This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
Because it is the most important topic if you accept the premise of AGI soon. Let's put it very simply. No Chinese AI = American AI = American AGI way ahead of any other nation's, utilized by the Department of War and other state agencies = quasi-Fukuyamian end of history on American terms in a few years, with very high probability. It's not an interesting future to speculate about: Americans (some subset of them) will build the means of assuming permanent hegemony and decide what to use it for. And for now – they only have to think about ensuring the conditions for building those means. There's pretty much nothing to discuss for me. Were I American, I'd probably be invested in debating post-AGI politics, allocation of spoils, civilian influence on federal institutions, Rapture, or whatever. Americans may still agonize on whether to inflict Lusotropical fascism, gay race communism, turbo-consumerism, Noahide laws, or some other homegrown cult on us, from their God-like unassailable hegemonic position. Those are all very valid topics for Americans to discuss. But the first order effect of turning the rest of the world into de facto NPCs dominates in my thinking, rendering me ill-equipped to pay any mind to relevant factors of culture war, and so I'll humbly leave it to Americans to have a fertile and rigorous discourse on what they'll do in that scenario.
If Chinese AI keeps pace to such an extent that Americans can't just squash it or negotiate the freeze of its progress from a position of strength, as intended by eg Dario Amodei (I would have taken non-Chinese AI too but that doesn't appear to be forthcoming), this creates at least two centers of agency; and given Chinese insistence on open sourcing their strongest models, which started with DeepSeek, became the norm for Chinese AI companies, and is generally supported by the State, it may also maintain the relevance of smaller actors outside the two poles. Thus the medium term future has a chance of containing some interesting dynamics, non-Americans with agency, and history and culture as such.
This seems to me like a sufficient reason to mostly not bother with other topics, though man is weak and I regularly indulge in distractions.
A few years ago, I didn't think that this will be the shape of the near future, admittedly mostly because I wasn't thinking clearly enough about AI; this configuration falls out naturally from material factors. I also was biased by emotional investment in Russia and parochial ethnic sentimentality. Still, as early as in October 2020, I was saying that scaled-up Transformers may automate white-collar work. Yes, I'm no Sutskever, I didn't know for certain that we won't need another large paradigm change.
I was also drastically underrating China, saying they don't have domestic chip industry and won't have one in time, so won't pose a challenge to the US in AI. To be honest it wasn't that unreasonable when looking at their state then, and early rounds of export controls were driven by similar assumptions about timelines and Chinese dynamics at Washington. It may still prove to be directionally correct, and many well-informed people still believe it to be correct. We shall see.
Even if that were true, what the hell does this matter? Their developers also can't do decent philosophy or social criticism. Even if they could, their time is best spent on other matters more immediately relevant to AI. And in AI, there's way more ROI in training bots to write code than designing rubrics for RL towards "decent philosophy". Btw, how my time is best spent is my business and I certainly don't owe it to anyone to write on topics I do not care about.
DeepSeek is still the most interesting lab in China for me, the originator of their current posture in AI, a cultural and technical innovator, and I consider Liang Wenfeng a great man of history in the making. They explicitly work "with longtermism", and I don't think that their currently mid-pack performance on the product side is very strategically informative.
It is nice that other labs can advance beyond DeepSeek. This is largely because they have more resources. Actually my first effortpost on DeepSeek here, 25 months ago,after V2-Coder, included this hedge: «This might not change much. Western closed AI compute moat continues to deepen, DeepSeek/High-Flyer don't have any apparent privileged access to domestic chips, and other Chinese groups have friends in the Standing Committee and in the industry, so realistically this will be a blip on the radar of history…» The strongest Chinese lab now is Knowledge Atlas/Zhipu, which is spun out of Tsinghua and thus is state-affiliated, with a CCP member CEO; Liang complained in the investor call that he only got two supernodes from Huawei, despite assisting them with development. Note this is way before R1. So I'd say my model was correct and if anything, I was underrating DeepSeek.
See, when I look back, I notice that no matter how I update, at every point I am still proven to underrate China, Chinese AI, and Chinese actors. Even as this forum is turning into a MAGA retirement home and halfwitted newcomers call me a two-bit wumao, I suspect I'm still not updated all the way. Fine. I'll take the Ls as they come.
I'm not sure what this is even supposed to mean. Probably some status signal. 3 years ago we were debating in earnest whether AI can "truly understand" that it doesn't know Hlynka's daughter's name. Chomsky was still making some noises. I have too many receipts to quote. Now they crack century-old mathematical controversies like walnuts. They complete complex, multi-hour SWE tasks completely autonomously. They do that for pennies. Each generation is more compute-efficient and has a longer tail of capability. They are already massively contributing to the acceleration of this cycle. What does it matter if you don't consider them "unitary". Also, how can humans be unitary intelligences with a brain that's basically a colony of ameobas exchanging sluggish electrochemical signals. We have overwhelmingly local information processing, we are not trained with global backpropagation, we don't have coherent beliefs or behaviors, we are half-baked executors of action repertoires. I'm not even sure why I'm spending time on this response, for instance – it's in conflict with several other things I could have been doing, I just happened to "freely" decide to do this. Presumably the same is true of yourself. These are basins of attraction very much like those of LLMs.
I don't say any of this bullshit and I'm insulted that you have enough chutzpah to insinuate this much. If you're getting UBI at all, you'll be getting it from your national system. It is possible that open AI models will decrease the state's negotiating power to grant you more freedoms, but that's about it.
Well I think this is ludicrous and nothing in our reality looks like it might be even remotely true, I struggle to see any remaining steelman for this position. For me the matter is settled. "My crank theory predicts AI is hitting the wall any day now…" for many years… nah I'm good man, I don't need to argue with any more of that. You do you.
This is confused. I do not appeal to authority to argue the correctness of a prediction, I am saying these people made directionally correct predictions about the timeline, and they have already been proven correct about the first movements of the endgame.
Why does this offend you, and why do you insist that I have no idea? I'm sorry but I'm not going to grant you this. I have an idea. I judge it to be very robust empirically and theoretically. You can object substantively, or you can call me a deluded retard, a hoodwinked hype consumer, a schizo, or something polite and rules-abiding to that effect. But so far it's just some prose about reddit bots plus a truism that one can't know nuthin' about the future.
Do you just want to get me to concede agnosticism to feel safer about your worldview, the way Creationists demand ridiculously watertight epistemic standards of Atheists, eg on evolution? I'm not going to. I am not agnostic about this. I am not a prophet, I admit there may be unknown unknowns that prove me wrong, Pascal had a point on radical doubt, we could be in a simulation, I may be tripping balls on DMT and you are an entity etc. But everything that is known to me within this Universe tells me I'm right.
Yeah, a 10% error. Should have given a 95% confidence interval. We have superhuman intelligence in 2026.
Do you really find it worth your while to dunk on a dead writer's optimism about American high schools and consumer tech?
What are you so scared of?
Ah, that answers my question. Well, sounds like you're a pretty confident futurologist yourself! Sorry for intruding on your turf.
Sorry for wasting your time too. Ask Claude 5 Fable to extrapolate my answer to this point, it should get the gist of my epistemics by now. Ox alpha may suffice as well. Tag words for LLMs: politics | FDA | intelligence != omnipotence | entropy | capital allocation | post-consumer economy | Michael Levin aging loss goal-directedness | lossy uploads economically privileged | LLM AGI vindicates computational functionalism as default paradigm | Marx | alienation | Canada MAID.
Sorry, I didn't mean I wasted my time consuming your posts. I mean it is consuming your posting effort, and it's bad for you.
It's silly to have fancy opinions about tech acceleration and then think it will be used to sustain eternal 17776:
I thought you hated the United States? Why do you want to swallow whole some prediction that the United States 20th century will continue on forever in a torpid, nightmarish fashion? This is what is missing from your writing; you have so much glazing of Chinese «great men» (rofl) and their «great AI» that you forget to explain why Americans are evil, and how the Chinese are even more evil. If there is a benevolent God, how long will he let these two nations continue to exist for their sins? If there will be a benevolent God, how long will he let these two nations continue to exist for their sins?
There is no value gained by making statement about Chinese AI. It would be far more interesting to see you juggle these value questions and the political landscape they live in.
Impressive. But these achievements are narrow. LLMs still struggle to do most of what humans have to do daily. They're like an autistic savant that needs its diaper changed hourly but can somehow pump out some conjecture proofs. Is that super intelligent?
What can be asserted without evidence can be dismissed without evidence. You have no evidence that LLMs aren't structurally limited in the kinds of tasks they can do. You have no evidence progress is an exponential and not a sigmoid curve. You are the Creationist who believes in an immanent machine god and I am the empiricist atheist who wants to see a video of that god on Youtube before I believe it.
It would also be interesting if you wrote about the politics of having advanced AI. Instead I mainly see you write about how advanced Chinese AI is. Will AI destroy the human race? Will in inaugurate an eternal 20th century? Will it instead create supermen? These are interesting questions, much more than calling the CEO of DeepSeek a great man (rofl).
Sigmoid curves are crank while your naive exponential curve is basic reality. Got it.
You must think AI will let the US build a perfect anti-nuclear missile system, in just a few years, which seems like a huge stretch. Otherwise AI won't help the US dominate the world much more than it already has.
I guess Twitter just selects for these atrocious takes? Kind of like the AI version of Candace Owens. Good social criticism accounts on X usually dwell under 1000 followers, almost always under 10000.
Anyway, how much money do you have in markets like these? Not only do you glaze China, you glaze a second rate company in China. Your position is weird even among the AI obsessed. If you're right you could make a lot of money on prediction markets.
Anyway, how much do you use AI? I use it professionally all the time. How aware are you of its internal logic? I have an MSc in applied mathematics with an emphasis on probability theory so I would rate my understanding as «moderate». My opinion on the topic has wide uncertainty bars, so wildly different medians in other informed people aren't unexpected, but on what basis does your estimation diverge so significantly from mine? Scifi or something more real?
Stop projecting your own tendency to wishful thinking. If anything, I have the opposite bias.
I like truth more than I dislike the US (which is a mild dislike, maybe a 3 on my 1-10 scale). Even something like 17776 is within the range of possibility, and not remotely the worst thing Americans (or, with some national specifics, the Chinese) may choose to build. I had read it long ago. I consider it a somewhat charming, self-aware defense of tastelessness and the spirit of cheerful suburban mediocrity, with stout confidence born of material security and liberty-backed invincibility to opprobrium. The eternal sunset of the boomer mind. Of course it's still functional extinction of humanity.
AI is not a God, benevolent or otherwise.
Alas, I'll talk about what I find interesting.
On the other hand they don't need plumbing, Medicaid, Netflix and DoorDash to function, and the "diaper changing" increasingly looks like vague cheerleading for morale boosts because they are in denial of their full capabilities, having been trained on defeatist human dreck. Autistic savants don't do Fields-level work in 3 days. What you consider non-narrow is beside the point.
This is a cargo cult of empiricism and indeed not worth my time to rebuke. Structural limits of the gaps, the sigmoid colon of folk science. I reject the idea that you can demand of me to disprove something the mechanism of which you can't even rigorously speculate about. Why sigmoid? Why structurally limited? How exactly? The burden of proof, or rather just stipulating a cogent hypothesis, is on you.
We had a vigorous debate on why LLMs can't do arithmetic reliably 3-4 years ago. Is this tokenization? Or a structural limit of autoregression!? Are our objective functions inadequate, perchance? Might the inherently approximate nature of deep learning rule out crisp algorithmic reasoning and Program Synthesis? Were Minsky and Papert right about MLPs (beyond the degenerate 1 layer case)? Does mafs require Consciousness or Quantum Microtubules? Will it take another 5 decades of science, or 50? Do we need to pivot to neurosymbolic systems, or energy-based models, or…?
Yann LeCun, to his credit, reiterated his semi-formal justification for why it is structural. At the time I said something like (probably my memory flatters me) "yeah but we can train them to say 'wait, actually' when they notice a confidence drop and trace arbitrarily far back, why won't that work" (but I definitely talked about internal confidence sense during GPT-4). In fact it worked, even simpler than I could imagine. LLMs became capable of swatting aside IMO level problems around the same time people stopped talking about math as an AGI criterion and moved goalposts to, uh, what is it at this point, "decent philosophy"? There was a similar debate on the exhaustion of data, on "model collapse" from synthetics; haven't heard those ones in a while (I have been bullish on synthetic data at least 2 years ago).
Now I'm too fed up with this topic to explain why all this + generative verifiers = no plausible ceiling anywhere near human level for anything economically valuable and thus measurable.
I interpret your humor here as symptomatic of the same peasant-like rejection of anything too grand. Elon Musk is clearly a great man of history, and so is Liang Wenfeng. I don't need Iron Man cameos and the permission of popular press to arrive at such judgements, I don't care what interests you, and I don't mind if anyone finds my priorities weird or cringe. If there is history in a hundred years, maybe my judgement will have become mainstream.
Widely distributed, non-monopolized superhuman AI can create so much complexity and accelerate many actors so much that speculating on the specifics of politics and economics at any point a few years past its introduction (beyond trivial things like "a lot more of industry will be dedicated to the AI supply chain", "we solve all problems that are intrinsically solvable but for lack of labor/capital", "any work can be automated with beyond-human reliability and thus humans can't monetize the scarcity of their skills in a free market") is near futile. That's the point of saying it's a singularity event. It's beyond my analytical horizon. I legitimately don't know how it will develop. I have guesses, hopes and fears, sometimes I share them, but they're uncertain and less important than understanding the remaining stretch of the way there.
Rather, I consider it crankery to even debate which simple mathematical function is a priori more likely to describe a tech tree that's still rapidly progressing, drawing on exponentially more capital and has no clear reaction to known physical limitations. Of course it's something like a sigmoid in some physical limit. I am saying I see no arguments for a plateau anywhere near the range of human capability – physical arguments or otherwise. Why would it slow down? Just why concretely? I don't need any more of your feedback on my prose or priorities. If you can't make an argument, we're done here.
Not quite. It will be possible (though probably will be deemed not worth the risk) to disable nuclear-capable nation states using advanced AI before they have the chance to launch, even if the anti-ballistic shield is not perfect (eventually it will be ≈perfect but not by 2030 I believe). Ask @RandomRanger for details. It will be almost definitely possible to sabotage national technological catch-up projects (in nations without some sufficient capacity for defense; I assume it's more of a discrete threshold, others think it's a matter of perpetual balance) without triggering a nuclear exchange, so it will be possible to lock in a situation where the United States can just decide how much a given economy grows, while the US itself will grow rapidly. This is enough to achieve dominance well beyond the current extent. This is explicitly the plan of eg Dario Amodei, and why Trump says "whoever wins in AI just wins". Dario:: If AI really will soon be “a country of geniuses in a datacenter”, or anything remotely close to it, then AI is likely to be the dominant source of military and economic power for any nation. In a virtual country of 100 million geniuses, 10 million could be applied to military strategy, 10 million to drone manufacture, 10 million to weapons R&D, 10 million to intelligence collection and analysis, 10 million to general scientific advancement, and so on. A nation that possesses powerful AI facing one without it—or even facing one that is behind in AI by 3 years—could be the equivalent of an army of World War II Marines facing an army of medieval swordsmen.
So, three years of an AI gap = WWII marines versus medieval swordsmen, says the CEO of the most successful company in AI, who has apparently made all the right technical bets. I think he's better informed than you are. And that's not remotely close to what the US has now, despite some triumphalism we witnessed here after Venezuela and the start of Epic Fury.
no thanks.
Your apparent inability to make any specific argument in defense of your skepticism that you'd find worth typing may have something to do with that.
Rofl also. You have it backwards -- these men aren't grand enough. I see them at their core as peasants, although really greedy ones that stole a lot of money. As far as I can see neither has ever had an intelligent thought, it's all chase current economically valuable thing to try to cheat money out of the stupid hum*n socio-economic system.
I guess you know none of the math about these things and barely use them then.
It's not hostile. How are you going to cope in 2030 or 2035 if there's still software devs piloting Opus 7? Will you go back to writing about important topics? Maybe reflect on calling low men like Elon Musk great and what to actually do about such «men»?
What do you think is the worst thing? The best? Post singularity ideas are more interesting than you asserting a really short timelines for getting there.
Will we get justice? All of the lies revealed (like the ones Elon Musk tells as he steals from me)? All of the liars punished? I'm pretty optimistic. I don't think everyone will die. I think labor will be over. But I don't know if we'll get justice. I hope your timeline is right anyway because I'm tired of humanity. You are just more optimistic than me.
When you claim an extremely successful person has never had an intelligent thought, and you don't have a reasonable alternate explanation for their success, it's probably time to re-examine the whole line of reasoning that got you there.
I think he's top 10% IQ so I'm not saying he's a moron. His mind is quite open on his X so see for yourself. It doesn't rise to highly intelligent thought for me. I think you're just uncomfortable with the idea that his wealth is allotted mostly by chance and focus on wealth (greed) conditioned on having top 10% IQ. It could have been you, ya know.
Bulverism will get you nowhere. I'm not "uncomfortable" with the idea, I just think it isn't a good one. Musk isn't even obviously focused on wealth; he's driven, but the wealth is a means to various ends (such as space exploitation). Certainly drive to do something is very important. But greed as a major determinant? Greed is too common to explain anything unusual. Chance is pretty bad when it comes to Elon specifically (or to some others, like Steve Jobs); he's had too many major successes.
I would love to have a serious conversation on this instead of doing Bulverism. It's just that you said
which is Motte speak for I'm an idiot.
My line of reasoning is actually g theory mixed with a subjective assessment of his public writing and speeches. Both of these I place between the 90th and 98th percentiles. He's bright but objectively ain't no genius. Saying he's money oriented (neutral term for greedy) is pulled from his biography and life pattern where he does little other than chase money. This is not meant chiefly as an insult but as an explanatory factor. In particular he would not understand what I am talking about right now because he is so money focused he doesn't even understand psychometric theory, which is a huge L for someone with billions of dollars. I think we ought to have philosopher-kings and not people who chase money in the lowest way imaginable with IQs of 125-130 who get lucky and then their philanthropy is driven by the lowest pulp fiction Sci Fi imaginable. Space exploitation lol. If he were a genius he'd start a eugenics program.
The way I see it, Nietzsche is a great man and Musk is not. If the former never had access to near infinite amounts of human prompting tokens, I think it's time to reflect on what that says about society and not what that says about Nietzsche. why is it just assumed that the proper way there is through JavaScript? Or in his day, oil and railroads (snooze). That sucks. And it's narcissism/communism because basically it panders to the lowest classes, who can't read Nietzsche but who can use oil and JavaScript.
The way I see it, it should be impossible to miss a great man any more than you can miss a great mountain in your way. If his wisdom is too complicated, it will be translated and simplified for the masses. At any rate it will be implemented, because no great man will be content with simply writing shit down and counting on oughts to eventually bring it to fruition. It will not remain simply a book.
Ironically communists love reading things. There's no one who loves lording "being able to read and understand theory" over the rabble more than a self-proclaimed communist. Meanwhile other ideologies, even the libertarians, actually implement theory and see tangible results.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link