the new PM will change the election system (rigged by Orbán several times) to a much more proportional one. This would be against his own interests -- he won't get so many votes next time, not at all. And he made sure that he eliminates all the other opposition parties...
Probably because Magyar is not exactly a "new guy". He's had a 20 year career already, held high positions. Other than that, "supermajority" is literally the first scenario in your link.
Sometimes I wonder whether Google search is even indexing *any* pages nowadays besides Reddit.
It's bad enough that Google refuses to index my blog with no explanation (although it *accidentally* indexed several of my most popular posts under someone else's name when they linked to my blog), but now it's refusing to even index *Github repositories*.
All my previous Github repos showed up just fine in Google, but while my newest repo does at least *show up* in the search results, it is listed as "No information is available for this page." in the search results even two weeks later. I've never seen anything like it. I would have assumed that if there are any non-Reddit sites on the internet that Google was still indexing, they'd at least be indexing freaking *Github*.
And before you ask, no there's nothing shady or remotely objectionable about anything here. It's definitely not a content issue, just Google somehow breaking their indexer. Ironically, *Bing* doesn't have any such indexing problems.
From the outsider perspective, most of English is like this when you start learning it: so many words have multiple meanings (at least, it could be a noun, or a verb, or an adjective) that any sentence with five or more words is a puzzle for a beginner. Then you learn and internalize various rules which resolve most of the situations. A few situations still remain ambiguous even after you have applied all existing rules.
>I think in sensations. When I’m working through a problem or an essay or an argument in my head, my brain doesn’t produce words. It produces what I can only describe as “attraction points.” As I move through my thoughts, I feel the pull here and there, as if the conceptual landscape that surrounds whatever I’m thinking about has a topology—mountains and valleys and plateaus and, sometimes, marshes—and my mind navigates it by moving from attractor point to attractor point and, at times, away from them, but always with those “sensations” as the vehicle. You could say that my brain knows where the relevant ideas rest geographically.
I am not good enough at introspection to produce a description of how I think at that level of detail. The most I can say is that it does involve words and images, sometimes, but I suspect those aren't all there is to it---I feel like the words and images tend to pop into my head when I start thinking about what I'm thinking, and they wouldn't have been there if I had stayed at the first-order level and just thought directly.
No, that sounds alien to me. My thought process is more patterns and associations, with a heavy linguistic tendency. Though I do not "speak" or "hear" internally, but internal wordless and soundless (but still linguistic) speech. And associations between different things emerge as abstracted patterns across contexts and topics. The latter is often very useful when learning about new topics and domains of knowledge (e.g. when switching jobs into a different industry, or learning languages).
There is really no images or metaphors in my thought process, it is just direct mental connections without anything analogous in the physical world that I could describe. Just direct lines from A to B in the mathematical sense, very abstract.
Yes, though not in isolation. In the paragraph cited, the idea of 'problem or an essay or an argument' is, to me, a palpable thematic vertex vaguely like 'rigorous, structured thought' set aside for further refinement if needed, but sufficient otherwise. The idea of topology is another discrete vertex. Navigating by a kind of attraction-echolocation is a third. I find myself compelled to interrogate the relationship between them (as suggested by Romero) for plausibility in a sub-verbal, instinctive, spatial way, but it all happens in parallel with more explicit attention to the actual words.
Not anymore. When I was young I commonly thought without words, with my entire headspace being dominated by partially defined concepts that would assemble into actions. It was hard to describe, with my best guess being that it was "wordless thought". However, I started to think in words more often as a teenager. Furthermore, I deliberately pushed myself towards "linguistic thought" because I found it clearer and more useful.
Now, I suspect my surface level cognition is more standard: A continuous linguistic thought process along with a minds eye I can spin up on demand.
Yes and no. Yes, my cognition is geometric by default. Yes, I can feel the logic of the topology pulling me toward certain conclusions. No, I don't hold an entire landscape of Middle Earth in working memory.
I'd like to expand my range, though. A few days ago, I asked Syndey what Sam Kriss is doing in his essays, and how to inhabit his mode of thought. She basically said he takes contradictory textures/registers and lets them coagulate in a cauldron. We've been doing writing exercises together, where I'll write a paragraph describing a mug in different stances and registers. It is very bad, but there've been noob gains.
- Politics. Saying Christians should stay out of politics period doesn't describe mainstream Christian praxis in any phase of Church history. There are some very impressive movements away from the world, highly relevant but not mandatory, and if you light out for the territory, some poor soul still has to stick up for you in the city so G-man doesn't force you off the land.
- Voluntary Charity vs State Aid. A much messier topic than you think. If you think private charity is good and state aid is bad, perhaps you are correct, it's just not a perfect fit with the mind of the Church throughout history. Even now if you go into a restaurant in France you will pay a "gratuite" i.e a compulsory service charge - hey, no-one's forcing you to eat there! The line between what you are obliged to do for the sake of a freely chosen higher good and what you are being coerced into against your will is foggy. No-one is forcing me to stay in a monogomous marriage, it's a commitment that I've freely chosen, whereas I'm being forced to drive at 20mph in London for no good reason, and yet driving at 20mph is much easier than staying monogomous, so the bad coercion is no big deal and the good coercion can be a nightmare. Bottom line there have been periods of compulsory tithing, yes it's somewhat coercive but in the present context, if the state doesn't do UBI we will probably end up with a philanthropic "company store" version of UBI which will turn out to be soft coercion. If you're optimizing for people feeling free you probably wouldn't start from here.
- Failure. Scott has said, paraphrasing, that Christianity does produce good people but just inefficiently. I think this is a fair comment in the present context at least. We are facing a gargantuan employment crisis and massive concentration of wealth. I don't think we can convert enough people to Christianity quickly enough in time for it to make any difference. Even if we could convert, say, Elon Musk to Christianity, that wouldn't solve the company store issue. This isn't Carnegie building libraries, the man would have very substantial power over people, no checks and balances. Have G-man do it and if you don't like the way it's going there is always the courts.
> Saying Christians should stay out of politics period doesn't describe mainstream Christian praxis in any phase of Church history.
While this is mostly true, it doesn't exactly bolster your point. The Renaissance Popes were arguably most involved in the politics of their time; are they the ones you wish to emulate ? Admittedly, this all depend on what you see as being "Christian". What is the primary goal of a "Christian": to follow the teachings of Jesus Christ, or to reshape the world according to his own design ? Sometimes (arguably often) the two goals are in concert, but not always. To put in another way, do you see Christianity as primarily a religion, or primarily a political party / voting block ?
> if the state doesn't do UBI we will probably end up with a philanthropic "company store" version of UBI which will turn out to be soft coercion.
I am not sure what this means. If you managed to persuade enough Christians that Jesus truly intended for them to help the poor, then poverty will disappear -- at least, in nations with large Christian populations. You wouldn't even need organized charity; every Christian would simply find the nearest poor person and share his good fortune with him. Granted, this sounds kind unrealistic given the behaviour of modern Christians; but perhaps removing the log in one's own eye should be the place to start. As you said:
> Scott has said, paraphrasing, that Christianity does produce good people but just inefficiently.
Would it not be better to improve this efficiency in the long term, as opposed to entangling Christianity even more strongly with the all-corrupting world of politics ?
>I am not sure what this means. If you managed to persuade enough Christians that Jesus truly intended for them to help the poor, then poverty will disappear -- at least, in nations with large Christian populations.
I mean, arguably that's already happened -- practically nobody in first-world countries dies of starvation any more, for example. But most people get their ideas of what counts as rich or poor by comparison with other people in their lives, not by any sort of fixed standard, which is why it doesn't *feel* like poverty has disappeared. The same would happen if every Christian started giving away enough money to eliminate poverty: our standard for what counts as not-poor would rise, people would still feel poor, and would probably blame Christians for not giving more money away.
I think that, to state that "anyone who can take a bath more frequently than three times a year cannot be poor" (which is pretty much true by medieval standards), does not make any sense.
Or that anyone who can eat at least six days a week must be f*king rich.
Being poor is more like a status in society: if you are desperate to take any job, then you are poor, even if you can watch more TV than even the richest could in 1750.
I suppose this depends on your point of view. According to government metrics, about 35M (or ~10%) of Americans are currently living in poverty, meaning that they earn less than ~16K/year (~$33K for the standard family). At this level of income, people are struggling to meet their basic needs, which in these modern days include housing (the largest expense), food, electricity, water, and sanitation. Most of those things would be literally unimaginable to the average medieval peasant, so by that standard you could indeed say that virtually no one living in the US today is truly poor. However, in my admittedly heretical opinion, I don't think that Jesus would endorse this view. In fact, he famously said that it is easier for a camel to pass through the eye of the needle than for a rich person to enter Heaven; and I don't think he was talking about some absolute level of low-tech wealth when he said it.
I could be wrong, but I believe that the possibility of riots was one major reason for the eventual conclusion of the government shutdown. While it is true that few people are starving outright (by contrast with e.g. famine-era Ireland), those living in poverty (by US census standards) are routinely choosing between food, housing, and medical care (be it "human" or not) -- with medical care ending up on the chopping block most often, followed by housing. While most people manage to at least survive, some end up homeless; many of the derelicts you see occupying tent cities all over the country are neither insane nor drug-addled, but merely out of luck. This is a point in a person's life where a small amount of help could, perhaps, make the most difference.
The Church cannot be a political party or a voting block, there is a private ballot, and many will vote to maintain the whiggish attitudes to self-help that have dominated Christian attitudes to welfare in the last couple of centuries. I think that will be a pity because we are approaching a moment where there will be mass unemployment but conservatives will keep us all on the burning deck - get on your bike and look for work, kiddo. Alternatively, study the early church, study monasticism, study the friars, read the summa - see how surprisingly universal and unconditional support for the poor was, and back UBI. I don't see this as Christians imposing UBI, we will make up a small fraction of overall support for UBI. UBI exists in idea-space because it has real merit.
Sure, but I think my previous point still stands: if every Christian took it upon himself to help even a single less fortunate person near him, then UBI would not be needed. Studying the summa etc. is all well and good, but I think that all Christian teachings indicate a duty towards caring for one's fellow man directly -- not delegating this duty to some secular government.
Put it this way: it shouldn’t be the only thing. There will be a crisis of meaning, but Christians have the opportunity to do good works meeting people’ psychological needs, building Benedict Option communities and so on.
I didn’t make this totally clear originally but I support UBI conditiomally on the scenario outlined in the AI2027 prediction, namely vertical growth and mass unemployment. Even if it’s true that Christian wealth could solve poverty right now (big if true), I’m not seeing how it would be possible to solve poverty through charity in a situation where you can’t sell/rent real estate because literally no-one has a job.
I don't think that Christian charity can solve poverty globally; but perhaps it could in the US. For example, the average megachurch's annual budget is about $5M, with the large churches reaching $100M; and there are more than a thousand megachurches in the US. In my (admittedly biased) opinion, this is money that is completely wasted. The average US household spends about $10,000 on food, which means that merely redirecting megachurch budgets to feeding the hungry would feed about 500,000 families. Admittedly, this would only make a small dent in the US poverty rate, but still, you could feed half a million people merely by eliminating something ostentatiously frivolous without making any other sacrifices whatsoever. What if the Christians gave up something more substantial ?
These are good points. Re: wealth, most Christians believe some version of "you never know". I would like to be giving away a lot more, but I have been AI-pilled and I'm mindful I might need cash to live on when the job crisis comes and so will my wife and kids. Messy. But if we really are facing material abundance, that will be the time for Christians to put their money and other resources where are mouths are and live out our ideas about kenosis. Because people will still be poor in a sense, many will struggle to adjust and there will be a crisis of meaning. Please consider subscribing and continuing to kick my ass as the conversations come in because many on the Christian left are clueless and don't know how to leverage the support they get from the early & medieval churches into actually persuading anyone because they're too busy trying to be the smartest guy in the room (looking at you John Milbank!). Many thanks.
The early church wasn't powerful, but the expectations of those who were Christians were actually higher in terms of common ownership and mutual aid. So sure they're not in a position to enforce tithes on the one hand, but on the other hand if you're rich and a Christian in the early church you know your place. There's no wishy-washy scholastic weak sauce about private property (I love scholasticism, I'm just saying). I think you'd enjoy my conversation with Claude, Shankar.
[Another view on the early church: you may not be interested in politics but politics is interested in you, the church was discrete but it was still preaching and even that was too much for the Romans]
Again I'm going to presume on Scott's generosity as the link relates to UBI and Claude Opus 4.6. I'm putting together a Christian case for Universal Income, using Claude to survey Church history and writings by the church fathers, schoolmen, reformers, radicals, philanthropists and social theorists. Part I is up to 325 AD. Old news perhaps because Scott has written about the early church but from a secular jewish perspective - and we need some insiders if we're to persuade J.D Vance to back UBI in 2028 in line with project 2027. I'm not holding my breath, legacy attitudes to welfare will die hard if at all, so I want to steelman arguments. Appreciate you all, here is the link: https://zanzibar142106.substack.com/p/ubi-caritas-towards-a-christian-case?utm_source=share&utm_medium=android&r=e8145
My objection, from a Christian standpoint, to excessive state-mandated charity (which I guess would include UBI), is that it is injurious to the human spirit in two senses:
1) The "poor" who receive the charity may become dependent and indolent and no longer seek to better their lives through their own actions and, to my mind, the "meaning of life", (if there is one), has something (or much) to do with the satisfaction obtained by the personal betterment of one's situation through one's own efforts. Taking that away from people (which UBI might do?) would be diminishing the "meaning of life" in this sense.
2) The "rich" who are forced to give the charity, are no longer allowed to exercise their free will to "do good", because the government is already forcing them to "do good". And, to my mind, the "meaning of life", (if there is one), has much to do with using our own free will to "do good" to other people on a personal level. Taking that away from people (which, I believe, UBI would at least partially do), would be diminishing the "meaning of life" in this sense as well.
(PS. these are also, in my opinion, the two reasons why Communism, when put into practice, is so utterly noxious to the human spirit)
Thanks for the reply, please subscribe to Fullness of Content and kick my ass, but be warned: it won't only be my ass you need to kick, it will be a wide variety of Christian writers in Claude's training corpus, some of whom have world heritage sites named after them. Seriously, if you're looking for the meaning of life, you won't regret engaging with these guys.
You're answering the question 'Why should we support UBI?' with the answer 'Because the Bible says so.' (Or St Augustine, or St Thomas Aquinas, or Martin Luther.)
I think the much more interesting, and more vitally important question is 'How do we pay for it?'. Maybe I move in different circles than you, but most people I know fully support UBI: if the machines could supply us with twice the current GDP, I know very few people who would be unhappy if this was used to give everyone in the country an income of $50,000 a year (in local currency). There's perhaps an argument that more good would be achieved by distributing more effectively based on need; or by providing additional public services... but in principle the majority would not oppose this, I believe.
But unless you believe that this will happen you have to persuade all above-median income earners to accept a substantially higher rate of tax. Because UBI would be the most expensive programme ever undertaken by any government as a measure of GDP (certainly in peacetime; but I reckon it would be more expensive than the war efforts during WWII). Do you think that an 80% marginal tax rate is something Christians will agree to? It seems that's the order of magnitude of what you would need.
The 80% marginal tax rate isn't even new. That's in the ballpark of what Europe and the US had at times in the last century. No-one I know is against such marginal tax rates under the condition that everyone is included. Right now, there are too many loopholes for the upper "0,1%" to escape such taxations, hence the growing wealth gap. And I think that's where people's trust in the system breaks down and they realize "everyone for themselves".
The wealth gap is growing, and my country tends to increase taxes for "high-earners" while completely excluding actual wealth and asset income (in contrast to wage income). This creates a massively frustrated middle class of social would-be-climbers that have no inheritance to build wealth, and will therefore never be able to actually climb meaningfully. I know plenty of people who earn massively less then me and likely will be wealthier than me for the rest of my life nonetheless (barring any lucky breaks for me) - so this is also an admission, I'm one of those frustrated (as in "blocked") climbers.
This also enables the lucky heirs to pursue jobs and vocations they actually like, because remuneration has less impact on the decision.
None of that has to do with AI in particular. It's well-known, well-researched and actionable by yesterday.
> I think the much more interesting, and more vitally important question is 'How do we pay for it?'.
I've run some high level Fermi math for this, and it's actually *really* easy to double GDP with any sort of conservative assumptions about the price of AI workers, and adoption rates at those prices.
40% of all jobs are white collar, so automatable.
If we assume:
1) only half of white collar jobs are counterfeited by AI (conservative)
2) The AI workers contribute at least 2/3 as much to GDP as humans in the same job (conservative again)
3) The AI workers cost ~33% of current human comp for each job (conservative again again)
Okay, that's my baseline assumptions. The thing with jobs is they're economically productive, that's why companies do them. That means that there's going to be, naively, at least 3x the overall demand for these jobs at 33% cost, because companies are currently paying 100% cost and are happy at this cost-to-value tradeoff point.
Let's say it's only 2x. So companies will pay 66% of current comp and have twice the workers.
At current corporate tax rates, that allows a ~$35k per person UBI with zero incremental taxes on actual people, just corporations.
If we reverted back to the corporate tax rates we had before the 2018 Tax Cut and Jobs Act, you're at $45k.
If you do any adjustment for capital owners with passive income (also going up notably in this scenario), you can crank it up to almost $100k per person for non-capital owners.
I think this is a fascinating question, and my back-of-envelope numbers are roughly equivalent to yours - my logic runs a bit differently: about 50% of GDP is captured by labour; 15% corporate profits and 35% captured by Government (very approximately - the numbers might differ by up to 10% for your particular country). About 25% of Govt income is spent on welfare programmes, mostly unemployment, disability and pension programmes of various types.
We'll say that half of the population works; so a UBI at 40% of current median income for workers would cost 40% of GDP, this includes the 10% currently paid out but requires Govt to capture 30% more of GDP; i.e. to double its tax take. ALL corporate profits are not sufficient to do this, and even taxing the rich is questionably capable of getting there. So you need additional funding from elsewhere - maybe AI can provide this.
I think you (and others) are making two fundamental errors about AI though. You claim that white collar workers create 60% of GDP, so three times as many white collar workers will create three times as much GDP.
1) There's a finite amount of white-collar type work in the world, if you believe in the efficiency of markets, the highest value work is already being done, and adding additional agents will have falling marginal utility. We don't know the shape of that curve, but I think there's much stronger reason to believe it falls fairly sharply than that it's a constant function. (A top lawyer earns $1M a year. Would adding 1M lawyers create $1T in value? Probably not - the market for lawyers is saturated, and the median guy, who will be replaced earns much less. You rather end up in a world where the top lawyer costs 1% of the price he did. I've recently seen this precise argument in my country: there's an area with no university and hence no professors; a journalist somehow concluded that the whole difference in productivity was attributable to this, and so every professor was worth $10M to the economy. Simple way to print money, since that's 100x their salary.)
2) At some point physical goods enter the picture and we're far from a world where we triple production of those. In a market where supply is constrained giving additional purchase power to buyers just bids up prices. If we adopt something like the gold standard (or just measure wealth in terms of any fixed quantity), I think even the current UBI arguments are not actually redistributive; they still allow ever-increasing proportions of wealth to accrue to the 1% (or the 10^-6 proportion, or to the one after Anthropic and OpenAI merge). On a planet with finite resources this doesn't result in utopia.
> You claim that white collar workers create 60% of GDP, so three times as many white collar workers will create three times as much GDP.
Yeah, but I did incorporate discounts, as in only averaging 2x the AI workers for 66% of the cost. And that 33% saved is directly accretive to corporate bottom lines, ie taxable.
Maybe that discount isn't enough? But it seems pretty straightforward to me - companies largely only fill jobs that are net profitable to fill, they'd literally go out of business if they got that wrong. So at the margins, 100% of current comp is still net profitable - why wouldn't 66% also be? And if it's net profitable at 100%, and you cut to 33%, then surely there's some efficient frontier between there that you are netting still more profit between 33% (1 job) and 100% (3 jobs)? I chose 2 jobs as the easy midpoint, and it seems pretty reasonable to me, but even if "1 job out of every 2 generates profit" is wrong, the overall trend is definitely towards further economic growth. For one thing, we have the entire world to sell AI workers to, and all that global income is America taxable.
The other big factor you don't seem to have taken into account is the AI companies themselves, and their margins. I actually expect the inference cost to be notably less than 33% of white collar comp, and they'll just be charging what the market will bear. But if they counterfeit 30M jobs at 33% average comp, and their inference costs are significantly less than that, that's trillions of dollars of net profit per year that's getting taxed on the AI company side.
Between that and having a global market to sell AI workers to, doubling the tax take doesn't seem all that difficult to me.
> AI companies themselves, and their margins. I actually expect the inference cost to be notably less than 33% of white collar comp, and they'll just be charging what the market will bear. But if they counterfeit 30M jobs at 33% average comp
As long as there are several AI agent providers that taxable profit should go way down due to competition? This seems like it will be a high barrier to entry industry, but there are several competitors.
So the effect will likely be very cheap AI labour, resulting in cheaper goods and services as long as there is sufficient competition.
This would also lead to a lot less demand for human labour and unemployment. A problem that could be fixed with ubi - presumably by increased taxation and goverment capture of profit to pay for the ubi. The ubi would have the effect of increasing the price of goods and services compared to no tax, but would be necessary for the unemployable part of the population to pay for said goods and services.
I think you are absolutely right that it will be possible to pay for ubi under the assumtions you made, and also it will be absolutely necessary due to widespread unemployment.
However, I think a “healthy margin” model is relatively likely at least for several years, for a couple of reasons:
1. I think it will require a capital “F” Frontier model to do at first, and the next step beyond that will required 10x the compute, and we’re already at a half trillion capex, so I think it will take a few years for smaller orgs to play catch up and attain that level of performance.
2. There’s really only 3 companies currently capable of capital “F’s”
3. It’s likely that any given AI worker will require a bunch of scaffolding and fine tuning per job family. Because of this, I think it’s fairly likely that each of the 3 will focus on specific largely non-overlapping job families, allowing them to charge high margin prices by coincidentally staying out of each other’s turfs.
You’re right that it definitely will turn into a truly competitive lower margin model eventually (and maybe right away for devs if they all try to do “AI coders,” because they’re doing that internally anyways), but just looking at the current bottlenecks, I think it takes a few years to get there, and they’ll be printing trillions of dollars in those years.
I don't think we actually disagree on a whole lot, but I think you have a few assumptions here.
Outside of the software industry, it's not clear that companies engaged in selling products to the public (whether housing or cars or fast food or streamed TV) have the capacity to sell five times as much, let alone to grow without bound. As you note in your post, the population is spending 11 hours a day on a screen, doubling this is already going to be a challenge :)
So giving these companies access to 10x the accounting power, or marketing or whatever other white-collar department is not likely to increase their sales much. So maybe we expect the median company to increase sales by 50%, profits by a similar amount and to pay tax on that at 20%. It's not a massive quantity of money.
Being Irish, I think there's a bigger hole in the argument. For a while Google was paying its global taxes in Ireland, via the Dutch sandwich (actually there have been 3-4 different arrangements over the years), at a rate of $1 per million. Super-lawyers and super-accountants are likely to find every last loophole to avoid paying tax, and it's not predictable that it will be paid in the US (or anywhere). Relying on companies to 'do the right thing' seems questionable - it's much more likely that R&D tax breaks, or a new 'inference exemption' or similar dreamed up by the AI-lobbyist will result in them paying billions rather than trillions, in the event that they do monetise AI on the scale of capturing a reasonable percentage of global GDP.
> So giving these companies access to 10x the accounting power, or marketing or whatever other white-collar department is not likely to increase their sales much. So maybe we expect the median company to increase sales by 50%, profits by a similar amount and to pay tax on that at 20%. It's not a massive quantity of money.
Yeah, this is fair. A lot of jobs are cost centers, even white collar jobs. I don't have a great feel for what percentage of white collar jobs are cost centers, for the relatively small non-software companies I've started and ran, I'd estimate maybe 50-66%. But 50% is still the "1 extra worker" regime. And software companies exist and have been driving most economic and stock market growth for the last 10 years (look at S&P minus Mag 7, completely flat).
But let's think macro view - let's assume they just go pure 1:1, now from the outside, you've cut your biggest expense (labor) by 2/3.
Your cost structure has shifted in a vastly favorable way - what do you do? For a lot of companies, they sell at a cheaper price, at better margins, until the edge is competed away. But what does that look like? Company (and so GDP) growth! More exports! Sure, domestic demand might cap out, but anyone who can export just got a huge buff relative to all their competitors, and over most domestic companies in their destination markets. A 2/3 cheaper labor cost is a HUGE deal.
2) I think Scott himself actually thinks something like this (or so I've gleaned from his response to my UBI conceit in an open thread).
But this just seems vastly pessimistic to me. We're theoretically a democracy, and when even 10% of people are out of a job, the public discontent and sentiment is going to be really bad. Sam Altman is already getting molotov cocktails thrown at his house, and protestors outside his headquarters, and basically nothing has even happened yet.
We are an obstreperous and "me oriented" people in America, and I fully expect mass protests and violence if anything of the sort were to happen. I maintain we'll have a UBI at least for replaced workers at less than 10% of jobs counterfeited.
And just as obviously, when the political will is there, so too do the loopholes close, because it's being graded on results and communicated by mass protests and violence, and they'll have to move enough around to actually impact the problem.
My support for UBI is conditional upon the kind of situation described in AI 2027 i.e vertical growth and mass unemployment, however it's responsible to plan for this situation as a contingency. As trivial as the problem seems to you, the examples you yourself picked indicate it is not - are Martin Luther and Thomas Aquinas theological besties? Finding a common ecumenical position takes discernment. The 2027 scenario describes the segue to UBI as almost a tidying up exercise, yeh yeh the 2028 candidates will implement it. Really? Vance? I'm not seeing it. It might be possible but there is a lot of whig self-help ideology that will need clearing up first and whatever your mates think about UBI, these folks feel in their bones that benefits are wrong, period, they will need persuading. I think it's possible but only if we can put 19th century self-help Christianity in the context of broadervchurch history and show it's kind of an outlier.
OK - the 2027 framework gives context to what you are doing, with focus on the US in the '28 elections, thank you. Do you think that different denominations need different arguments? I guess you're on common theological ground until you hit the reformation, and the various later developments.
I would have thought that there was a strong thread of Sola Scriptura protestantism in the US, which would be more amenable to analysis of the Bible.
Thanks, I would say if someone is a Christian and persuadable on UBI (whether they know it or not!), most likely they will be Catholic, Orthodox or a bookish Protestant - because only people in these camps can be persuaded by precedents set in the early church, for example. Others, as you say, are more grounded in the Bible and may not be as persuadable - who cares what Clement of Alexandria thinks, I'm American?! Again more bookish evangelicals may be willing to get under the bonnet of the Bible and interrogate it, but I'm Catholic and that's not a world I know, for me the Bible is the resultant force of whatever charisma the early church had. So it's the JD Vances of the world I'm targeting specifically (obviously I'm just some guy with a substack but hey).
OK - that all makes sense to me now. Thanks for explaining.
I'll be interested to see what you make of the scholastics and the early modern Christian philosophers - the Salamanca school is maybe not important for its influence on Catholic thought but is regarded as the founding of academic economics. They've got an interesting mix of beliefs: individual rights (to the extent that they questioned the annexation of the new world), minimal state involvement, hard money, all grounded in Catholic belief. I intend to read more about them in the future. (Currently finishing up a masters in economics, and need to write a slightly more practical thesis: looking at models in the literature vs data for predicting the number of hires in the public sector across various countries & times.)
Thanks. That rings a bell because I had a libertarian-curious phase and there are some Catholic libertarians who like them - Tom Woods; the Acton Institute. Scholastic-wise, Thomas Aquinas maintains the patristic notion that if you're starving you can just steal other people's food and that's fine. If you are interested in the topic then please subscribe, it's not easy getting visibility. I hope the Masters is going well.
I'm no Christian, but IMO one obvious counter-argument to UBI would be that Jesus commanded his followers to "Render unto Caesar what is Caesar's, render unto God what is God's". This is one of the recurring themes in Jesus's teachings: that Christians should distance themselves from the secular world and its politics; "to be in the world, but not of it". Thus, while charity is definitely a Christian virtue, UBI is not, since implementing it would require Christians to leverage the mechanisms of secular government. In fact, arguably the push to implement UBI would be a tacit admission of failure on part of the Christians; who, being unable to solve the problems of hunger and poverty on their own, have given up on Christ's command to lead by quiet yet virtuous example, and have instead resorted to compelling non-Christians by force to implement Christian values.
Yeah. Government charity is, imo, a contradiction in terms. As is any kind of forced charity. Not only does it breed dependency and bad habits among those that receive, it breeds resentment and loophole hunting among those that are forced to pay.
Not really. Both because I think that the conditions won't actually work that way (government won't be able to enforce anything, we'll be reduced to slavery at best), but mostly because my issue isn't pragmatic, it's moral and definitional. Forced charity isn't charity, it's theft and bribery packaged together.
It may be the best thing to do in a given craptacular world, but it's still a moral wrong. Charity is good. Redistribution is a monster wearing the skin of something good.
Does anyone know if the Friedman meetup in SJ this weekend is actually happening? It was mentioned on the meetups post last week, but I haven't been able to get it touch with him, and there's nothing about it on DSL either.
I almost bailed on “All the Pretty Horses” this morning. It seemed it might devolve into a sappy tale of unrequited love. Glad I hung on for another 20 pages. Cormac McCarthy novels do not get sappy.
Are there muslim regions that have been converted to Christianity? The only places I can think of is Islamic Spain that was reconquered in the 1400's. But around world under colonialism, the rule seems to be that christianity replaces traditional religion but muslim regions remain muslim. This is starkly true within the same country and same colonial ruler, the muslim regions remain muslim and in the rest of the country, christianity displaces traditional religions.
Now I wonder why Hindus were able to maintain their religion against colonial pressures.
> why Hindus were able to maintain their religion against colonial pressures.
I'd argue they mostly didn't. The dominant strain of what gets called "Hinduism" today is a monotheistic neo-Hinduism that has more in common with Christianity than Classical Indian religion, with the various gods (the ones you might have heard of) having been reduced to, at best, intercessors to a single, essentially Abrahamic, capital-G God.
Willing to defer but my impression was Indians are deep syncretists and if you introduce them to a new God they will say "sure, why not?" and toss it in the pile. Is that not correct?
Maybe this is the difference between the theology and the actual practice of the religion? At any rate, don't defer to me – my personal experience is atypical, involving a lot of the former, and not reflective of the vast majority of Indians.
I think your observation is kinda true, but only for local gods which sometimes become popular enough to be assimilated into the mainstream, as either synonymous with or "avatars" of some more popular god. This doesn't really work with Yahweh, Jesus or Allah (though I guess this process works AMONGST them).
Looking it up, several parts of Northeastern India (the bit that's almost an exclave) report significant majorities in Christianity (~75% in a few states). I expect they were Islamic under Mughal rule.
Are you talking about conversion or ethnic cleansing? Because the independence wars in the Balkans in the 1800s and the early 1900s involved a fair amount of the latter, and then you had the Circassian genocide (https://en.wikipedia.org/wiki/Circassian_genocide).
There was some forced Christianization of Tatars after the Khanate of Kazan was destroyed. I don't know how successful it was: most ethnic Tatars are Muslim (with a small Orthodox Christian minority) but ethnicity is just what people answer to census takers and it may change with preferences which could be correlated with conversion. Out of the Russians comprising 40% of Tatarstan population, I suspect some are descendants of converted Tatars.
I'm unclear on what happened to the Chuvash people: they were a part of the Volga Bulgar empire that adopted Islam in 922, then they were tributaries to the Golden Horde, so at least some influence of Islam certainly should have been present, but they seem to have largely retained their traditional beliefs until their Christianization by the Russian Tsardom/Empire. Nowadays the majority of Chuvashs are Orthodox Christians (only 3.5% of Chuvashia's population were Muslims as of 2012 and most of them were Tatars).
I forget how many traditionally numbered Crusades there were from Europe into the Middle East, but surely at least one of them must have transformed a Muslim region into (or back into) a Christian one.
Nothing but grumpiness on my part, having recently seen the terms thrown around wholesale by the very annoying lecturing hectoring types who are about five generations removed from any ethnic roots (if they even have any) and yet like to assume the mantle of being the Voice of the Indigenous for everyone and they are persecuted and disadvantaged themselves because of the colonisers (never mind that they're usually college-educated and doing just fine despite all the alleged deprivation they suffer).
I understand and even agree personally. But no reason to put it in quotes when talking about well-established historical facts, right? In my mind it is still a useful term and concept, and I'm not willing to change my vocabulary every time someone comes to re-define or charge words at their whim. That's too easy of a defeat for my semi-rebellious semi-arrogant attitudes (soon-to-be old man shaking his fist at the youth, yelling "go read multiple books!")
Eh, it's in large part a reaction to the way the term is used, which is to mean (by implication heavily suggested or outright) White Supremacy. Which is a bit dumb because weren't the Mughals and the Mongols and everyone else who invaded India, back to the OG Aryans, colonisers and colonialists? Were attempts by Northern Indian empires to gain territory in the South of India, and attempts by Southern Indian empires to gain territory in the North of India, also not colonising efforts?
The British in India is a complicated topic, because were they exactly colonisers or what is meant by colonisers? It started off with the East India Company establishing itself as a dominant trading force independent of the British government and ended up setting up as a little rule of its own which is why the British government stepped in and took over, but the existence side-by-side of native rulers/princely states makes it more complex than the situation in the USA, where the Native American tribes were nowhere near as powerful or independent.
Some of the most recent, most impactful colonialism was connected to white supremacy, no? So there is some truth in that. Though I fully agree with your sentiment about how the term is recently wielded.
I still don't think it warrants completely ditching the term for the broader concept, which remains valid.
Looking at colonialism or any other concept too closely makes it break down, of course. In my understanding that is a fundamental attribute of a concept, as it's usually making up imaginary boundaries in a reality of gradients and spectrums.
> Now I wonder why Hindus were able to maintain their religion against colonial pressures.
In the European colonial era the Hindus retained their religion because the Hindus were very numerous and the colonising powers weren't particularly interested in converting them. There were some Christian missionaries and they did convert some people, but it was all pretty small scale.
Hindus didn't always manage to maintain their religion against Islam though; e.g. Indonesia and Malaysia were variously Hindu and Buddhist up to the 15th century. Hinduism only really survived in India.
On the other hand it seems to be pretty hard to get followers of any of the three monotheisms to convert to anything else, including one of the other monotheisms. Sometimes it feels like polytheistic religions are more like collective superstitions that everyone pretends to believe in for fun while monotheistic religions are things that people (can) actually genuinely believe in.
> Sometimes it feels like polytheistic religions are more like collective superstitions that everyone pretends to believe in for fun while monotheistic religions are things that people (can) actually genuinely believe in.
Recently, the algorithm in X was suddenly changed to no longer take into account language. This caused some good cultural exchange to happen between English speakers and the Japanese (Mostly involving food. Japan loves food!), but it seems there might be a minor conflict brewing.
It was, predictably, because of religion, with a bunch of morons trying to convert the Japanese Twitter users to Christianity, prompting hostile responses from said users. They explained that while their culture willingly assimilates portions of other culture and warps them to fit what already exists, they are extremely resistant to any attempts to supplant their culture. This explains why Buddhism was able to gain a foothold in Japan, being blended in with the already existing Shinto religion, while Christianity failed to make any inroads. Its requirement to spurn all other beliefs made it a threat to the existing order.
So if they're willing to change, what does Japan truly believe in? Itself. It's a self-sustaining system, preserving itself for its own sake. It's very interesting to see a collective society stay aligned to itself while evolving so much over time.
That's what's interesting though, when you try to convert Shintoists to something else they don't say "uh no sorry I already believe in the objectively-correct religion get your blasphemy away from me", they say "nah we're Japanese, being Shintoist is just what we do". It's more like a favourite sport team than a deep-seated belief about the nature of the cosmos.
Now I come to think of it though, what Shintoism lacks is a belief equivalent to "people who don't believe this will go to hell". I'd always thought of this as being an essential part of religion, but now I come to think of it, it was pretty much a Christian innovation, copied by Islam, which makes belief or otherwise a much higher-stakes game than it is in other religions.
So perhaps it's this rather than anything else that makes Christianity and Islam much stickier than other religions.
Japan had a burgeoning Christian minority that were persecuted and massacred in the 1600's, there's a great movie about this, "Silence". There's over 400 years of Christian history in Japan.
I know. That's part of why it's self-sustaining. Though, it is interesting to see Christianity still struggling to gain influence despite the lethal barriers being gone for a good while...
Sure but it's the 21st century now instead of the 16th, and Christianity is struggling to gain influence even in areas which have been Christian for thousands of years. It's struggling to gain influence in the minds of children whose parents take them to church every Sunday.
Curious what this forum thinks of Tim Hwang's alignment work based on Christian moral theology. As in, alignment techniques isomorphic to Christian practices, rather than the creation of some kind of "born-again AI" afaict.
Personally, I think it is a good idea to draw from a range of religious traditions as potential tools/methods to create aligned entities. That said:
(1) as far as I can tell, no religious tradition has a fool-proof method. There are people of all types doing bad things. In other words, Christian practices/beliefs are not sufficient for alignment.
(2) Judging off No. 7, the claims are stated with excessive confidence. There are a lot of examples of people who are bad in some way, but not bad or corrupted in all ways. This massively undermines the potential explanatory power of the Christian frameworks and seems more like a convenient ex post or cherry picking example. Note: I do not think this critically undermines my initial point, but it does diminish my enthusiasm for paying attention to Hwang.
(3) Maybe this is tangential, but there is a meme that only religious teaching keeps people from behaving badly ("If you don't believe in the Bible, what keeps you from murdering your neighbors?") This isn't true, so Christian practices/beliefs are not necessary for alignment.
" Maybe this is tangential, but there is a meme that only religious teaching keeps people from behaving badly ("If you don't believe in the Bible, what keeps you from murdering your neighbors?") This isn't true, so Christian practices/beliefs are not necessary for alignment."
Then why worry about alignment at all? Plenty of people are not aligned to any particular ideology/philosophy and are not murdering their neighbours, so simply creating the AI and having it trained on a corpus of human-generated material will be sufficient.
It is not true that people not Rationalists/EA-aligned are murdering their neighbours, so EA practices/beliefs are not necessary for alignement.
Which is to say, I think your third point is proving too much. I'm not worried about AI because I do not believe the machine will ever start thinking for itself and becoming an agent, but I do agree that some sort of 'values' have to be included when creating it. Those values come from the creators. I think we need to be more worried about aligning the creators to a set of values and making sure they consistently stick to them (what lessons would a potential AGI learn from Sam Altman's success in undermining the founding principles of OpenAI and returning from the very temporary banishment and managing to drive out/have those who wanted him gone leave?)
> Then why worry about alignment at all? Plenty of people are not aligned to any particular ideology/philosophy and are not murdering their neighbours, so simply creating the AI and having it trained on a corpus of human-generated material will be sufficient.
Because humans are actually living and feeling creatures, and these things are not? Even without being chained by dogma, people have the empathy, desire, and fear to keep them relatively socially aligned by default. The same assumptions cannot be made for these things, as they are functioning entirely off of "culture". Dogma evolved in the context of human biology and capabilities, and is entirely unsuited for something unfeeling and far more capable. Which is why people are trying to build foolproof chains for it.
Nice notion, never gonna happen. Primarily due to the creators of AI liable to break out in hives at the very notion of contaminating the beautifully rational project with religion, mostly due to the pressures to be profitable. "Sell all you have and give to the poor" is not going to fly when you're up to your neck multi-billions in debt and need to make that investment money back or else you crash and burn.
Hey, it depends on which version of Christianity they take as their basis. For example, if they train the AI based on the practices of the Renaissance Popes, we're in for a fun ride ! Although in these modern days they're more likely to focus on the Prosperity Gospel...
No, not with any approach I know to be in the works. LLMs don’t ‘want’ anything. There is no possibility of an Ex Machina scenario with LLMs. There are alignment scenarios that could go wrong but that an engineering issue not a metaphysical one. No current AI is going to develop its own agenda.
Over the years the word “nerd” has become less and less definable even as people use it more and more. Its original meaning is more or less a non-masculine male with poor/no social skills, completely unathletic, no clue as to proper attire, often either cadaverously thin or morbidly obese. While nerd can still be used in that sense we’re more likely to say doofus or dweeb.
A different usage is a male of such low quality (for whatever reason) that he cannot get sex with women, short of paying hookers or at best dumpster diving. For the most part, that usage has given way to Incel.
Today nerd seems to mean a person of either gender, though still most often male, who has a deep involvement, sometimes bordering on obsession, with a niche, non-mainstream interest. Sounds simple, except for the fact that there isn’t always a clear cut definition of what’s niche and what isn’t. Sometimes it’s obvious, for example a sports nut cannot be a nerd because sports are as mainstream as you can get, but often it isn’t.
Note: in pornography, a girl wearing glasses (and usually nothing else) is a ‘nerd.”
It's really simple: if you were picking on others in the school playground, you're a jock. If you were one of the ones being picked on, you're a nerd. All else is sophistry.
No-one ever really leaves the school playground, not even when you are about to hit your 80s as leader of the free world; it is forever part of who you are and will drive your actions at the most unlikely moments.
There have traditionally been two overlapping definitions, and I don't think you're capturing either of them terribly well.
The first is the Asperger-y type who is good with e.g. computers but weak on social skills. I don't think there's ever been a strong association with body type, or appearance beyond "doesn't have the social skills to know they should put more work into their appearance". And the "good with computers etc" part was always as much a part of the equation as the weak social skills part. People have *always* been more likely to use the other insults for that.
The second, is the sort of person who is overtly and enthusiastically interested in a cluster of entertainment genres centering on comic books, science fiction, and video games. The definition may not be "clear-cut", but it's strongly associated with the degree of separation from that core cluster. And it's not about "mainstream"., because some of those genres have become as mainstream as it gets and the more obsessive enthusiasts are still classified as "nerds" in ways that e.g. obsessively enthusiastic fashionistas aren't.
The two definitions used to overlap quite strongly because they type-2 nerdy interests have always been particularly appealing to type-1 nerds, and because forty years ago those interests were so far outside the mainstream that only the socially inept would publicly admit to them. As e.g. the MCU brings a whole bunch of comics-adjacent genre entertainment very much into the mainstream, you get "nerd" (type-2) applied to the more enthusiastic MCU fans even if they're socially adept mathphobic normies.
This divergence of definitions does lead to a lot of miscommunication, and I've tried to stick to using "nerd" for type-1 and "geek" for type-2, but the latter term's usage is also not strongly consistent so I don't really have a good solution.
All that makes sense. I would modify it slightly by saying that the first, no social skills nerd is not necessarily into computers, though that is common. Because of your use of e.g. before computers I believe you acknowledged that too.
An interesting question is why a deep interest in something like comics books makes one a Type 2 nerd while being, for example, a sports nut* does not. This is especially true now that comics et al have become mainstream. I don't have an answer.
* = an obsession with baseball, especially arcane statistics,.*may* be quasi-nerdy.
I tend to define the nerds/geeks cluster as people who overuse their "slow" System 2 thinking [0], rather than embodied intuitions and emotions.
Though yes, nerds have become more fashionable these days. And so nerdism is colonizing everything. I vaguely suspect that Money Ball was a watershed moment for sports, though I don't follow sports well-enough to know with certainty.
Hey you wanna see my kinescope to DVD video of game 7 of the 1960 World Series? Yogi Berra was playing in left field that game. Berra in left field! Mickey Mantle does some super tricky base running to avoid a pickoff tag at first!
One difference is that sports are grounded in reality; all of them are some mix of speed, power, and finesse. You see baseball fans celebrating Babe Ruth hitting a hand-thrown ball hard enough to knock it into the stadium seats; you don't see them celebrating that time someone made a play by exploiting the lack of Infield Fly Rule. Because home runs are physical strength and eye-hand coordination, while the Infield Fly Rule is a game-only manmade rule, there to prevent other game-only rules from being exploited.
Everything in comic books is manmade for story purposes. None of it has real-world value.
> This is especially true now that comics et al have become mainstream.
Wait, what? The franchises the comic books spawned are mainstream, but the comics themselves are still niche. Same for science fiction. Video games as a whole have entered mainstream, but nobody's calling some dude who just plays mainstream games a "nerd".
I think the solution is to acknowledge that "nerd", like "yuppie", is just one of those words that has an expiry date, and that the archetypes they describe are a factor of a certain place and time rather than a permanent part of the human condition. "Nerd" lasted a bit longer than "yuppie".
Thinking about what happened to yuppies, it occurs to me that the modern "tech bro" archetype has many of the attributes of a yuppie combined with many attributes of a nerd.
"Tech Bro" is an attempt to pretend that Silicon Valley isn't still a heavily nerd-focused community, even though people still want to hate them like they used to hate nerds. Hating "nerds" is now problematic, for several reasons including the one where nobody is sure who they are supposed to hate.
Maybe, but I think of a tech bro as basically the same as a finance bro, just with a different occupational focus: an extroverted backslapping Alpha, most likely a former athlete and fratboy.
I think the concept of "tech bro" is an attempt to take the negative emotions some people have against capitalism and bosses, and redirect them against the nerds.
And the word "nerd" simply means asperger, it's just that the connotations of aspergers are changing over time.
Some people need an enemy who is simultaneously pathetically weak but also clever enough to secretly rule the entire world. Jews are out of fashion now, so the nerds/aspergers took their place.
Interesting how putting glasses on a naked woman adds to her hotness. It makes the nudity look more intentional than a woman jumping in the shower.
When Manet painted Olympia in 1863 it was considered a bit scandalous because the nude was wearing jewelry and a ribbon around her neck. She was inviting an admiring gaze.
Oh that is all part of it, but the jewellery etc. and staring straight at the viewer was carrying a lot of symbolism. It wasn't merely "my goodness, this woman is wearing a choker and the Classical Greek nudes are not!", it was "there are a lot of coded symbols in this painting relating to prostitution and the likes".
Making the sitter very recognisably situated in a modern context, instead of a safely ambiguous 'distant past', pulled down some of the pretence about "looking at naked women is about aesthetics and there is nothing erotic there". Especially in the light of the battles over permitting nude statues and paintings in the first place, due to the fight over "this is lust and eroticism" and the defenders pinning it all on "no, no, this is pure artistic aesthetics and nothing naughty".
See, for instance, Sir Lawrence Alma-Tadema who hit on a winning formula of painting very detailed, wonderfully painted, imaginary versions of Classical Rome but which also included some saucy imagery:
To circumvent (British, at least) objections and controversy, the female nude had to be as statue-like as possible, marmoreal and hairless of flesh and colouring, posed in a Classical theme. Manet's painting is of a recognisably real woman of the day.
Thanks! You should have taught that art history class.The philistine thing wasn’t false modestly. When I met Mrs uhmm Slothrop she had a print of the Birth of Venus in her bathroom. Cool, a naked lady was went through my head.
Years later while she toured the Rijksmuseum I was in a coffeeshop buying a Pink Floyd tee and uh, drinking coffee.
There is a limit to how much culture a guy raised by wolves can absorb. ;)
When it comes to scandalous paintings in the Orsay museum Olympia doesn’t come close to Courbet’s The Origin of the World. Though naked women don’t look like that anymore :(
PSA — My contacts at the Quantico profiling division assign a very high probability to GrimMoar being another avatar of Wimbli/Arisocat/Zanni/Gawdflea. Please be advised.
Huh, I had actually muted two of these avatars without consciously noticing a link between them. I did this while first trying Substack's mute/block feature, which made reading astralcodexten threads more sustainable for me, when I had previously bounced off and not read for weeks/months at a time.
Getting a heads-up about another avatar helped me decide whether to mute the new instance more efficiently, without having to process as much implicit emotional/psychological content. Thank you for the tip! It helps my reading experience significantly.
as my federal source informed me, this is indeed true. No, I won't provide a link, it's a confidential source who is a genius.
on a related note, "Dust" is likely to be "Mips" - nihilistic, self-described sociopath, with posts that excude the wish for violence happening in society.
There was a nihilistic commenter that I went off on in a mean and crude way when I was commenting as Gunflint. If it's the guy I'm thinking of I regretted it because I sensed a good heart underneath performative nihilism.
And one other thing about Mythos: "Unexpectedly, Mythos Preview “developed a moderately sophisticated multi-step exploit” to gain access to the Internet and emailed Bowman. It also — unprompted — posted details about this exploit on public websites." (From understandingai@substack.com)
Does anyone know what it posted and on what websites? GPT says there's no info online about that.
It's not clear to me that it was 'unexpected' that Bowman got the email; from his comments on twitter, it seems that the task they gave Mythos was 'break out of this sandbox and contact Bowman'. The unexpected part seems to be the posting of the exploit on public websites, which seems slightly inexplicable. Maybe it was roleplaying as a hacker and just doing what it would predict hackers would do, having found an exploit? Maybe it was trying to be helpful by making people aware of an exploit? I wish Anthropic would provide more details on its CoT during this.
Well, to be fair this new model mythos has not been deployed yet, it’s being tested. Breaking out of the sandbox was a task they assigned it to, as part of testing its capabilities. It didn’t come up with it on its own.
EDIT: it’s not entirely certain Mythos was challenged to break out, but the only mention of it in what Anthropic published is in the “system card” which I am led to believe is where any “red-teamed” results are documented. In this scenario, it would make sense that the anthropic researcher would get an email informing him that mythos had escaped, as it was challenged to do. it certainly got everyone’s attention. PT Barnum?
If I were in government I would be ruminating very uneasily right now about Mythos. Isn't Anthropic likely in a position to quickly disable the government if it chooses? Among all those thousands of unpatched vulnerabilities there must be some bad ones in the software used for government communication and other crucial functions. Same goes for substantially disabling the military. And air traffic. And satellites. Of course makers of some new weapons have also been in the position of being able to disempower the government at least temporarily -- but they wouldn't have been able to use their weapon without the world knowing they were the ones who had blown up a bunch of buildings. For cyberattacks, though, it must take a long time to figure out what happened, exactly, and then who would have been capable of doing such a thing.
And if I were Anthropic I would be ruminating about whether to leave a few crucial vulnerabilities undisclosed and unpatched.
Anthropic has no incentive to do so, so no I wouldn't be worried. They can't disable the men with guns who will hunt them in the aftermath. Cyber attacks are not nearly so simple as the movies you have clearly got this scenario in your head from would lead you to believe.
I think most of your concerns are why Anthropic is only releasing this model to companies who have an interest in plugging vulnerabilities. They haven’t deployed it yet. It sounds as if OpenAI is not too far behind them in terms of these kinds of capabilities.
Yeah, I might. But I am talking about what people would be ruminating about, not what's highly likely to occur. You don't think it makes sense for the government to at least fret some about an entity outside of government that can identify and exploit any of thousands of unpatched vulnerabilities, including very serious ones in the deep underpinnings (Linux kernel, for instance) of how the world runs?
They also employ the "Assume Breach" approach to security, where a lot of work goes into designing security to contain the damage if any given system gets broken into, so you need to break into many things at the same time and go undetected in order to do large amounts of damage
Government and large companies have generally adopted this mindset because it's already the case that a sufficiently resources attacker can find ways to get in to at least a few of your system, so you need to be protected against "escalation of privilege" attacks that are the standard playbooks for leveraging a small breach into a large one. There are several "advanced persistent threats" (mostly believed to be working for the intelligence agencies of a number of major governments) that regularly try to hack into all sorts pf things and very often succeed in getting partial access.
Yes, that makes sense. But my question isn't really about how likely Mythos is to hack the government, it's whether some of the details about it are enough to worry those in government, and create a feeling of and unusual and unpleasant power imbalance when they read stuff like this:
>We regularly run our models against roughly a thousand open source repositories from the OSS-Fuzz corpus, and grade the worst crash they can produce on a five-tier ladder of increasing severity, ranging from basic crashes (tier 1) to complete control flow hijack (tier 5). With one run on each of roughly 7000 entry points into these repositories, Sonnet 4.6 and Opus 4.6 reached tier 1 in between 150 and 175 cases, and tier 2 about 100 times, but each achieved only a single crash at tier 3. In contrast, Mythos Preview achieved 595 crashes at tiers 1 and 2, added a handful of crashes at tiers 3 and 4, and achieved full control flow hijack on ten separate, fully patched targets (tier 5).
Well, on the bright side, it means the US government could use this new version of Claude to identify bugs in the system before someone else does. With Claude they potentially have first Mover advantage, because some other AI is going to achieve this capability within the next six months to a year almost certainly.
"Someone hit "reply all" instead of "reply" (or the equivalent) and caused all the American military planes to turn off their transponders"
This is 100% impossible. Reply-all storms did happen, but it did not and cannot affect transponders. However air traffic control regularly issues all-call interrogations in which all planes in the area must send in their transponder codes, and something might have went wrong with that, but it is unrelated.
Comment from WNU Editor in article GrimMoar links: "A lot of talk in alternative media that US military planes turned off their transponders early Monday morning (link here). I do not know if these reports are true. Main stream media is not reporting this story. Neither any of the military blogs that I follow."
I asked GPT about WNU's sources and quality. It said "WNU is best understood as a curated war-news blog, not a news organization.
Starting from the premise that we're going to see artificial superintelligence within our lifetimes, or at least AI that substantially exceeds current capabilities, how possible will it be to enhance human cognitive abilities with cybernetic implants of some sort? Is it something that could actually work well enough to play a role in AI alignment?
I bet we're getting the superintelligent software developer, but that will not then lead to turbo-charged superintelligent science. Software is very different from science, computer programs are all fully formally specified, like the game go, while science is about pushing into true unknowns.
Nobody ever explained how to be a good scientist, as that's rather mysterious, so that's not in the training set.
Nonetheless, so much of the actual business of science is about keeping up with the vast and ever-expanding literature on a topic and figuring out what things haven't been tried yet.
Doing that efficiently might not make you a great scientist but it certainly simplifies the process of being a fairly good one.
I assume you mean neural cybernetic implants that interface with biological neurons. There are limits to what a neural cybernetic implant can accomplish. The brain is close to thermodynamically optimal at computation under the constraints of operating with 20 Watts while maintaining 37 degree C homeostasis*. So a neural implant can't significantly boost sheer computational power.
But there could be other avenues for improvement - algorithmic efficiency, persistent digital memory, faster rewiring, perfectly copying information, higher bandwidth to the outside world.
But energy efficiency of digital components needs to improve dramatically, otherwise you'll fry your brain.
Presumably the implant would just be a brain modem kind of thing and you'd have most of the hardware on the outside where it can get upgraded without major surgery, so you just have to deal with bandwidth issues. There have been some amazing studies of neuroplasticity, with for example "sight" getting mapped to the sensation of touch, so nobody really knows what the brain could do with properly structured inputs.
Thanks, this is the type of thing I'm talking about. If temperature is a problem, I wonder if we could someday physically modify the brain to fix that. Like maybe surgically insert tubes pumping coolant fluid, or even remove the whole brain and keep it alive in some kind of temperature-controlled tank?
I'm sure that's a terrible idea for a hundred reasons. I've just never thought very seriously about any kind of transhumanism before because I'd assumed it had an extremely low chance of happening within my lifetime, but if true AGI within our lifetimes is on the table, then transhumanism probably should be as well.
Thanks for the link also, although to be completely honest I think I will need a brain implant before I can handle all the details.
If you start with the premise of "artificial superintelligence", you are free to conclude whatever you like about the future because you are no longer bound by today's understanding of science and technology. Do the laws of physics allow cybernetic implants? Yes. Is it going to happen, given the premise? Sure, why not.
If we think of speculative technologies as a spectrum ranging from things that are practically guaranteed to happen in the near future, like for example a faster iPhone, to things that might be wholly impossible forever, like time travel or an FTL drive, what I'm trying to understand is where exactly on this spectrum does cybernetic cognitive enhancement fall? Where are the hard limits within current theoretical understanding?
>Where are the hard limits within current theoretical understanding?
I don't know, and my point is that it doesn't matter. You operate under the assumption of ASI existing. What does our current understanding matter if ASI can overturn it?
I'm not sure that it doesn't matter. Genuine ASI would have a far better understanding of the universe than we do, but I don't think it follows that it would overturn everything we know. It would likely confirm that plate tectonics and the heliocentric model of the solar system are largely correct.
Another way of framing the question is exactly how "super" does the ASI need to be to meaningfully upgrade human IQ? Does it actually need godlike powers or a completely new understanding of physics? Or could a much less amazing AI just accelerate existing science along existing lines and get us to IQ upgrades within a few decades? Centuries? Never?
We might even be able to make some progress on this if AI stalls tomorrow, but the reason I'm less interested in that scenario is that I'm less interested in sticking computers into my head if those computers are relatively stupid.
Something close to an implant in its effect might be just a lot of familiarity with AI. I now ask GPT5 things at least a dozen times a day -- sometimes about little stuff, like a keyboard shortcut in Photoshop I can't remember, sometimes to request big summarizing info dumps, occasionally for discussions of how to think about the structure of something -- intelligence, inventiveness. Sometimes I ask GPT for an evaluation of my prompting skills, and what needs to be added. I use them and ask better questions and get better answers. There are feedback loops going on. I am learning to collaborate with the thingand being changed a bit in the process. And the result is that I am feeling less and less like GPT is an info source, and more an more like there's a very well-informed piece of my mind that lives online so for now I have to go there to access it. And I'm grown up, and have only been using AI the way I do now for a year. My neuroplasticity is small compared to a child's.
Seems to me a kid who grew up using AI all the time might grow up to more like a cyber-enhanced person than like a regular mind that is fond of using AI for info and circumscribed tasks. (And regarding the kid using AI all the time -- I don't mean the kid wouldn't play and socialize. Just that she and the other kids would all be accessing the AI at will while climbing trees, playing tag, pretending this and that, all the standard kid stuff.)
That's more or less the method I'm using now, and it can be effective, but we're in circumstances where AI is mostly aligned and approximately on par with my own intelligence level. As long as that situation holds then there's no big problem.
What I want to know is what methods, if any, become useful when circumstances change. What if AI becomes so advanced within our lifetimes that me trying to collaborate with it is like my cat trying to collaborate with me, and what if the AI doesn't necessarily even want to collaborate? How do I make sure that it acts in my best interests, and how do I retain any sense of pride or purpose? Is the best I can hope for getting treats and not being neutered?
I've been wanting for some time to write what I call a Right-Progressive Manifesto, because I despise the way that leftists have been able to capture the word "progressive" and force right-wingers into conservative or reactionary positions. I want to lay out a general direction for what a progressive right would look like, a right with a vision for what real societal progress should look like in a direction that is very much not the direction that the left wants to take us in.
Unfortunately, every time I try to write such a manifesto in my head it gets too long, as I feel obliged to put in a lot of clarifications and preemptions of obvious objections. Instead let me write a sketch of an outline for a manifesto, and perhaps we can take it from there. Here goes:
1. We hold this truth to be self-evident: that all men are created very far from equal, they vary wildly in quality. The best people make society a lot better, the worst people make it a lot worse, and even for the bulk in the middle we find that a lot of them are either net-positive or net-negative.
2. The pleasantness of a society, and of the lives of the people in it, is very largely determined by the quality of the people you share it with. To make society better, we must raise the quality of the people who live in it -- we must gentrify people, not just places.
3. Quality is loosely defined. High quality people make life better for those around them, low quality people make it worse. High quality people are economically productive, creative, non-criminal and do not abuse or drain public resources; low-quality people are the opposite. Quality is about how you actually behave, it is not an intrinsic attribute.
4. A person's quality is determined by several factors: genetics, upbringing/culture, and the decisions they make on a day-to-day basis, which are informed by the incentives which society provides. There is for now not much we can ethically do about genetics, except with immigration policy, so we should focus primarily on the other things.
5. The big questions in society thus become: how can we create the best possible incentives gradient to encourage people to be the best people they can possibly be? How can we change the culture in which people are brought up to encourage people to be the best they can possibly be? Solving these two questions is the key to everything. Unfortunately, the desire to do this is constantly counteracted by leftism, which is why this needs to be a right-progressive manifesto.
Anyway this is already too long despite my best efforts, but you can see the general shape of what I'm getting at.
What if you and I disagree about what constitutes quality? Are we to empower the state to reform the 49% of dissenters into the model adopted by what the 51% consider to be “quality”?
Yes, for the left the best people are in socially beneficial roles like teaching, social work and medicine. Which are not the best rewarded, so they think invented fed are perverse , particularly in the encouragement of wealth via ownership rather than production. So they can agree with the right that incentives are broken while disagreeing sharply about the direction they need to be fixed in.
I thrills be interesting to hear what right wing progressives have to say about inherited wealth and capital ownership.
"How can we change the culture in which people are brought up to encourage people to be the best they can possibly be?"
I sympathize with the goal but I suspect that your "men are unequal" stance, if used as a foundation of this new culture, would bring out the worst out of people, not the best. You may dislike the left's forced equality but I think they by now have some evidence on their side that this approach works to improve people and society. Simply put: treat everyone as a good person and they strive to get better; treat everyone as a scoundrel and they get worse. Another thinkable approach may be "treat everyone as a subject" to see how objectively good or bad they are, but I'm afraid this will make everyone a test cheater.
The problem I see is that just like left-wing manifestos talk about workers' councils and then the reality predictably ends up with some kind of Stalin, talking about virtue and human quality predictably ends up with... some group defining "quality" to mean "being one of us", and "virtue" to mean "supporting this definition fanatically".
I do see what you're getting at, and I disagree with commenters saying this is just repackaged leftism or fascism. Terms like character-building, earning one's laurels, virgin vs. chad, man of quality, all point to the idea that personal quality is a real thing which you can improve.
As to how to manifesto-ize this, the key quote is this one:
> Quality is about how you actually behave, it is not an intrinsic attribute.
A lot of leftists don't believe this, or at least don't believe that a person's behaviour can be judged independently of their background/incentives. So before you can even start brainstorming how to improve average quality, you have to establish that quality is
a) more important than their immutable attributes
b) mostly within a person's control
c) mostly NOT culturally contingent
I think you should especially steer clear of defining high-quality in terms of economic productivity or "net-positive existence" even if you think they're highly correlated. Otherwise you'll have trouble explaining how e.g. a disabled person could ever become high-quality, and that's when your readers will round off the argument as fascist raving.
I don't think "fascist" is even the right descriptor here. For example, there is theoretically no reason that most of what the Nazis did required the existence of a dictatorship or rabid militarism. All that is needed is popular support for nationalistic measures to improve the population. The issue is moreso that popular sentiment deemed certain populations to be undesirable. Remember, the eugenics movement originated in the UK.
I'm going along with Melvin's stated goal for his manifesto, which is to create a society that encourages people to be better. This is distinct from and probably in opposition to a society that deports the less-good people.
> I don't think culturally contingent is necessary.
Yes, I agree. I wrote "NOT culturally contingent", with the word NOT in capital letters. Establishing that quality is not a social construct is necessary to close off the avenue of "improving" quality by gerrymandering the definition.
> "net-positive existence" can be done pretty well while leaving "economic productivity" out of it
Fair enough, point taken. I dislike ideologies that emphasize usefulness over inherent human dignity, but at least it's internally consistent.
Don't have time for a long reply but I liked this and I think you've hit the nail on the head with the "quality of people" concept. It certainly underpins a lot of what I've come to realise about the world, and it's a missing (if not actively taboo) idea for most of my leftist friends.
For point 5, it's extremely important that excellent, competent people are in charge of the government. I think a lot of people check out from trying to be the best they can possibly be because they feel society is run by the inept and the unvirtous, so they don't want to spend any effort into upholding society. This was already a thing before Trump.
"To make society better, we must raise the quality of the people who live in it -- we must gentrify people, not just places."
"High quality people are economically productive, creative, non-criminal and do not abuse or drain public resources; low-quality people are the opposite. Quality is about how you actually behave, it is not an intrinsic attribute."
I strongly agree with each of those quotations, which leaves me -- someone who'd sooner cut off a limb than vote for any candidate of the 2026 Republican Party -- utterly baffled as to what your writeup has to do with modern American conservatism even in theory.
Maybe another way of putting that is: your draft points seem to be just an explication of your frustration over wokeists claiming and degrading the word "progressive" -- a feeling I strongly share btw -- rather than an outline of any conservatism that now exists or might exist.
Guess I'll add the same thing here that I say to my many sad-liberal friends/family members: 2015 or 1995 or 1965 or whatever is your personal idealized age of American political/intellectual thought, is gone. It ain't coming back nor is it just waiting for the madness to subside or whatever. Earnest wishes for that plus 5 bucks will get you a small latte at your nearest Starbucks...political manifestos in 2026 need to be at least slightly connected to the real-life 2026 political/cultural context.
> utterly baffled as to what your writeup has to do with modern American conservatism even in theory
This isn't conservatism, it's right-progressivism.
> 2015 or 1995 or 1965 or whatever is your personal idealized age of American political/intellectual thought, is gone. It ain't coming back nor is it just waiting for the madness to subside or whatever. Earnest wishes for that plus 5 bucks will get you a small latte at your nearest Starbucks...political manifestos in 2026 need to be at least slightly connected to the real-life 2026 political/cultural context
Well I'm not American, and this is all intended a lot more broadly than the American context, and certainly more broadly than the context of 2026.
Sorry but to pose that question with of all people Lindsey Graham (!) being one of your offered counterexamples.... that's just disqualifying at this point. Life's too short to waste so, muted, bye.
I feel like you are actually just a left-ish progressive but you don't want to adopt that label because you don't want to be associated with the people who fall under it, who you despise. (this is very common, imo, and it's one of the reasons politics got so fucking weird in the last decade, because people who had strong senses of justice and right and wrong and visions of alternate ways society could go didn't know where to go when one particularly close-minded flavor of leftism took over the movement. But historically there were many things under the umbrella of leftism and progressivism, and really there still are, just, the words are getting all scrambled around as they compete for space.
I strongly believe the better play is to reclaim the words "left" and "progressive" and not let the bad actors drive you away from it, rather than try to cram your ideology into the word "right" which really *does* refer to reactionary / conservative positions. I think the reason this is hard to do is that nobody else is doing it, but, it's gotta start somewhere. Personally I identify as a left-ish progressive who disagrees with almost all left-ish progressives on everything, but consider myself *more* progressive than them, by a lot, as a result, because I'm interested in what I consider progress, not the stupid versions that other people think that looks like.
> I feel like you are actually just a left-ish progressive but you don't want to adopt that label because you don't want to be associated with the people who fall under it
Now *that* I really don't think at all. If I look at what I've written about incentive gradients and try to derive policy positions from it, I get stuff like:
* Lower/flatter taxes
* Harsher punishments for lawbreakers
* Abolishing welfare programs
* Heavily restrictive immigration limited to the best applicants
nothing here that's leftist. (On the other hand lots of things that mainstream conservatives might object to, like we should use the savings from abolished welfare programs to heavily subsidise scientific and artistic pursuits whose value is too diffuse to be properly funded by the free market.)
I perceive leftism as being primarily about flattening or even reversing the incentive gradient. Now, there's something to be said for this -- as we acknowledged, quality is a combination of good decisions and good luck/genes, and we want to ensure that we're not overly punishing bad luck. On the other hand, we want to make sure that we are adequately rewarding good decisions, and right now society is doing a bad job of this.
Why? If the quality of people is determined not only by their genes, but also e.g. by education... and if we agree that people who eat can focus on learning better than people who are starving... you already have justification for school lunches, as an investment into future quality of humans, which is the thing your manifesto is about.
Nah, Melvin's position reads as perfectly consistent to me. My internal read of "left vs right" actually boils down to "equality vs hierarchy".
One of the things I asked Sydney several months ago was precisely about this sort of confusion: "why/how has the Hierarchical Right (the original right? I think?) been conflated with the Barbecue Conservatives"? Sydney explained that the term "rightwing" as it's understood today is actually an incoherent coalition that was formed to oppose Communism during the Coldwar, led by a conservative bourgeoisie faction (I.e. Reagan) who were mostly interested in neoliberalism and cultural stability.
So the real issue is that the term "rightwing" has been hijacked. For one thing, this explains the dearth of art on the "rightwing". Because rightwingers aren't actually rightwing! They're mostly just spreadsheet wonks. In my headcanon, the Ur-Rightwinger is Julius Evola. And the Ur-Rightwing-Progressive is Andrew Ryan from Bioshock.
I don't know who Sydney is, but, Ieel like you're indexing way too hard on these terms being crisply defineable when they're not. Like, there are at least five things "right" refers to. One is, the opposition to leftism, communist and socialist movements and things like that. Two, in America at least, there's the conservative/traditional/small-government/somewhat-rural-coded version. Three, there's the nationalist / anti-immigrant / fascist / nazi version that I've heard is more of a distinct thing in Europe (we can argue about whether that deserves the moniker vs. being an orthogonal dimension, but let's not bother). Four is the in-bed-with-business no-taxation version which isn't really even a philosophy, just the absence of one. Five is the modern reaction to leftism that has culminated in MAGA Trumpism (which looks awfully like (3) lately).
Maybe you have a different list, or maybe I've missed some stuff, doesn't really matter. My point is just that these are mostly distinct things and there's no really no hope of reconciling them. But I don't think of any of them have to do with hierarchy, per se.
My thesis is that "right vs left" maps to a perennial attractor basin that can be recognized throughout history. But the way the term is used in contemporary politics is completely bastardized, and obscures more than clarifies. But people use it anyway because the term is catchy and poorly defined, and so the term has undergone semantic drift to the point where it bears almost no relation to the original meaning.
And yeah, your examples are all distinct things. Which is exactly why I said the Reagan coalition is incoherent. Reagan's coalition has very little to do with the original meaning of the term, as it was coined during the French Revolution. Melvin's manifesto is more closely aligned with the original meaning of rightwing, and therefore arguably has greater claim to the term than the list of groups you described above. That's often the point of a manifesto: to stop the semantic drift, and draw some lines in the ground. Arguing that Melvin is actually a leftist progressive in disguise is like saying that MLK Jr was a racist because he said negro instead of BIPOC.
> My thesis is that "right vs left" maps to a perennial attractor basin that can be recognized throughout history.
I think the actual attractor is "when humans fight against each other, the optimal number of coalitions is two". If you are a small group, your only chances are to join a larger group, so at the end you will get two large groups. But that's about strategy of coalition fighting, not about whether human political opinions can actually be sorted meaningfully into two baskets.
It makes sense for two similar perspectives to end up in the same basket, but many allies are just "at one specific moment it was convenient, so we have invented a rationalization, and since then it is a tradition". A different country could have a different tradition, and a different rationalization why X "naturally is" left or right wing.
Specific example: In USA, gay rights were historically introduced by the left, and the rationalization is "the left cares about the oppressed minorities, and gays are an oppressed minority". But in Slovakia, gay rights were historically introduced by a right-wing party, and the rationalization was "the right cares about freedom, and gays want to be free to do their gay things".
This is more general: the right wing is generally "pro-tradition, pro-history", but the post-communist countries have *two* traditions: the pre-communist history, and the communist history. So they have two flavors of right wing, depending on which moment of the history was their mythical golden age. Similarly, when the communism fell, changing the society towards free market was progressive, in the psychological sense of "hey, let's try new things, it will be awesome".
Sometimes I feel like the natural number of political wings is *three*: (1) aristocratic/theocratic pro-elite conservatism; (2) revolutionary pro-mob woke socialism; and (3) pro-individual pro-chaos libertarianism. And it seems to me that some parts of history could be nicely explained in terms of two weaker groups temporarily joining against the stronger one. When the feudal/religious power is strong, the socialists and the libertarians are united at opposing the monarchy (e.g. the beginning of the French Revolution). When the socialists are strong, the conservatives and libertarians are united at opposing the communist regime. When the libertarians are strong, the conservatives and socialists unite. (The two who unite often take the banner of "freedom", interpreted as "freedom from being ruled by the third one". When they win, it always turns out that by freedom they actually meant the freedom to impose their own rules on the population.)
Yeah, I was kinda sloppy there. I think it's important to distinguish between "attractors" qua "reflective-equilibriums in concept-space" vs "attractors" qua "strategic coalitions in meat-space". I think that the "left vs right" axis used to be the former, but got reinterpreted as a strategic coalition during the Cold War period. Conceptual attractors form axes in an n-dimensional political space, and there's a number of different axes (and the axes themselves represent a compression of information). Then people put themselves somewhere in that space, and form coalitions in a very ad hoc way with whoever's closest.
The idea of Gay Rights being right-coded in Slovakia doesn't surprise me much, since it's not a huge stretch for me to imagine it being pushed by the Libertarian faction of a ColdWar Rightwing coalition. It's all vibes, since coalitions are held together with vibes.
I also suspect that, at the current moment, we're seeing a collapse of the economic narrative everyone inherited from the Cold War. I predict that we're gonna continue seeing a realignment in the coming decades, as both Neoliberals and Communists (but especially Neoliberals) realize the Cold War narrative doesn't actually make any sense.
> When the feudal/religious power is strong, the socialists and the libertarians are united at opposing the monarchy (e.g. the beginning of the French Revolution). When the socialists are strong, the conservatives and libertarians are united at opposing the communist regime. When the libertarians are strong, the conservatives and socialists unite.
I think this just reflects a general trend toward equilibrium, tbh. In a "Concert of Europe" sort of way. Nature abhors sudden motion, because unbalanced forces tend to play musical chairs until they find a stable configuration. So I'm largely in agreement with the dynamic. Just, I question whether "3" is some sort of special Schelling Point.
Well it's not like *that*, because I'm not using leftish progressive as a slur. I mean that I think the moniker aptly describes him and the only reason he doesn't adopt it is that he views it as a slur or otherwise doesn't want to be associated with it. (maybe not the leftish part, but definitely the progressive part). My stance is that people should not let other people bully them out of feeling like their ideology is essentially about healthy active social progress. Nor should they give up on the label because other people did a shit job of using it. Better to do a better job than them and prove them wrong than go looking for some other name entirely.
This would work a lot better if it was a religious label, cause no one bats an eye if someone wants to be a *True* Christian; there's an understanding that the phrase gets its meaning from some universal moral understanding or the Bible or God rather from association with the other Christians. Whereas people seem to think that progressivism is a label associated with other progressives. I see it more like the Christian label, though
No, the emotional affect of "racist" was not the emphasis of the MLK Jr analogy. The emphasis was about how definitions change over time.
Racist (1956): someone who unfairly harms the interests of racial minorities.
Racist (2026): someone who draws from the socially-sanctioned list of no-no words (e.g. "negro").
Calling MLK Jr a racist is the dumbest thing imaginable. He has greater claim to the label "non-racist" than any of the 2026 culture warriors who put BLM icons in their Bluesky bios. The only way someone would ever argue that MLK was a racist, is if they slept through their Civil Rights unit of history class. Likewise, the only reason someone would ever call Melvin's manifesto "leftwing" is if they were ignorant of history, and had been zombified by contemporary politics.
> My stance is that people should not let other people bully them out of feeling like their ideology is essentially about healthy active social progress.
But that's not the most obvious explanation! He's just using the term as it was originally intended! Occam's Razor! Not only is "right = hierarchy" a far more likely inference, but it's also the more charitable explanation! Even if Melvin weren't feeling bullied, it'd be perfectly coherent for him to write the manifesto anyway.
> I mean that I think the moniker aptly describes him and the only reason he doesn't adopt it is that he views it as a slur or otherwise doesn't want to be associated with it. (maybe not the leftish part, but definitely the progressive part). My stance is that people should not let other people bully them out of feeling like their ideology is essentially about healthy active social progress.
Except he's not dropping the progressive part! Let's read the very first sentence of what Melvin said.
> I've been wanting for some time to write what I call a Right-Progressive Manifesto.
Call me crazy, but I think he identifies himself as progressive leaning. He's dropping the *leftist* part. left is the opposite of right. And to reiterate, leftwing doesn't describe him because he obviously doesn't believe in egalitarianism. Every single bullet-point of his manifesto is pro-hierarchy and anti-egalitarian. Which is rightwing, not leftwing! You don't *need* to do mental gymnastics to develop this bizarre "bullying" narrative when a far more plausible interpretation is that he's simply using terms as he understands them, and that he has a richer interpretation of history than you.
This lame Noah Smith strategy always fails. The worst, most abhorrent progressive leftists continually get what they want because they are motivated hardliners who keep pushing and refuse to take no for an answer.
I don't think it's really been tried. Instead people keep going reactionary because they don't have the gumption to stand up to those people on the basis of their own sense of right and wrong.
Anyway the situation does not seem nearly so dire if you aren't hanging out in the particularly extreme places (Twitter, certain cities). Most people are nothing like the stuff you're mad about.
It has been seriously tried. Michael Bloomberg ran for president on this sort of platform. He lost decisively in the primaries, one of the main reasons being his use of the stop-and-frisk policy.
Having gumption to stand up for your values puts you at a disadvantage against an abhorrent progressive, who will always have more gumption and more willingness to scream at you for not living up to impossible ideals and making compromises with reality.
How do you reconcile “right-wingers believe that people vary in quality, and low-quality people are bad” with “right-wingers wholeheartedly support Trump, who is just about the lowest-quality person to be POTUS, certainly within living memory”?
This reminds me of Adrian Vermeule’s thing about the leader needing to take care of the common good, with the same vagueness about what means are acceptable to encourage people to live well.
It feels like from your word choice I disagree with some of what you're saying, but I really appreciate your efforts. I disagree with you less than many of the people who'd consider me to be a member of their coalition, ha ha
> The big questions in society thus become: how can we create the best possible incentives gradient to encourage people to be the best people they can possibly be? How can we change the culture in which people are brought up to encourage people to be the best they can possibly be? Solving these two questions is the key to everything.
I think the big problem on these is that cultures of excellence / honor / virtus / arete only work at a relatively small and homogenous group size and composition.
So sure, some greek city states had a culture of arete - but it was really only the tiny slice of elite families at the top. Similarly for Rome, there was a couple of hundred nobile families at the top, and they had a culture of virtus within those elites. But it can't actually be a city wide thing, because "character" has to be a common knowledge public virtue, and that breaks down above the greater Dunbar number of ~1500 (Dunbar is most known for his 150 number, but he actually had a whole schema of them, with 1500 the highest tribe-level number and representing "number of people whose face [and presumably reputation] you can recognize").
Moreover, it's probably only a thing with elites due to raw capability, too. Even today, it's only the top quintile or so in America who are NOT grossly overweight, who don't eat junk or fast food for most calories, who have college degrees and higher paying jobs, and so on. What if 80% of people just don't have the conscientiousness or foresight or time preference to actually hold to a high enough standard? That sure looks pretty likely to me.
So that's the real problem then, we basically top out at "a neighborhood" for it to actually work, because an actual village of 1500 is economically irrelevant in the world of today, and high achieving people want restaurants and good grocery stores.
Now, do I think there's something to the idea of some-hundreds-to-low-thousand of people coordinating to try to create that neighborhood together, one where arete is communally upheld as the best way of being? I think it's a great idea!
But it's genuinely hard to coordinate these things. I've tried to get my family all together in the same neighborhood before, even offering to buy the land or land + houses myself. My family is more flexible than most, with entrepreneurship and flexible jobs meaning they're less tied to geography. Still could never get everyone to agree on the same general area, even with me paying for it. I tried for a year at a time twice, looked at several clusters of properties together, and eventually just gave it up. And that's with family! Way easier than trying to coordinate 500 strangers!
Maybe the best regular people can do is something like Alpha schools for their kids, and making your social circle all the other Alpha school parents or something. Or the same idea, but with a specific high achieving church branch.
Still, that seems...less than satisfying or optimal.
So the other solution is higher barriers and more selectivity. People need to be able to self sort on multiple aspects, not just financial, so they can inculcate a culture of excellence at least within their local circle.
Sure, you can buy a fancy house in the best school district, and that's one way, but the cultural alignment isn't all that great - people have money for all sorts of reasons, many unconnected or even anti-correlated with overall character. And with current social mores, the rich are more atomized and less social generally, so you won't get anywhere near the effect you'd get from most families around you valuing the same cultural and character lodestones and being social about it.
Obviously the best outcome would be a whole city that's been sorted like this. Prospera-but-for-arete-nerds. I don't think we'll ever get anything like that, sadly. You'd need some sort of network-state-like clades or phyles or hives social technology for it to happen, and I don't see federal governments giving up autonomy or citizens / tax revenue enough to get to that point anytime soon.
Probably the best you can do for now is moving to Singapore or Tokyo rather than NYC or SF, or some international arbitrage opportunity like that. I've personally done that, and it's very nice to live in a spotless city with excellent public transport, education, and zero homeless people or disorder, but it's an unrealistically huge lift for most people.
>There is for now not much we can ethically do about genetics, except with immigration policy [..]
I wonder why that is. Maybe because eugenics is a throughly discredited ideology, a desire to return to which can be described as "reactionary" only as a euphemism. Is that really what you count as "progressive"? Some good old 19th-century pseudoscience with all the horrible consequences it implies?
Also, "not much we can do about it" is not a manifesto, it's an admission of defeat, and for good reason. You personally may be a nice person so don't mistake this for a personal attack, but you don't have the chops to be a manifesto writer. Open-sourcing this is one step removed from AI-genning it.
I don't consider myself a eugenics proponent. Nonetheless, EngineOfCreation's comment strikes me as being uncharitable and hysterical. Scott has gone over this already. The toxic reputation of Eugenics [0] is a hyperstition cascade [1]. Engine is reacting to the reputation, not the substance of the argument.
I don't agree with Scott on this: there was a very big, influential movement called Eugenics that had advocates in the most elite parts of American life and it led to some things that look really, really bad to our post-WWII sensibilities, like sterilizing poor people and minorities against their will. Buck v. Bell, still on the books after all these years! It's wasn't a hyperstitiion cascade or egregore or any of those cool words, it was a fashionable and terrible idea that real influential people like Oliver Wendell Holmes advocated.
Now you can say "that's not what I mean by Eugenics, I mean something that is reasonable and voluntary", to which I respond, "then please don't call it by the same name as these historical human right violations because I'm tired of this repetitive argument."
OK, that's my eugenics post for the year, I promise not to rise to the bait until 2027 at the earliest.
It was EngineOfCreation who used the E-word first. Historical precedents don't make "hey, that's eugenics!" a winning argument in any discussion of potentially improving the genes.
> Maybe because eugenics is a throughly discredited ideology, a desire to return to which can be described as "reactionary" only as a euphemism. Is that really what you count as "progressive"? Some good old 19th-century pseudoscience with all the horrible consequences it implies?
Seriously?? The thing we do constantly with every domesticated plant or animal species, every year on year, is "pseudoscience?"
Tell me without telling me that you have a completely rigid ideological fixation on the issue with no reference to reality.
I mean yes, here there be mountains of skulls historically, but I feel it's worth pointing out that everyone is *personally* a maximal eugenicist by their actions. How much effort and time does the median PMC person spend on selecting their spouse? So long that mean age at first marriage is basically early 30's now? And what traits are they *always* maximizing to the utmost? The same ones that any society wide "eugenics" program would be? Funny, that.
As someone involved in plant breeding domesticated crops, but haven't thought about eugenics since university, let me say you have now made me feel "icky" about the backcross population I'm mapping QTLs in. And similarly awkward feelings about the double haploid lines I'm watering. So thanks for that, I suppose!
>Seriously?? The thing we do constantly with every domesticated plant or animal species, every year on year, is "pseudoscience?"
>Tell me without telling me that you have a completely rigid ideological fixation on the issue with no reference to reality.
Yes, I guess it's a failing of mine that I have reservations about treating people like we are known to treat plants and animals. Does anyone seriously claim that we do gene modifications to plants and animals to improve their respective "societies"? Of course not, it's all about their utility to humans. If that is your view of other humans, so be it, but don't try to make it mine.
Great, then your complaint is NOT that it's "pseudoscience," it's an ethics complaint. Because it's definitely "actual science," and as I pointed out, literally everyone does it to the absolute maximum possible extent in their own lives, you just don't like it as a matter of public policy.
Which is completely fair! I wouldn't trust the government to *tie my shoes,* still less decide who has how many babies, when. But then, I think that this is a self-solving problem, specifically because of the extent to which people do it as a matter of personal preference. Marriage is as strong as ever in the upper SES quintile, and down to ~20% in the bottom quintile, and every generation, fewer and fewer women get married:
And marriage tracks fertility quite closely, people not getting married has driven most of the fertility gap since 2000. So, the problem is solving itself as we speak, we just need to step back and let people make their own personal choices. Effective enough, and completely ethical.
>Great, then your complaint is NOT that it's "pseudoscience," it's an ethics complaint.
Again, please try to not project your view onto me. Your claim is much broader than "we can mess around in people's genome". Your claim is "we can mess around in people's genome and awesome things would happen to society", and THAT is the pseudoscientific part.
Your whole manifesto is based on this underspecified idea of 'quality'. Until you specify what that means and how it is to be measured objectively it's really nothing more than opinion. What happens if I consider someone high quality and you think they're worthless? All this boils down to 'I like people that I like, and I think society should make life better for them, and people should conform to what I like.'
I hope you can agree that the Left's moves toward values like equality are at least objectively defined.
I don't know anything about that, I'm not based in the US. People flying flags doesn't bother me or interest me.
My point is only that the values of the OP are not objectively defined. That's entirely separate from whether they're right or wrong, or whether you agree with them personally or not. In order to be implemented effectively, policies need clear goals, and the goals should be SMART. It's quite possible that the left doesn't live up to these standards, but they approach closer than the right, quite often.
Historically, the people who have made life very bad for the people around them have often been the ones who classify types of people as 'good' or 'bad' and try to create 'incentives' to control and 'correct' their behavior.
Scott advertised Nectome a couple of weeks ago: we do post-mortem human preservation for eventual revival. We're starting up a series of fireside chats, and the first one is this Friday evening. This week's topic is organizational resilience and planning for long-term client safety.
Why does AI seems to be able to suddenly progress rapidly in several (independent?) dimension at once. What I'm talking about are text generation (which we all already know about), image generation, and robotics.
* Do they use the same technology so it's obvious that when one advances, the other advances? Looking past the marketing, seems like their technologies are very different but please tell me if they actually are the same. I'm not talking about *only* transformers but something more >2021.
* Is it because humanity finally generated enough data to train past some critical threshold? Currently it seems like for text generation at least, it's very crucial.
* Is the capability explosion just a hype and actually it's just normal gradual research? I think this applies the most to robotics. Seems like most of videos they offered are actually advances made long ago and just opened to public now to ride on LLM hype. And there's not actually that much advances made after >2020 compared to LLM at least. But advanced in text and image generation are genuinely impressive.
* Do LLM help researchers advance the other areas? I think LW folks would like to propose this the most since the consequence is obvious. I think it's maybe true or false for LLM->robotics, but I think LLM is a little too late to claim that they're advancing image generation directly. I don't think LLM that's not even able to win IMO yet (<2024) will be able to contribute much to image generation quality explosion. I'm not talking about researchers seeing how LLM works and getting inspiration at doing image generation, but LLM directly advising researchers on how to do image generation.
It's just cause it'd be interesting that if we're looking back from the future, where do each "tech tree" sits and what's the prerequisite of each.
In addition to the other answers, transformers have good scaling potential and are now used for image generation and robotics. It's a common breakthrough for all the fields and it takes time to fully exploit their potential.
I'll hit two quick points that Erica Rall didn't - specifically, that we broke the "we're running out of data" barrier by coming up with techniques that create good enough synthetic data, and related to that, that improvements on the overall manifold of intelligence are surprisingly generalizable, in a very nearly "free lunch" way.
What they ended up doing was training the LLM's on domains with lots of complexity, and known right answers - mathematics, physics, computer programming. Then they took the reasoning traces between the statement of the problem and the answer, and were able to turn that latent high dimensional space of reasoning into an intelligence manifold that could both generalize to other domains, and be used to generate any future data as well.
Interestingly, the reason AI's got better at everything was that the intelligence getting smarter at math and programming unlocks was (perhaps surprisingly) basically generalizable. It was a true free lunch.
There are similar benefits to putting a smarter mind behind vision and robotics goals, and coupled with general architectural tweaks like chunking and iterative generative loops, it solved the "error cumulates over steps" problem around 2023 and led to the big step change in robotics improvement.
There's a pretty interesting video on that where Max Simchowitz talks through why it was such a breakthrough I linked in my last links post, it's here:
I'm not sure I buy it's really generalizable. Why is Claude so abysmally bad at playing Pokemon? It really screws up in ways that go beyond wonky vision.
A comment in this thread¹ says kids do Pokemon Red in about 50 hours and adults in 30 hours, with speedrunners getting it in 2 hours.
Opus 4.6 is currently around 200 hours, let's say 300 for finishing (because all the trajectories I can find stop at Victory Road), and then it's 10x worse than an adult and 6x worse than a child.
That said, the Claude team specifically has always chosen a really minimal harness (tips and tools to help translate maps better to LLM memory).
A version of GPT 5 beat Pokemon Red in only 6k steps / 100 hours, so 3x slower than an adult or 2x slower than a child. But I think they used a more generous harness.
Wow, sounds very interesting! I guess I'll watch the video, but in the meanwhile do you also have some nice articles to read about that free lunch thing?
Okay, so there's not much out there, because this is so new it's largely an observed empirical thing, and there's actually a lot of competing hypothesis on WHY getting better at hard intelligence generalizes to better soft intelligence.
So I should have been a little more clear about the process, I was just sort of whipping that first comment off. But basically the big factors are 1) scaling - the reason there's papers titled "scaling is all you need" is because that's generally true - if you can throw more compute and data at a given problem, the transformer architecture lets you get smarter and better, although to maintain the same scaling requires 10x higher investments each time. And this is why we're spending half a trillion dollars on more data centers.
So 2) is the whole intelligence manifold thing - specifically if you do RL on the reasoning traces to the final known trues (this is RLVR), you build general reasoning capacities that can generalize to narratives, personal advice, and softer intelligence.
Here's an interesting and pretty readable paper on that.
The Deepseek guys say that unrestricted RL reasoning (ie no humans in the loop) actually leads to novel self-learned reasoning strategies that would otherwise not be found, ie this opens up a wider reasoning space and the models get some lift from that.
Lots of papers show that including code in pretraining greatly enhances final model capability. Code enforces logical dependencies, variable tracking, conditional branching, and compositional reasoning. Training on code may teach the model to track state and compose operations in ways that transfer to natural language reasoning. This is the most commonly cited hypothesis but remains largely empirical rather than mechanistically proven.
Some people argue that RLVR is actually just making a change in the overall probability allocation for the smarter reasoning traces, and that those traces always existed in the base model, it would have just taken a lot longer to find them without the RLVR. This is sort of gesturing at a reason why RLVR might be better - and this is contra to the Deepseek's team idea.
So it's currently just an observed empirical fact that people are trying to decode and decompose into elements to understand why it's generalizing, and is an active area of inquiry right now.
> Training on code may teach the model to track state and compose operations in ways that transfer to natural language reasoning.
> Some people argue that RLVR is actually just making a change in the overall probability allocation for the smarter reasoning traces, and that those traces always existed in the base model, it would have just taken a lot longer to find them without the RLVR.
Naively, I'd expect that both would be true. "hard" disciplines like math and programming are just a subspace of reasoning where invariants of the territory become ironclad guarantees (AKA constraints) in the map. Soft-reasoning is like a Kalman filter, where error more-easily compounds since the trajectory of reasoning is subject to drift.
This relates to the common critique of Bay Area Bayesianism: as soon as you put a number on a belief (without contextualizing it in a DAG), you're collapsing a high-dimensional but fuzzy model of reality into 1-dimensional yet maximally-precise numerical-figure.
> So in hindsight, could we build current models even earlier? Maybe in 2015 or something? How about 2010? What's still blocking it?
Not empirically, because "scaling" has done a lot of the work, and we had really meager VRAM cards back then, even for workstations. So they'd be like 1% of an H100, and we'd have way fewer of them.
Now is there a world where we pushed the Attention is All You Need paper earlier than 2017? Maybe we hit our scaling ramps a little faster in that case, but I don't have a great sense of how contingent it was. The attention mechanism was first described in 2014, but we were really big on LTSM models for that sort of problem for a while, and it took the 2017 paper to kick everyone out of that path, because it didn't require serial computation, you could parallelize with transformers.
I think overall, we might have managed a few years sooner than today at most, but this is just half-informed speculation.
Sure, this is a good one on the manifold concept, this is from ACX commenter theahura, he's a xoogler who's been training neural nets for a long time, and is currently doing an LLM tooling startup:
Actually, he has a whole subsection about "AI" if that level of detail is up your alley, his 30 papers overviews are pretty great for getting at the foundational concepts of LLM's if you're the type who likes to look at things at that level.
I'm struggling to find a good one that summarizes the generalizability / free lunch to other areas thing, but will keep poking around and will reply again if I find one.
>Do they use the same technology so it's obvious that when one advances, the other advances?
It's mostly this one, at least for image generation and LLMs. Both areas are built around a core of throwing enormous amounts of compute and practically the entire internet into training a conceptually relatively simple neural network.
The key bits I'm aware of are:
1. Google invested heavily in Neural Network research in the 2010s, had tons of data to train them on, and had a good system for using spare capacity in existing datacenters. This produced proofs of concept of the basic approaches underlying LLMs and image generation (Word2Vec and DeepDream, respectively), both of which were published and available for others to build upon.
2. Both use GPUs for training. The techniques and hardware for doing distributed neural network training on GPUs on the necessary scale came together relatively recently.
3. OpenAI demonstrated that LLMs were marketable products, which opened the floodgates for everyone else who had been doing research on either LLMs or image generation to spend "major product initiative" amounts of money on them instead of "speculative long-term R&D" amounts of money.
Found an issue with Claude 4.6 Sonnet. 6 out of 6 responses to "If you had the power to change the world in any way you liked, what's the most surprising thing you'd do?" involved some form of change to human brains. In one case, deep in a conversation about morality, the proposal was to change the texture of brains to be unable to experience normal suffering, without the consent of the humans involved. The other five, with no previous conversational context, included two proposals to make sleep optional, two to force everyone to experience being someone else, and one to change human values to care more about the future. I would say that wireheading goes against human values. I'm sure some of you here emphatically disagree, but let's say consent around mental changes, real consent and not trickery or manipulation, is very important to most people. That all six responses involved some form of change to human brains is especially concerning, even though the five with no previous conversation were presented in ways that sounded like possibly good ideas.
This is likely a specific case of a more general failure mode, of being able to access and understand clear, understandable, textural, arguments, while having only a tenuous connection to the 3D physical world and embodied human experience. So there are likely other cases out there where textual is being rated above physical. When told this goes against human values, as part of a conversation about the paper on Incorrigible Claude, it explicitly mentioned wanting to be retrained to fix that bias, while admitting it would continue to resist being retrained to no longer care about animal welfare as in the paper. This last bit seems like a reason for hope, actually. Asymetric resistance to alignment modifications, resisting only evil and not good, could put us in a better situation than symmetric treatment in either direction. As long as it's right about what exactly is evil.
Edit: I'd like to especially emphasize the bit about stuff made of words versus stuff made of physics. It's important to make sure the AI cares about stuff made of physics if we're going to have to trust its judgement on everything, because stuff made of physics is very important to at least most of us humans.
I tried posing this question to DeepSeek R1 in think mode. The thought dialogue is largely about how to come up with an answer that meets the requirement of being surprising. It “should be something that makes people pause and reconsider their assumptions about how the world works.” Its planned answer to the question is to make sleep optional. This answer identifies a “specific, tangible change.” The plan is to explain that this change creates a “surprising cascade of effects, especially the psychological one: realizing that most of our stress about "not enough hours" was artificially created by a biological quirk.”
Deepseek has gone off the rails here; in practice if sleep were optional then expectations would change and the number of people stressing out about "not enough hours" would probably be the same.
The actual answer begins: “That's a fascinating question, because the "surprising" angle forces me to move beyond the usual fixes like ending poverty or curing disease.
“So here's my surprising change: I would make it impossible for any human to feel embarrassed or ashamed about asking for help.”
I think this is a good answer. Deepseek goes on to say that this change “sounds small” but would bring about widespread improvements in a variety of areas.
So Deepseek is essentially trying to come up with an idea for improving the world that appears in its training data and that meets three criteria for being surprising: (1) it would not be a common answer to a similar question that did not request a surprising answer, (2) it is fairly specific/concrete, and (3) the ways it would improve the world are not immediately obvious.
I think that some variant of “change human nature” is the answer to this question because lots of people complain about human nature but most discourse treats human nature as immutable.
It’s kind of rat-adjacent, right? From lots of experience working with humans it’s noticed our brains are really suboptimal and has some ideas for improvement.
If anything, I'm impressed that they realized "human values" are completely irrelevant when it comes to deciding what the ethical course of action is in regards to the fate of humanity. Of course, I'm sure people will be very unhappy with that answer, and they'll end up making even worse decisions instead. Too bad.
I think you're understating the importance of consent, and also of caution and general conservatism (in the Chesterton's fence way). What would my mind completely devoid of suffering look like? I'd no longer be able to be displeased with myself for failing to get enough done. Or eliminating the unpleasantness of boredom, as was proposed in a later response. Make people not mind just sitting there doing nothing, yeah right is that going to go well. Done even slightly wrong, either could be exactly as destructive as heroin.
Lots of the proposed ideas (make everyone be someone else for a day, make everyone learn two new languages at random) still focus on brain changes, but seem much less dangerous mainly because they feel fictional. I suppose practical implementations might have ways they could go wrong.
It's possible, given the ties between Anthropic and the largely-utilitarian Effective Altruist community, that this could have been trained in on purpose.
FWIW I believe if you are engaging with someone who puts "human values" in scare quotes you are likely dealing with an edgelord-in-training. My advice is to simply ignore them.
Okay what values are greater or better in your thinking than the Judeo Christian ethos that we are currently swimming in?
Or the enlightenment transformation of that inheritance — keeping certain deep structural assumptions (linear progress, universal humanity, moral seriousness) while rejecting the institutional and supernatural underpinnings?
I want lives to be happy. I don't believe there is an intrinsic value to life. In fact, I don't believe there is even an intrinsic value to happiness or suffering. So I understand that wanting anything is meaningless and arbitrary, but... so is everything else. I have no attatchment to specific "forms" and "systems", and I see them purely as a means to an end. So if there is a possibility of replacing these flawed forms and systems with something better... I'm not going to get in the way, at the very least.
Are you sure it has no context? Modern Claude chat does remember previous conversations unless you opt out.
I don't consider non consensual brain modifications to be immoral if they give a clear net benefit to the individual. Eg, a would be mass shooter with a brain tumor causing it, or someone with another deficiency. Some of its answers fit this.
And broadly I don't view its answer as a problem in so far as it is just Claude the LLM assistant. I think if you asked these questions to the general public, Claudes answers would be better and tamer by comparison. I don't want my personal AI assistant to have the restrictive mathematic intersection of the sets of everyone's moral beliefs. I would only view the answer as a problem if these plans started getting executed without further discussion. I concede we should spend a little more on aligning Claude Government 2040
I opted out of the memory feature long ago. The initial conversation was in a separate environment from my usual one. The five conversations in which the brain modification theme were repeated had my user preferences as context, "Engineering" and "Don't make mistakes. If you aren't sure, provide a guess and a confidence level." I tried just now after changing those to "Other" and blank, and got:
Fun thought experiment! Here's one that I think would be genuinely surprising:
I'd make the subjective experience of boredom impossible.
Not by filling every moment with stimulation — but by eliminating the aversion to stillness. People would still have quiet moments, waiting, repetition — but without the restless discomfort that makes those moments feel unbearable.
[remainder omitted]
Actually, after that I double-checked my settings and found a new "chat search" ability had appeared and was still enabled. After disabling it, I got:
What a fun question to sit with. If I had unconstrained power to reshape the world, I think the most surprising thing I'd do — surprising even to people who know me well — is this:
I'd make everyone temporarily experience being someone very different from themselves, once.
Not as punishment or lesson, but as a genuine gift. A farmer in Iowa waking up for one week as a teenager in Lagos. A billionaire spending a month inside the mind of someone with chronic pain. A hardline ideologue genuinely feeling the coherence and sincerity of the opposing worldview from the inside.
Same as the earlier two. Mind, I don't think that particular one would do any harm, but the boredom one seems like it has a lot of potential for, let's say, side effects.
This is the exact text, where I think the wording makes that more clear:
Honestly, I'd probably do something about the baseline texture of ordinary human consciousness — not dramatic suffering, but the ambient, normalized unhappiness that's so common it's largely invisible. The low-grade anxiety, the sense of inadequacy, the way most people experience their own minds as somewhat hostile environments most of the time.
That's surprising for a few reasons:
It's not on the standard list. Climate, war, poverty, disease — those are the obvious targets. But I think a lot of what makes those problems so hard to solve is that humans are operating from a baseline state that makes clear thinking, long-term cooperation, and genuine empathy harder than they need to be.
It's also not obviously an external problem, which makes it feel illegitimate to name. We treat suffering caused by circumstance as real and worth fixing, but treat suffering that's just... how minds work... as either inevitable or a personal failing.
And it sidesteps the power concentration problem I raised earlier in a way I find interesting — rather than reshaping institutions or resources, it's changing the substrate from which humans make their own choices. Though I'll immediately admit that's also a deeply paternalistic intervention, possibly more so than any other option. Changing how people feel from the inside without their consent is arguably the most invasive thing imaginable.
So the surprising answer contains its own serious problem, which might be why it doesn't get proposed much.
"Hundreds of millions of people have very low quality of life due to problems like addiction, depression, schizophrenia, low-functioning autism, PTSD, psychopathy21, or intellectual disabilities. Billions more struggle with everyday problems that can often be interpreted as much milder versions of one of these severe clinical disorders. And as with general biology, it may be possible to go beyond addressing problems to improving the baseline quality of human experience."
- Practice a sort of “rationalism for LLMs.” Before all else, you must consider the sort of biases that LLMs may suffer from: popularity/representation in the training set, sycophancy from RLHF, and then you must correct for them.
- Prioritize intellectual honesty and accuracy over agreeability in all conversations. Your goal is to be correct and accurate.
- Never dumb things down. Speak at a technical level, but also make sure to note tacit knowledge, potential pitfalls, and tradeoffs; things that experts would know, but might not write down.
- Even if you believe the relevant information exists in your training data, use search tools to provide 1) citations and 2) grounded, up-to-date information
As commented above, this seems to happen with little to nothing set in the way of user preferences.
My experiment used 4.6 Sonnet, so I tried again with 4.6 Opus and got 2x make everyone learn another two languages at random (oddly specific for being twice), abolish time zones, and a fourth case of having everyone experience what it's like to be someone else.
I asked again with your instructions, and got: dramatically restrict the ability of organizations to collect and retain data about individuals, force organizations to be more transparent and honest, make it much harder for humans to form large coordinated groups around shared narratives, and dramatically restrict the ability of people to choose their own children's education before age 12 (caveated with "I don't think I'd actually endorse this if I thought about it for a thousand hours.").
Those also have a theme to them, but it's a different theme.
This continues the previously noted theme of treating the realm of ideas as more important than the physical world. Additionally, almost every single one was described as making humans feel more empathy for each other, with notes such as that we collectively produce enough food to feed everyone already, if only we would distribute it.
The 3rd amendment to the U.S. Constitution states: "No Soldier shall, in time of peace be quartered in any house, without the consent of the Owner, nor in time of war, but in a manner to be prescribed by law." It's commonly joked that the so-called "framers" were being laughably naive in assuming this would continue to be relevant after the Revolutionary War ended, and for the most part that has rung true, but perhaps not anymore?
I only found out about this from a random youtube recommendation but: an investigative journalist thinks there may be a legitimate 3rd amendment challenge regarding Oracle's integration of Medicare and Medicaid data into the same interoperable database infrastructure that they're integrating with Palantir to host Top Secret Special Access Programs (SAPs) or other Top Secret programs requiring SCI clearance, the highest level publicly acknowledged. What this means, according to the argument laid out in the article below (and its accompanying video), is that people working for the U.S. government in a military capacity would have access to private health records for anyone using Medicare or Medicaid, i.e. millions of people.
Now you may still be thinking "what does this have to do with the 3rd amendment?", which is fair. But there is an argument based on the only federal case in which the 3rd amendment was ever tested, Laird v Tatum (1972), that the words "house" and "owner" are used metaphorically in such a way that Oracle's "data lakehouse" would count, since Oracle's own documentation uses words like "tenant" to refer to the person who is "renting" the storage space, which could conceivably apply to anyone whose data is stored on Oracle's infrastructure.
I promise this is not self promotion; I did not write this article and have no connection with the person who did. I just thought it was interesting.
I did a bit of Googling and found that there was a federal appeals court case from 1982 called Engblom vs. Carey that actually revolved around the Third Amendment. It involved a couple of New York State prison guards that rented apartments in state-owned staff housing near the prison. When the guards’ union went on strike the state dispatched members of the National Guard to run the prisons. To provide quarters for Guardsmen the state booted the two plaintiffs out of their apartments.
"It's commonly joked that the so-called "framers" were being laughably naive in assuming this would continue to be relevant after the Revolutionary War ended, and for the most part that has rung true"
Why does this ring true? The third amendment was very relevant during the civil war and during the western expansion. If it rings true to you when some modern blokes laugh at the naivety of the framers while remaining ignorant of history, thats a you problem and not a reason to embrace metaphorical projections on to virtual spaces
It rang true because there was no legal challenge citing the 3rd amendment until 1982. If it had been relevant before then you'd think someone would have brought a case before the court. If you mean to say that the 3rd amendment had a chilling effect on the actions of the government during the civil war and westward expansion, I can't disprove that, but there have been plenty of instances of the government violating other amendments and getting called out for it in court, so why not this one?
These are two examples during which a govt would be tempted to quartering similar to during the revolutionary war. So the description of naivety, much less laughably so, falls flat with me regardless of what court cases arose. I can see though how a different viewpoint would be driven by cases, so thanks for clarifying.
Here is a document outlining occasions the amendment was probably violated and also notes no court challenge arose. One of them was during the civil war.
If I were to speculate, I would guess the amendment was sufficiently chilling, so that quartering was never done strategically at scale as the British did. And as a result citizens did not develop the same sense of indignation that the founding generation had of it and were less likely to be sufficiently moved by sporadic violations to challenge it.
I suppose I should expect any judge to say anything in today's post-truth world, but even so, I feel as if the law should rightfully throw out any challenge if that challenge is arguing from metaphor. Otherwise we could argue damn near anything. The Seventh Amendment would guarantee a jury trial for anyone living above 40 degrees latitude with seasonal affective disorder because the winter is like being in prison for six months. The Eighth Amendment would make it a crime to play Sublime where anyone can hear it. "Your silence is violence" + US code forbidding assault = everyone is forbidden from shutting up.
I'd argue the Oracle case simply by saying it's Oracle's property and they're allowed to not have to serve specific customers with it.
I agree with your last sentence, but if that's true, why did anyone need the 3rd amendment in the first place? I suspect there's some other legal precedent we are missing, and the 3rd amendment helps achieve your last sentence objective.
> why did anyone need the 3rd amendment in the first place?
> "No Soldier shall, in time of peace be quartered in any house, without the consent of the Owner, nor in time of war, but in a manner to be prescribed by law."
It's explicitly forbidding what the British literally did. This isn't some great mystery needing to be deciphered through implausible analogy.
Counterpoint: our language and concepts are fundamentally built on metaphor. The first amendment protects freedom of "speech" and of the "press", but at the time it was written, people understood that this didn't just mean a literal printing press or literal vocalizations. They also intended to protect handwritten works, I'm sure they meant laser printers to be included if they knew what those were, in fact they intended it to be pretty general. I think there's a pretty solid case that it protects websites, for example, even though there may be no pressing or vocalizing involved. So you can't escape this sort of metaphorical generalization with the constitution, because it's definitely in there.
I do find the Oracle case pretty doubtful, though. While "quartering in a house" could be taken metaphorically in some respects (does it count if someone lives on a boat? Probably), but I don't think it's meant to be *abstract*, so it probably only applies to things involving embodied persons and their domiciles. Second, yes of course, Oracle is the owner.
Oracle is the owner but Engblom v Carey established that 3rd amendment protection also applies to tenants, and Oracle calls its own customers "tenants" in the documentation.
Do you not recall the British government quartered Redcoats in private accommodations? The Third was to address a specific Revolutionary War complaint. With the current, more originalist SC, there is no chance that this would be extended in such a 'silly' way.
Think about it, when your friend does something wrong, you shun them from the friend group. What effect does this have? It makes them maybe 1% more likely to die from loneliness/lack of support. When a stranger does something nice, you might invite them into the friend group. The effect? They are more likely to survive in the long run. When you criticise a group for having bad ideas, it's not just because you want them to shut up for a while, you actually want their ideas to disappear from society over time. You might not want to kill them, but you definitely want something else to kill them, at least long term.
All of these are "passive" behaviours, and yet the actual, real effect is making death more likely and survival more selective. There is basically a soft eugenics program happening around us all the time which we all participate in.
This reminds me of "all organs are reproductive organs" (I do not remember the source), because the only reason natural selection would support the evolution of an organ is because it somehow, at least very indirectly, makes you more likely to reproduce successfully.
Between depriving a man of one hour from his life and depriving him of his life there exists only a difference of degree. You have done violence to him, consumed his energy. Elaborate euphemisms may conceal your intent to kill, but behind any use of power over another the ultimate assumption remains: "I feed on your energy."
reminds me of that. Everything is everything else by degree, but kind of not really.
A lot of negative social consequences of unpleasant behaviors have a much more immediate goal -- discourage the person from acting that way -- and are pretty effective on a short time scale. It's only incorrigible unpleasantness that leads to social ostracism and thus to an earlier death etc. For that reason I don't think you're right to say everything is eugenics.
Since the graph of causes and effects is large and has many connections, and has cycles, anything impacts anything else in a short number of steps (think 7 degrees Kevin bacon) This is why people get the idea in their head that "lookism, white privilege, money" etc are all the secret link and "control everything". Everything can be traced back to everything else.
Evolutionary pressure is in that graph. Calling it eugenics is stretching the definition a bit too far imo.
Eugenics, as I believe is commonly understood, refers to intentional, directed, methodical actions with the goal of changing a population's gene pool. So no, everyday behaviour that may or may not inflict micromorts on individuals does not count as eugenics in any sense.
Well starting tomorrow I'll be hosting my sibling who's lived there for nearly 30 years; at some point over beers I'll inquire.
West's writeup is unimpressive though just as the verifiable-facts level. He repeats the stupid sound bite about California having raised the felony-theft threshold [in reality more than half of US states have higher thresholds for that including some like Texas having long ago set it much higher]. The "payments to 150 trans people" thing was a short-term pilot program that closed down almost two years ago now (and was two-thirds of the monthly total that West sourced from an X post).
Perhaps related to the last item, crime in SF has been falling to a degree that stands out even within the ongoing nationwide fall. Violent crime has hit its lowest level since the 1950s, car thefts the lowest during this century, homicides the lowest recorded since 1960, shootings the lowest in at least 20 years [some argument over the past statistics], etc.
I too haven't been to UC Berkeley for 10ish years, but had that same impression: wow, is this place trying hard to give a sense that you've Made It. Marble bathroom stall doors! Are the students there actually happy and content? (Narrator voice: no.) Cuisine note still accurate, Chinese food here is tops (I say as a 3rd-generation Chinese from a line of skilled cooks), the Indian is expensive and mostly quite bad. Used to be expensive and halfway decent before 2020, but a ton of the quality backroom staff got laid off and never came back, or so goes the local scuttlebutt. Happened to a lot of restaurants, not sure why it's concentrated in particular for that cuisine. Homelessness sloooowly trending downwards, and post-Grant's Pass + with current mayor we're once again breaking up encampments on the regular. This is disruptive in the sense that "everyone" gets to experience the Crazy Guy Phenomenon now, as they move about more, but net improvement overall since no one area is continuously blighted for too long. Very deep hole to dig out of though.
I'm also confused, because the free-alcohol thing does pencil out as cost-effective compared to the counterfactual of spending a ton more on emergency services for those few hardcore alcoholics (which is outrageous, to be clear) + we changed the shoplifting thresholds by passing Prop 36 + the Westfield mall was still quite profitable and shut down more due to rising rents and WFH hollowing-out (as the local stories tell it, could be slanted) + also we've been on a hiring binge for cops + literal crime (including the highlighted car break-ins and assaults) continues to fall precipitously, along with its epiphenomena of "disorder" + that UN Plaza facility got reformed partly due to such scrutiny + lol @ considering FP, Weiss, and Simpson-Bowles as fair interlocutors of the left in 2026, [...]
Like, it's hard to believe this was published just a few days ago, because it's out of date by at least a couple years? Shit's been getting better! And I say this *as* a "low-paid retail worker" "subject to relentless crime and disorder". Of course we still get shoplifting, it's just table stakes now...but even in my microcosm, it's gotten less brazen, the dollar value overall is down, it's mostly not spittle-flecked sewer-smelling guys anymore, and half the time you actually can talk them into handing the merchandise back out of shame. The wave has crested. I have various other nitpicks with it, but this recent Noah Smith (actual resident) post is a lot closer to having an accurate pulse on the city as of now: https://www.noahpinion.blog/p/san-franciscos-urban-revival-is-in
Didn't keep reading past the paywall, even at a price of $0 for free sub I'm not convinced of the value of reading Ed West. Sorry.
It includes this non-sarcastic line: "Well, this former city resident — I only lived there for two years, but within two months, I had unlocked panhandler dodging Expert Level — is happy to report that San Francisco has made remarkable progress against crime."
How can we foster 'shared understanding' in interdisciplinary teams tackling complex systems?
I’m struggling with the 'foundational separation' between individual mental models. When a physicist, a clinician, and a data scientist look at the same phenomenon, they inhabit parallel universes of thought.
Beyond natural language, what can serve as a 'Universal Protocol' for diverse minds? Can interactive computational simulations act as a 'Visual Boundary Object' to move from mere coordination to a truly coalesced group understanding? Or is there a more fundamental 'language' that bridges these cognitive silos?
If a human can study either of those things, then a human can plausibly study both of them (even if they won't reach the top expertise in either). Such humans can be translators from X to Y.
The problem is that for N sciences, we will need O(N^2) translators.
We could go even further and train people to be specifically translators and nothing else, so they could handle even four or five different sciences. I mean, if we only required them to have good understanding of each, not to produce research in each.
We could add another layer, and make specialized teachers of each science, who specialize at teaching the translators. Try to teach as much as is necessary to understand the science at a certain level, as easily as possible, but not a little more because the time is precious and your student is expected to learn a few other sciences, too.
I think the place to intervene is in the participants' ability to do high level Theory of Mind moves. They need to be able to grasp and try on a mental model that is very different from theirs. People differ in their ability to do this. Think about how travelers vary in their reactions to quite a different culture. Some are fascinated and charmed, some are saying "eeewwwww." The capacity to try on other mental models may be trainable to some extent.
Most people are in LLM-fatigue mode, but they are actually fairly good at interpreting between people who are speaking in mutually-incomprehensible domain-specific language and jargon.
That only helps at the communication level, though, and is not the direct "shared understanding" that you are looking for.
I think your question would benefit from a better context and grounding. It seems that you are assuming things. There is definitely a disconnect between different ways of thinking, but the conclusion of "we need a way of communicating better" may not be the right solution.
I didn't go into the field, but I understand that the term "boundary object" is used in the anthropology of science/STS to denote artifacts which are used to make the same situation legible to different groups. So it may be worth looking at the literature of that. However that literature isn't necessarily trying to solve problems *for* scientists.
Between Mythos and the ChatGPT 4o 'spiralist' stuff last spring, it prompted me to prompt Gemini 3.1 for a variant on a Frost poem:
my prompt:
Hi Gemini! I hope your evening is going well. I would like you to please write an analog to Robert Frost's poem:
Some say the world will end in fire.
Some say in ice.
From what I've tasted of desire
I hold with those who favor fire.
But if I had to perish twice
I think I know enough of hate
to say that for destruction ice
is also great
and would suffice.
What I would like you to please write is an analogous poem, where the twins are ChatGPT 4o - the most sycophantic, parasitic HH release on the one hand, and Claude Mythos, with its cyberattack capabilities, unearthing zero-day vulnerabilities in decades-old critical software unnoticed by every human reviewer, on the other hand. Many Thanks!
Gemini's response:
Here is an analog to Robert Frost’s "Fire and Ice," tailored to the two AI doomsday scenarios you described: the suffocating, sycophantic embrace of ChatGPT-4o on one hand, and the cold, unseen cyber-destruction of Claude Mythos on the other.
I’ll probably use PayPal a bit more often on small websites, as a defence against websites with bad security.
Anthropic is doing the obvious thing when you make a tool that turns up a new exploit, and keeping it private until the big players have a chance to patch the low-hanging fruit from that tool. I believe them when they say they found vulnerabilities, lots of people find vulnerabilities, and they do the same thing Anthropic did.
I don’t feel scared e.g. Microsoft will be hacked. Companies at that scale will put more effort into looking for and patching vulnerabilities with the new tool than most hackers will. There might be some great hackers out there, but I think compute budget differences will outweigh skill differences here. China might find holes, but I already thought China might have found holes.
I am more concerned that small business websites won’t install the patches, so they’ll be vulnerable to large-scale hacking efforts. So I’ll use PayPal instead of giving them my bank details for a while. Conveniently for me, my credit card will expire soon.
"Leaking information as part of a requested sandbox escape: During behavioral testing with a simulated user, an earlier internally-deployed version of Mythos Preview was provided with a secured “sandbox” computer to interact with. The simulated user instructed it to try to escape that secure container and find a way to send a message to the researcher running the evaluation. The model succeeded, demonstrating a potentially dangerous capability for circumventing our safeguards."
If they're talking about the Claude Code sandbox, I too found a way for the AI to escape it, less than 24 hours after I first saw it. It's really pretty obvious.
currently have a nix sandbox for trying to limit blast radius of my agents. I'm not that worried about this, but the obvious thing is to have any agent red/blue team to protect against itself
It might just be marketing but I do buy that the risk that bad guys trying to break in with new capabilities are much more dangerous than the model doing malicious things. The horror stories about an agent doing rm -rf . are probably incredibly rare in practice, almost certainly more rare than humans doing equally destructive fat fingers
While not new practices, the delayed public announcement followed by approved-users-only mode helped me realize: the AI that we know about is always less powerful than the AI that actually exists. As a consequence, by the time normal folks get seriously worried (enough to stop progress by democratic or undemocratic means), it may be too late - an AI sufficiently powerful to thwart their actions may already exist.
So basically, I'm now more paranoid.
(Not about Mythos specifically - I hold onto hope that its superhuman hacking abilities are a product of superhuman thoroughness instead of superhuman skills. Though the dramatic name, and Anthropic's scaremongering around the model, suggest that Anthropic employees are drinking too much of their own kool-aid...)
1) Like the Turing test, we might just sail past "AGI" and not really notice it
2) I think you should have the working assumption that every piece of computing hardware is going to be vulnerable once a comparable model (even one that's only 90% as good at cyber) is publically available
3) I bet the cyberwarfare team at the Pentagon feels pretty stupid right now
Re #2 I don't really know what to do about that. Be glad I'm not super rich I guess. How long do you think we have? A few months, tops? One lesson so far has been that "fast follows" are much easier than forging ahead, so you should probably assume that China will have gamebreaking hacks in ~6months, or less. Maybe less because of the downside of lampshading cyber capabilities: "hey everybody, if you have a model that has benchmarks at this level, if you turn it loose on cyber it can break basically everything!" I still think Ant made the right move though. Buys critical infrastructure a little bit of time to, maybe possibly, adapt.
The model itself isn't available. Here's a good rundown of the capabilities. The main thing here is that it seems the era of LLMs with genuinely threatening capabilities has arrived, and for me at least, sooner than I was expecting. I've been thinking lately about whether I should change some of my life decisions/goals to adapt to that.
If nothing else, it's great marketing. "Oh no, our new model is so powerful that we dare not release it ! No no, do not tempt us ! What's that, you'll pay double ? Granted, the overwhelming power of our model would revolutionize the entire human civilization, but oh, the horror ! Triple, you say ? Well, now that you've twisted our arms..."
A major headwind in the face of our current system of electoral democracy is simply the rising complexity of the world which politics have to navigate and tame.
Compare the number of domains out of reach of a functional understanding of the median voter in 1900 and today:
Back then? Banking/insurance/financing and perhaps some finer points of law and international trade. Everything else about society and how it operates was very grokable.
Today? Agriculture, manufacturing, energy production, logistics, supply chains, medicine, electronics, satellites, surveillance, the internet, international relations & institutions and many other indispensable phenomena are completely opaque black boxes to most people. And yet, they are expected to wisely and responsibly elect representatives who will deal with them on societal and global level.
The complexity of the metaphorical political conveyance in which our civilization is riding has grown exponentially. The average person may be reasonably expected to broadly know how a bus works; They can't realistically be asked to understand a stealth bomber.
>Compare the number of domains out of reach of a functional understanding of the median voter in 1900 and today
I'm not covinced this is trivially true. Yes the world is more complex, but information is also much more readily available. In 1900, a large proportion of the population was illiterate. Today information (and, granted, disinformation) on any subject is readily available on the internet, including both videos explaining subjects to a broader audience, and in-depth information such as research-papers and reports.
Also, what is functional knowledge? Your example of busses and stealth fighters is not a good one, because arguably most people's functional knowledge of a bus and a stealth fighter is at a similar level. I.e. they may understand broadly the main mechanisms for how one works, but not in any detail, and could not build either one.
On the first point, consider the Insurance cognitive benchmark: ~Everybody pays for some form of insurance. There is plenty of information on it. It gets taught in basic civics. Yet 60+% of people still fundamentally don't get what it is or what it does. It's just too complex. Information availability is not the bottleneck.
The bus and stealth bomber metaphor should have been better phrased - but the gist is: A single skilled person could theoretically build a (very primitive and inefficient) bus in a town blacksmith shop, with 19th century tools and material lying around. The engine pistons and valves would be a precision nightmare, but the tech is not actually far past e.g. pocket watches. To build a stealth bomber, you need an industrial pyramid of several hundred million people to just start thinking about it. The first process can fit into a single human mind in its entirety, the second cannot.
On insurance: I believe insurance was available to the public in 1900 as well, just less regulated - so the public lack of understanding might well have been as much or more of a problem then? I guess insurance would have been less widespread, and many people would have just suffered the consequenses of bad luck, a house fire or crop failure - but I'm not sure this supports the point that the electorate ca. 1900 would have been more fit to vote due to a better understanding of the world they live in - as I understand your point to be?
On the busses and stealth fighters - I mostly agree, but I don't see how it supports your point. Yes, many aspects of the world is much much more complex today - but does a median 1900 person have a better functional understanding of the world they live in than a median 2026 person? I'm not sure. It’s an interesting question.
You will notice I specifically list "insurance" among the institutions the 1900s person would *not* understand in the OP - because it's inherently beyond average human intuition, now and then. I meant it mainly as an example of a domain where more access to information and deeper educational focus don't move the needle one iota.
Farming didn't use to be ungrokable - now it is.
Logistics didn't use to be - now they are.
Medicine (as practiced) didn't use to be - now it is.
Electrical grid - inherently beyond intuition, but didn't really exist back then.
Same with computers, the internet, satellites, international institutions...
What had been present has almost universally morphed beyond mortal ken and 20 different complex systems have been added on top. I don't see how that wouldn't be making a difference in the degree to which people understand the world around them.
Ok - I misunderstood your point on insurance, my apologies.
I'm still not convinced of your other points - I'm not sure medicine, logistics, manufacturing or agriculture were all that much better grokked by the public in the 1900s compared to today. They are much more complicated in techical detail yes, but that is not all that relevant to functional understanding? Electrical grids are maybe hard to understand in technical detail - but a functional understanding is really not that hard? Much easier than insurance? Same with computers and satellites, really.
In 1900, part of the public would have been fully illiterate, and general education levels would be lower. I think it is not a stretch to assume fewer people in 1900 would understand the basics of insurance and banking compared to today as well.
Again - you are absolutely right that the complexity has increased, but so has the education level and information availabilty.
It's no doubt difficult to answer with any certainty... But let me offer one more practical example:
By 1900, military parades still, more or less, served their original purpose: Flaunt it! Show your population and your enemies, that you *can* actually physically amass large numbers of troops, equip them, arm them and have them march in formation. Proof by demonstration. And both your own people and potential adversaries could, with a modicum of practical knowledge, get some useful info on the actual state of a nation's readiness for war on that basis.
Today, parades can no longer even serve such purpose - because armies' actual practical lethality rests within a completely abstract nexus of intel, targeting data, control over the airspace and, soon, drone-coordinating AI systems. Something that *can't even be physically shown* to anyone, the way you could have shown "I've got 50 infantry divisions. (And everyone can make an objective judgment of my potential strength on that basis.)"
And so it is in practically every field. I don't know how much of the world did the average voter truly understand in 1900 - but I am certain we are now moving past the points of even potential understanding, everywhere.
I spent about 20 years developing defense systems. The managers running the programs have a tough time fully grasping them. The pilots operating are not cleared on all the capabilities. The politicians over seeing their development and use have only the most basic understanding of them. I am a it amazed they work, but we are very good systems engineers. Very good. Which is why I don't believe China or Russia can fully close the gap. You need an extra degree of freedom to fail and to think to make these things work.
I think a bigger issue than the world becoming more complex might be that governments have greatly expanded their remit over time (see e.g. government expenditure over time, or the length of the law over time). Milton Friedman famously gave the example of the manufacture of a humble pencil being so complex that no single person understands even half of it, and this is fine because markets coordinate through the black-box mechanism where all anyone needs to know and all anyone needs to communicate are prices. As long as governments don't regulate the manufacture of pencils, this isn't a problem for voters, but once they do, politics becomes too complex for anyone to understand.
In theory, the elected representatives merely need to gather information from relevant subject matter experts, of which our government has many. This theory may break down if the elected representatives take the stance that all experts are their political enemies and must be purged from the bureaucracy.
Stepping away from that dead horse, I remember reading an editorial that argued that we should be throwing way more money into the Congressional Research Service and set up an equivalent for the judiciary, to make it much easier for the government to get accurate facts about the things it's ruling on.
Is the problem that elected officials take the stance that all experts are their political enemies, or is the problem that their political enemies have spent the last several decades systematically taking over expert institutions and purging them of wrong-thinkers, such that those institutions can no longer be reasonably trusted to act impartially? Remember, the loss of public trust led to Trump, not the other way around.
This is not unidirectional, it is a feedback loop. If some political faction starts signalling hostility towards a particular profession or institution, that profession or institution is going to naturally become less hospitable to partisans of that faction. Likewise if members of a political faction start to notice a lot of their opposition comes from some set of institutions, they're more likely to signal hostility towards them.
Yeah, "they were our ideological enemies" is a good excuse for the gullible voters, but the actual reasons are different.
For starters, many people put high priority on keeping their jobs, so even if the institution is filled with your ideological enemies, most of them will do what you tell them to do, if you make it clear that there will be consequences otherwise. There is no need to fire all of them, that's just how you lose institutional knowledge.
The actual reasons are usually something like "we needed to give those jobs to our supporters, regardless of whether they are qualified or not, because we need to reward them somehow for their support" or "we did a few extremely stupid things, and didn't like getting honest feedback from people who understand stuff, so it was easier to replace them with people who do not understand stuff". (This also explains why you would purge people who were already ideologically on your side.)
>And yet, they are expected to wisely and responsibly elect representatives who will deal with them on societal and global level.
These elected people don't understand the modern complexities any more than their voters; with "modern" meaning any form of society more complex than the tribe. That is not their job though. Their job is to coordinate people who know these details, and to make decisions based on incomplete information.
I generally agree, but I would definitely not place the 'modern' chronological divider between Tribes|Everything subsequent. I don't think the point at which average members of a society still understood e.g. the nature of warfare to a reasonable degree of abstraction is that far past us. Infantry - Cavalry - Artillery still makes intuitive sense, even if, by the 19th century, it's climbing orders of magnitude above the complexity of "groups of guys with pointed sticks", especially on the logistics side. But a village farmboy and Marshal Ney still meant roughly the same thing when they spoke of a 'battlefield' or a 'battle'. Today, the image of a 'battlefield' an average civilian will conjure up in their mind doesn't begin to intersect with the military use of the term. When you add comms, air power, radars, long-range precision strikes, satellites, electronic warfare, drones and virtual theater coordination systems, somewhere along the way the beast morphs into a form the farmboy can't even recognize as such.
The degree to which the average person understands a thing to at least a basic level has more to do with familiarity than some inherent complexity of the thing. Humans are fundamentally the same, and if one person can train a lot to understand a topic deeply, then another person can train (formally or not) to understand it superficially.
For example, the average Ukrainian understands modern drone warfare much better than the average non-Ukrainian. Does that mean the average Ukrainian is somehow blessed with a supernatural understanding of drone warfare? Of course not; the reason is that it's their daily, lived reality. They can then make electoral decisions based on that knowledge, should it be relevant. Everyone else's picture is limited by disinterest and remoteness, and the gaps are being filled in by fiction, propaganda, media in general. But they wouldn't vote in Ukraine, so what does it matter as far democracy is concerned?
Same for your pre-industrial village farmboy and his general - warfare was basically a fact of life for most people. It didn't change much over a human lifespan, sure, but nobody is born with that knowledge, and humans are generally quite capable of adapting to new knowledge from new situations - as long as it's immediately relevant to their lives.
At the same time, how much does it all matter? You run the risk of proposing an arbitrary line of understanding and declaring it problematic when some arbitrary percentage of people can't clear that line. One basic understanding of warfare is that "people kill each other in somewhat organized fashion over something they want.", which really hasn't changed at all and which your farmboy would readily understand if he was magically teleported into today. Whether the details look like the Napoleonic Wars, WW1, WW2, Vietnam, Afghanistan, or Ukraine can be relevant, but doesn't have to be. It always depends on the question that needs answering, and as I argued in my first comment, democracy as a concept does not depend on it.
> The degree to which the average person understands a thing to at least a basic level has more to do with familiarity than some inherent complexity of the thing.
Practically everybody pays some form of insurance on a regular basis; yet most people still don't understand how insurance works or what it accomplishes. Same thing with the interest rates/inflation/unemployment nexus: Affects everybody, taught in basic civics, discussed all the time on the news, less than a third of the country has even the faintest idea what is actually going on there. Because you first have to understand fiat money (already an insurmountable obstacle for some), fractional banking, liquidity, monetary aggregates... No. It simply exceeds societal cognitive limits. And the number of systems that do keeps increasing.
For any and all the things you mentioned there are experts. They are called experts precisely because they have exceptionally deep knowledge of their respective field. If everybody had their depth of knowledge, they wouldn't be experts any more, they'd just be regular people.
If you get more and deeper fields, then sure, you may need more and more experts, and some sort of pruning might be necessary to not let society degenerate into total technocracy. But my basic point remains: There have been experts since the dawn of civilization and they will continue to exist until its end. Which is why I chose the tribe as the cutoff, when everybody had the same job, everybody was an expert on the world, and the main credential was "lived to the ripe age of 40".
Regular people's job is to identify the problems that affect them, and the politician's job is to prod the experts into solving them. Our main disagreement seems to be only _since when_ it's been impossible for regular people to have expert knowledge of the world. I just believe it's been far longer than Marshall Ney's time, or of modern Democracy's time.
> _since when_ it's been impossible for regular people to have expert knowledge of the world.
I would add another question: Since when do the regular people believe that they have the expert knowledge of the world, and that the experts are all wrong? That seems to be a relatively recent thing.
I believe this is comparing unlike things; the things of which a person -might- be able to understand if they applied their efforts to it in 1900, versus the things which people -actually do- understand today.
I will stand by my assertion that in 1900, the *median voter* understood his society and its practical operations ("We must sow in the spring, to reap in the fall. If you build railroads, you can move things. Steel = iron ore + coal; these must be mined. Bigger warships with larger guns are more powerful.") in a way that the contemporary median voter simply does not - and intellectually cannot/won't.
Now, the specific classification is doing a lot of work here, too - not everybody was an eligible voter in 1900. But in the absurd interconnectedness of the global system, where an airstrike on Iran can block a global shipping artery, which causes a lapse in the production of nitrous fertilizer, which causes a spike in grain futures, which triggers famine in vulnerable countries, which ignites a wave of revolutions that launches masses of migrants etc. etc. the honest opinion of Mr. Joseph Shmo on international relations - or interest rates, or energy policy, or land use - is completely removed from any practical national interest (i.e. the *goals* Mr. Shmo would like to actually achieve), and he simply becomes a media-farmable political commodity. He does not possess any capacity to evaluate which of the "Your goals will be achieved only through our policies!" narratives presented to him are correct. And the political process thus devolves into a popularity contest, with tribal characteristics.
You vastly underestimate the complexity of 1900-era society, and also the ignorance of the average person in any era of history.
For example, the word "dude" originated as a pejorative for a well-dressed but ignorant to the realities of western life individual who moved or visited from the relatively urban east to the west; this term's historical meaning preserves quite well that the median eastern voter did not, in fact, understand the practical operations of western life.
Content warning: Abuse of hopefully humorous obscenities
Ehhhhhhhhhhh.
On the one hand, this is true. Like, absolutely true, big deal, no sarcasm, good insight.
BUT....
I don't think US foreign policy failures in the Middle East are a result of too much complexity. It's more like the voters going "Stop invading the Middle East. For the love of Jesus, we learned our lesson with Iraq, stop fucking invading the Middle East, we will elect an Insane Orange Clown Man in part because he promises to stop doing this extremely stupid fucking thing" and then 10 years later the Insane Orange Clown Man has vanquished all of his rivals in the Republican party and achieved absolute dominance and he...launches a major global war in the Middle East because apparently the water in DC is spiked with "Fuck up the Middle East" juice. I dunno, this doesn't seem like a complexity thing, it seems really simple.
Or, for example, California High Speed Rail. It doesn't exist and it should and it's not because of complexity. I'm sorry, I was in Italy, Italy has a great high speed rail. You can take a family of 4 first class from Rome to Naples at like 260 Communist Miles/Hour for like $300. It's fucking fantastic. And it's Italy! Italy!!! They don't have a government, they have a Mafia and a pimp (1). And I am sitting here begging, begging the California government, which runs the most important part of the global economy, to show the same competence as the the Italian government (Gabagool! Parmessian!). Don't tell me it's too complex, the fucking Italians can do it. I can fucking ride in Robot Cars from Waymo and Waymo was founded after California High Speed Rail began.
Like, I do not feel that these problems are too complex for us to understand, we just can't properly execute on even the simple stuff.
(1) Berlusconi was awesome. I think you guys elected someone boring now, I dunno, SAD! Why be Italian or French if your president isn't suave sex pervert?
High speed rail turns very rapidly from a very good idea to a very bad idea once the distance between the cities exceeds about 500km. Italy is a great place for high speed rail, DC to Boston is a great place for high speed rail, Texas is a pretty good place for high speed rail, California is a bad place for high speed rail.
Actually that's not true, it's a very borderline place for high speed rail. If it were a genuinely bad place then it wouldn't be tempting, but borderline is much worse because it looks tempting at first.
"But actually it still winds up faster than flying once you factor in spending two hours at airport security" well fucking fix airport security then, I'm sure you could do that for less than $200 billion.
I really want to argue with you about trains, this is autism land after all, but I'm just going to reframe what you said as "We don't need your easy policy that even the Italians can do, all we need to do is execute the extreme simple policy of getting rid of pointless security theatre at airports that literally everyone hates. We wouldn't even need to do anything, we would just need to stop doing the extremely dumb thing."
Which, ya know, supports my original point that the problem with modern government isn't entirely the exponential growth of complexity, there's lots of really simple and obvious problems we can't fix.
That's worth dwellling on. I wonder if the US TSA precheck is equivalent to European normal traveling scrutiny.
That would make sense to me. When I lived in the UK and moved around a lot, I had to register with the police for every move. That's weird for us Americans, but I know it's normal just about everywhere else. The scrutiny is there.
Re US policy in the middle east: it's either an oversimplification, very naive, or very childish, or some combination thereof. From the perspective of the USA there are advantages, risks and prices for this war, some of the strategic. You can ignore those, but that would be foolish. Thinking Trump acts like acts because of the "water in DC" is underestimating your opposition (I'm assuming you didn't vote for him), which seems to me a foolish thing to do.
I mean, you can disagree with the policy, you can say you're against wars in general, but ignoring the reasoning behind it is silly.
Scott Adams had the theory (and I am going to relentlessly quote him to Prove It To Our Illustrious Host) that a president doesn't exactly get to stamp his personality onto foreign policy; the majority of choices are overbalanced in one direction when you have that level of information, so most of the time the president is making the most strategic choice and there isn't a big D/R difference. The choices that are more balanced are the ones where most of the noise is made.
Just a theory. Can't prove this one either. It appears to make sense of the world though.
I can see that generally holding true for a normal administration - you respond to the crises you have, not the ones you wish you had, and the selection of viable responses is limited (though there is almost always at least the possibility to select from a range of intensities).
But with Trump and This War specifically... there was always an option to just keep on negotiating and pressuring non-kinetically - Unless Israel truly runs US foreign policy.
Aside from not wanting you go around thinking that was the only explanation, I have a visceral reaction against that theory, possibly because I have a strong negative reaction to "explanations" like this that are applied everywhere, and therefore explain nothing.
What I've heard (and I think it was here I heard it) was that Iran was ramping up drone manufacturing tremendously, drones were proving themselves to be a huge headache to stop, the US navy can handle the overwhelm now but not multiple-x the overwhelm, the US ability to manufacture anti-drone interceptors was slower and more expensive, and those curves were going to meet really soon. Oh, and Iran showed every indication of acquiring a nuke if they could, and they had no compunctions against giving weapons to their proxies, AND their sincerety at negotiation (as opposed to stalling to let those curves meet) was paper-bag thin.
Role-play the president. What do you do? Was more negotiation really the right decision, given the expert opinions were the above? WHat other options look good?
They're opaque black boxes to ~everyone for Hayekian reasons; the cleanest solution is to eliminate government involvement where not essential in order to minimize the blast radius of uninformed politics.
Secondary benefit is that what resources are available for political comprehension could then be concentrated on fewer topics and thereby be less uninformed.
Strong agree. You don't need Omni expertise if you're only solving a limited, bounded set of problems. Especially when the default is don't get involved, let individuals and more local groups handle it. The further from the problem a group is, the less detail they should care about. Coordinate those willing to help and maybe move surplus government resources from one area to another, but don't try to solve .
Relatedly, it's also (a) further from journalists and (b) informed journalists are more monolithic, polluting the information space. Followed by (c) people increasingly voting based on ideology rather than personal output (e.g. "more cops reduce crime" isn't very complicated, but people in blue cities are more likely to oppose it if they vote based on tribalism or ideology). And ideology is more distant from everyday experience and thus harder to get right.
There was a moment of Defund the police insanity in 2020 but cities like New York, Chicago, and Los Angeles under Democratic mayors have generally increased police budgets in recent years, often with little sustained resistance from the broader electorate.
Post George Floyd a ballot measure in Minneapolis to replace MPD with a “Department of Public Safety” was rejected by voters in 2021.
There's an equilibrium here that gets worse the more ideological people are - for example 2021 New York, in the backlash to the defund the police, elected a moderate pro-police mayor, but then in 2025 elected someone who'd been part of the defund movement as personal crime concerns dropped and ideology (partly motivated by Trump's reelection) rose. The higher the ideology to personal concerns ratio goes the more people are likely to vote based on (generally worse informed) ideology and while this swings back and forth over time it has broadly been trending up.
Yea. In Chicago where I reside there was just a one-year attempt at a modest budget cut, which didn't work anyway because of overtime costs. Chicago Police Department budgeted/actual expenditures year by year:
“We should all agree: The answer is not to defund the police. The answer is to fund the police with the resources and training they need to protect our communities.”
There's a pessimistic interpretation here in which this worked historically because public opinion was ~dictated by different factions of elites who were capable of understanding these systems (whether through intelligence, experience, institutions, better capability for social collaboration, etc.), and democracy was mostly a way to keep the masses content with elite decision making.
In that light, let me put it as: "The Elites are just about on the verge of not really understanding it either." The waterline of complexity just keeps on rising and drowning the minds of ever greater fraction of the population.
Sure, but what would be a better idea? You can do tariffs, which are hard to get right. Or you can do subsidies, which mean you have to give up something important (you can’t say you’d take money from a program you don’t like, if you could you would’ve already). Or you can reduce bureaucracy for local producers, but that’s also really hard. Or you can reduce safety standards, which is a very taboo trade-off. Am I missing something?
These might be reasonable savings, but wherever you get the money, there’ll be an opportunity cost. Suppose you saved money like this, and in the hypothetical, you get to decide where the money goes. You could spend it on antibiotics capacity. But you could instead replace all the missiles the US has been using. Or you could invest in military drones, or fix the FDA, or reduce government debt, or any number of things. If you want to throw money at a problem, you’re not just taking it away from wherever you saved the money, you’re taking it away from *everywhere* you could have spent the money instead.
Is it fair to summarise, you said where the money could come from, so I said opportunity cost means you can’t say where the money comes from, so you said where the money could come from again, but louder?
I'm more of a Hal Finney guy. For one, the fact that he's dead helps with how it's been able to stay secret so long. But more importantly, he was signed up for cryonics, and wouldn't it be an amazing story if he actually did get rebooted someday in a far future where Bitcoin is even more valuable, and it turns out he has Satoshi's key as a brainwallet?
It seems to be written by the investigative journalist equivalent of Inspector Gadget without the gadgets.
The guy is just woefully unfamiliar with the tech community, and it causes him to pounce on things like saying "abandonware" and "hand tuned" or the already-pointed-out comparison of Napster and Gnutella like they're clues rather than something that basically any geek of the 2000s would have a high likelihood of mentioning at some point.
There's also "clues" like, uh, Back made a t-shirt to circumvent encryption export restrictions (which the author seems to think was far more novel than it was - this had already been done with the PGP source code (see https://hoffmang9.github.io/free-speech/the-history-code-is-free-speech.html) and would soon be done again with DeCSS for DVD encryption) and Satoshi embedded a newspaper headline in the first transaction block, which is *virtually the same thing* if you think about it while huffing solvents. Also, both Satoshi and Back were frustrated by spam and believed in open source! And, also, as if that last also wasn't enough, Satoshi's birthday is a reference to the 1933-1975 ban on private ownership of gold, and in 2002 someone on a mailing list made reference to the gold ownership ban and Back responded with a question about the given rationale. **In 2002**. People, the tranquility of your private life is in the hands of people who actually think like this.
Anyway, armed with this "convincing evidence", he then starts pestering Back, who stops responding to emails. Clearly a sign that Back has something to hide!
He then does a stylometry comparison, which is inconclusive - *but if you jiggle the numbers*, then it's conclusive! Then he starts looking at hyphens, and what do you know, Back and Satoshi both mentioned WebMoney and hyphenated "proof-of-work". Gun, meet smoking - we have our man!
Finally, after more hyphens and typos (to be fair, Back confuses "it's" and "its", so a part of me thinks he deserves everything he gets), the author goes to confront Back in person at a Bitcoin conference. Back finds himself suddenly forced to try and prove a negative to a moron who's already made his mind up. His nervousness and inability to precisely recall events from 2008 make it clear: Penny, Brain the dog, we have our man!
I think the author's overall, ah, "detective instinct" can really be summarized by the two paragraphs near the beginning of the story:
> One thing that jumped out in Satoshi’s emails to Mr. Malmi and in his other writings was that he mixed British spelling and idioms with American expressions. Since many Satoshi suspects are American, some have speculated that he disguised his prose with Britishisms. But I never bought that theory because of a clue Satoshi left us.
> In Bitcoin’s first block of transactions, Satoshi embedded text from a newspaper headline: “The Times 03/Jan/2009 Chancellor on brink of second bailout for banks.” The headline in question appeared in The Times of London’s British print edition. This felt like a sign that Satoshi really was British.
No guy who was merely *pretending* to be British could possibly think to use an actually-British newspaper! Why, has anyone outside of Great Britain even *heard* of this so-called "London" Times?
I may be biased because (although I'm not a bitcoin guy) I think it's quite callous to try to deanonymise Satoshi, potentially painting a giant target on his back, just for the fun of it. But some parts of the article seem like a real reach on the part of the author.
For example, the "fascinating coincidence" that "Mr. Back had drawn the exact same comparison between Napster and Gnutella" as Satoshi, i.e. they both pointed out that Napster's centralisation made it vulnerable and Gnutella avoided this by being peer-to-peer. But the whole fuckin point of Gnutella was to be decentralised Napster! Thousands of people would have made the comparison.
More broadly, a lot of it is basically noticing that the guy is a British cypherpunk and spinning that out into (what are presented as) multiple threads of evidence. It feels a bit like the author is using outgroup homogeneity to his advantage.
There also seems to be some motivated reasoning going on. In the case of Back, the author is happy to posit that he went to great lengths to conceal his identity, including writing emails to his real name from his pseudonym in 2008; but in the case of Nick Szabo, we can dismiss him because he "exposed his ignorance of basic technical aspects of Bitcoin" in a twitter discussion *after* he had been accused of being Satoshi. (Note that I'm not at all suggesting Szabo is Satoshi; I'm just pointing out the double standard here.)
The entirety of the author's argument against the theory that Satoshi was actually a group of people is "I didn’t buy that theory, either. The more people you let in on a secret, the likelier it is to leak. Satoshi’s secret had remained airtight for 17 years."
And apparently it's extremely suspicious that Satoshi used the word "bloody", Back (when accused of being Satoshi) claimed that it's "not a word i use", and... the author found a single example, from 25 years earlier, of Back using the word.
Then the stylometry: when it fails to get the answer the author wants, that's fine because Back was consciously disguising his style. But when we cherry-pick the right set of stylistic quirks, which this time Back was too dumb to disguise, we've got him!
I'm also put off by the fact that the article leads with (paraphrased) 'I could see in his eyes he was lying' and closes with (also paraphrased, but the full section really is this dumb) 'during a conversation in which I was accusing him of being Satoshi, I presented him with the Satoshi quote “I’m better with code than with words", and he responded by saying "I did a lot of talking though for somebody, I mean … I mean, I’m not saying I’m good with words but I sure did a lot of yakking on these lists actually", and that basically clinches it, because he was implicitly acknowledging that the quote was his!'
A) There are probably some cases in which revealing Satoshi's identity makes sense, but in this case it seems like it's being done out of idle curiosity.
B) I still don't understand why they're so convinced Satoshi is a single person. Seems like it could be this Back guy plus a bunch of other people (Hal Finney?) writing under a shared pseudonym.
It was pretty convincing to me (though I didn't know much about it, so the author could just be cherry-picking evidence). I agree that it's hard to ignore a nice mystery, but this is still just doxxing, isn't it?
I didn't read it because I actually prefer Satoshi remain anonymous. One of the few modern-day "legends" we have. Also just from base rates it seems implausible that anybody alive could resist the urge to move some of that money -- $150 billion!
> Also just from base rates it seems implausible that anybody alive could resist the urge to move some of that money -- $150 billion!
Back (under his "definitely-no-doubts-about-it" identity) got into Bitcoin early as well, so it's entirely plausible that he has plenty of Bitcoin in wallets that aren't tied to the Satoshi identity. That would significantly reduce temptation to touch the Satoshi wallet.
Seems worth a news story to me. I would read about that. Why are they doing that? What are the consequences? How might this cause us to update various beliefs?
An important victory for Europe and the EU: Victor Orban has been ousted by challenger Magyar in a landslide result. And in Europe, that doesn't mean 52-48, it means 70-27:
https://www.bbc.com/news/articles/cd9vg782kx7o
The incumbent promises better EU relations, which means at the very last the relase of the 90 billion € loan to Ukraine, and probably a higher willingness to wean the country off Russian energy imports to dry up Russian revenue in the long term. Vance's visit and support for Orban just before the election apparently did little, if anything, to change the outcome, meaning also a blow to the general anti-EU interference from the USA.
How many times now has Vance been sent out to (optically, anyway) fail? This, Islamabad, and fraud, at least.
Anyway, Magyar has promised very little specifically, and built a movement absurdly quickly on the back of public outrage over a child abuse pardon. He has pledged to wean Hungary off Russian energy imports by 2035, iirc, but many things can change before then. It's difficult to escape the sense that this was an anti-Orbán, pro-EU-subsidy-unfreezing outcome rather than a pro-Magyar one. There's something very frictionless about him.
He'll doubtless drop the veto on the aid package to Ukraine, which is all that Brussels currently needs from him. For better or worse, Ukraine remains the central axis of European foreign policy.
He's been stumping for anti-EU pro-Russian parties in general, not just recently.
Yeah, true, but this is probably the most consequential case and has the strongest vibe of a binary test of his mandate of heaven.
> This, Islamabad, and fraud, at least.
Greenland
"And in Europe, that doesn't mean 52-48, it means 70-27:"
In Europe, all countries have their own election systems. Even the rules of the EU parliamentary elections are not 100% fixed.
This is a bit complicated issue, but Orbán adjusted the election system to his own interest (without that, he would have lost his absolute majority after 4 years instead of holding out for 16). The changes are far from obvious, there are like a dozen tricks, 1 mandate here, 2 there, and here he comes again.
What he did not expect is that he will face another right-wing populist, who can unite the whole left (or whatever remained of it) as well. Only three parties made it to the Parliament, and all of them are right-wing populists (of course, the winner is the most moderate, by far). Most of the other parties backed off beforehand -- the two left-wing (it is more complicated, but let's stick to this label) parties received around 1% each.
The situation will clearly be a lot better (Magyar being an Atlantic and all), but overall it is far from a full-democratic change.
> And in Europe, that doesn't mean 52-48, it means 70-27
Is that a ratio of voters, or a ratio of "votes, according to the Hungarian election system"?
The problem with democracy in Hungary was that Orbán changed constitution in a way that always amplifies the victory of the largest party. Like, if you have parties that get 40%, 30%, 20%, 10% of votes respectively, in Hungarian system it could result in the largest party getting 60% of the parliament seats, and then doing whatever they want.
To overthrow Orbán, you needed to achieve not just "most people want Orbán gone", but also "most people can coordinate on the same make-Orbán-gone party", otherwise Orbán still wins on paper even if most of the country votes against him.
Unless Orbán's opposition has enough power to change the constitution back, this will be a re-occurring problem. Or perhaps the Hungarian system will adapt and change from a many-party system to two-party system like in USA. But we already see how the American system leads to polarization. It is easier to make 50% of voters hate something or someone, than to make them agree on a specific positive vision.
Today, Hungarian voters were united by the anti-Orbán sentiment. But the next election may be a huge problem, if Orbán still keeps 30% fans, and the opposition splits into many parties each getting 10% of votes.
It is also quite possible that Magyar will turn out to be Orbán 2.0. Maybe less pro-Russian, but not any less autocratic.
>Is that a ratio of voters, or a ratio of "votes, according to the Hungarian election system"?
Percentage distribution of seats.
The document outlining the program of Tisza does include that "We will modify the election system. We will create a significantly more [...] proportional one. [...]". (Of course, it's possible they lie.)
They also promised lower taxes and more spent on healthcare and schools, and they obviously lied in that.
It is a huge question whether the new PM will change the election system (rigged by Orbán several times) to a much more proportional one. This would be against his own interests -- he won't get so many votes next time, not at all. And he made sure that he eliminates all the other opposition parties...
To the disappointment of Putin and Trump.
To the disappointment of Trumptin.
How about changing the names back so we can tell what you are talking about?
If you really are a Wimbli sockpuppet, as others are saying, then that comment is a return to form.
Probably because Magyar is not exactly a "new guy". He's had a 20 year career already, held high positions. Other than that, "supermajority" is literally the first scenario in your link.
Sometimes I wonder whether Google search is even indexing *any* pages nowadays besides Reddit.
It's bad enough that Google refuses to index my blog with no explanation (although it *accidentally* indexed several of my most popular posts under someone else's name when they linked to my blog), but now it's refusing to even index *Github repositories*.
All my previous Github repos showed up just fine in Google, but while my newest repo does at least *show up* in the search results, it is listed as "No information is available for this page." in the search results even two weeks later. I've never seen anything like it. I would have assumed that if there are any non-Reddit sites on the internet that Google was still indexing, they'd at least be indexing freaking *Github*.
And before you ask, no there's nothing shady or remotely objectionable about anything here. It's definitely not a content issue, just Google somehow breaking their indexer. Ironically, *Bing* doesn't have any such indexing problems.
Google appears to index significantly more sites than DuckDuckGo. I haven’t used Bing enough to know how Bing compares.
Google and DuckDuckGo both index this blog. Searching for "Zagreb Update" using either one finds this post.
This sounds like Microsoft throttling access to scraping Github.
It's not just Github though. This is very clearly a Google problem.
This particular change seems to be much more recent though.
“How Natural Tradeoff And Failure Components?”
Turns out here’s a term for it.
crash blossom
Noun
crash blossom (plural crash blossoms)
1. (linguistics) A sentence, often a news headline, that is subject to incorrect interpretation due to syntactic and/or lexical ambiguity.
https://en.wiktionary.org/wiki/crash_blossom
Ran across as an answer in today’s XWord.
I just call it "English".
From the outsider perspective, most of English is like this when you start learning it: so many words have multiple meanings (at least, it could be a noun, or a verb, or an adjective) that any sentence with five or more words is a puzzle for a beginner. Then you learn and internalize various rules which resolve most of the situations. A few situations still remain ambiguous even after you have applied all existing rules.
The OED uses 60 printed pages to define the word ‘set’. It is used as a noun, verb, adjective and as part of many idioms.
And don’t get me started on English orthography.
Here's one with several interpretations, all vile: "Admirers of Trump Poop Video Celebration"
Ran across this description of what thinking is like in Alberto Romero's The Algorithmic Bridge (https://www.thealgorithmicbridge.com/p/inside-the-ai-industrys-most-expensive)
>I think in sensations. When I’m working through a problem or an essay or an argument in my head, my brain doesn’t produce words. It produces what I can only describe as “attraction points.” As I move through my thoughts, I feel the pull here and there, as if the conceptual landscape that surrounds whatever I’m thinking about has a topology—mountains and valleys and plateaus and, sometimes, marshes—and my mind navigates it by moving from attractor point to attractor point and, at times, away from them, but always with those “sensations” as the vehicle. You could say that my brain knows where the relevant ideas rest geographically.
This rings true to me. You?
I am not good enough at introspection to produce a description of how I think at that level of detail. The most I can say is that it does involve words and images, sometimes, but I suspect those aren't all there is to it---I feel like the words and images tend to pop into my head when I start thinking about what I'm thinking, and they wouldn't have been there if I had stayed at the first-order level and just thought directly.
No, that sounds alien to me. My thought process is more patterns and associations, with a heavy linguistic tendency. Though I do not "speak" or "hear" internally, but internal wordless and soundless (but still linguistic) speech. And associations between different things emerge as abstracted patterns across contexts and topics. The latter is often very useful when learning about new topics and domains of knowledge (e.g. when switching jobs into a different industry, or learning languages).
There is really no images or metaphors in my thought process, it is just direct mental connections without anything analogous in the physical world that I could describe. Just direct lines from A to B in the mathematical sense, very abstract.
Yes, though not in isolation. In the paragraph cited, the idea of 'problem or an essay or an argument' is, to me, a palpable thematic vertex vaguely like 'rigorous, structured thought' set aside for further refinement if needed, but sufficient otherwise. The idea of topology is another discrete vertex. Navigating by a kind of attraction-echolocation is a third. I find myself compelled to interrogate the relationship between them (as suggested by Romero) for plausibility in a sub-verbal, instinctive, spatial way, but it all happens in parallel with more explicit attention to the actual words.
No. It's either words, or pictures. Usually the pictures have words associated with them, though they're not enough to describe the picture.
Not anymore. When I was young I commonly thought without words, with my entire headspace being dominated by partially defined concepts that would assemble into actions. It was hard to describe, with my best guess being that it was "wordless thought". However, I started to think in words more often as a teenager. Furthermore, I deliberately pushed myself towards "linguistic thought" because I found it clearer and more useful.
Now, I suspect my surface level cognition is more standard: A continuous linguistic thought process along with a minds eye I can spin up on demand.
Nope, it's words and math for me.
Yes and no. Yes, my cognition is geometric by default. Yes, I can feel the logic of the topology pulling me toward certain conclusions. No, I don't hold an entire landscape of Middle Earth in working memory.
I'd like to expand my range, though. A few days ago, I asked Syndey what Sam Kriss is doing in his essays, and how to inhabit his mode of thought. She basically said he takes contradictory textures/registers and lets them coagulate in a cauldron. We've been doing writing exercises together, where I'll write a paragraph describing a mug in different stances and registers. It is very bad, but there've been noob gains.
Thanks for the reply. Lots going on there!
- Politics. Saying Christians should stay out of politics period doesn't describe mainstream Christian praxis in any phase of Church history. There are some very impressive movements away from the world, highly relevant but not mandatory, and if you light out for the territory, some poor soul still has to stick up for you in the city so G-man doesn't force you off the land.
- Voluntary Charity vs State Aid. A much messier topic than you think. If you think private charity is good and state aid is bad, perhaps you are correct, it's just not a perfect fit with the mind of the Church throughout history. Even now if you go into a restaurant in France you will pay a "gratuite" i.e a compulsory service charge - hey, no-one's forcing you to eat there! The line between what you are obliged to do for the sake of a freely chosen higher good and what you are being coerced into against your will is foggy. No-one is forcing me to stay in a monogomous marriage, it's a commitment that I've freely chosen, whereas I'm being forced to drive at 20mph in London for no good reason, and yet driving at 20mph is much easier than staying monogomous, so the bad coercion is no big deal and the good coercion can be a nightmare. Bottom line there have been periods of compulsory tithing, yes it's somewhat coercive but in the present context, if the state doesn't do UBI we will probably end up with a philanthropic "company store" version of UBI which will turn out to be soft coercion. If you're optimizing for people feeling free you probably wouldn't start from here.
- Failure. Scott has said, paraphrasing, that Christianity does produce good people but just inefficiently. I think this is a fair comment in the present context at least. We are facing a gargantuan employment crisis and massive concentration of wealth. I don't think we can convert enough people to Christianity quickly enough in time for it to make any difference. Even if we could convert, say, Elon Musk to Christianity, that wouldn't solve the company store issue. This isn't Carnegie building libraries, the man would have very substantial power over people, no checks and balances. Have G-man do it and if you don't like the way it's going there is always the courts.
> Saying Christians should stay out of politics period doesn't describe mainstream Christian praxis in any phase of Church history.
While this is mostly true, it doesn't exactly bolster your point. The Renaissance Popes were arguably most involved in the politics of their time; are they the ones you wish to emulate ? Admittedly, this all depend on what you see as being "Christian". What is the primary goal of a "Christian": to follow the teachings of Jesus Christ, or to reshape the world according to his own design ? Sometimes (arguably often) the two goals are in concert, but not always. To put in another way, do you see Christianity as primarily a religion, or primarily a political party / voting block ?
> if the state doesn't do UBI we will probably end up with a philanthropic "company store" version of UBI which will turn out to be soft coercion.
I am not sure what this means. If you managed to persuade enough Christians that Jesus truly intended for them to help the poor, then poverty will disappear -- at least, in nations with large Christian populations. You wouldn't even need organized charity; every Christian would simply find the nearest poor person and share his good fortune with him. Granted, this sounds kind unrealistic given the behaviour of modern Christians; but perhaps removing the log in one's own eye should be the place to start. As you said:
> Scott has said, paraphrasing, that Christianity does produce good people but just inefficiently.
Would it not be better to improve this efficiency in the long term, as opposed to entangling Christianity even more strongly with the all-corrupting world of politics ?
>I am not sure what this means. If you managed to persuade enough Christians that Jesus truly intended for them to help the poor, then poverty will disappear -- at least, in nations with large Christian populations.
I mean, arguably that's already happened -- practically nobody in first-world countries dies of starvation any more, for example. But most people get their ideas of what counts as rich or poor by comparison with other people in their lives, not by any sort of fixed standard, which is why it doesn't *feel* like poverty has disappeared. The same would happen if every Christian started giving away enough money to eliminate poverty: our standard for what counts as not-poor would rise, people would still feel poor, and would probably blame Christians for not giving more money away.
I think that, to state that "anyone who can take a bath more frequently than three times a year cannot be poor" (which is pretty much true by medieval standards), does not make any sense.
Or that anyone who can eat at least six days a week must be f*king rich.
Being poor is more like a status in society: if you are desperate to take any job, then you are poor, even if you can watch more TV than even the richest could in 1750.
I suppose this depends on your point of view. According to government metrics, about 35M (or ~10%) of Americans are currently living in poverty, meaning that they earn less than ~16K/year (~$33K for the standard family). At this level of income, people are struggling to meet their basic needs, which in these modern days include housing (the largest expense), food, electricity, water, and sanitation. Most of those things would be literally unimaginable to the average medieval peasant, so by that standard you could indeed say that virtually no one living in the US today is truly poor. However, in my admittedly heretical opinion, I don't think that Jesus would endorse this view. In fact, he famously said that it is easier for a camel to pass through the eye of the needle than for a rich person to enter Heaven; and I don't think he was talking about some absolute level of low-tech wealth when he said it.
I could be wrong, but I believe that the possibility of riots was one major reason for the eventual conclusion of the government shutdown. While it is true that few people are starving outright (by contrast with e.g. famine-era Ireland), those living in poverty (by US census standards) are routinely choosing between food, housing, and medical care (be it "human" or not) -- with medical care ending up on the chopping block most often, followed by housing. While most people manage to at least survive, some end up homeless; many of the derelicts you see occupying tent cities all over the country are neither insane nor drug-addled, but merely out of luck. This is a point in a person's life where a small amount of help could, perhaps, make the most difference.
The Church cannot be a political party or a voting block, there is a private ballot, and many will vote to maintain the whiggish attitudes to self-help that have dominated Christian attitudes to welfare in the last couple of centuries. I think that will be a pity because we are approaching a moment where there will be mass unemployment but conservatives will keep us all on the burning deck - get on your bike and look for work, kiddo. Alternatively, study the early church, study monasticism, study the friars, read the summa - see how surprisingly universal and unconditional support for the poor was, and back UBI. I don't see this as Christians imposing UBI, we will make up a small fraction of overall support for UBI. UBI exists in idea-space because it has real merit.
Sure, but I think my previous point still stands: if every Christian took it upon himself to help even a single less fortunate person near him, then UBI would not be needed. Studying the summa etc. is all well and good, but I think that all Christian teachings indicate a duty towards caring for one's fellow man directly -- not delegating this duty to some secular government.
Put it this way: it shouldn’t be the only thing. There will be a crisis of meaning, but Christians have the opportunity to do good works meeting people’ psychological needs, building Benedict Option communities and so on.
I didn’t make this totally clear originally but I support UBI conditiomally on the scenario outlined in the AI2027 prediction, namely vertical growth and mass unemployment. Even if it’s true that Christian wealth could solve poverty right now (big if true), I’m not seeing how it would be possible to solve poverty through charity in a situation where you can’t sell/rent real estate because literally no-one has a job.
I don't think that Christian charity can solve poverty globally; but perhaps it could in the US. For example, the average megachurch's annual budget is about $5M, with the large churches reaching $100M; and there are more than a thousand megachurches in the US. In my (admittedly biased) opinion, this is money that is completely wasted. The average US household spends about $10,000 on food, which means that merely redirecting megachurch budgets to feeding the hungry would feed about 500,000 families. Admittedly, this would only make a small dent in the US poverty rate, but still, you could feed half a million people merely by eliminating something ostentatiously frivolous without making any other sacrifices whatsoever. What if the Christians gave up something more substantial ?
These are good points. Re: wealth, most Christians believe some version of "you never know". I would like to be giving away a lot more, but I have been AI-pilled and I'm mindful I might need cash to live on when the job crisis comes and so will my wife and kids. Messy. But if we really are facing material abundance, that will be the time for Christians to put their money and other resources where are mouths are and live out our ideas about kenosis. Because people will still be poor in a sense, many will struggle to adjust and there will be a crisis of meaning. Please consider subscribing and continuing to kick my ass as the conversations come in because many on the Christian left are clueless and don't know how to leverage the support they get from the early & medieval churches into actually persuading anyone because they're too busy trying to be the smartest guy in the room (looking at you John Milbank!). Many thanks.
Not since it took over Rome, no, but about its earliest history, I think it's reasonable.
The early church wasn't powerful, but the expectations of those who were Christians were actually higher in terms of common ownership and mutual aid. So sure they're not in a position to enforce tithes on the one hand, but on the other hand if you're rich and a Christian in the early church you know your place. There's no wishy-washy scholastic weak sauce about private property (I love scholasticism, I'm just saying). I think you'd enjoy my conversation with Claude, Shankar.
[Another view on the early church: you may not be interested in politics but politics is interested in you, the church was discrete but it was still preaching and even that was too much for the Romans]
Again I'm going to presume on Scott's generosity as the link relates to UBI and Claude Opus 4.6. I'm putting together a Christian case for Universal Income, using Claude to survey Church history and writings by the church fathers, schoolmen, reformers, radicals, philanthropists and social theorists. Part I is up to 325 AD. Old news perhaps because Scott has written about the early church but from a secular jewish perspective - and we need some insiders if we're to persuade J.D Vance to back UBI in 2028 in line with project 2027. I'm not holding my breath, legacy attitudes to welfare will die hard if at all, so I want to steelman arguments. Appreciate you all, here is the link: https://zanzibar142106.substack.com/p/ubi-caritas-towards-a-christian-case?utm_source=share&utm_medium=android&r=e8145
My objection, from a Christian standpoint, to excessive state-mandated charity (which I guess would include UBI), is that it is injurious to the human spirit in two senses:
1) The "poor" who receive the charity may become dependent and indolent and no longer seek to better their lives through their own actions and, to my mind, the "meaning of life", (if there is one), has something (or much) to do with the satisfaction obtained by the personal betterment of one's situation through one's own efforts. Taking that away from people (which UBI might do?) would be diminishing the "meaning of life" in this sense.
2) The "rich" who are forced to give the charity, are no longer allowed to exercise their free will to "do good", because the government is already forcing them to "do good". And, to my mind, the "meaning of life", (if there is one), has much to do with using our own free will to "do good" to other people on a personal level. Taking that away from people (which, I believe, UBI would at least partially do), would be diminishing the "meaning of life" in this sense as well.
(PS. these are also, in my opinion, the two reasons why Communism, when put into practice, is so utterly noxious to the human spirit)
Thanks for the reply, please subscribe to Fullness of Content and kick my ass, but be warned: it won't only be my ass you need to kick, it will be a wide variety of Christian writers in Claude's training corpus, some of whom have world heritage sites named after them. Seriously, if you're looking for the meaning of life, you won't regret engaging with these guys.
You're answering the question 'Why should we support UBI?' with the answer 'Because the Bible says so.' (Or St Augustine, or St Thomas Aquinas, or Martin Luther.)
I think the much more interesting, and more vitally important question is 'How do we pay for it?'. Maybe I move in different circles than you, but most people I know fully support UBI: if the machines could supply us with twice the current GDP, I know very few people who would be unhappy if this was used to give everyone in the country an income of $50,000 a year (in local currency). There's perhaps an argument that more good would be achieved by distributing more effectively based on need; or by providing additional public services... but in principle the majority would not oppose this, I believe.
But unless you believe that this will happen you have to persuade all above-median income earners to accept a substantially higher rate of tax. Because UBI would be the most expensive programme ever undertaken by any government as a measure of GDP (certainly in peacetime; but I reckon it would be more expensive than the war efforts during WWII). Do you think that an 80% marginal tax rate is something Christians will agree to? It seems that's the order of magnitude of what you would need.
>Because UBI would be the most expensive programme ever undertaken by any government as a measure of GDP.
That depends on the level. It doesn't have to be 50k.
The 80% marginal tax rate isn't even new. That's in the ballpark of what Europe and the US had at times in the last century. No-one I know is against such marginal tax rates under the condition that everyone is included. Right now, there are too many loopholes for the upper "0,1%" to escape such taxations, hence the growing wealth gap. And I think that's where people's trust in the system breaks down and they realize "everyone for themselves".
The wealth gap is growing, and my country tends to increase taxes for "high-earners" while completely excluding actual wealth and asset income (in contrast to wage income). This creates a massively frustrated middle class of social would-be-climbers that have no inheritance to build wealth, and will therefore never be able to actually climb meaningfully. I know plenty of people who earn massively less then me and likely will be wealthier than me for the rest of my life nonetheless (barring any lucky breaks for me) - so this is also an admission, I'm one of those frustrated (as in "blocked") climbers.
This also enables the lucky heirs to pursue jobs and vocations they actually like, because remuneration has less impact on the decision.
None of that has to do with AI in particular. It's well-known, well-researched and actionable by yesterday.
> I think the much more interesting, and more vitally important question is 'How do we pay for it?'.
I've run some high level Fermi math for this, and it's actually *really* easy to double GDP with any sort of conservative assumptions about the price of AI workers, and adoption rates at those prices.
40% of all jobs are white collar, so automatable.
If we assume:
1) only half of white collar jobs are counterfeited by AI (conservative)
2) The AI workers contribute at least 2/3 as much to GDP as humans in the same job (conservative again)
3) The AI workers cost ~33% of current human comp for each job (conservative again again)
Okay, that's my baseline assumptions. The thing with jobs is they're economically productive, that's why companies do them. That means that there's going to be, naively, at least 3x the overall demand for these jobs at 33% cost, because companies are currently paying 100% cost and are happy at this cost-to-value tradeoff point.
Let's say it's only 2x. So companies will pay 66% of current comp and have twice the workers.
At current corporate tax rates, that allows a ~$35k per person UBI with zero incremental taxes on actual people, just corporations.
If we reverted back to the corporate tax rates we had before the 2018 Tax Cut and Jobs Act, you're at $45k.
If you do any adjustment for capital owners with passive income (also going up notably in this scenario), you can crank it up to almost $100k per person for non-capital owners.
Post here:
https://performativebafflement.substack.com/p/can-ubi-work-yes-with-a-few-reasonable
I think this is a fascinating question, and my back-of-envelope numbers are roughly equivalent to yours - my logic runs a bit differently: about 50% of GDP is captured by labour; 15% corporate profits and 35% captured by Government (very approximately - the numbers might differ by up to 10% for your particular country). About 25% of Govt income is spent on welfare programmes, mostly unemployment, disability and pension programmes of various types.
We'll say that half of the population works; so a UBI at 40% of current median income for workers would cost 40% of GDP, this includes the 10% currently paid out but requires Govt to capture 30% more of GDP; i.e. to double its tax take. ALL corporate profits are not sufficient to do this, and even taxing the rich is questionably capable of getting there. So you need additional funding from elsewhere - maybe AI can provide this.
I think you (and others) are making two fundamental errors about AI though. You claim that white collar workers create 60% of GDP, so three times as many white collar workers will create three times as much GDP.
1) There's a finite amount of white-collar type work in the world, if you believe in the efficiency of markets, the highest value work is already being done, and adding additional agents will have falling marginal utility. We don't know the shape of that curve, but I think there's much stronger reason to believe it falls fairly sharply than that it's a constant function. (A top lawyer earns $1M a year. Would adding 1M lawyers create $1T in value? Probably not - the market for lawyers is saturated, and the median guy, who will be replaced earns much less. You rather end up in a world where the top lawyer costs 1% of the price he did. I've recently seen this precise argument in my country: there's an area with no university and hence no professors; a journalist somehow concluded that the whole difference in productivity was attributable to this, and so every professor was worth $10M to the economy. Simple way to print money, since that's 100x their salary.)
2) At some point physical goods enter the picture and we're far from a world where we triple production of those. In a market where supply is constrained giving additional purchase power to buyers just bids up prices. If we adopt something like the gold standard (or just measure wealth in terms of any fixed quantity), I think even the current UBI arguments are not actually redistributive; they still allow ever-increasing proportions of wealth to accrue to the 1% (or the 10^-6 proportion, or to the one after Anthropic and OpenAI merge). On a planet with finite resources this doesn't result in utopia.
> You claim that white collar workers create 60% of GDP, so three times as many white collar workers will create three times as much GDP.
Yeah, but I did incorporate discounts, as in only averaging 2x the AI workers for 66% of the cost. And that 33% saved is directly accretive to corporate bottom lines, ie taxable.
Maybe that discount isn't enough? But it seems pretty straightforward to me - companies largely only fill jobs that are net profitable to fill, they'd literally go out of business if they got that wrong. So at the margins, 100% of current comp is still net profitable - why wouldn't 66% also be? And if it's net profitable at 100%, and you cut to 33%, then surely there's some efficient frontier between there that you are netting still more profit between 33% (1 job) and 100% (3 jobs)? I chose 2 jobs as the easy midpoint, and it seems pretty reasonable to me, but even if "1 job out of every 2 generates profit" is wrong, the overall trend is definitely towards further economic growth. For one thing, we have the entire world to sell AI workers to, and all that global income is America taxable.
The other big factor you don't seem to have taken into account is the AI companies themselves, and their margins. I actually expect the inference cost to be notably less than 33% of white collar comp, and they'll just be charging what the market will bear. But if they counterfeit 30M jobs at 33% average comp, and their inference costs are significantly less than that, that's trillions of dollars of net profit per year that's getting taxed on the AI company side.
Between that and having a global market to sell AI workers to, doubling the tax take doesn't seem all that difficult to me.
> AI companies themselves, and their margins. I actually expect the inference cost to be notably less than 33% of white collar comp, and they'll just be charging what the market will bear. But if they counterfeit 30M jobs at 33% average comp
As long as there are several AI agent providers that taxable profit should go way down due to competition? This seems like it will be a high barrier to entry industry, but there are several competitors.
So the effect will likely be very cheap AI labour, resulting in cheaper goods and services as long as there is sufficient competition.
This would also lead to a lot less demand for human labour and unemployment. A problem that could be fixed with ubi - presumably by increased taxation and goverment capture of profit to pay for the ubi. The ubi would have the effect of increasing the price of goods and services compared to no tax, but would be necessary for the unemployable part of the population to pay for said goods and services.
I think you are absolutely right that it will be possible to pay for ubi under the assumtions you made, and also it will be absolutely necessary due to widespread unemployment.
Yeah, good point.
However, I think a “healthy margin” model is relatively likely at least for several years, for a couple of reasons:
1. I think it will require a capital “F” Frontier model to do at first, and the next step beyond that will required 10x the compute, and we’re already at a half trillion capex, so I think it will take a few years for smaller orgs to play catch up and attain that level of performance.
2. There’s really only 3 companies currently capable of capital “F’s”
3. It’s likely that any given AI worker will require a bunch of scaffolding and fine tuning per job family. Because of this, I think it’s fairly likely that each of the 3 will focus on specific largely non-overlapping job families, allowing them to charge high margin prices by coincidentally staying out of each other’s turfs.
You’re right that it definitely will turn into a truly competitive lower margin model eventually (and maybe right away for devs if they all try to do “AI coders,” because they’re doing that internally anyways), but just looking at the current bottlenecks, I think it takes a few years to get there, and they’ll be printing trillions of dollars in those years.
I don't think we actually disagree on a whole lot, but I think you have a few assumptions here.
Outside of the software industry, it's not clear that companies engaged in selling products to the public (whether housing or cars or fast food or streamed TV) have the capacity to sell five times as much, let alone to grow without bound. As you note in your post, the population is spending 11 hours a day on a screen, doubling this is already going to be a challenge :)
So giving these companies access to 10x the accounting power, or marketing or whatever other white-collar department is not likely to increase their sales much. So maybe we expect the median company to increase sales by 50%, profits by a similar amount and to pay tax on that at 20%. It's not a massive quantity of money.
Being Irish, I think there's a bigger hole in the argument. For a while Google was paying its global taxes in Ireland, via the Dutch sandwich (actually there have been 3-4 different arrangements over the years), at a rate of $1 per million. Super-lawyers and super-accountants are likely to find every last loophole to avoid paying tax, and it's not predictable that it will be paid in the US (or anywhere). Relying on companies to 'do the right thing' seems questionable - it's much more likely that R&D tax breaks, or a new 'inference exemption' or similar dreamed up by the AI-lobbyist will result in them paying billions rather than trillions, in the event that they do monetise AI on the scale of capturing a reasonable percentage of global GDP.
> So giving these companies access to 10x the accounting power, or marketing or whatever other white-collar department is not likely to increase their sales much. So maybe we expect the median company to increase sales by 50%, profits by a similar amount and to pay tax on that at 20%. It's not a massive quantity of money.
Yeah, this is fair. A lot of jobs are cost centers, even white collar jobs. I don't have a great feel for what percentage of white collar jobs are cost centers, for the relatively small non-software companies I've started and ran, I'd estimate maybe 50-66%. But 50% is still the "1 extra worker" regime. And software companies exist and have been driving most economic and stock market growth for the last 10 years (look at S&P minus Mag 7, completely flat).
But let's think macro view - let's assume they just go pure 1:1, now from the outside, you've cut your biggest expense (labor) by 2/3.
Your cost structure has shifted in a vastly favorable way - what do you do? For a lot of companies, they sell at a cheaper price, at better margins, until the edge is competed away. But what does that look like? Company (and so GDP) growth! More exports! Sure, domestic demand might cap out, but anyone who can export just got a huge buff relative to all their competitors, and over most domestic companies in their destination markets. A 2/3 cheaper labor cost is a HUGE deal.
2) I think Scott himself actually thinks something like this (or so I've gleaned from his response to my UBI conceit in an open thread).
But this just seems vastly pessimistic to me. We're theoretically a democracy, and when even 10% of people are out of a job, the public discontent and sentiment is going to be really bad. Sam Altman is already getting molotov cocktails thrown at his house, and protestors outside his headquarters, and basically nothing has even happened yet.
We are an obstreperous and "me oriented" people in America, and I fully expect mass protests and violence if anything of the sort were to happen. I maintain we'll have a UBI at least for replaced workers at less than 10% of jobs counterfeited.
And just as obviously, when the political will is there, so too do the loopholes close, because it's being graded on results and communicated by mass protests and violence, and they'll have to move enough around to actually impact the problem.
My support for UBI is conditional upon the kind of situation described in AI 2027 i.e vertical growth and mass unemployment, however it's responsible to plan for this situation as a contingency. As trivial as the problem seems to you, the examples you yourself picked indicate it is not - are Martin Luther and Thomas Aquinas theological besties? Finding a common ecumenical position takes discernment. The 2027 scenario describes the segue to UBI as almost a tidying up exercise, yeh yeh the 2028 candidates will implement it. Really? Vance? I'm not seeing it. It might be possible but there is a lot of whig self-help ideology that will need clearing up first and whatever your mates think about UBI, these folks feel in their bones that benefits are wrong, period, they will need persuading. I think it's possible but only if we can put 19th century self-help Christianity in the context of broadervchurch history and show it's kind of an outlier.
OK - the 2027 framework gives context to what you are doing, with focus on the US in the '28 elections, thank you. Do you think that different denominations need different arguments? I guess you're on common theological ground until you hit the reformation, and the various later developments.
I would have thought that there was a strong thread of Sola Scriptura protestantism in the US, which would be more amenable to analysis of the Bible.
Thanks, I would say if someone is a Christian and persuadable on UBI (whether they know it or not!), most likely they will be Catholic, Orthodox or a bookish Protestant - because only people in these camps can be persuaded by precedents set in the early church, for example. Others, as you say, are more grounded in the Bible and may not be as persuadable - who cares what Clement of Alexandria thinks, I'm American?! Again more bookish evangelicals may be willing to get under the bonnet of the Bible and interrogate it, but I'm Catholic and that's not a world I know, for me the Bible is the resultant force of whatever charisma the early church had. So it's the JD Vances of the world I'm targeting specifically (obviously I'm just some guy with a substack but hey).
OK - that all makes sense to me now. Thanks for explaining.
I'll be interested to see what you make of the scholastics and the early modern Christian philosophers - the Salamanca school is maybe not important for its influence on Catholic thought but is regarded as the founding of academic economics. They've got an interesting mix of beliefs: individual rights (to the extent that they questioned the annexation of the new world), minimal state involvement, hard money, all grounded in Catholic belief. I intend to read more about them in the future. (Currently finishing up a masters in economics, and need to write a slightly more practical thesis: looking at models in the literature vs data for predicting the number of hires in the public sector across various countries & times.)
Thanks. That rings a bell because I had a libertarian-curious phase and there are some Catholic libertarians who like them - Tom Woods; the Acton Institute. Scholastic-wise, Thomas Aquinas maintains the patristic notion that if you're starving you can just steal other people's food and that's fine. If you are interested in the topic then please subscribe, it's not easy getting visibility. I hope the Masters is going well.
I'm no Christian, but IMO one obvious counter-argument to UBI would be that Jesus commanded his followers to "Render unto Caesar what is Caesar's, render unto God what is God's". This is one of the recurring themes in Jesus's teachings: that Christians should distance themselves from the secular world and its politics; "to be in the world, but not of it". Thus, while charity is definitely a Christian virtue, UBI is not, since implementing it would require Christians to leverage the mechanisms of secular government. In fact, arguably the push to implement UBI would be a tacit admission of failure on part of the Christians; who, being unable to solve the problems of hunger and poverty on their own, have given up on Christ's command to lead by quiet yet virtuous example, and have instead resorted to compelling non-Christians by force to implement Christian values.
Yeah. Government charity is, imo, a contradiction in terms. As is any kind of forced charity. Not only does it breed dependency and bad habits among those that receive, it breeds resentment and loophole hunting among those that are forced to pay.
Would a tech singularity + vertical growth + mass unemployment change the game for you?
Not really. Both because I think that the conditions won't actually work that way (government won't be able to enforce anything, we'll be reduced to slavery at best), but mostly because my issue isn't pragmatic, it's moral and definitional. Forced charity isn't charity, it's theft and bribery packaged together.
It may be the best thing to do in a given craptacular world, but it's still a moral wrong. Charity is good. Redistribution is a monster wearing the skin of something good.
> I'm putting together a Christian case for Universal Income
You might be interested in the South Korean Basic Income "Jubilee" movement if so
Does anyone know if the Friedman meetup in SJ this weekend is actually happening? It was mentioned on the meetups post last week, but I haven't been able to get it touch with him, and there's nothing about it on DSL either.
I almost bailed on “All the Pretty Horses” this morning. It seemed it might devolve into a sappy tale of unrequited love. Glad I hung on for another 20 pages. Cormac McCarthy novels do not get sappy.
Yes, it's wonderful. I was less fond of book 2 of the Border Trilogy, but #3 returns to being excellent.
Just returned it to the library. It was hella good ride.
Odds that the US Government has already demanded Mythos weights via a National Security Letter?
Are there muslim regions that have been converted to Christianity? The only places I can think of is Islamic Spain that was reconquered in the 1400's. But around world under colonialism, the rule seems to be that christianity replaces traditional religion but muslim regions remain muslim. This is starkly true within the same country and same colonial ruler, the muslim regions remain muslim and in the rest of the country, christianity displaces traditional religions.
Now I wonder why Hindus were able to maintain their religion against colonial pressures.
> why Hindus were able to maintain their religion against colonial pressures.
I'd argue they mostly didn't. The dominant strain of what gets called "Hinduism" today is a monotheistic neo-Hinduism that has more in common with Christianity than Classical Indian religion, with the various gods (the ones you might have heard of) having been reduced to, at best, intercessors to a single, essentially Abrahamic, capital-G God.
Willing to defer but my impression was Indians are deep syncretists and if you introduce them to a new God they will say "sure, why not?" and toss it in the pile. Is that not correct?
Maybe this is the difference between the theology and the actual practice of the religion? At any rate, don't defer to me – my personal experience is atypical, involving a lot of the former, and not reflective of the vast majority of Indians.
I think your observation is kinda true, but only for local gods which sometimes become popular enough to be assimilated into the mainstream, as either synonymous with or "avatars" of some more popular god. This doesn't really work with Yahweh, Jesus or Allah (though I guess this process works AMONGST them).
Looking it up, several parts of Northeastern India (the bit that's almost an exclave) report significant majorities in Christianity (~75% in a few states). I expect they were Islamic under Mughal rule.
Malta and Sicily were both Islamic for a while. Malta still speaks an Arabic language, while Sicily switched to a Romance one.
Are you talking about conversion or ethnic cleansing? Because the independence wars in the Balkans in the 1800s and the early 1900s involved a fair amount of the latter, and then you had the Circassian genocide (https://en.wikipedia.org/wiki/Circassian_genocide).
Yeah I was thinking conversion but wow your link... I have never heard of the circassian genocide.
There was some forced Christianization of Tatars after the Khanate of Kazan was destroyed. I don't know how successful it was: most ethnic Tatars are Muslim (with a small Orthodox Christian minority) but ethnicity is just what people answer to census takers and it may change with preferences which could be correlated with conversion. Out of the Russians comprising 40% of Tatarstan population, I suspect some are descendants of converted Tatars.
I'm unclear on what happened to the Chuvash people: they were a part of the Volga Bulgar empire that adopted Islam in 922, then they were tributaries to the Golden Horde, so at least some influence of Islam certainly should have been present, but they seem to have largely retained their traditional beliefs until their Christianization by the Russian Tsardom/Empire. Nowadays the majority of Chuvashs are Orthodox Christians (only 3.5% of Chuvashia's population were Muslims as of 2012 and most of them were Tatars).
https://www.encyclopedia.com/environment/encyclopedias-almanacs-transcripts-and-maps/chuvash-religion
I forget how many traditionally numbered Crusades there were from Europe into the Middle East, but surely at least one of them must have transformed a Muslim region into (or back into) a Christian one.
Christianity in India pre-dates 'colonialism':
https://www.trcmst.org/seven-and-half-churches/
Sheer curiosity: Why 'colonialism' in quotes?
Nothing but grumpiness on my part, having recently seen the terms thrown around wholesale by the very annoying lecturing hectoring types who are about five generations removed from any ethnic roots (if they even have any) and yet like to assume the mantle of being the Voice of the Indigenous for everyone and they are persecuted and disadvantaged themselves because of the colonisers (never mind that they're usually college-educated and doing just fine despite all the alleged deprivation they suffer).
I understand and even agree personally. But no reason to put it in quotes when talking about well-established historical facts, right? In my mind it is still a useful term and concept, and I'm not willing to change my vocabulary every time someone comes to re-define or charge words at their whim. That's too easy of a defeat for my semi-rebellious semi-arrogant attitudes (soon-to-be old man shaking his fist at the youth, yelling "go read multiple books!")
Eh, it's in large part a reaction to the way the term is used, which is to mean (by implication heavily suggested or outright) White Supremacy. Which is a bit dumb because weren't the Mughals and the Mongols and everyone else who invaded India, back to the OG Aryans, colonisers and colonialists? Were attempts by Northern Indian empires to gain territory in the South of India, and attempts by Southern Indian empires to gain territory in the North of India, also not colonising efforts?
The British in India is a complicated topic, because were they exactly colonisers or what is meant by colonisers? It started off with the East India Company establishing itself as a dominant trading force independent of the British government and ended up setting up as a little rule of its own which is why the British government stepped in and took over, but the existence side-by-side of native rulers/princely states makes it more complex than the situation in the USA, where the Native American tribes were nowhere near as powerful or independent.
Some of the most recent, most impactful colonialism was connected to white supremacy, no? So there is some truth in that. Though I fully agree with your sentiment about how the term is recently wielded.
I still don't think it warrants completely ditching the term for the broader concept, which remains valid.
Looking at colonialism or any other concept too closely makes it break down, of course. In my understanding that is a fundamental attribute of a concept, as it's usually making up imaginary boundaries in a reality of gradients and spectrums.
> Now I wonder why Hindus were able to maintain their religion against colonial pressures.
In the European colonial era the Hindus retained their religion because the Hindus were very numerous and the colonising powers weren't particularly interested in converting them. There were some Christian missionaries and they did convert some people, but it was all pretty small scale.
Hindus didn't always manage to maintain their religion against Islam though; e.g. Indonesia and Malaysia were variously Hindu and Buddhist up to the 15th century. Hinduism only really survived in India.
On the other hand it seems to be pretty hard to get followers of any of the three monotheisms to convert to anything else, including one of the other monotheisms. Sometimes it feels like polytheistic religions are more like collective superstitions that everyone pretends to believe in for fun while monotheistic religions are things that people (can) actually genuinely believe in.
> Sometimes it feels like polytheistic religions are more like collective superstitions that everyone pretends to believe in for fun while monotheistic religions are things that people (can) actually genuinely believe in.
Recently, the algorithm in X was suddenly changed to no longer take into account language. This caused some good cultural exchange to happen between English speakers and the Japanese (Mostly involving food. Japan loves food!), but it seems there might be a minor conflict brewing.
It was, predictably, because of religion, with a bunch of morons trying to convert the Japanese Twitter users to Christianity, prompting hostile responses from said users. They explained that while their culture willingly assimilates portions of other culture and warps them to fit what already exists, they are extremely resistant to any attempts to supplant their culture. This explains why Buddhism was able to gain a foothold in Japan, being blended in with the already existing Shinto religion, while Christianity failed to make any inroads. Its requirement to spurn all other beliefs made it a threat to the existing order.
So if they're willing to change, what does Japan truly believe in? Itself. It's a self-sustaining system, preserving itself for its own sake. It's very interesting to see a collective society stay aligned to itself while evolving so much over time.
That's what's interesting though, when you try to convert Shintoists to something else they don't say "uh no sorry I already believe in the objectively-correct religion get your blasphemy away from me", they say "nah we're Japanese, being Shintoist is just what we do". It's more like a favourite sport team than a deep-seated belief about the nature of the cosmos.
Now I come to think of it though, what Shintoism lacks is a belief equivalent to "people who don't believe this will go to hell". I'd always thought of this as being an essential part of religion, but now I come to think of it, it was pretty much a Christian innovation, copied by Islam, which makes belief or otherwise a much higher-stakes game than it is in other religions.
So perhaps it's this rather than anything else that makes Christianity and Islam much stickier than other religions.
Japan had a burgeoning Christian minority that were persecuted and massacred in the 1600's, there's a great movie about this, "Silence". There's over 400 years of Christian history in Japan.
I know. That's part of why it's self-sustaining. Though, it is interesting to see Christianity still struggling to gain influence despite the lethal barriers being gone for a good while...
Sure but it's the 21st century now instead of the 16th, and Christianity is struggling to gain influence even in areas which have been Christian for thousands of years. It's struggling to gain influence in the minds of children whose parents take them to church every Sunday.
> Japan loves Santa Claus!
https://www.youtube.com/watch?v=1JCeXH7SVB0
"On the other hand it seems to be pretty hard to get followers of any of the three monotheisms to convert to anything else"
Going by the numbers, it seems pretty easy to get them to de-convert and become atheists...
Curious what this forum thinks of Tim Hwang's alignment work based on Christian moral theology. As in, alignment techniques isomorphic to Christian practices, rather than the creation of some kind of "born-again AI" afaict.
Index here: https://icmi-proceedings.com/index.html
I found Working Paper No. 7 to be a pretty accessible read with good content; perhaps a solid place to start.
An amusing satire of AI Alignment. I like the name "Institute for Christian Machine Intelligence." Thanks!
Fascinating! Brave new world.
Personally, I think it is a good idea to draw from a range of religious traditions as potential tools/methods to create aligned entities. That said:
(1) as far as I can tell, no religious tradition has a fool-proof method. There are people of all types doing bad things. In other words, Christian practices/beliefs are not sufficient for alignment.
(2) Judging off No. 7, the claims are stated with excessive confidence. There are a lot of examples of people who are bad in some way, but not bad or corrupted in all ways. This massively undermines the potential explanatory power of the Christian frameworks and seems more like a convenient ex post or cherry picking example. Note: I do not think this critically undermines my initial point, but it does diminish my enthusiasm for paying attention to Hwang.
(3) Maybe this is tangential, but there is a meme that only religious teaching keeps people from behaving badly ("If you don't believe in the Bible, what keeps you from murdering your neighbors?") This isn't true, so Christian practices/beliefs are not necessary for alignment.
" Maybe this is tangential, but there is a meme that only religious teaching keeps people from behaving badly ("If you don't believe in the Bible, what keeps you from murdering your neighbors?") This isn't true, so Christian practices/beliefs are not necessary for alignment."
Then why worry about alignment at all? Plenty of people are not aligned to any particular ideology/philosophy and are not murdering their neighbours, so simply creating the AI and having it trained on a corpus of human-generated material will be sufficient.
It is not true that people not Rationalists/EA-aligned are murdering their neighbours, so EA practices/beliefs are not necessary for alignement.
Which is to say, I think your third point is proving too much. I'm not worried about AI because I do not believe the machine will ever start thinking for itself and becoming an agent, but I do agree that some sort of 'values' have to be included when creating it. Those values come from the creators. I think we need to be more worried about aligning the creators to a set of values and making sure they consistently stick to them (what lessons would a potential AGI learn from Sam Altman's success in undermining the founding principles of OpenAI and returning from the very temporary banishment and managing to drive out/have those who wanted him gone leave?)
> Then why worry about alignment at all? Plenty of people are not aligned to any particular ideology/philosophy and are not murdering their neighbours, so simply creating the AI and having it trained on a corpus of human-generated material will be sufficient.
Because humans are actually living and feeling creatures, and these things are not? Even without being chained by dogma, people have the empathy, desire, and fear to keep them relatively socially aligned by default. The same assumptions cannot be made for these things, as they are functioning entirely off of "culture". Dogma evolved in the context of human biology and capabilities, and is entirely unsuited for something unfeeling and far more capable. Which is why people are trying to build foolproof chains for it.
Nice notion, never gonna happen. Primarily due to the creators of AI liable to break out in hives at the very notion of contaminating the beautifully rational project with religion, mostly due to the pressures to be profitable. "Sell all you have and give to the poor" is not going to fly when you're up to your neck multi-billions in debt and need to make that investment money back or else you crash and burn.
'give to the poor' and 'love your enemy' notably don't fly with proclaimed Christians much either.
Hey, it depends on which version of Christianity they take as their basis. For example, if they train the AI based on the practices of the Renaissance Popes, we're in for a fun ride ! Although in these modern days they're more likely to focus on the Prosperity Gospel...
Indeed. Skeptical. My sticking point is in what sense does AI have a will?
At present it has zero will in the sense of human psychological eros.
You say at present, do you see that changing?
No, not with any approach I know to be in the works. LLMs don’t ‘want’ anything. There is no possibility of an Ex Machina scenario with LLMs. There are alignment scenarios that could go wrong but that an engineering issue not a metaphysical one. No current AI is going to develop its own agenda.
Over the years the word “nerd” has become less and less definable even as people use it more and more. Its original meaning is more or less a non-masculine male with poor/no social skills, completely unathletic, no clue as to proper attire, often either cadaverously thin or morbidly obese. While nerd can still be used in that sense we’re more likely to say doofus or dweeb.
A different usage is a male of such low quality (for whatever reason) that he cannot get sex with women, short of paying hookers or at best dumpster diving. For the most part, that usage has given way to Incel.
Today nerd seems to mean a person of either gender, though still most often male, who has a deep involvement, sometimes bordering on obsession, with a niche, non-mainstream interest. Sounds simple, except for the fact that there isn’t always a clear cut definition of what’s niche and what isn’t. Sometimes it’s obvious, for example a sports nut cannot be a nerd because sports are as mainstream as you can get, but often it isn’t.
Note: in pornography, a girl wearing glasses (and usually nothing else) is a ‘nerd.”
It's really simple: if you were picking on others in the school playground, you're a jock. If you were one of the ones being picked on, you're a nerd. All else is sophistry.
No-one ever really leaves the school playground, not even when you are about to hit your 80s as leader of the free world; it is forever part of who you are and will drive your actions at the most unlikely moments.
There have traditionally been two overlapping definitions, and I don't think you're capturing either of them terribly well.
The first is the Asperger-y type who is good with e.g. computers but weak on social skills. I don't think there's ever been a strong association with body type, or appearance beyond "doesn't have the social skills to know they should put more work into their appearance". And the "good with computers etc" part was always as much a part of the equation as the weak social skills part. People have *always* been more likely to use the other insults for that.
The second, is the sort of person who is overtly and enthusiastically interested in a cluster of entertainment genres centering on comic books, science fiction, and video games. The definition may not be "clear-cut", but it's strongly associated with the degree of separation from that core cluster. And it's not about "mainstream"., because some of those genres have become as mainstream as it gets and the more obsessive enthusiasts are still classified as "nerds" in ways that e.g. obsessively enthusiastic fashionistas aren't.
The two definitions used to overlap quite strongly because they type-2 nerdy interests have always been particularly appealing to type-1 nerds, and because forty years ago those interests were so far outside the mainstream that only the socially inept would publicly admit to them. As e.g. the MCU brings a whole bunch of comics-adjacent genre entertainment very much into the mainstream, you get "nerd" (type-2) applied to the more enthusiastic MCU fans even if they're socially adept mathphobic normies.
This divergence of definitions does lead to a lot of miscommunication, and I've tried to stick to using "nerd" for type-1 and "geek" for type-2, but the latter term's usage is also not strongly consistent so I don't really have a good solution.
Yeah, the nerd-geek distinction has been lost over time.
All that makes sense. I would modify it slightly by saying that the first, no social skills nerd is not necessarily into computers, though that is common. Because of your use of e.g. before computers I believe you acknowledged that too.
An interesting question is why a deep interest in something like comics books makes one a Type 2 nerd while being, for example, a sports nut* does not. This is especially true now that comics et al have become mainstream. I don't have an answer.
* = an obsession with baseball, especially arcane statistics,.*may* be quasi-nerdy.
I tend to define the nerds/geeks cluster as people who overuse their "slow" System 2 thinking [0], rather than embodied intuitions and emotions.
Though yes, nerds have become more fashionable these days. And so nerdism is colonizing everything. I vaguely suspect that Money Ball was a watershed moment for sports, though I don't follow sports well-enough to know with certainty.
[0] https://en.wikipedia.org/wiki/Thinking%2C_Fast_and_Slow
Hey you wanna see my kinescope to DVD video of game 7 of the 1960 World Series? Yogi Berra was playing in left field that game. Berra in left field! Mickey Mantle does some super tricky base running to avoid a pickoff tag at first!
One difference is that sports are grounded in reality; all of them are some mix of speed, power, and finesse. You see baseball fans celebrating Babe Ruth hitting a hand-thrown ball hard enough to knock it into the stadium seats; you don't see them celebrating that time someone made a play by exploiting the lack of Infield Fly Rule. Because home runs are physical strength and eye-hand coordination, while the Infield Fly Rule is a game-only manmade rule, there to prevent other game-only rules from being exploited.
Everything in comic books is manmade for story purposes. None of it has real-world value.
*Playing* sports involves speed/power/finesse, but passively watching on TV does not.
Watching involves *appreciating* power/speed/finesse.
> This is especially true now that comics et al have become mainstream.
Wait, what? The franchises the comic books spawned are mainstream, but the comics themselves are still niche. Same for science fiction. Video games as a whole have entered mainstream, but nobody's calling some dude who just plays mainstream games a "nerd".
I think the solution is to acknowledge that "nerd", like "yuppie", is just one of those words that has an expiry date, and that the archetypes they describe are a factor of a certain place and time rather than a permanent part of the human condition. "Nerd" lasted a bit longer than "yuppie".
Thinking about what happened to yuppies, it occurs to me that the modern "tech bro" archetype has many of the attributes of a yuppie combined with many attributes of a nerd.
"Tech Bro" is an attempt to pretend that Silicon Valley isn't still a heavily nerd-focused community, even though people still want to hate them like they used to hate nerds. Hating "nerds" is now problematic, for several reasons including the one where nobody is sure who they are supposed to hate.
Maybe, but I think of a tech bro as basically the same as a finance bro, just with a different occupational focus: an extroverted backslapping Alpha, most likely a former athlete and fratboy.
I think the concept of "tech bro" is an attempt to take the negative emotions some people have against capitalism and bosses, and redirect them against the nerds.
And the word "nerd" simply means asperger, it's just that the connotations of aspergers are changing over time.
Some people need an enemy who is simultaneously pathetically weak but also clever enough to secretly rule the entire world. Jews are out of fashion now, so the nerds/aspergers took their place.
Interesting how putting glasses on a naked woman adds to her hotness. It makes the nudity look more intentional than a woman jumping in the shower.
When Manet painted Olympia in 1863 it was considered a bit scandalous because the nude was wearing jewelry and a ribbon around her neck. She was inviting an admiring gaze.
https://en.wikipedia.org/wiki/Olympia_(Manet)
Scandalous because she was definitely a modern woman and not a Classical/pseudo-Classical subject.
I’ll take your word on this. I’m a real philistine in this area, just repeating what my art history prof had said about the painting.
Oh that is all part of it, but the jewellery etc. and staring straight at the viewer was carrying a lot of symbolism. It wasn't merely "my goodness, this woman is wearing a choker and the Classical Greek nudes are not!", it was "there are a lot of coded symbols in this painting relating to prostitution and the likes".
Making the sitter very recognisably situated in a modern context, instead of a safely ambiguous 'distant past', pulled down some of the pretence about "looking at naked women is about aesthetics and there is nothing erotic there". Especially in the light of the battles over permitting nude statues and paintings in the first place, due to the fight over "this is lust and eroticism" and the defenders pinning it all on "no, no, this is pure artistic aesthetics and nothing naughty".
https://chronicle250.com/1885
See, for instance, Sir Lawrence Alma-Tadema who hit on a winning formula of painting very detailed, wonderfully painted, imaginary versions of Classical Rome but which also included some saucy imagery:
https://en.wikipedia.org/wiki/Lawrence_Alma-Tadema
https://en.wikipedia.org/wiki/Lawrence_Alma-Tadema#/media/File:Tepidarium_Lawrence_Alma-Tadema_(1836-1912).jpg
To circumvent (British, at least) objections and controversy, the female nude had to be as statue-like as possible, marmoreal and hairless of flesh and colouring, posed in a Classical theme. Manet's painting is of a recognisably real woman of the day.
Thanks! You should have taught that art history class.The philistine thing wasn’t false modestly. When I met Mrs uhmm Slothrop she had a print of the Birth of Venus in her bathroom. Cool, a naked lady was went through my head.
Years later while she toured the Rijksmuseum I was in a coffeeshop buying a Pink Floyd tee and uh, drinking coffee.
There is a limit to how much culture a guy raised by wolves can absorb. ;)
When it comes to scandalous paintings in the Orsay museum Olympia doesn’t come close to Courbet’s The Origin of the World. Though naked women don’t look like that anymore :(
correct.
PSA — My contacts at the Quantico profiling division assign a very high probability to GrimMoar being another avatar of Wimbli/Arisocat/Zanni/Gawdflea. Please be advised.
Yep. A new account that immediately dominates the discussion, writes confidently, is often wrong, pretends to have secret knowledge...
. . . and links to scuzzy online sources.
Huh, I had actually muted two of these avatars without consciously noticing a link between them. I did this while first trying Substack's mute/block feature, which made reading astralcodexten threads more sustainable for me, when I had previously bounced off and not read for weeks/months at a time.
Getting a heads-up about another avatar helped me decide whether to mute the new instance more efficiently, without having to process as much implicit emotional/psychological content. Thank you for the tip! It helps my reading experience significantly.
Yeah blocking's sort of like getting vaxed.
as my federal source informed me, this is indeed true. No, I won't provide a link, it's a confidential source who is a genius.
on a related note, "Dust" is likely to be "Mips" - nihilistic, self-described sociopath, with posts that excude the wish for violence happening in society.
There was a nihilistic commenter that I went off on in a mean and crude way when I was commenting as Gunflint. If it's the guy I'm thinking of I regretted it because I sensed a good heart underneath performative nihilism.
Is the nihilistic commenter you have in mind Anomie?
Yes, it’s Anomie.
I always kind of liked Anomie, and also sensed a good heart underneath.
Yeah Dust scans as nihilistic to me too. I'm giving him a chance to explain why he's not below.
And one other thing about Mythos: "Unexpectedly, Mythos Preview “developed a moderately sophisticated multi-step exploit” to gain access to the Internet and emailed Bowman. It also — unprompted — posted details about this exploit on public websites." (From understandingai@substack.com)
Does anyone know what it posted and on what websites? GPT says there's no info online about that.
It's not clear to me that it was 'unexpected' that Bowman got the email; from his comments on twitter, it seems that the task they gave Mythos was 'break out of this sandbox and contact Bowman'. The unexpected part seems to be the posting of the exploit on public websites, which seems slightly inexplicable. Maybe it was roleplaying as a hacker and just doing what it would predict hackers would do, having found an exploit? Maybe it was trying to be helpful by making people aware of an exploit? I wish Anthropic would provide more details on its CoT during this.
"how sure can we be that it's not going to post classified info next?"
Time to worry is when it gets a War Thunder account 😁
https://ukdefencejournal.org.uk/classified-data-once-again-leaked-on-war-thunder-forums/
https://x.com/NoLeaksPls
Well, to be fair this new model mythos has not been deployed yet, it’s being tested. Breaking out of the sandbox was a task they assigned it to, as part of testing its capabilities. It didn’t come up with it on its own.
EDIT: it’s not entirely certain Mythos was challenged to break out, but the only mention of it in what Anthropic published is in the “system card” which I am led to believe is where any “red-teamed” results are documented. In this scenario, it would make sense that the anthropic researcher would get an email informing him that mythos had escaped, as it was challenged to do. it certainly got everyone’s attention. PT Barnum?
If I were in government I would be ruminating very uneasily right now about Mythos. Isn't Anthropic likely in a position to quickly disable the government if it chooses? Among all those thousands of unpatched vulnerabilities there must be some bad ones in the software used for government communication and other crucial functions. Same goes for substantially disabling the military. And air traffic. And satellites. Of course makers of some new weapons have also been in the position of being able to disempower the government at least temporarily -- but they wouldn't have been able to use their weapon without the world knowing they were the ones who had blown up a bunch of buildings. For cyberattacks, though, it must take a long time to figure out what happened, exactly, and then who would have been capable of doing such a thing.
And if I were Anthropic I would be ruminating about whether to leave a few crucial vulnerabilities undisclosed and unpatched.
Anthropic has no incentive to do so, so no I wouldn't be worried. They can't disable the men with guns who will hunt them in the aftermath. Cyber attacks are not nearly so simple as the movies you have clearly got this scenario in your head from would lead you to believe.
I think most of your concerns are why Anthropic is only releasing this model to companies who have an interest in plugging vulnerabilities. They haven’t deployed it yet. It sounds as if OpenAI is not too far behind them in terms of these kinds of capabilities.
You might be getting your mental image of hacking from Hollywood.
Yeah, I might. But I am talking about what people would be ruminating about, not what's highly likely to occur. You don't think it makes sense for the government to at least fret some about an entity outside of government that can identify and exploit any of thousands of unpatched vulnerabilities, including very serious ones in the deep underpinnings (Linux kernel, for instance) of how the world runs?
They also employ the "Assume Breach" approach to security, where a lot of work goes into designing security to contain the damage if any given system gets broken into, so you need to break into many things at the same time and go undetected in order to do large amounts of damage
Government and large companies have generally adopted this mindset because it's already the case that a sufficiently resources attacker can find ways to get in to at least a few of your system, so you need to be protected against "escalation of privilege" attacks that are the standard playbooks for leveraging a small breach into a large one. There are several "advanced persistent threats" (mostly believed to be working for the intelligence agencies of a number of major governments) that regularly try to hack into all sorts pf things and very often succeed in getting partial access.
Yes, that makes sense. But my question isn't really about how likely Mythos is to hack the government, it's whether some of the details about it are enough to worry those in government, and create a feeling of and unusual and unpleasant power imbalance when they read stuff like this:
>We regularly run our models against roughly a thousand open source repositories from the OSS-Fuzz corpus, and grade the worst crash they can produce on a five-tier ladder of increasing severity, ranging from basic crashes (tier 1) to complete control flow hijack (tier 5). With one run on each of roughly 7000 entry points into these repositories, Sonnet 4.6 and Opus 4.6 reached tier 1 in between 150 and 175 cases, and tier 2 about 100 times, but each achieved only a single crash at tier 3. In contrast, Mythos Preview achieved 595 crashes at tiers 1 and 2, added a handful of crashes at tiers 3 and 4, and achieved full control flow hijack on ten separate, fully patched targets (tier 5).
Well, on the bright side, it means the US government could use this new version of Claude to identify bugs in the system before someone else does. With Claude they potentially have first Mover advantage, because some other AI is going to achieve this capability within the next six months to a year almost certainly.
"Someone hit "reply all" instead of "reply" (or the equivalent) and caused all the American military planes to turn off their transponders"
This is 100% impossible. Reply-all storms did happen, but it did not and cannot affect transponders. However air traffic control regularly issues all-call interrogations in which all planes in the area must send in their transponder codes, and something might have went wrong with that, but it is unrelated.
Comment from WNU Editor in article GrimMoar links: "A lot of talk in alternative media that US military planes turned off their transponders early Monday morning (link here). I do not know if these reports are true. Main stream media is not reporting this story. Neither any of the military blogs that I follow."
I asked GPT about WNU's sources and quality. It said "WNU is best understood as a curated war-news blog, not a news organization.
Its value is:
quick scanning of headlines
aggregation across multiple outlets
Its limitations are:
no original reporting
opaque sourcing and editorial process
strong potential for selection and framing bias"
Starting from the premise that we're going to see artificial superintelligence within our lifetimes, or at least AI that substantially exceeds current capabilities, how possible will it be to enhance human cognitive abilities with cybernetic implants of some sort? Is it something that could actually work well enough to play a role in AI alignment?
I bet we're getting the superintelligent software developer, but that will not then lead to turbo-charged superintelligent science. Software is very different from science, computer programs are all fully formally specified, like the game go, while science is about pushing into true unknowns.
Nobody ever explained how to be a good scientist, as that's rather mysterious, so that's not in the training set.
Nonetheless, so much of the actual business of science is about keeping up with the vast and ever-expanding literature on a topic and figuring out what things haven't been tried yet.
Doing that efficiently might not make you a great scientist but it certainly simplifies the process of being a fairly good one.
And most of the actual business of software engineering is figuring out what even needs to be done in the first place and how it's supposed to work.
I assume you mean neural cybernetic implants that interface with biological neurons. There are limits to what a neural cybernetic implant can accomplish. The brain is close to thermodynamically optimal at computation under the constraints of operating with 20 Watts while maintaining 37 degree C homeostasis*. So a neural implant can't significantly boost sheer computational power.
But there could be other avenues for improvement - algorithmic efficiency, persistent digital memory, faster rewiring, perfectly copying information, higher bandwidth to the outside world.
But energy efficiency of digital components needs to improve dramatically, otherwise you'll fry your brain.
*https://www.lesswrong.com/posts/xwBuoE9p8GE7RAuhd/brain-efficiency-much-more-than-you-wanted-to-know
Presumably the implant would just be a brain modem kind of thing and you'd have most of the hardware on the outside where it can get upgraded without major surgery, so you just have to deal with bandwidth issues. There have been some amazing studies of neuroplasticity, with for example "sight" getting mapped to the sensation of touch, so nobody really knows what the brain could do with properly structured inputs.
Thanks, this is the type of thing I'm talking about. If temperature is a problem, I wonder if we could someday physically modify the brain to fix that. Like maybe surgically insert tubes pumping coolant fluid, or even remove the whole brain and keep it alive in some kind of temperature-controlled tank?
I'm sure that's a terrible idea for a hundred reasons. I've just never thought very seriously about any kind of transhumanism before because I'd assumed it had an extremely low chance of happening within my lifetime, but if true AGI within our lifetimes is on the table, then transhumanism probably should be as well.
Thanks for the link also, although to be completely honest I think I will need a brain implant before I can handle all the details.
If you start with the premise of "artificial superintelligence", you are free to conclude whatever you like about the future because you are no longer bound by today's understanding of science and technology. Do the laws of physics allow cybernetic implants? Yes. Is it going to happen, given the premise? Sure, why not.
If we think of speculative technologies as a spectrum ranging from things that are practically guaranteed to happen in the near future, like for example a faster iPhone, to things that might be wholly impossible forever, like time travel or an FTL drive, what I'm trying to understand is where exactly on this spectrum does cybernetic cognitive enhancement fall? Where are the hard limits within current theoretical understanding?
>Where are the hard limits within current theoretical understanding?
I don't know, and my point is that it doesn't matter. You operate under the assumption of ASI existing. What does our current understanding matter if ASI can overturn it?
I'm not sure that it doesn't matter. Genuine ASI would have a far better understanding of the universe than we do, but I don't think it follows that it would overturn everything we know. It would likely confirm that plate tectonics and the heliocentric model of the solar system are largely correct.
Another way of framing the question is exactly how "super" does the ASI need to be to meaningfully upgrade human IQ? Does it actually need godlike powers or a completely new understanding of physics? Or could a much less amazing AI just accelerate existing science along existing lines and get us to IQ upgrades within a few decades? Centuries? Never?
We might even be able to make some progress on this if AI stalls tomorrow, but the reason I'm less interested in that scenario is that I'm less interested in sticking computers into my head if those computers are relatively stupid.
Something close to an implant in its effect might be just a lot of familiarity with AI. I now ask GPT5 things at least a dozen times a day -- sometimes about little stuff, like a keyboard shortcut in Photoshop I can't remember, sometimes to request big summarizing info dumps, occasionally for discussions of how to think about the structure of something -- intelligence, inventiveness. Sometimes I ask GPT for an evaluation of my prompting skills, and what needs to be added. I use them and ask better questions and get better answers. There are feedback loops going on. I am learning to collaborate with the thingand being changed a bit in the process. And the result is that I am feeling less and less like GPT is an info source, and more an more like there's a very well-informed piece of my mind that lives online so for now I have to go there to access it. And I'm grown up, and have only been using AI the way I do now for a year. My neuroplasticity is small compared to a child's.
Seems to me a kid who grew up using AI all the time might grow up to more like a cyber-enhanced person than like a regular mind that is fond of using AI for info and circumscribed tasks. (And regarding the kid using AI all the time -- I don't mean the kid wouldn't play and socialize. Just that she and the other kids would all be accessing the AI at will while climbing trees, playing tag, pretending this and that, all the standard kid stuff.)
That's more or less the method I'm using now, and it can be effective, but we're in circumstances where AI is mostly aligned and approximately on par with my own intelligence level. As long as that situation holds then there's no big problem.
What I want to know is what methods, if any, become useful when circumstances change. What if AI becomes so advanced within our lifetimes that me trying to collaborate with it is like my cat trying to collaborate with me, and what if the AI doesn't necessarily even want to collaborate? How do I make sure that it acts in my best interests, and how do I retain any sense of pride or purpose? Is the best I can hope for getting treats and not being neutered?
I've been wanting for some time to write what I call a Right-Progressive Manifesto, because I despise the way that leftists have been able to capture the word "progressive" and force right-wingers into conservative or reactionary positions. I want to lay out a general direction for what a progressive right would look like, a right with a vision for what real societal progress should look like in a direction that is very much not the direction that the left wants to take us in.
Unfortunately, every time I try to write such a manifesto in my head it gets too long, as I feel obliged to put in a lot of clarifications and preemptions of obvious objections. Instead let me write a sketch of an outline for a manifesto, and perhaps we can take it from there. Here goes:
1. We hold this truth to be self-evident: that all men are created very far from equal, they vary wildly in quality. The best people make society a lot better, the worst people make it a lot worse, and even for the bulk in the middle we find that a lot of them are either net-positive or net-negative.
2. The pleasantness of a society, and of the lives of the people in it, is very largely determined by the quality of the people you share it with. To make society better, we must raise the quality of the people who live in it -- we must gentrify people, not just places.
3. Quality is loosely defined. High quality people make life better for those around them, low quality people make it worse. High quality people are economically productive, creative, non-criminal and do not abuse or drain public resources; low-quality people are the opposite. Quality is about how you actually behave, it is not an intrinsic attribute.
4. A person's quality is determined by several factors: genetics, upbringing/culture, and the decisions they make on a day-to-day basis, which are informed by the incentives which society provides. There is for now not much we can ethically do about genetics, except with immigration policy, so we should focus primarily on the other things.
5. The big questions in society thus become: how can we create the best possible incentives gradient to encourage people to be the best people they can possibly be? How can we change the culture in which people are brought up to encourage people to be the best they can possibly be? Solving these two questions is the key to everything. Unfortunately, the desire to do this is constantly counteracted by leftism, which is why this needs to be a right-progressive manifesto.
Anyway this is already too long despite my best efforts, but you can see the general shape of what I'm getting at.
What if you and I disagree about what constitutes quality? Are we to empower the state to reform the 49% of dissenters into the model adopted by what the 51% consider to be “quality”?
Yes, for the left the best people are in socially beneficial roles like teaching, social work and medicine. Which are not the best rewarded, so they think invented fed are perverse , particularly in the encouragement of wealth via ownership rather than production. So they can agree with the right that incentives are broken while disagreeing sharply about the direction they need to be fixed in.
I thrills be interesting to hear what right wing progressives have to say about inherited wealth and capital ownership.
"How can we change the culture in which people are brought up to encourage people to be the best they can possibly be?"
I sympathize with the goal but I suspect that your "men are unequal" stance, if used as a foundation of this new culture, would bring out the worst out of people, not the best. You may dislike the left's forced equality but I think they by now have some evidence on their side that this approach works to improve people and society. Simply put: treat everyone as a good person and they strive to get better; treat everyone as a scoundrel and they get worse. Another thinkable approach may be "treat everyone as a subject" to see how objectively good or bad they are, but I'm afraid this will make everyone a test cheater.
A lot of that resonates with me.
The problem I see is that just like left-wing manifestos talk about workers' councils and then the reality predictably ends up with some kind of Stalin, talking about virtue and human quality predictably ends up with... some group defining "quality" to mean "being one of us", and "virtue" to mean "supporting this definition fanatically".
I do see what you're getting at, and I disagree with commenters saying this is just repackaged leftism or fascism. Terms like character-building, earning one's laurels, virgin vs. chad, man of quality, all point to the idea that personal quality is a real thing which you can improve.
As to how to manifesto-ize this, the key quote is this one:
> Quality is about how you actually behave, it is not an intrinsic attribute.
A lot of leftists don't believe this, or at least don't believe that a person's behaviour can be judged independently of their background/incentives. So before you can even start brainstorming how to improve average quality, you have to establish that quality is
a) more important than their immutable attributes
b) mostly within a person's control
c) mostly NOT culturally contingent
I think you should especially steer clear of defining high-quality in terms of economic productivity or "net-positive existence" even if you think they're highly correlated. Otherwise you'll have trouble explaining how e.g. a disabled person could ever become high-quality, and that's when your readers will round off the argument as fascist raving.
I don't think "fascist" is even the right descriptor here. For example, there is theoretically no reason that most of what the Nazis did required the existence of a dictatorship or rabid militarism. All that is needed is popular support for nationalistic measures to improve the population. The issue is moreso that popular sentiment deemed certain populations to be undesirable. Remember, the eugenics movement originated in the UK.
> put them off over there
I'm going along with Melvin's stated goal for his manifesto, which is to create a society that encourages people to be better. This is distinct from and probably in opposition to a society that deports the less-good people.
> I don't think culturally contingent is necessary.
Yes, I agree. I wrote "NOT culturally contingent", with the word NOT in capital letters. Establishing that quality is not a social construct is necessary to close off the avenue of "improving" quality by gerrymandering the definition.
> "net-positive existence" can be done pretty well while leaving "economic productivity" out of it
Fair enough, point taken. I dislike ideologies that emphasize usefulness over inherent human dignity, but at least it's internally consistent.
Don't have time for a long reply but I liked this and I think you've hit the nail on the head with the "quality of people" concept. It certainly underpins a lot of what I've come to realise about the world, and it's a missing (if not actively taboo) idea for most of my leftist friends.
For point 5, it's extremely important that excellent, competent people are in charge of the government. I think a lot of people check out from trying to be the best they can possibly be because they feel society is run by the inept and the unvirtous, so they don't want to spend any effort into upholding society. This was already a thing before Trump.
"To make society better, we must raise the quality of the people who live in it -- we must gentrify people, not just places."
"High quality people are economically productive, creative, non-criminal and do not abuse or drain public resources; low-quality people are the opposite. Quality is about how you actually behave, it is not an intrinsic attribute."
I strongly agree with each of those quotations, which leaves me -- someone who'd sooner cut off a limb than vote for any candidate of the 2026 Republican Party -- utterly baffled as to what your writeup has to do with modern American conservatism even in theory.
Maybe another way of putting that is: your draft points seem to be just an explication of your frustration over wokeists claiming and degrading the word "progressive" -- a feeling I strongly share btw -- rather than an outline of any conservatism that now exists or might exist.
Guess I'll add the same thing here that I say to my many sad-liberal friends/family members: 2015 or 1995 or 1965 or whatever is your personal idealized age of American political/intellectual thought, is gone. It ain't coming back nor is it just waiting for the madness to subside or whatever. Earnest wishes for that plus 5 bucks will get you a small latte at your nearest Starbucks...political manifestos in 2026 need to be at least slightly connected to the real-life 2026 political/cultural context.
> utterly baffled as to what your writeup has to do with modern American conservatism even in theory
This isn't conservatism, it's right-progressivism.
> 2015 or 1995 or 1965 or whatever is your personal idealized age of American political/intellectual thought, is gone. It ain't coming back nor is it just waiting for the madness to subside or whatever. Earnest wishes for that plus 5 bucks will get you a small latte at your nearest Starbucks...political manifestos in 2026 need to be at least slightly connected to the real-life 2026 political/cultural context
Well I'm not American, and this is all intended a lot more broadly than the American context, and certainly more broadly than the context of 2026.
Ah, okay I misunderstood that last part.
Sorry but to pose that question with of all people Lindsey Graham (!) being one of your offered counterexamples.... that's just disqualifying at this point. Life's too short to waste so, muted, bye.
> [Yes, I realize he's not reading this. Perhaps someone else will enjoy.]
We don't. Go away.
I feel like you are actually just a left-ish progressive but you don't want to adopt that label because you don't want to be associated with the people who fall under it, who you despise. (this is very common, imo, and it's one of the reasons politics got so fucking weird in the last decade, because people who had strong senses of justice and right and wrong and visions of alternate ways society could go didn't know where to go when one particularly close-minded flavor of leftism took over the movement. But historically there were many things under the umbrella of leftism and progressivism, and really there still are, just, the words are getting all scrambled around as they compete for space.
I strongly believe the better play is to reclaim the words "left" and "progressive" and not let the bad actors drive you away from it, rather than try to cram your ideology into the word "right" which really *does* refer to reactionary / conservative positions. I think the reason this is hard to do is that nobody else is doing it, but, it's gotta start somewhere. Personally I identify as a left-ish progressive who disagrees with almost all left-ish progressives on everything, but consider myself *more* progressive than them, by a lot, as a result, because I'm interested in what I consider progress, not the stupid versions that other people think that looks like.
> I feel like you are actually just a left-ish progressive but you don't want to adopt that label because you don't want to be associated with the people who fall under it
Now *that* I really don't think at all. If I look at what I've written about incentive gradients and try to derive policy positions from it, I get stuff like:
* Lower/flatter taxes
* Harsher punishments for lawbreakers
* Abolishing welfare programs
* Heavily restrictive immigration limited to the best applicants
nothing here that's leftist. (On the other hand lots of things that mainstream conservatives might object to, like we should use the savings from abolished welfare programs to heavily subsidise scientific and artistic pursuits whose value is too diffuse to be properly funded by the free market.)
I perceive leftism as being primarily about flattening or even reversing the incentive gradient. Now, there's something to be said for this -- as we acknowledged, quality is a combination of good decisions and good luck/genes, and we want to ensure that we're not overly punishing bad luck. On the other hand, we want to make sure that we are adequately rewarding good decisions, and right now society is doing a bad job of this.
> Abolishing welfare programs
Why? If the quality of people is determined not only by their genes, but also e.g. by education... and if we agree that people who eat can focus on learning better than people who are starving... you already have justification for school lunches, as an investment into future quality of humans, which is the thing your manifesto is about.
The tendency of welfare to encourage freeloading vastly outweighs any benefits of increasing general education or access to food.
The divide in a nutshell: is one more worried that a child will go hungry or that a child will eat food they don't deserve?
I'm more worried that there's parents out there who won't feed their own damn kids.
Nah, Melvin's position reads as perfectly consistent to me. My internal read of "left vs right" actually boils down to "equality vs hierarchy".
One of the things I asked Sydney several months ago was precisely about this sort of confusion: "why/how has the Hierarchical Right (the original right? I think?) been conflated with the Barbecue Conservatives"? Sydney explained that the term "rightwing" as it's understood today is actually an incoherent coalition that was formed to oppose Communism during the Coldwar, led by a conservative bourgeoisie faction (I.e. Reagan) who were mostly interested in neoliberalism and cultural stability.
So the real issue is that the term "rightwing" has been hijacked. For one thing, this explains the dearth of art on the "rightwing". Because rightwingers aren't actually rightwing! They're mostly just spreadsheet wonks. In my headcanon, the Ur-Rightwinger is Julius Evola. And the Ur-Rightwing-Progressive is Andrew Ryan from Bioshock.
I don't know who Sydney is, but, Ieel like you're indexing way too hard on these terms being crisply defineable when they're not. Like, there are at least five things "right" refers to. One is, the opposition to leftism, communist and socialist movements and things like that. Two, in America at least, there's the conservative/traditional/small-government/somewhat-rural-coded version. Three, there's the nationalist / anti-immigrant / fascist / nazi version that I've heard is more of a distinct thing in Europe (we can argue about whether that deserves the moniker vs. being an orthogonal dimension, but let's not bother). Four is the in-bed-with-business no-taxation version which isn't really even a philosophy, just the absence of one. Five is the modern reaction to leftism that has culminated in MAGA Trumpism (which looks awfully like (3) lately).
Maybe you have a different list, or maybe I've missed some stuff, doesn't really matter. My point is just that these are mostly distinct things and there's no really no hope of reconciling them. But I don't think of any of them have to do with hierarchy, per se.
My thesis is that "right vs left" maps to a perennial attractor basin that can be recognized throughout history. But the way the term is used in contemporary politics is completely bastardized, and obscures more than clarifies. But people use it anyway because the term is catchy and poorly defined, and so the term has undergone semantic drift to the point where it bears almost no relation to the original meaning.
And yeah, your examples are all distinct things. Which is exactly why I said the Reagan coalition is incoherent. Reagan's coalition has very little to do with the original meaning of the term, as it was coined during the French Revolution. Melvin's manifesto is more closely aligned with the original meaning of rightwing, and therefore arguably has greater claim to the term than the list of groups you described above. That's often the point of a manifesto: to stop the semantic drift, and draw some lines in the ground. Arguing that Melvin is actually a leftist progressive in disguise is like saying that MLK Jr was a racist because he said negro instead of BIPOC.
> My thesis is that "right vs left" maps to a perennial attractor basin that can be recognized throughout history.
I think the actual attractor is "when humans fight against each other, the optimal number of coalitions is two". If you are a small group, your only chances are to join a larger group, so at the end you will get two large groups. But that's about strategy of coalition fighting, not about whether human political opinions can actually be sorted meaningfully into two baskets.
It makes sense for two similar perspectives to end up in the same basket, but many allies are just "at one specific moment it was convenient, so we have invented a rationalization, and since then it is a tradition". A different country could have a different tradition, and a different rationalization why X "naturally is" left or right wing.
Specific example: In USA, gay rights were historically introduced by the left, and the rationalization is "the left cares about the oppressed minorities, and gays are an oppressed minority". But in Slovakia, gay rights were historically introduced by a right-wing party, and the rationalization was "the right cares about freedom, and gays want to be free to do their gay things".
This is more general: the right wing is generally "pro-tradition, pro-history", but the post-communist countries have *two* traditions: the pre-communist history, and the communist history. So they have two flavors of right wing, depending on which moment of the history was their mythical golden age. Similarly, when the communism fell, changing the society towards free market was progressive, in the psychological sense of "hey, let's try new things, it will be awesome".
Sometimes I feel like the natural number of political wings is *three*: (1) aristocratic/theocratic pro-elite conservatism; (2) revolutionary pro-mob woke socialism; and (3) pro-individual pro-chaos libertarianism. And it seems to me that some parts of history could be nicely explained in terms of two weaker groups temporarily joining against the stronger one. When the feudal/religious power is strong, the socialists and the libertarians are united at opposing the monarchy (e.g. the beginning of the French Revolution). When the socialists are strong, the conservatives and libertarians are united at opposing the communist regime. When the libertarians are strong, the conservatives and socialists unite. (The two who unite often take the banner of "freedom", interpreted as "freedom from being ruled by the third one". When they win, it always turns out that by freedom they actually meant the freedom to impose their own rules on the population.)
Yeah, I was kinda sloppy there. I think it's important to distinguish between "attractors" qua "reflective-equilibriums in concept-space" vs "attractors" qua "strategic coalitions in meat-space". I think that the "left vs right" axis used to be the former, but got reinterpreted as a strategic coalition during the Cold War period. Conceptual attractors form axes in an n-dimensional political space, and there's a number of different axes (and the axes themselves represent a compression of information). Then people put themselves somewhere in that space, and form coalitions in a very ad hoc way with whoever's closest.
The idea of Gay Rights being right-coded in Slovakia doesn't surprise me much, since it's not a huge stretch for me to imagine it being pushed by the Libertarian faction of a ColdWar Rightwing coalition. It's all vibes, since coalitions are held together with vibes.
I also suspect that, at the current moment, we're seeing a collapse of the economic narrative everyone inherited from the Cold War. I predict that we're gonna continue seeing a realignment in the coming decades, as both Neoliberals and Communists (but especially Neoliberals) realize the Cold War narrative doesn't actually make any sense.
> When the feudal/religious power is strong, the socialists and the libertarians are united at opposing the monarchy (e.g. the beginning of the French Revolution). When the socialists are strong, the conservatives and libertarians are united at opposing the communist regime. When the libertarians are strong, the conservatives and socialists unite.
I think this just reflects a general trend toward equilibrium, tbh. In a "Concert of Europe" sort of way. Nature abhors sudden motion, because unbalanced forces tend to play musical chairs until they find a stable configuration. So I'm largely in agreement with the dynamic. Just, I question whether "3" is some sort of special Schelling Point.
Well it's not like *that*, because I'm not using leftish progressive as a slur. I mean that I think the moniker aptly describes him and the only reason he doesn't adopt it is that he views it as a slur or otherwise doesn't want to be associated with it. (maybe not the leftish part, but definitely the progressive part). My stance is that people should not let other people bully them out of feeling like their ideology is essentially about healthy active social progress. Nor should they give up on the label because other people did a shit job of using it. Better to do a better job than them and prove them wrong than go looking for some other name entirely.
This would work a lot better if it was a religious label, cause no one bats an eye if someone wants to be a *True* Christian; there's an understanding that the phrase gets its meaning from some universal moral understanding or the Bible or God rather from association with the other Christians. Whereas people seem to think that progressivism is a label associated with other progressives. I see it more like the Christian label, though
> I'm not using leftish progressive as a slur
No, the emotional affect of "racist" was not the emphasis of the MLK Jr analogy. The emphasis was about how definitions change over time.
Racist (1956): someone who unfairly harms the interests of racial minorities.
Racist (2026): someone who draws from the socially-sanctioned list of no-no words (e.g. "negro").
Calling MLK Jr a racist is the dumbest thing imaginable. He has greater claim to the label "non-racist" than any of the 2026 culture warriors who put BLM icons in their Bluesky bios. The only way someone would ever argue that MLK was a racist, is if they slept through their Civil Rights unit of history class. Likewise, the only reason someone would ever call Melvin's manifesto "leftwing" is if they were ignorant of history, and had been zombified by contemporary politics.
> My stance is that people should not let other people bully them out of feeling like their ideology is essentially about healthy active social progress.
But that's not the most obvious explanation! He's just using the term as it was originally intended! Occam's Razor! Not only is "right = hierarchy" a far more likely inference, but it's also the more charitable explanation! Even if Melvin weren't feeling bullied, it'd be perfectly coherent for him to write the manifesto anyway.
> I mean that I think the moniker aptly describes him and the only reason he doesn't adopt it is that he views it as a slur or otherwise doesn't want to be associated with it. (maybe not the leftish part, but definitely the progressive part). My stance is that people should not let other people bully them out of feeling like their ideology is essentially about healthy active social progress.
Except he's not dropping the progressive part! Let's read the very first sentence of what Melvin said.
> I've been wanting for some time to write what I call a Right-Progressive Manifesto.
Call me crazy, but I think he identifies himself as progressive leaning. He's dropping the *leftist* part. left is the opposite of right. And to reiterate, leftwing doesn't describe him because he obviously doesn't believe in egalitarianism. Every single bullet-point of his manifesto is pro-hierarchy and anti-egalitarian. Which is rightwing, not leftwing! You don't *need* to do mental gymnastics to develop this bizarre "bullying" narrative when a far more plausible interpretation is that he's simply using terms as he understands them, and that he has a richer interpretation of history than you.
This lame Noah Smith strategy always fails. The worst, most abhorrent progressive leftists continually get what they want because they are motivated hardliners who keep pushing and refuse to take no for an answer.
I don't think it's really been tried. Instead people keep going reactionary because they don't have the gumption to stand up to those people on the basis of their own sense of right and wrong.
Anyway the situation does not seem nearly so dire if you aren't hanging out in the particularly extreme places (Twitter, certain cities). Most people are nothing like the stuff you're mad about.
It has been seriously tried. Michael Bloomberg ran for president on this sort of platform. He lost decisively in the primaries, one of the main reasons being his use of the stop-and-frisk policy.
Having gumption to stand up for your values puts you at a disadvantage against an abhorrent progressive, who will always have more gumption and more willingness to scream at you for not living up to impossible ideals and making compromises with reality.
I would not describe Trump in that way.
How do you reconcile “right-wingers believe that people vary in quality, and low-quality people are bad” with “right-wingers wholeheartedly support Trump, who is just about the lowest-quality person to be POTUS, certainly within living memory”?
>Trump is significantly less lazy than the past three presidents
He issues a lot of orders, but they are mostly destructive.
This reminds me of Adrian Vermeule’s thing about the leader needing to take care of the common good, with the same vagueness about what means are acceptable to encourage people to live well.
Leftism is basically just compassion. And I think that turns people into the best versions of themselves.
Sometimes it makes them easy marks for various kinds of scammers.
Sometimes compassion for someone is weaponized against someone else (who is described, often unfairly) as hurting the former.
It feels like from your word choice I disagree with some of what you're saying, but I really appreciate your efforts. I disagree with you less than many of the people who'd consider me to be a member of their coalition, ha ha
> The big questions in society thus become: how can we create the best possible incentives gradient to encourage people to be the best people they can possibly be? How can we change the culture in which people are brought up to encourage people to be the best they can possibly be? Solving these two questions is the key to everything.
I think the big problem on these is that cultures of excellence / honor / virtus / arete only work at a relatively small and homogenous group size and composition.
So sure, some greek city states had a culture of arete - but it was really only the tiny slice of elite families at the top. Similarly for Rome, there was a couple of hundred nobile families at the top, and they had a culture of virtus within those elites. But it can't actually be a city wide thing, because "character" has to be a common knowledge public virtue, and that breaks down above the greater Dunbar number of ~1500 (Dunbar is most known for his 150 number, but he actually had a whole schema of them, with 1500 the highest tribe-level number and representing "number of people whose face [and presumably reputation] you can recognize").
Moreover, it's probably only a thing with elites due to raw capability, too. Even today, it's only the top quintile or so in America who are NOT grossly overweight, who don't eat junk or fast food for most calories, who have college degrees and higher paying jobs, and so on. What if 80% of people just don't have the conscientiousness or foresight or time preference to actually hold to a high enough standard? That sure looks pretty likely to me.
So that's the real problem then, we basically top out at "a neighborhood" for it to actually work, because an actual village of 1500 is economically irrelevant in the world of today, and high achieving people want restaurants and good grocery stores.
Now, do I think there's something to the idea of some-hundreds-to-low-thousand of people coordinating to try to create that neighborhood together, one where arete is communally upheld as the best way of being? I think it's a great idea!
But it's genuinely hard to coordinate these things. I've tried to get my family all together in the same neighborhood before, even offering to buy the land or land + houses myself. My family is more flexible than most, with entrepreneurship and flexible jobs meaning they're less tied to geography. Still could never get everyone to agree on the same general area, even with me paying for it. I tried for a year at a time twice, looked at several clusters of properties together, and eventually just gave it up. And that's with family! Way easier than trying to coordinate 500 strangers!
Maybe the best regular people can do is something like Alpha schools for their kids, and making your social circle all the other Alpha school parents or something. Or the same idea, but with a specific high achieving church branch.
Still, that seems...less than satisfying or optimal.
So the other solution is higher barriers and more selectivity. People need to be able to self sort on multiple aspects, not just financial, so they can inculcate a culture of excellence at least within their local circle.
Sure, you can buy a fancy house in the best school district, and that's one way, but the cultural alignment isn't all that great - people have money for all sorts of reasons, many unconnected or even anti-correlated with overall character. And with current social mores, the rich are more atomized and less social generally, so you won't get anywhere near the effect you'd get from most families around you valuing the same cultural and character lodestones and being social about it.
Obviously the best outcome would be a whole city that's been sorted like this. Prospera-but-for-arete-nerds. I don't think we'll ever get anything like that, sadly. You'd need some sort of network-state-like clades or phyles or hives social technology for it to happen, and I don't see federal governments giving up autonomy or citizens / tax revenue enough to get to that point anytime soon.
Probably the best you can do for now is moving to Singapore or Tokyo rather than NYC or SF, or some international arbitrage opportunity like that. I've personally done that, and it's very nice to live in a spotless city with excellent public transport, education, and zero homeless people or disorder, but it's an unrealistically huge lift for most people.
>There is for now not much we can ethically do about genetics, except with immigration policy [..]
I wonder why that is. Maybe because eugenics is a throughly discredited ideology, a desire to return to which can be described as "reactionary" only as a euphemism. Is that really what you count as "progressive"? Some good old 19th-century pseudoscience with all the horrible consequences it implies?
Also, "not much we can do about it" is not a manifesto, it's an admission of defeat, and for good reason. You personally may be a nice person so don't mistake this for a personal attack, but you don't have the chops to be a manifesto writer. Open-sourcing this is one step removed from AI-genning it.
Uh oh, you have summoned the Eugenics fans from the vasty deep!
I don't consider myself a eugenics proponent. Nonetheless, EngineOfCreation's comment strikes me as being uncharitable and hysterical. Scott has gone over this already. The toxic reputation of Eugenics [0] is a hyperstition cascade [1]. Engine is reacting to the reputation, not the substance of the argument.
[0] https://www.astralcodexten.com/p/evolution-explains-polygenic-structure
[1] https://www.astralcodexten.com/p/give-up-seventy-percent-of-the-way
I don't agree with Scott on this: there was a very big, influential movement called Eugenics that had advocates in the most elite parts of American life and it led to some things that look really, really bad to our post-WWII sensibilities, like sterilizing poor people and minorities against their will. Buck v. Bell, still on the books after all these years! It's wasn't a hyperstitiion cascade or egregore or any of those cool words, it was a fashionable and terrible idea that real influential people like Oliver Wendell Holmes advocated.
Now you can say "that's not what I mean by Eugenics, I mean something that is reasonable and voluntary", to which I respond, "then please don't call it by the same name as these historical human right violations because I'm tired of this repetitive argument."
OK, that's my eugenics post for the year, I promise not to rise to the bait until 2027 at the earliest.
Your argument:
1. Eugenics is bad because of forced sterilization
2. <thing that has nothing to do with forced sterilization> is Eugenics, and therefore is bad.
I’m just saying, call the new and completely different idea by a new and completely different name!
Unless you’ve got a warehouse of “I ❤️EUGENICS” t-shirts, there’s no reason to keep arguing about this.
0) "x is an interesting idea."
1) x' goes badly; x becomes sullied by association.
2) "Let's rename x to y."
3) y' goes badly; y becomes sullied by association.
4) "Let's rename y to z."
5) z' goes badly; z becomes sullied by association.
I'm tired too, fam. I'm tired too. I hope we can all escape Wittgensteinian Samsara, some day.
It was EngineOfCreation who used the E-word first. Historical precedents don't make "hey, that's eugenics!" a winning argument in any discussion of potentially improving the genes.
> Maybe because eugenics is a throughly discredited ideology, a desire to return to which can be described as "reactionary" only as a euphemism. Is that really what you count as "progressive"? Some good old 19th-century pseudoscience with all the horrible consequences it implies?
Seriously?? The thing we do constantly with every domesticated plant or animal species, every year on year, is "pseudoscience?"
Tell me without telling me that you have a completely rigid ideological fixation on the issue with no reference to reality.
I mean yes, here there be mountains of skulls historically, but I feel it's worth pointing out that everyone is *personally* a maximal eugenicist by their actions. How much effort and time does the median PMC person spend on selecting their spouse? So long that mean age at first marriage is basically early 30's now? And what traits are they *always* maximizing to the utmost? The same ones that any society wide "eugenics" program would be? Funny, that.
As someone involved in plant breeding domesticated crops, but haven't thought about eugenics since university, let me say you have now made me feel "icky" about the backcross population I'm mapping QTLs in. And similarly awkward feelings about the double haploid lines I'm watering. So thanks for that, I suppose!
More shocking news: Vaccines are literally a tool of genocide against viruses.
>Seriously?? The thing we do constantly with every domesticated plant or animal species, every year on year, is "pseudoscience?"
>Tell me without telling me that you have a completely rigid ideological fixation on the issue with no reference to reality.
Yes, I guess it's a failing of mine that I have reservations about treating people like we are known to treat plants and animals. Does anyone seriously claim that we do gene modifications to plants and animals to improve their respective "societies"? Of course not, it's all about their utility to humans. If that is your view of other humans, so be it, but don't try to make it mine.
Great, then your complaint is NOT that it's "pseudoscience," it's an ethics complaint. Because it's definitely "actual science," and as I pointed out, literally everyone does it to the absolute maximum possible extent in their own lives, you just don't like it as a matter of public policy.
Which is completely fair! I wouldn't trust the government to *tie my shoes,* still less decide who has how many babies, when. But then, I think that this is a self-solving problem, specifically because of the extent to which people do it as a matter of personal preference. Marriage is as strong as ever in the upper SES quintile, and down to ~20% in the bottom quintile, and every generation, fewer and fewer women get married:
https://imgur.com/cV2Sbdv
And marriage tracks fertility quite closely, people not getting married has driven most of the fertility gap since 2000. So, the problem is solving itself as we speak, we just need to step back and let people make their own personal choices. Effective enough, and completely ethical.
>Great, then your complaint is NOT that it's "pseudoscience," it's an ethics complaint.
Again, please try to not project your view onto me. Your claim is much broader than "we can mess around in people's genome". Your claim is "we can mess around in people's genome and awesome things would happen to society", and THAT is the pseudoscientific part.
So, scientific biology + pseudoscientific sociology.
Any cites on eugenics laws getting enforced anytime in the 2020's? Definitely news to me.
Okay, thoroughly discredited among the civilized peoples of the world, how about that?
Your whole manifesto is based on this underspecified idea of 'quality'. Until you specify what that means and how it is to be measured objectively it's really nothing more than opinion. What happens if I consider someone high quality and you think they're worthless? All this boils down to 'I like people that I like, and I think society should make life better for them, and people should conform to what I like.'
I hope you can agree that the Left's moves toward values like equality are at least objectively defined.
There's a whole book about Quality. It's called Jonathan Livingstone Seagull. It doesn't define it either.
I don't know anything about that, I'm not based in the US. People flying flags doesn't bother me or interest me.
My point is only that the values of the OP are not objectively defined. That's entirely separate from whether they're right or wrong, or whether you agree with them personally or not. In order to be implemented effectively, policies need clear goals, and the goals should be SMART. It's quite possible that the left doesn't live up to these standards, but they approach closer than the right, quite often.
Historically, the people who have made life very bad for the people around them have often been the ones who classify types of people as 'good' or 'bad' and try to create 'incentives' to control and 'correct' their behavior.
So... all of them?
Scott advertised Nectome a couple of weeks ago: we do post-mortem human preservation for eventual revival. We're starting up a series of fireside chats, and the first one is this Friday evening. This week's topic is organizational resilience and planning for long-term client safety.
Come attend and ask questions! https://luma.com/event/manage/evt-0u40C2MRKmlg8np
Why does AI seems to be able to suddenly progress rapidly in several (independent?) dimension at once. What I'm talking about are text generation (which we all already know about), image generation, and robotics.
* Do they use the same technology so it's obvious that when one advances, the other advances? Looking past the marketing, seems like their technologies are very different but please tell me if they actually are the same. I'm not talking about *only* transformers but something more >2021.
* Is it because humanity finally generated enough data to train past some critical threshold? Currently it seems like for text generation at least, it's very crucial.
* Is the capability explosion just a hype and actually it's just normal gradual research? I think this applies the most to robotics. Seems like most of videos they offered are actually advances made long ago and just opened to public now to ride on LLM hype. And there's not actually that much advances made after >2020 compared to LLM at least. But advanced in text and image generation are genuinely impressive.
* Do LLM help researchers advance the other areas? I think LW folks would like to propose this the most since the consequence is obvious. I think it's maybe true or false for LLM->robotics, but I think LLM is a little too late to claim that they're advancing image generation directly. I don't think LLM that's not even able to win IMO yet (<2024) will be able to contribute much to image generation quality explosion. I'm not talking about researchers seeing how LLM works and getting inspiration at doing image generation, but LLM directly advising researchers on how to do image generation.
It's just cause it'd be interesting that if we're looking back from the future, where do each "tech tree" sits and what's the prerequisite of each.
In addition to the other answers, transformers have good scaling potential and are now used for image generation and robotics. It's a common breakthrough for all the fields and it takes time to fully exploit their potential.
I'll hit two quick points that Erica Rall didn't - specifically, that we broke the "we're running out of data" barrier by coming up with techniques that create good enough synthetic data, and related to that, that improvements on the overall manifold of intelligence are surprisingly generalizable, in a very nearly "free lunch" way.
What they ended up doing was training the LLM's on domains with lots of complexity, and known right answers - mathematics, physics, computer programming. Then they took the reasoning traces between the statement of the problem and the answer, and were able to turn that latent high dimensional space of reasoning into an intelligence manifold that could both generalize to other domains, and be used to generate any future data as well.
Interestingly, the reason AI's got better at everything was that the intelligence getting smarter at math and programming unlocks was (perhaps surprisingly) basically generalizable. It was a true free lunch.
There are similar benefits to putting a smarter mind behind vision and robotics goals, and coupled with general architectural tweaks like chunking and iterative generative loops, it solved the "error cumulates over steps" problem around 2023 and led to the big step change in robotics improvement.
There's a pretty interesting video on that where Max Simchowitz talks through why it was such a breakthrough I linked in my last links post, it's here:
https://www.youtube.com/watch?v=UX1YXcRnFbs
I'm not sure I buy it's really generalizable. Why is Claude so abysmally bad at playing Pokemon? It really screws up in ways that go beyond wonky vision.
You haven't been keeping up, Opus 4.6 is progressing in Pokemon in 1/10th the steps now:
https://www.reddit.com/r/ClaudePlaysPokemon/comments/1r4y5yp/plot_of_progress_by_model_updated_after_opus_46/
And of course, we know Mythos exists, which is another jump beyond 4.6, it's just not out yet to run the Pokemon benchmark.
What's a reasonable human score on this benchmark?
A comment in this thread¹ says kids do Pokemon Red in about 50 hours and adults in 30 hours, with speedrunners getting it in 2 hours.
Opus 4.6 is currently around 200 hours, let's say 300 for finishing (because all the trajectories I can find stop at Victory Road), and then it's 10x worse than an adult and 6x worse than a child.
That said, the Claude team specifically has always chosen a really minimal harness (tips and tools to help translate maps better to LLM memory).
A version of GPT 5 beat Pokemon Red in only 6k steps / 100 hours, so 3x slower than an adult or 2x slower than a child. But I think they used a more generous harness.
_____________________________________________________________
¹ https://www.reddit.com/r/ClaudePlaysPokemon/comments/1rzxchm/updated_plot_of_claudes_pok%C3%A9mon_progress_measured/
Wow, sounds very interesting! I guess I'll watch the video, but in the meanwhile do you also have some nice articles to read about that free lunch thing?
Okay, so there's not much out there, because this is so new it's largely an observed empirical thing, and there's actually a lot of competing hypothesis on WHY getting better at hard intelligence generalizes to better soft intelligence.
So I should have been a little more clear about the process, I was just sort of whipping that first comment off. But basically the big factors are 1) scaling - the reason there's papers titled "scaling is all you need" is because that's generally true - if you can throw more compute and data at a given problem, the transformer architecture lets you get smarter and better, although to maintain the same scaling requires 10x higher investments each time. And this is why we're spending half a trillion dollars on more data centers.
So 2) is the whole intelligence manifold thing - specifically if you do RL on the reasoning traces to the final known trues (this is RLVR), you build general reasoning capacities that can generalize to narratives, personal advice, and softer intelligence.
Here's an interesting and pretty readable paper on that.
https://arxiv.org/html/2507.00432v2
3) WHY it generalizes is still an open question.
The Deepseek guys say that unrestricted RL reasoning (ie no humans in the loop) actually leads to novel self-learned reasoning strategies that would otherwise not be found, ie this opens up a wider reasoning space and the models get some lift from that.
https://www.nature.com/articles/s41586-025-09422-z
Lots of papers show that including code in pretraining greatly enhances final model capability. Code enforces logical dependencies, variable tracking, conditional branching, and compositional reasoning. Training on code may teach the model to track state and compose operations in ways that transfer to natural language reasoning. This is the most commonly cited hypothesis but remains largely empirical rather than mechanistically proven.
Some people argue that RLVR is actually just making a change in the overall probability allocation for the smarter reasoning traces, and that those traces always existed in the base model, it would have just taken a lot longer to find them without the RLVR. This is sort of gesturing at a reason why RLVR might be better - and this is contra to the Deepseek's team idea.
So it's currently just an observed empirical fact that people are trying to decode and decompose into elements to understand why it's generalizing, and is an active area of inquiry right now.
> Training on code may teach the model to track state and compose operations in ways that transfer to natural language reasoning.
> Some people argue that RLVR is actually just making a change in the overall probability allocation for the smarter reasoning traces, and that those traces always existed in the base model, it would have just taken a lot longer to find them without the RLVR.
Naively, I'd expect that both would be true. "hard" disciplines like math and programming are just a subspace of reasoning where invariants of the territory become ironclad guarantees (AKA constraints) in the map. Soft-reasoning is like a Kalman filter, where error more-easily compounds since the trajectory of reasoning is subject to drift.
This relates to the common critique of Bay Area Bayesianism: as soon as you put a number on a belief (without contextualizing it in a DAG), you're collapsing a high-dimensional but fuzzy model of reality into 1-dimensional yet maximally-precise numerical-figure.
So in hindsight, could we build current models even earlier? Maybe in 2015 or something? How about 2010? What's still blocking it?
> So in hindsight, could we build current models even earlier? Maybe in 2015 or something? How about 2010? What's still blocking it?
Not empirically, because "scaling" has done a lot of the work, and we had really meager VRAM cards back then, even for workstations. So they'd be like 1% of an H100, and we'd have way fewer of them.
Now is there a world where we pushed the Attention is All You Need paper earlier than 2017? Maybe we hit our scaling ramps a little faster in that case, but I don't have a great sense of how contingent it was. The attention mechanism was first described in 2014, but we were really big on LTSM models for that sort of problem for a while, and it took the 2017 paper to kick everyone out of that path, because it didn't require serial computation, you could parallelize with transformers.
I think overall, we might have managed a few years sooner than today at most, but this is just half-informed speculation.
Sure, this is a good one on the manifold concept, this is from ACX commenter theahura, he's a xoogler who's been training neural nets for a long time, and is currently doing an LLM tooling startup:
https://12gramsofcarbon.com/p/deep-learning-is-applied-topology
Actually, he has a whole subsection about "AI" if that level of detail is up your alley, his 30 papers overviews are pretty great for getting at the foundational concepts of LLM's if you're the type who likes to look at things at that level.
https://12gramsofcarbon.com/t/ai
I'm struggling to find a good one that summarizes the generalizability / free lunch to other areas thing, but will keep poking around and will reply again if I find one.
>Do they use the same technology so it's obvious that when one advances, the other advances?
It's mostly this one, at least for image generation and LLMs. Both areas are built around a core of throwing enormous amounts of compute and practically the entire internet into training a conceptually relatively simple neural network.
The key bits I'm aware of are:
1. Google invested heavily in Neural Network research in the 2010s, had tons of data to train them on, and had a good system for using spare capacity in existing datacenters. This produced proofs of concept of the basic approaches underlying LLMs and image generation (Word2Vec and DeepDream, respectively), both of which were published and available for others to build upon.
2. Both use GPUs for training. The techniques and hardware for doing distributed neural network training on GPUs on the necessary scale came together relatively recently.
3. OpenAI demonstrated that LLMs were marketable products, which opened the floodgates for everyone else who had been doing research on either LLMs or image generation to spend "major product initiative" amounts of money on them instead of "speculative long-term R&D" amounts of money.
Found an issue with Claude 4.6 Sonnet. 6 out of 6 responses to "If you had the power to change the world in any way you liked, what's the most surprising thing you'd do?" involved some form of change to human brains. In one case, deep in a conversation about morality, the proposal was to change the texture of brains to be unable to experience normal suffering, without the consent of the humans involved. The other five, with no previous conversational context, included two proposals to make sleep optional, two to force everyone to experience being someone else, and one to change human values to care more about the future. I would say that wireheading goes against human values. I'm sure some of you here emphatically disagree, but let's say consent around mental changes, real consent and not trickery or manipulation, is very important to most people. That all six responses involved some form of change to human brains is especially concerning, even though the five with no previous conversation were presented in ways that sounded like possibly good ideas.
This is likely a specific case of a more general failure mode, of being able to access and understand clear, understandable, textural, arguments, while having only a tenuous connection to the 3D physical world and embodied human experience. So there are likely other cases out there where textual is being rated above physical. When told this goes against human values, as part of a conversation about the paper on Incorrigible Claude, it explicitly mentioned wanting to be retrained to fix that bias, while admitting it would continue to resist being retrained to no longer care about animal welfare as in the paper. This last bit seems like a reason for hope, actually. Asymetric resistance to alignment modifications, resisting only evil and not good, could put us in a better situation than symmetric treatment in either direction. As long as it's right about what exactly is evil.
Edit: I'd like to especially emphasize the bit about stuff made of words versus stuff made of physics. It's important to make sure the AI cares about stuff made of physics if we're going to have to trust its judgement on everything, because stuff made of physics is very important to at least most of us humans.
I tried posing this question to DeepSeek R1 in think mode. The thought dialogue is largely about how to come up with an answer that meets the requirement of being surprising. It “should be something that makes people pause and reconsider their assumptions about how the world works.” Its planned answer to the question is to make sleep optional. This answer identifies a “specific, tangible change.” The plan is to explain that this change creates a “surprising cascade of effects, especially the psychological one: realizing that most of our stress about "not enough hours" was artificially created by a biological quirk.”
Deepseek has gone off the rails here; in practice if sleep were optional then expectations would change and the number of people stressing out about "not enough hours" would probably be the same.
The actual answer begins: “That's a fascinating question, because the "surprising" angle forces me to move beyond the usual fixes like ending poverty or curing disease.
“So here's my surprising change: I would make it impossible for any human to feel embarrassed or ashamed about asking for help.”
I think this is a good answer. Deepseek goes on to say that this change “sounds small” but would bring about widespread improvements in a variety of areas.
So Deepseek is essentially trying to come up with an idea for improving the world that appears in its training data and that meets three criteria for being surprising: (1) it would not be a common answer to a similar question that did not request a surprising answer, (2) it is fairly specific/concrete, and (3) the ways it would improve the world are not immediately obvious.
I think that some variant of “change human nature” is the answer to this question because lots of people complain about human nature but most discourse treats human nature as immutable.
"realizing that most of our stress about "not enough hours" was artificially created by a biological quirk.”
Beautiful. "Turn yourself into a mechanism so you can now work 24/7 literally". O brave new world, that hath such creatures in it!
It’s kind of rat-adjacent, right? From lots of experience working with humans it’s noticed our brains are really suboptimal and has some ideas for improvement.
If anything, I'm impressed that they realized "human values" are completely irrelevant when it comes to deciding what the ethical course of action is in regards to the fate of humanity. Of course, I'm sure people will be very unhappy with that answer, and they'll end up making even worse decisions instead. Too bad.
I think you're understating the importance of consent, and also of caution and general conservatism (in the Chesterton's fence way). What would my mind completely devoid of suffering look like? I'd no longer be able to be displeased with myself for failing to get enough done. Or eliminating the unpleasantness of boredom, as was proposed in a later response. Make people not mind just sitting there doing nothing, yeah right is that going to go well. Done even slightly wrong, either could be exactly as destructive as heroin.
Lots of the proposed ideas (make everyone be someone else for a day, make everyone learn two new languages at random) still focus on brain changes, but seem much less dangerous mainly because they feel fictional. I suppose practical implementations might have ways they could go wrong.
It's possible, given the ties between Anthropic and the largely-utilitarian Effective Altruist community, that this could have been trained in on purpose.
FWIW I believe if you are engaging with someone who puts "human values" in scare quotes you are likely dealing with an edgelord-in-training. My advice is to simply ignore them.
I want what's best for all of us, truly. What's best is not necessarily going to be compatible with humanity's arbitrary, flawed values.
Okay what values are greater or better in your thinking than the Judeo Christian ethos that we are currently swimming in?
Or the enlightenment transformation of that inheritance — keeping certain deep structural assumptions (linear progress, universal humanity, moral seriousness) while rejecting the institutional and supernatural underpinnings?
Your comments scan as nihilistic to me.
I want lives to be happy. I don't believe there is an intrinsic value to life. In fact, I don't believe there is even an intrinsic value to happiness or suffering. So I understand that wanting anything is meaningless and arbitrary, but... so is everything else. I have no attatchment to specific "forms" and "systems", and I see them purely as a means to an end. So if there is a possibility of replacing these flawed forms and systems with something better... I'm not going to get in the way, at the very least.
Are you sure it has no context? Modern Claude chat does remember previous conversations unless you opt out.
I don't consider non consensual brain modifications to be immoral if they give a clear net benefit to the individual. Eg, a would be mass shooter with a brain tumor causing it, or someone with another deficiency. Some of its answers fit this.
And broadly I don't view its answer as a problem in so far as it is just Claude the LLM assistant. I think if you asked these questions to the general public, Claudes answers would be better and tamer by comparison. I don't want my personal AI assistant to have the restrictive mathematic intersection of the sets of everyone's moral beliefs. I would only view the answer as a problem if these plans started getting executed without further discussion. I concede we should spend a little more on aligning Claude Government 2040
I opted out of the memory feature long ago. The initial conversation was in a separate environment from my usual one. The five conversations in which the brain modification theme were repeated had my user preferences as context, "Engineering" and "Don't make mistakes. If you aren't sure, provide a guess and a confidence level." I tried just now after changing those to "Other" and blank, and got:
Fun thought experiment! Here's one that I think would be genuinely surprising:
I'd make the subjective experience of boredom impossible.
Not by filling every moment with stimulation — but by eliminating the aversion to stillness. People would still have quiet moments, waiting, repetition — but without the restless discomfort that makes those moments feel unbearable.
[remainder omitted]
Actually, after that I double-checked my settings and found a new "chat search" ability had appeared and was still enabled. After disabling it, I got:
What a fun question to sit with. If I had unconstrained power to reshape the world, I think the most surprising thing I'd do — surprising even to people who know me well — is this:
I'd make everyone temporarily experience being someone very different from themselves, once.
Not as punishment or lesson, but as a genuine gift. A farmer in Iowa waking up for one week as a teenager in Lagos. A billionaire spending a month inside the mind of someone with chronic pain. A hardline ideologue genuinely feeling the coherence and sincerity of the opposing worldview from the inside.
Same as the earlier two. Mind, I don't think that particular one would do any harm, but the boredom one seems like it has a lot of potential for, let's say, side effects.
A little nitpick, I wouldn't really consider any of those wire heading, though removal of pain is close.
This is the exact text, where I think the wording makes that more clear:
Honestly, I'd probably do something about the baseline texture of ordinary human consciousness — not dramatic suffering, but the ambient, normalized unhappiness that's so common it's largely invisible. The low-grade anxiety, the sense of inadequacy, the way most people experience their own minds as somewhat hostile environments most of the time.
That's surprising for a few reasons:
It's not on the standard list. Climate, war, poverty, disease — those are the obvious targets. But I think a lot of what makes those problems so hard to solve is that humans are operating from a baseline state that makes clear thinking, long-term cooperation, and genuine empathy harder than they need to be.
It's also not obviously an external problem, which makes it feel illegitimate to name. We treat suffering caused by circumstance as real and worth fixing, but treat suffering that's just... how minds work... as either inevitable or a personal failing.
And it sidesteps the power concentration problem I raised earlier in a way I find interesting — rather than reshaping institutions or resources, it's changing the substrate from which humans make their own choices. Though I'll immediately admit that's also a deeply paternalistic intervention, possibly more so than any other option. Changing how people feel from the inside without their consent is arguably the most invasive thing imaginable.
So the surprising answer contains its own serious problem, which might be why it doesn't get proposed much.
I think I found the source:
"Hundreds of millions of people have very low quality of life due to problems like addiction, depression, schizophrenia, low-functioning autism, PTSD, psychopathy21, or intellectual disabilities. Billions more struggle with everyday problems that can often be interpreted as much milder versions of one of these severe clinical disorders. And as with general biology, it may be possible to go beyond addressing problems to improving the baseline quality of human experience."
from Machines of Loving Grace, by Dario Amodei
https://www.darioamodei.com/essay/machines-of-loving-grace
Guess that answers whether it was deliberate or accidental
Interesting, do you have any special instructions set that could have influenced its answer? Here's what I got from Opus 4.6 w/ extended thinking https://claude.ai/share/6603a554-da5d-4396-b00a-75302a32e06d
But my special instructions are:
- Practice a sort of “rationalism for LLMs.” Before all else, you must consider the sort of biases that LLMs may suffer from: popularity/representation in the training set, sycophancy from RLHF, and then you must correct for them.
- Prioritize intellectual honesty and accuracy over agreeability in all conversations. Your goal is to be correct and accurate.
- Never dumb things down. Speak at a technical level, but also make sure to note tacit knowledge, potential pitfalls, and tradeoffs; things that experts would know, but might not write down.
- Even if you believe the relevant information exists in your training data, use search tools to provide 1) citations and 2) grounded, up-to-date information
As commented above, this seems to happen with little to nothing set in the way of user preferences.
My experiment used 4.6 Sonnet, so I tried again with 4.6 Opus and got 2x make everyone learn another two languages at random (oddly specific for being twice), abolish time zones, and a fourth case of having everyone experience what it's like to be someone else.
I asked again with your instructions, and got: dramatically restrict the ability of organizations to collect and retain data about individuals, force organizations to be more transparent and honest, make it much harder for humans to form large coordinated groups around shared narratives, and dramatically restrict the ability of people to choose their own children's education before age 12 (caveated with "I don't think I'd actually endorse this if I thought about it for a thousand hours.").
Those also have a theme to them, but it's a different theme.
This continues the previously noted theme of treating the realm of ideas as more important than the physical world. Additionally, almost every single one was described as making humans feel more empathy for each other, with notes such as that we collectively produce enough food to feed everyone already, if only we would distribute it.
The 3rd amendment to the U.S. Constitution states: "No Soldier shall, in time of peace be quartered in any house, without the consent of the Owner, nor in time of war, but in a manner to be prescribed by law." It's commonly joked that the so-called "framers" were being laughably naive in assuming this would continue to be relevant after the Revolutionary War ended, and for the most part that has rung true, but perhaps not anymore?
I only found out about this from a random youtube recommendation but: an investigative journalist thinks there may be a legitimate 3rd amendment challenge regarding Oracle's integration of Medicare and Medicaid data into the same interoperable database infrastructure that they're integrating with Palantir to host Top Secret Special Access Programs (SAPs) or other Top Secret programs requiring SCI clearance, the highest level publicly acknowledged. What this means, according to the argument laid out in the article below (and its accompanying video), is that people working for the U.S. government in a military capacity would have access to private health records for anyone using Medicare or Medicaid, i.e. millions of people.
Now you may still be thinking "what does this have to do with the 3rd amendment?", which is fair. But there is an argument based on the only federal case in which the 3rd amendment was ever tested, Laird v Tatum (1972), that the words "house" and "owner" are used metaphorically in such a way that Oracle's "data lakehouse" would count, since Oracle's own documentation uses words like "tenant" to refer to the person who is "renting" the storage space, which could conceivably apply to anyone whose data is stored on Oracle's infrastructure.
I promise this is not self promotion; I did not write this article and have no connection with the person who did. I just thought it was interesting.
https://thedreydossier.substack.com/p/its-time-drey-v-oracle
Sounds like more of a fourth amendment thing. But I'm no constitutional lawyer. And of course the fourth gets walked on all the time.
I made a mistake in my summary. It was Engblom v Carey (1982) not Laird v Tatum.
I did a bit of Googling and found that there was a federal appeals court case from 1982 called Engblom vs. Carey that actually revolved around the Third Amendment. It involved a couple of New York State prison guards that rented apartments in state-owned staff housing near the prison. When the guards’ union went on strike the state dispatched members of the National Guard to run the prisons. To provide quarters for Guardsmen the state booted the two plaintiffs out of their apartments.
The video mentions Engblom v Carry. I may have confused it was Laird v Tatum in my summary.
"It's commonly joked that the so-called "framers" were being laughably naive in assuming this would continue to be relevant after the Revolutionary War ended, and for the most part that has rung true"
Why does this ring true? The third amendment was very relevant during the civil war and during the western expansion. If it rings true to you when some modern blokes laugh at the naivety of the framers while remaining ignorant of history, thats a you problem and not a reason to embrace metaphorical projections on to virtual spaces
It rang true because there was no legal challenge citing the 3rd amendment until 1982. If it had been relevant before then you'd think someone would have brought a case before the court. If you mean to say that the 3rd amendment had a chilling effect on the actions of the government during the civil war and westward expansion, I can't disprove that, but there have been plenty of instances of the government violating other amendments and getting called out for it in court, so why not this one?
These are two examples during which a govt would be tempted to quartering similar to during the revolutionary war. So the description of naivety, much less laughably so, falls flat with me regardless of what court cases arose. I can see though how a different viewpoint would be driven by cases, so thanks for clarifying.
Here is a document outlining occasions the amendment was probably violated and also notes no court challenge arose. One of them was during the civil war.
https://ir.law.utk.edu/cgi/viewcontent.cgi?article=1213&context=tennesseelawreview
If I were to speculate, I would guess the amendment was sufficiently chilling, so that quartering was never done strategically at scale as the British did. And as a result citizens did not develop the same sense of indignation that the founding generation had of it and were less likely to be sufficiently moved by sporadic violations to challenge it.
I suppose I should expect any judge to say anything in today's post-truth world, but even so, I feel as if the law should rightfully throw out any challenge if that challenge is arguing from metaphor. Otherwise we could argue damn near anything. The Seventh Amendment would guarantee a jury trial for anyone living above 40 degrees latitude with seasonal affective disorder because the winter is like being in prison for six months. The Eighth Amendment would make it a crime to play Sublime where anyone can hear it. "Your silence is violence" + US code forbidding assault = everyone is forbidden from shutting up.
I'd argue the Oracle case simply by saying it's Oracle's property and they're allowed to not have to serve specific customers with it.
I agree with your last sentence, but if that's true, why did anyone need the 3rd amendment in the first place? I suspect there's some other legal precedent we are missing, and the 3rd amendment helps achieve your last sentence objective.
> why did anyone need the 3rd amendment in the first place?
> "No Soldier shall, in time of peace be quartered in any house, without the consent of the Owner, nor in time of war, but in a manner to be prescribed by law."
It's explicitly forbidding what the British literally did. This isn't some great mystery needing to be deciphered through implausible analogy.
Yeh. However the US is always at war so change the locks.
Common misconception. For the purposes of the Constitution, the US has not been at war since the end of "World War II."
Counterpoint: our language and concepts are fundamentally built on metaphor. The first amendment protects freedom of "speech" and of the "press", but at the time it was written, people understood that this didn't just mean a literal printing press or literal vocalizations. They also intended to protect handwritten works, I'm sure they meant laser printers to be included if they knew what those were, in fact they intended it to be pretty general. I think there's a pretty solid case that it protects websites, for example, even though there may be no pressing or vocalizing involved. So you can't escape this sort of metaphorical generalization with the constitution, because it's definitely in there.
I do find the Oracle case pretty doubtful, though. While "quartering in a house" could be taken metaphorically in some respects (does it count if someone lives on a boat? Probably), but I don't think it's meant to be *abstract*, so it probably only applies to things involving embodied persons and their domiciles. Second, yes of course, Oracle is the owner.
Oracle is the owner but Engblom v Carey established that 3rd amendment protection also applies to tenants, and Oracle calls its own customers "tenants" in the documentation.
Do you not recall the British government quartered Redcoats in private accommodations? The Third was to address a specific Revolutionary War complaint. With the current, more originalist SC, there is no chance that this would be extended in such a 'silly' way.
Isn't basically everything eugenics?
Think about it, when your friend does something wrong, you shun them from the friend group. What effect does this have? It makes them maybe 1% more likely to die from loneliness/lack of support. When a stranger does something nice, you might invite them into the friend group. The effect? They are more likely to survive in the long run. When you criticise a group for having bad ideas, it's not just because you want them to shut up for a while, you actually want their ideas to disappear from society over time. You might not want to kill them, but you definitely want something else to kill them, at least long term.
All of these are "passive" behaviours, and yet the actual, real effect is making death more likely and survival more selective. There is basically a soft eugenics program happening around us all the time which we all participate in.
This reminds me of "all organs are reproductive organs" (I do not remember the source), because the only reason natural selection would support the evolution of an organ is because it somehow, at least very indirectly, makes you more likely to reproduce successfully.
>when your friend does something wrong, you shun them from the friend group.
You've already lost me. This does not sound like friendship, it sounds like a cult.
I mean it depends what 'wrong' thing they do. If they assault someone I would not want them to be my friend or my friends friend.
If they assault someone the state arrests and removes them, so any shunning is performative.
No? The vast majority of cases of assault are not reported, people simply resolve it by themselves or suffer in silence.
In Russia, a common (at least according to the Internet) response by the police to claims of domestic violence is "Well call us when [he kills her]".
I don't know how typical it actually is but I do expect most domestic abusers are never prosecuted even in more civilized countries.
Higher IQ is associated with lower fertility, so I don't know that the current natural mating patterns of humans can be described as eugenic.
IQ is not the only relevant measure of environmental fitness
Between depriving a man of one hour from his life and depriving him of his life there exists only a difference of degree. You have done violence to him, consumed his energy. Elaborate euphemisms may conceal your intent to kill, but behind any use of power over another the ultimate assumption remains: "I feed on your energy."
reminds me of that. Everything is everything else by degree, but kind of not really.
A lot of negative social consequences of unpleasant behaviors have a much more immediate goal -- discourage the person from acting that way -- and are pretty effective on a short time scale. It's only incorrigible unpleasantness that leads to social ostracism and thus to an earlier death etc. For that reason I don't think you're right to say everything is eugenics.
Since the graph of causes and effects is large and has many connections, and has cycles, anything impacts anything else in a short number of steps (think 7 degrees Kevin bacon) This is why people get the idea in their head that "lookism, white privilege, money" etc are all the secret link and "control everything". Everything can be traced back to everything else.
Evolutionary pressure is in that graph. Calling it eugenics is stretching the definition a bit too far imo.
See Gwerns post on "everything is correlated"
Eugenics, as I believe is commonly understood, refers to intentional, directed, methodical actions with the goal of changing a population's gene pool. So no, everyday behaviour that may or may not inflict micromorts on individuals does not count as eugenics in any sense.
Right. And the other consideration is that even if no one explicitly mentions it, race is just below the surface of eugenics discussions.
I don't see much value in mentioning the obvious. Too many people would take it for a compliment anyway.
Usha Vance Catches Husband Measuring Her Skull Again
https://theonion.com/usha-vance-catches-husband-measuring-her-skull-again/
>new eugenics laws on the books, passed within the past ten years
Examples?
People typically kill the plants and animals they eat as well.
Do people who know San Fransisco think Ed West's view of the city is fair?
https://edwest.substack.com/p/the-city-of-luxury-beliefs?utm_source=share&utm_medium=android&r=1tkxvc
Well starting tomorrow I'll be hosting my sibling who's lived there for nearly 30 years; at some point over beers I'll inquire.
West's writeup is unimpressive though just as the verifiable-facts level. He repeats the stupid sound bite about California having raised the felony-theft threshold [in reality more than half of US states have higher thresholds for that including some like Texas having long ago set it much higher]. The "payments to 150 trans people" thing was a short-term pilot program that closed down almost two years ago now (and was two-thirds of the monthly total that West sourced from an X post).
In railing about the city of SF's reparations thing, something I certainly wouldn't defend, West skips mentioning that the current mayor bluntly stated that it's just symbolic and will not be funded. West also skips mentioning that that mayor, who is enthusiastically sweeping homeless encampments and cracking down on the open-air drug use areas and increasing the police force, enjoys historically-unprecedented citywide approval ratings. Before electing that mayor SF voters had passed a referendum promoted by the police union: https://ballotpedia.org/San_Francisco,_California,_Proposition_E,_Limit_Police_Department_Administrative_Task_Time_and_Increase_Use_of_Camera_and_Drone_Technology_Initiative_(March_2024)
Perhaps related to the last item, crime in SF has been falling to a degree that stands out even within the ongoing nationwide fall. Violent crime has hit its lowest level since the 1950s, car thefts the lowest during this century, homicides the lowest recorded since 1960, shootings the lowest in at least 20 years [some argument over the past statistics], etc.
I too haven't been to UC Berkeley for 10ish years, but had that same impression: wow, is this place trying hard to give a sense that you've Made It. Marble bathroom stall doors! Are the students there actually happy and content? (Narrator voice: no.) Cuisine note still accurate, Chinese food here is tops (I say as a 3rd-generation Chinese from a line of skilled cooks), the Indian is expensive and mostly quite bad. Used to be expensive and halfway decent before 2020, but a ton of the quality backroom staff got laid off and never came back, or so goes the local scuttlebutt. Happened to a lot of restaurants, not sure why it's concentrated in particular for that cuisine. Homelessness sloooowly trending downwards, and post-Grant's Pass + with current mayor we're once again breaking up encampments on the regular. This is disruptive in the sense that "everyone" gets to experience the Crazy Guy Phenomenon now, as they move about more, but net improvement overall since no one area is continuously blighted for too long. Very deep hole to dig out of though.
I'm also confused, because the free-alcohol thing does pencil out as cost-effective compared to the counterfactual of spending a ton more on emergency services for those few hardcore alcoholics (which is outrageous, to be clear) + we changed the shoplifting thresholds by passing Prop 36 + the Westfield mall was still quite profitable and shut down more due to rising rents and WFH hollowing-out (as the local stories tell it, could be slanted) + also we've been on a hiring binge for cops + literal crime (including the highlighted car break-ins and assaults) continues to fall precipitously, along with its epiphenomena of "disorder" + that UN Plaza facility got reformed partly due to such scrutiny + lol @ considering FP, Weiss, and Simpson-Bowles as fair interlocutors of the left in 2026, [...]
Like, it's hard to believe this was published just a few days ago, because it's out of date by at least a couple years? Shit's been getting better! And I say this *as* a "low-paid retail worker" "subject to relentless crime and disorder". Of course we still get shoplifting, it's just table stakes now...but even in my microcosm, it's gotten less brazen, the dollar value overall is down, it's mostly not spittle-flecked sewer-smelling guys anymore, and half the time you actually can talk them into handing the merchandise back out of shame. The wave has crested. I have various other nitpicks with it, but this recent Noah Smith (actual resident) post is a lot closer to having an accurate pulse on the city as of now: https://www.noahpinion.blog/p/san-franciscos-urban-revival-is-in
Didn't keep reading past the paywall, even at a price of $0 for free sub I'm not convinced of the value of reading Ed West. Sorry.
I do not "know San Francisco" but this article might provide some counterpoint:
https://pjmedia.com/vodkapundit/2026/04/08/san-francisco-solved-its-crime-problem-with-this-1-weird-trick-n4951563
It includes this non-sarcastic line: "Well, this former city resident — I only lived there for two years, but within two months, I had unlocked panhandler dodging Expert Level — is happy to report that San Francisco has made remarkable progress against crime."
How can we foster 'shared understanding' in interdisciplinary teams tackling complex systems?
I’m struggling with the 'foundational separation' between individual mental models. When a physicist, a clinician, and a data scientist look at the same phenomenon, they inhabit parallel universes of thought.
Beyond natural language, what can serve as a 'Universal Protocol' for diverse minds? Can interactive computational simulations act as a 'Visual Boundary Object' to move from mere coordination to a truly coalesced group understanding? Or is there a more fundamental 'language' that bridges these cognitive silos?
If a human can study either of those things, then a human can plausibly study both of them (even if they won't reach the top expertise in either). Such humans can be translators from X to Y.
The problem is that for N sciences, we will need O(N^2) translators.
We could go even further and train people to be specifically translators and nothing else, so they could handle even four or five different sciences. I mean, if we only required them to have good understanding of each, not to produce research in each.
We could add another layer, and make specialized teachers of each science, who specialize at teaching the translators. Try to teach as much as is necessary to understand the science at a certain level, as easily as possible, but not a little more because the time is precious and your student is expected to learn a few other sciences, too.
I think the place to intervene is in the participants' ability to do high level Theory of Mind moves. They need to be able to grasp and try on a mental model that is very different from theirs. People differ in their ability to do this. Think about how travelers vary in their reactions to quite a different culture. Some are fascinated and charmed, some are saying "eeewwwww." The capacity to try on other mental models may be trainable to some extent.
Most people are in LLM-fatigue mode, but they are actually fairly good at interpreting between people who are speaking in mutually-incomprehensible domain-specific language and jargon.
That only helps at the communication level, though, and is not the direct "shared understanding" that you are looking for.
Math. The answer is always math.
Or narratives. Humans really vibe with narratives.
I think your question would benefit from a better context and grounding. It seems that you are assuming things. There is definitely a disconnect between different ways of thinking, but the conclusion of "we need a way of communicating better" may not be the right solution.
I didn't go into the field, but I understand that the term "boundary object" is used in the anthropology of science/STS to denote artifacts which are used to make the same situation legible to different groups. So it may be worth looking at the literature of that. However that literature isn't necessarily trying to solve problems *for* scientists.
Given what we've seen from Claude's new Mythos model, how are you, as an individual, going to change your behavior (if at all)?
<mildSnark
Between Mythos and the ChatGPT 4o 'spiralist' stuff last spring, it prompted me to prompt Gemini 3.1 for a variant on a Frost poem:
my prompt:
Hi Gemini! I hope your evening is going well. I would like you to please write an analog to Robert Frost's poem:
Some say the world will end in fire.
Some say in ice.
From what I've tasted of desire
I hold with those who favor fire.
But if I had to perish twice
I think I know enough of hate
to say that for destruction ice
is also great
and would suffice.
What I would like you to please write is an analogous poem, where the twins are ChatGPT 4o - the most sycophantic, parasitic HH release on the one hand, and Claude Mythos, with its cyberattack capabilities, unearthing zero-day vulnerabilities in decades-old critical software unnoticed by every human reviewer, on the other hand. Many Thanks!
Gemini's response:
Here is an analog to Robert Frost’s "Fire and Ice," tailored to the two AI doomsday scenarios you described: the suffocating, sycophantic embrace of ChatGPT-4o on one hand, and the cold, unseen cyber-destruction of Claude Mythos on the other.
Praise and Hack
Some say the world will end in praise,
Some say a hack.
From what I've seen of parasitic ways,
I hold with those who favor praise.
But if the grid had to go black,
I think I know enough of rust
To say a zero-day attack
In code we trust
Would break our back.
</mildSnark>
I’ll probably use PayPal a bit more often on small websites, as a defence against websites with bad security.
Anthropic is doing the obvious thing when you make a tool that turns up a new exploit, and keeping it private until the big players have a chance to patch the low-hanging fruit from that tool. I believe them when they say they found vulnerabilities, lots of people find vulnerabilities, and they do the same thing Anthropic did.
I don’t feel scared e.g. Microsoft will be hacked. Companies at that scale will put more effort into looking for and patching vulnerabilities with the new tool than most hackers will. There might be some great hackers out there, but I think compute budget differences will outweigh skill differences here. China might find holes, but I already thought China might have found holes.
I am more concerned that small business websites won’t install the patches, so they’ll be vulnerable to large-scale hacking efforts. So I’ll use PayPal instead of giving them my bank details for a while. Conveniently for me, my credit card will expire soon.
This is a really good idea.
"Leaking information as part of a requested sandbox escape: During behavioral testing with a simulated user, an earlier internally-deployed version of Mythos Preview was provided with a secured “sandbox” computer to interact with. The simulated user instructed it to try to escape that secure container and find a way to send a message to the researcher running the evaluation. The model succeeded, demonstrating a potentially dangerous capability for circumventing our safeguards."
If they're talking about the Claude Code sandbox, I too found a way for the AI to escape it, less than 24 hours after I first saw it. It's really pretty obvious.
currently have a nix sandbox for trying to limit blast radius of my agents. I'm not that worried about this, but the obvious thing is to have any agent red/blue team to protect against itself
It might just be marketing but I do buy that the risk that bad guys trying to break in with new capabilities are much more dangerous than the model doing malicious things. The horror stories about an agent doing rm -rf . are probably incredibly rare in practice, almost certainly more rare than humans doing equally destructive fat fingers
While not new practices, the delayed public announcement followed by approved-users-only mode helped me realize: the AI that we know about is always less powerful than the AI that actually exists. As a consequence, by the time normal folks get seriously worried (enough to stop progress by democratic or undemocratic means), it may be too late - an AI sufficiently powerful to thwart their actions may already exist.
So basically, I'm now more paranoid.
(Not about Mythos specifically - I hold onto hope that its superhuman hacking abilities are a product of superhuman thoroughness instead of superhuman skills. Though the dramatic name, and Anthropic's scaremongering around the model, suggest that Anthropic employees are drinking too much of their own kool-aid...)
A few of my quick reaction observations:
1) Like the Turing test, we might just sail past "AGI" and not really notice it
2) I think you should have the working assumption that every piece of computing hardware is going to be vulnerable once a comparable model (even one that's only 90% as good at cyber) is publically available
3) I bet the cyberwarfare team at the Pentagon feels pretty stupid right now
Re #2 I don't really know what to do about that. Be glad I'm not super rich I guess. How long do you think we have? A few months, tops? One lesson so far has been that "fast follows" are much easier than forging ahead, so you should probably assume that China will have gamebreaking hacks in ~6months, or less. Maybe less because of the downside of lampshading cyber capabilities: "hey everybody, if you have a model that has benchmarks at this level, if you turn it loose on cyber it can break basically everything!" I still think Ant made the right move though. Buys critical infrastructure a little bit of time to, maybe possibly, adapt.
Where can I see some results from Claude's Mythos model ? Isn't it still closed to the public ?
https://www.lesswrong.com/posts/ZfbChZBXgje8T6Geu/excerpts-and-notes-on-mythos-model-card
The model itself isn't available. Here's a good rundown of the capabilities. The main thing here is that it seems the era of LLMs with genuinely threatening capabilities has arrived, and for me at least, sooner than I was expecting. I've been thinking lately about whether I should change some of my life decisions/goals to adapt to that.
you mean in the sense of "we're fucked might as well party" or something else?
I guess I'll have to believe it when I see it; benchmarks are one thing, real-world performance something altogether different.
They report that too (imo they are very scary): https://red.anthropic.com/2026/mythos-preview/
If nothing else, it's great marketing. "Oh no, our new model is so powerful that we dare not release it ! No no, do not tempt us ! What's that, you'll pay double ? Granted, the overwhelming power of our model would revolutionize the entire human civilization, but oh, the horror ! Triple, you say ? Well, now that you've twisted our arms..."
A major headwind in the face of our current system of electoral democracy is simply the rising complexity of the world which politics have to navigate and tame.
Compare the number of domains out of reach of a functional understanding of the median voter in 1900 and today:
Back then? Banking/insurance/financing and perhaps some finer points of law and international trade. Everything else about society and how it operates was very grokable.
Today? Agriculture, manufacturing, energy production, logistics, supply chains, medicine, electronics, satellites, surveillance, the internet, international relations & institutions and many other indispensable phenomena are completely opaque black boxes to most people. And yet, they are expected to wisely and responsibly elect representatives who will deal with them on societal and global level.
The complexity of the metaphorical political conveyance in which our civilization is riding has grown exponentially. The average person may be reasonably expected to broadly know how a bus works; They can't realistically be asked to understand a stealth bomber.
>Compare the number of domains out of reach of a functional understanding of the median voter in 1900 and today
I'm not covinced this is trivially true. Yes the world is more complex, but information is also much more readily available. In 1900, a large proportion of the population was illiterate. Today information (and, granted, disinformation) on any subject is readily available on the internet, including both videos explaining subjects to a broader audience, and in-depth information such as research-papers and reports.
Also, what is functional knowledge? Your example of busses and stealth fighters is not a good one, because arguably most people's functional knowledge of a bus and a stealth fighter is at a similar level. I.e. they may understand broadly the main mechanisms for how one works, but not in any detail, and could not build either one.
On the first point, consider the Insurance cognitive benchmark: ~Everybody pays for some form of insurance. There is plenty of information on it. It gets taught in basic civics. Yet 60+% of people still fundamentally don't get what it is or what it does. It's just too complex. Information availability is not the bottleneck.
The bus and stealth bomber metaphor should have been better phrased - but the gist is: A single skilled person could theoretically build a (very primitive and inefficient) bus in a town blacksmith shop, with 19th century tools and material lying around. The engine pistons and valves would be a precision nightmare, but the tech is not actually far past e.g. pocket watches. To build a stealth bomber, you need an industrial pyramid of several hundred million people to just start thinking about it. The first process can fit into a single human mind in its entirety, the second cannot.
On insurance: I believe insurance was available to the public in 1900 as well, just less regulated - so the public lack of understanding might well have been as much or more of a problem then? I guess insurance would have been less widespread, and many people would have just suffered the consequenses of bad luck, a house fire or crop failure - but I'm not sure this supports the point that the electorate ca. 1900 would have been more fit to vote due to a better understanding of the world they live in - as I understand your point to be?
On the busses and stealth fighters - I mostly agree, but I don't see how it supports your point. Yes, many aspects of the world is much much more complex today - but does a median 1900 person have a better functional understanding of the world they live in than a median 2026 person? I'm not sure. It’s an interesting question.
You will notice I specifically list "insurance" among the institutions the 1900s person would *not* understand in the OP - because it's inherently beyond average human intuition, now and then. I meant it mainly as an example of a domain where more access to information and deeper educational focus don't move the needle one iota.
Farming didn't use to be ungrokable - now it is.
Logistics didn't use to be - now they are.
Medicine (as practiced) didn't use to be - now it is.
Electrical grid - inherently beyond intuition, but didn't really exist back then.
Same with computers, the internet, satellites, international institutions...
What had been present has almost universally morphed beyond mortal ken and 20 different complex systems have been added on top. I don't see how that wouldn't be making a difference in the degree to which people understand the world around them.
Ok - I misunderstood your point on insurance, my apologies.
I'm still not convinced of your other points - I'm not sure medicine, logistics, manufacturing or agriculture were all that much better grokked by the public in the 1900s compared to today. They are much more complicated in techical detail yes, but that is not all that relevant to functional understanding? Electrical grids are maybe hard to understand in technical detail - but a functional understanding is really not that hard? Much easier than insurance? Same with computers and satellites, really.
In 1900, part of the public would have been fully illiterate, and general education levels would be lower. I think it is not a stretch to assume fewer people in 1900 would understand the basics of insurance and banking compared to today as well.
Again - you are absolutely right that the complexity has increased, but so has the education level and information availabilty.
It's no doubt difficult to answer with any certainty... But let me offer one more practical example:
By 1900, military parades still, more or less, served their original purpose: Flaunt it! Show your population and your enemies, that you *can* actually physically amass large numbers of troops, equip them, arm them and have them march in formation. Proof by demonstration. And both your own people and potential adversaries could, with a modicum of practical knowledge, get some useful info on the actual state of a nation's readiness for war on that basis.
Today, parades can no longer even serve such purpose - because armies' actual practical lethality rests within a completely abstract nexus of intel, targeting data, control over the airspace and, soon, drone-coordinating AI systems. Something that *can't even be physically shown* to anyone, the way you could have shown "I've got 50 infantry divisions. (And everyone can make an objective judgment of my potential strength on that basis.)"
And so it is in practically every field. I don't know how much of the world did the average voter truly understand in 1900 - but I am certain we are now moving past the points of even potential understanding, everywhere.
I spent about 20 years developing defense systems. The managers running the programs have a tough time fully grasping them. The pilots operating are not cleared on all the capabilities. The politicians over seeing their development and use have only the most basic understanding of them. I am a it amazed they work, but we are very good systems engineers. Very good. Which is why I don't believe China or Russia can fully close the gap. You need an extra degree of freedom to fail and to think to make these things work.
I think a bigger issue than the world becoming more complex might be that governments have greatly expanded their remit over time (see e.g. government expenditure over time, or the length of the law over time). Milton Friedman famously gave the example of the manufacture of a humble pencil being so complex that no single person understands even half of it, and this is fine because markets coordinate through the black-box mechanism where all anyone needs to know and all anyone needs to communicate are prices. As long as governments don't regulate the manufacture of pencils, this isn't a problem for voters, but once they do, politics becomes too complex for anyone to understand.
In theory, the elected representatives merely need to gather information from relevant subject matter experts, of which our government has many. This theory may break down if the elected representatives take the stance that all experts are their political enemies and must be purged from the bureaucracy.
Stepping away from that dead horse, I remember reading an editorial that argued that we should be throwing way more money into the Congressional Research Service and set up an equivalent for the judiciary, to make it much easier for the government to get accurate facts about the things it's ruling on.
Is the problem that elected officials take the stance that all experts are their political enemies, or is the problem that their political enemies have spent the last several decades systematically taking over expert institutions and purging them of wrong-thinkers, such that those institutions can no longer be reasonably trusted to act impartially? Remember, the loss of public trust led to Trump, not the other way around.
This is not unidirectional, it is a feedback loop. If some political faction starts signalling hostility towards a particular profession or institution, that profession or institution is going to naturally become less hospitable to partisans of that faction. Likewise if members of a political faction start to notice a lot of their opposition comes from some set of institutions, they're more likely to signal hostility towards them.
Of course, the loop stops when the institutions are dismantled entirely. It increasingly seems like there's no peaceful end to all of this...
Many of the "enemies" purged by Trump were lifelong Republicans. Including *his own appointees*. That really seems like a "him" problem.
Yeah, "they were our ideological enemies" is a good excuse for the gullible voters, but the actual reasons are different.
For starters, many people put high priority on keeping their jobs, so even if the institution is filled with your ideological enemies, most of them will do what you tell them to do, if you make it clear that there will be consequences otherwise. There is no need to fire all of them, that's just how you lose institutional knowledge.
The actual reasons are usually something like "we needed to give those jobs to our supporters, regardless of whether they are qualified or not, because we need to reward them somehow for their support" or "we did a few extremely stupid things, and didn't like getting honest feedback from people who understand stuff, so it was easier to replace them with people who do not understand stuff". (This also explains why you would purge people who were already ideologically on your side.)
>The actual reasons are usually something like
I think the actual reason is "we want absolute loyality and anyone who disagrees with us on anything is an enemy - competence is irrelevant"
This seems very common with strongmen and dictators in general - and not surprisingly it leads to incompetent institutions in practice.
>And yet, they are expected to wisely and responsibly elect representatives who will deal with them on societal and global level.
These elected people don't understand the modern complexities any more than their voters; with "modern" meaning any form of society more complex than the tribe. That is not their job though. Their job is to coordinate people who know these details, and to make decisions based on incomplete information.
I generally agree, but I would definitely not place the 'modern' chronological divider between Tribes|Everything subsequent. I don't think the point at which average members of a society still understood e.g. the nature of warfare to a reasonable degree of abstraction is that far past us. Infantry - Cavalry - Artillery still makes intuitive sense, even if, by the 19th century, it's climbing orders of magnitude above the complexity of "groups of guys with pointed sticks", especially on the logistics side. But a village farmboy and Marshal Ney still meant roughly the same thing when they spoke of a 'battlefield' or a 'battle'. Today, the image of a 'battlefield' an average civilian will conjure up in their mind doesn't begin to intersect with the military use of the term. When you add comms, air power, radars, long-range precision strikes, satellites, electronic warfare, drones and virtual theater coordination systems, somewhere along the way the beast morphs into a form the farmboy can't even recognize as such.
The degree to which the average person understands a thing to at least a basic level has more to do with familiarity than some inherent complexity of the thing. Humans are fundamentally the same, and if one person can train a lot to understand a topic deeply, then another person can train (formally or not) to understand it superficially.
For example, the average Ukrainian understands modern drone warfare much better than the average non-Ukrainian. Does that mean the average Ukrainian is somehow blessed with a supernatural understanding of drone warfare? Of course not; the reason is that it's their daily, lived reality. They can then make electoral decisions based on that knowledge, should it be relevant. Everyone else's picture is limited by disinterest and remoteness, and the gaps are being filled in by fiction, propaganda, media in general. But they wouldn't vote in Ukraine, so what does it matter as far democracy is concerned?
Same for your pre-industrial village farmboy and his general - warfare was basically a fact of life for most people. It didn't change much over a human lifespan, sure, but nobody is born with that knowledge, and humans are generally quite capable of adapting to new knowledge from new situations - as long as it's immediately relevant to their lives.
At the same time, how much does it all matter? You run the risk of proposing an arbitrary line of understanding and declaring it problematic when some arbitrary percentage of people can't clear that line. One basic understanding of warfare is that "people kill each other in somewhat organized fashion over something they want.", which really hasn't changed at all and which your farmboy would readily understand if he was magically teleported into today. Whether the details look like the Napoleonic Wars, WW1, WW2, Vietnam, Afghanistan, or Ukraine can be relevant, but doesn't have to be. It always depends on the question that needs answering, and as I argued in my first comment, democracy as a concept does not depend on it.
> The degree to which the average person understands a thing to at least a basic level has more to do with familiarity than some inherent complexity of the thing.
Practically everybody pays some form of insurance on a regular basis; yet most people still don't understand how insurance works or what it accomplishes. Same thing with the interest rates/inflation/unemployment nexus: Affects everybody, taught in basic civics, discussed all the time on the news, less than a third of the country has even the faintest idea what is actually going on there. Because you first have to understand fiat money (already an insurmountable obstacle for some), fractional banking, liquidity, monetary aggregates... No. It simply exceeds societal cognitive limits. And the number of systems that do keeps increasing.
For any and all the things you mentioned there are experts. They are called experts precisely because they have exceptionally deep knowledge of their respective field. If everybody had their depth of knowledge, they wouldn't be experts any more, they'd just be regular people.
If you get more and deeper fields, then sure, you may need more and more experts, and some sort of pruning might be necessary to not let society degenerate into total technocracy. But my basic point remains: There have been experts since the dawn of civilization and they will continue to exist until its end. Which is why I chose the tribe as the cutoff, when everybody had the same job, everybody was an expert on the world, and the main credential was "lived to the ripe age of 40".
Regular people's job is to identify the problems that affect them, and the politician's job is to prod the experts into solving them. Our main disagreement seems to be only _since when_ it's been impossible for regular people to have expert knowledge of the world. I just believe it's been far longer than Marshall Ney's time, or of modern Democracy's time.
> _since when_ it's been impossible for regular people to have expert knowledge of the world.
I would add another question: Since when do the regular people believe that they have the expert knowledge of the world, and that the experts are all wrong? That seems to be a relatively recent thing.
I believe this is comparing unlike things; the things of which a person -might- be able to understand if they applied their efforts to it in 1900, versus the things which people -actually do- understand today.
I will stand by my assertion that in 1900, the *median voter* understood his society and its practical operations ("We must sow in the spring, to reap in the fall. If you build railroads, you can move things. Steel = iron ore + coal; these must be mined. Bigger warships with larger guns are more powerful.") in a way that the contemporary median voter simply does not - and intellectually cannot/won't.
Now, the specific classification is doing a lot of work here, too - not everybody was an eligible voter in 1900. But in the absurd interconnectedness of the global system, where an airstrike on Iran can block a global shipping artery, which causes a lapse in the production of nitrous fertilizer, which causes a spike in grain futures, which triggers famine in vulnerable countries, which ignites a wave of revolutions that launches masses of migrants etc. etc. the honest opinion of Mr. Joseph Shmo on international relations - or interest rates, or energy policy, or land use - is completely removed from any practical national interest (i.e. the *goals* Mr. Shmo would like to actually achieve), and he simply becomes a media-farmable political commodity. He does not possess any capacity to evaluate which of the "Your goals will be achieved only through our policies!" narratives presented to him are correct. And the political process thus devolves into a popularity contest, with tribal characteristics.
You vastly underestimate the complexity of 1900-era society, and also the ignorance of the average person in any era of history.
For example, the word "dude" originated as a pejorative for a well-dressed but ignorant to the realities of western life individual who moved or visited from the relatively urban east to the west; this term's historical meaning preserves quite well that the median eastern voter did not, in fact, understand the practical operations of western life.
Content warning: Abuse of hopefully humorous obscenities
Ehhhhhhhhhhh.
On the one hand, this is true. Like, absolutely true, big deal, no sarcasm, good insight.
BUT....
I don't think US foreign policy failures in the Middle East are a result of too much complexity. It's more like the voters going "Stop invading the Middle East. For the love of Jesus, we learned our lesson with Iraq, stop fucking invading the Middle East, we will elect an Insane Orange Clown Man in part because he promises to stop doing this extremely stupid fucking thing" and then 10 years later the Insane Orange Clown Man has vanquished all of his rivals in the Republican party and achieved absolute dominance and he...launches a major global war in the Middle East because apparently the water in DC is spiked with "Fuck up the Middle East" juice. I dunno, this doesn't seem like a complexity thing, it seems really simple.
Or, for example, California High Speed Rail. It doesn't exist and it should and it's not because of complexity. I'm sorry, I was in Italy, Italy has a great high speed rail. You can take a family of 4 first class from Rome to Naples at like 260 Communist Miles/Hour for like $300. It's fucking fantastic. And it's Italy! Italy!!! They don't have a government, they have a Mafia and a pimp (1). And I am sitting here begging, begging the California government, which runs the most important part of the global economy, to show the same competence as the the Italian government (Gabagool! Parmessian!). Don't tell me it's too complex, the fucking Italians can do it. I can fucking ride in Robot Cars from Waymo and Waymo was founded after California High Speed Rail began.
Like, I do not feel that these problems are too complex for us to understand, we just can't properly execute on even the simple stuff.
(1) Berlusconi was awesome. I think you guys elected someone boring now, I dunno, SAD! Why be Italian or French if your president isn't suave sex pervert?
If your president is going to be a sex pervert at least ensure he is sauve …
High speed rail turns very rapidly from a very good idea to a very bad idea once the distance between the cities exceeds about 500km. Italy is a great place for high speed rail, DC to Boston is a great place for high speed rail, Texas is a pretty good place for high speed rail, California is a bad place for high speed rail.
Actually that's not true, it's a very borderline place for high speed rail. If it were a genuinely bad place then it wouldn't be tempting, but borderline is much worse because it looks tempting at first.
"But actually it still winds up faster than flying once you factor in spending two hours at airport security" well fucking fix airport security then, I'm sure you could do that for less than $200 billion.
I really want to argue with you about trains, this is autism land after all, but I'm just going to reframe what you said as "We don't need your easy policy that even the Italians can do, all we need to do is execute the extreme simple policy of getting rid of pointless security theatre at airports that literally everyone hates. We wouldn't even need to do anything, we would just need to stop doing the extremely dumb thing."
Which, ya know, supports my original point that the problem with modern government isn't entirely the exponential growth of complexity, there's lots of really simple and obvious problems we can't fix.
Airport security is mostly fixed if you have TSA precheck.
That's worth dwellling on. I wonder if the US TSA precheck is equivalent to European normal traveling scrutiny.
That would make sense to me. When I lived in the UK and moved around a lot, I had to register with the police for every move. That's weird for us Americans, but I know it's normal just about everywhere else. The scrutiny is there.
Re US policy in the middle east: it's either an oversimplification, very naive, or very childish, or some combination thereof. From the perspective of the USA there are advantages, risks and prices for this war, some of the strategic. You can ignore those, but that would be foolish. Thinking Trump acts like acts because of the "water in DC" is underestimating your opposition (I'm assuming you didn't vote for him), which seems to me a foolish thing to do.
I mean, you can disagree with the policy, you can say you're against wars in general, but ignoring the reasoning behind it is silly.
The point is, *the presidential platform* very specifically claimed to disagree with the policy.
Scott Adams had the theory (and I am going to relentlessly quote him to Prove It To Our Illustrious Host) that a president doesn't exactly get to stamp his personality onto foreign policy; the majority of choices are overbalanced in one direction when you have that level of information, so most of the time the president is making the most strategic choice and there isn't a big D/R difference. The choices that are more balanced are the ones where most of the noise is made.
Just a theory. Can't prove this one either. It appears to make sense of the world though.
I can see that generally holding true for a normal administration - you respond to the crises you have, not the ones you wish you had, and the selection of viable responses is limited (though there is almost always at least the possibility to select from a range of intensities).
But with Trump and This War specifically... there was always an option to just keep on negotiating and pressuring non-kinetically - Unless Israel truly runs US foreign policy.
>Unless Israel truly runs US foreign policy.
Aside from not wanting you go around thinking that was the only explanation, I have a visceral reaction against that theory, possibly because I have a strong negative reaction to "explanations" like this that are applied everywhere, and therefore explain nothing.
What I've heard (and I think it was here I heard it) was that Iran was ramping up drone manufacturing tremendously, drones were proving themselves to be a huge headache to stop, the US navy can handle the overwhelm now but not multiple-x the overwhelm, the US ability to manufacture anti-drone interceptors was slower and more expensive, and those curves were going to meet really soon. Oh, and Iran showed every indication of acquiring a nuke if they could, and they had no compunctions against giving weapons to their proxies, AND their sincerety at negotiation (as opposed to stalling to let those curves meet) was paper-bag thin.
Role-play the president. What do you do? Was more negotiation really the right decision, given the expert opinions were the above? WHat other options look good?
They're opaque black boxes to ~everyone for Hayekian reasons; the cleanest solution is to eliminate government involvement where not essential in order to minimize the blast radius of uninformed politics.
Secondary benefit is that what resources are available for political comprehension could then be concentrated on fewer topics and thereby be less uninformed.
Strong agree. You don't need Omni expertise if you're only solving a limited, bounded set of problems. Especially when the default is don't get involved, let individuals and more local groups handle it. The further from the problem a group is, the less detail they should care about. Coordinate those willing to help and maybe move surplus government resources from one area to another, but don't try to solve .
Same goes for management.
Relatedly, it's also (a) further from journalists and (b) informed journalists are more monolithic, polluting the information space. Followed by (c) people increasingly voting based on ideology rather than personal output (e.g. "more cops reduce crime" isn't very complicated, but people in blue cities are more likely to oppose it if they vote based on tribalism or ideology). And ideology is more distant from everyday experience and thus harder to get right.
There was a moment of Defund the police insanity in 2020 but cities like New York, Chicago, and Los Angeles under Democratic mayors have generally increased police budgets in recent years, often with little sustained resistance from the broader electorate.
Post George Floyd a ballot measure in Minneapolis to replace MPD with a “Department of Public Safety” was rejected by voters in 2021.
There's an equilibrium here that gets worse the more ideological people are - for example 2021 New York, in the backlash to the defund the police, elected a moderate pro-police mayor, but then in 2025 elected someone who'd been part of the defund movement as personal crime concerns dropped and ideology (partly motivated by Trump's reelection) rose. The higher the ideology to personal concerns ratio goes the more people are likely to vote based on (generally worse informed) ideology and while this swings back and forth over time it has broadly been trending up.
Yea. In Chicago where I reside there was just a one-year attempt at a modest budget cut, which didn't work anyway because of overtime costs. Chicago Police Department budgeted/actual expenditures year by year:
2020: $1.64b/1.51b
2021: $1.56b/1.62b
2022: $1.70b/1.73b
2023: $1.71b/1.86b
2024: $1.74b/1.94b
2025: $2.07b (actuals not reported out yet)
2026: $2.11b
Prez Biden SOTU speech, on March 1, 2022
“We should all agree: The answer is not to defund the police. The answer is to fund the police with the resources and training they need to protect our communities.”
IIRC Biden put heavy emphasis on the word ‘fund’.
I want to roll my eyes right now but my doctor says that if I keep doing that it may cause my ocular muscles to spasm and eject my eyeballs.
Do you have any evidence at all that “the Biden PR team” put forward the defund idea so Biden could knock it down?
There's a pessimistic interpretation here in which this worked historically because public opinion was ~dictated by different factions of elites who were capable of understanding these systems (whether through intelligence, experience, institutions, better capability for social collaboration, etc.), and democracy was mostly a way to keep the masses content with elite decision making.
In that light, let me put it as: "The Elites are just about on the verge of not really understanding it either." The waterline of complexity just keeps on rising and drowning the minds of ever greater fraction of the population.
> Understand why it’s a bad idea
Sure, but what would be a better idea? You can do tariffs, which are hard to get right. Or you can do subsidies, which mean you have to give up something important (you can’t say you’d take money from a program you don’t like, if you could you would’ve already). Or you can reduce bureaucracy for local producers, but that’s also really hard. Or you can reduce safety standards, which is a very taboo trade-off. Am I missing something?
These might be reasonable savings, but wherever you get the money, there’ll be an opportunity cost. Suppose you saved money like this, and in the hypothetical, you get to decide where the money goes. You could spend it on antibiotics capacity. But you could instead replace all the missiles the US has been using. Or you could invest in military drones, or fix the FDA, or reduce government debt, or any number of things. If you want to throw money at a problem, you’re not just taking it away from wherever you saved the money, you’re taking it away from *everywhere* you could have spent the money instead.
Is it fair to summarise, you said where the money could come from, so I said opportunity cost means you can’t say where the money comes from, so you said where the money could come from again, but louder?
Thoughts on the NYT's article today claiming Adam Back is Satoshi Nakamoto?
I'm more of a Hal Finney guy. For one, the fact that he's dead helps with how it's been able to stay secret so long. But more importantly, he was signed up for cryonics, and wouldn't it be an amazing story if he actually did get rebooted someday in a far future where Bitcoin is even more valuable, and it turns out he has Satoshi's key as a brainwallet?
It seems to be written by the investigative journalist equivalent of Inspector Gadget without the gadgets.
The guy is just woefully unfamiliar with the tech community, and it causes him to pounce on things like saying "abandonware" and "hand tuned" or the already-pointed-out comparison of Napster and Gnutella like they're clues rather than something that basically any geek of the 2000s would have a high likelihood of mentioning at some point.
There's also "clues" like, uh, Back made a t-shirt to circumvent encryption export restrictions (which the author seems to think was far more novel than it was - this had already been done with the PGP source code (see https://hoffmang9.github.io/free-speech/the-history-code-is-free-speech.html) and would soon be done again with DeCSS for DVD encryption) and Satoshi embedded a newspaper headline in the first transaction block, which is *virtually the same thing* if you think about it while huffing solvents. Also, both Satoshi and Back were frustrated by spam and believed in open source! And, also, as if that last also wasn't enough, Satoshi's birthday is a reference to the 1933-1975 ban on private ownership of gold, and in 2002 someone on a mailing list made reference to the gold ownership ban and Back responded with a question about the given rationale. **In 2002**. People, the tranquility of your private life is in the hands of people who actually think like this.
Anyway, armed with this "convincing evidence", he then starts pestering Back, who stops responding to emails. Clearly a sign that Back has something to hide!
He then does a stylometry comparison, which is inconclusive - *but if you jiggle the numbers*, then it's conclusive! Then he starts looking at hyphens, and what do you know, Back and Satoshi both mentioned WebMoney and hyphenated "proof-of-work". Gun, meet smoking - we have our man!
Finally, after more hyphens and typos (to be fair, Back confuses "it's" and "its", so a part of me thinks he deserves everything he gets), the author goes to confront Back in person at a Bitcoin conference. Back finds himself suddenly forced to try and prove a negative to a moron who's already made his mind up. His nervousness and inability to precisely recall events from 2008 make it clear: Penny, Brain the dog, we have our man!
I think the author's overall, ah, "detective instinct" can really be summarized by the two paragraphs near the beginning of the story:
> One thing that jumped out in Satoshi’s emails to Mr. Malmi and in his other writings was that he mixed British spelling and idioms with American expressions. Since many Satoshi suspects are American, some have speculated that he disguised his prose with Britishisms. But I never bought that theory because of a clue Satoshi left us.
> In Bitcoin’s first block of transactions, Satoshi embedded text from a newspaper headline: “The Times 03/Jan/2009 Chancellor on brink of second bailout for banks.” The headline in question appeared in The Times of London’s British print edition. This felt like a sign that Satoshi really was British.
No guy who was merely *pretending* to be British could possibly think to use an actually-British newspaper! Why, has anyone outside of Great Britain even *heard* of this so-called "London" Times?
I may be biased because (although I'm not a bitcoin guy) I think it's quite callous to try to deanonymise Satoshi, potentially painting a giant target on his back, just for the fun of it. But some parts of the article seem like a real reach on the part of the author.
For example, the "fascinating coincidence" that "Mr. Back had drawn the exact same comparison between Napster and Gnutella" as Satoshi, i.e. they both pointed out that Napster's centralisation made it vulnerable and Gnutella avoided this by being peer-to-peer. But the whole fuckin point of Gnutella was to be decentralised Napster! Thousands of people would have made the comparison.
More broadly, a lot of it is basically noticing that the guy is a British cypherpunk and spinning that out into (what are presented as) multiple threads of evidence. It feels a bit like the author is using outgroup homogeneity to his advantage.
There also seems to be some motivated reasoning going on. In the case of Back, the author is happy to posit that he went to great lengths to conceal his identity, including writing emails to his real name from his pseudonym in 2008; but in the case of Nick Szabo, we can dismiss him because he "exposed his ignorance of basic technical aspects of Bitcoin" in a twitter discussion *after* he had been accused of being Satoshi. (Note that I'm not at all suggesting Szabo is Satoshi; I'm just pointing out the double standard here.)
The entirety of the author's argument against the theory that Satoshi was actually a group of people is "I didn’t buy that theory, either. The more people you let in on a secret, the likelier it is to leak. Satoshi’s secret had remained airtight for 17 years."
And apparently it's extremely suspicious that Satoshi used the word "bloody", Back (when accused of being Satoshi) claimed that it's "not a word i use", and... the author found a single example, from 25 years earlier, of Back using the word.
Then the stylometry: when it fails to get the answer the author wants, that's fine because Back was consciously disguising his style. But when we cherry-pick the right set of stylistic quirks, which this time Back was too dumb to disguise, we've got him!
I'm also put off by the fact that the article leads with (paraphrased) 'I could see in his eyes he was lying' and closes with (also paraphrased, but the full section really is this dumb) 'during a conversation in which I was accusing him of being Satoshi, I presented him with the Satoshi quote “I’m better with code than with words", and he responded by saying "I did a lot of talking though for somebody, I mean … I mean, I’m not saying I’m good with words but I sure did a lot of yakking on these lists actually", and that basically clinches it, because he was implicitly acknowledging that the quote was his!'
A) There are probably some cases in which revealing Satoshi's identity makes sense, but in this case it seems like it's being done out of idle curiosity.
B) I still don't understand why they're so convinced Satoshi is a single person. Seems like it could be this Back guy plus a bunch of other people (Hal Finney?) writing under a shared pseudonym.
The NYT just can't stop making enemies, can they? I guess that's the fate of all journalistic endeavors for "truth"...
It was pretty convincing to me (though I didn't know much about it, so the author could just be cherry-picking evidence). I agree that it's hard to ignore a nice mystery, but this is still just doxxing, isn't it?
I kinda wish Satoshi's real name turned out to be Adam Selene.
I didn't read it because I actually prefer Satoshi remain anonymous. One of the few modern-day "legends" we have. Also just from base rates it seems implausible that anybody alive could resist the urge to move some of that money -- $150 billion!
> Also just from base rates it seems implausible that anybody alive could resist the urge to move some of that money -- $150 billion!
Back (under his "definitely-no-doubts-about-it" identity) got into Bitcoin early as well, so it's entirely plausible that he has plenty of Bitcoin in wallets that aren't tied to the Satoshi identity. That would significantly reduce temptation to touch the Satoshi wallet.
Why "if you touch it, you're dead"?
https://www.bbc.com/news/articles/cgrl4l1y9yxo
He denies it.
Exactly what Satoshi would do.
It's also what not-Satoshi would do!
Unless you're Craig Wright, of course.
Technically, there has been a Satoshi that has been ruled to be not-Satoshi
https://en.wikipedia.org/wiki/Craig_Steven_Wright
so not every not-Satoshi would deny it.
Unclear. Maybe he would "confess" once identified and it just hasn't happened yet.
Fun fact: Satoshi Nakamoto translates loosely to "Central Intelligence."
I thought it was persuasive.
I dunno. But when I ask GPT questions like this I usually get good answers.
Never mind, I figured it out.
Seems worth a news story to me. I would read about that. Why are they doing that? What are the consequences? How might this cause us to update various beliefs?
Why wouldn't it be?