Reverse centaurs are the answer to the AI paradox (2025) | Hacker News Reader

Reverse centaurs are the answer to the AI paradox (2025)

pluralistic.net

113 pointsjason_s71 comments

(2025)

Doctorow has a book-length version of this, "The Reverse Centaur's Guide to Life After AI", which I just read. It's not one of his better works. It's all conventional anti-AI wisdom. Too much of it is about the impact of AI on the punditry and writing industries. I expected better from Doctorow.

I would have expected more focus on the relationship between AI and power. Doctorow has been down that road in his fiction. There's worry about government oppression, but government oppression has been around for most of recorded history, and probably peaked with East Germany. What he doesn't get into is how AI empowers corporations to be more obnoxious. Which is strange, because he's been into that in his fiction. See his "Unauthorized Bread".

One thing AI does is to make corporate control of individuals cheaper. The classic Big Brother peaked, as mentioned, with East Germany, where about 10% of the population was involved with the Stasi. Only a government that didn't have to make a profit could do that. Corporations couldn't go in for that level of surveillance because it wasn't cost-effective when it was labor-intensive. With modern surveillance technology to collect data, and AI to analyze it, it now pays. See Flock, Google, etc.

Does anyone know what incentivised citizens to cooperate with the Stasi?

They were arguably the most repressive secret police regime in the world. They disappeared people and their families, and used torture, intimidation, and the biggest network of informants ever assembled (in some cities almost 20% of the population were informants!).

People collaborated because they had to, on top of the usual reasons for joining a military or intelligence service.

Probably the same as for ICE today.

Eichmann in Jerusalem: A Report on the Banality of Evil

Hitler's Willing Executioners: Ordinary Germans and the Holocaust

Ordinary Men: Reserve Police Battalion 101 and the Final Solution in Poland

Popular Opinion and Political Dissent in the Third Reich

The Stasi and the Nazis were two different groups in different time periods. I am only wondering about the Stasi.

The core is the same - totalitarianism.

Same list of things that compels people to become an asset for any intelligence agency, adjusted for the time in history.

is there any good ai book worth reading ? I dont mean things discussing building a tranformer or whatever . I mean like high level ones.

Heard good things of Empire of AI By Karen Hao

> This is a political act of resistance. Margaret Thatcher's motto, after all, was "There is no alternative," by which she meant, "Stop trying to think of alternatives." The bully's trick is to present your defeat as a fait accompli: "Resistance is futile."

More broadly, that trick is the most effective and important today: No matter the issue, people - including on HN - will tell you how powerless they are and how pointless it is. It's such a social norm that they will actively resist and attack anyone who violates it and suggests otherwise - goodness forbid you have hope, an idea, or want to do something. I see many adopt bizarrely false perspectives and opinions for which the only sensible explanation (IMHO) is they are strong defenses against action and empowerment - many/most won't even talk about serious issues like climate change (see - I bet you're triggered by the suggestion of discussing it).

People never accomplish anything without believing in it first. It's old, basic military psyops to preach hopelessness to the enemy soldiers, in dropped pamphlets, radio broadcasts, through propaganda campaigns. Sorry, I forgot - we can't talk about those things either.

Until people believe, nothing will change. When they believe, everything will change. That's why there is so much impetus toward despair.

> Until people believe, nothing will change.

I'm with you here.

> When they believe, everything will change.

Taken literally, this is absurd. Belief might be necessary for change, but it's not sufficient!

> That's why there is so much impetus toward despair.

Sometimes people really are in hopeless situations. It seems cruel to blame their despair on not believing enough.

> Taken literally, this is absurd. Belief might be necessary for change, but it's not sufficient!

I don't read this as "if you just believe, you can do anything!", but rather "when the populace as a whole starts believing that they can fight back against an oppressive minority, they can definitely do that."

> Sometimes people really are in hopeless situations.

The fight for despair is surreal; at least people believe they can really be hopeless and won't let anyone stop them.

I don't find empty platitudes like "you can just do things" all that inspiring. I prefer seeing a specific example of something you could do and an explanation about why it might help.

Ok, here goes:

You could look at your situation, think about it as objectively as you can (ignoring, as much as possible, any media driven mind worms), pick any particular problem you face that looks solvable, and try to solve it. If none of the problems seems particularly solvable, break them into pieces and look at the resulting sub-problems in the same way. Regardless of the outcome, iterate, since it's ultimately a numbers game.

The reason this might help is that 1) you might solve some fraction of your problems, 2) to the extent that you're suffering from learned hopelessness, you'll be unlearning it, 3) regardless of either of the above, you'll be likely to learn something, as this is a very common side effect of trying to do stuff.

Hope that helps.

That’s a generic algorithm, not an example. I mean specific examples. Like, people sharing things they did.

No, it's a specific thing you can do.

You can literally just sit down, right now, and do this.

But if you insist on hearing about a specific example of something I did, here's one that's appropriate. I was frustrated that all the time I spent reading/hearing/watching about other people solving their problems left me oddly dissatisfied. I realized that that was because it wasn't actually solving any of mine (except perhaps assuaging boredom a little). Thinking about it, I realized that that was because I was wasting the impetus to improve my situation on voyeurism.

> I prefer seeing a specific example of something you could do and an explanation about why it might help.

Great! Come up with an example!

I mean, that's an extremely broad request.

You can install solar panels on your house, and a heat pump. This will help reduce fossil fuel energy usage.

You can self-host Mastodon and help move your friends and family members still using Xitter onto it. This will help reduce the support being given to fascist oligarchs.

You can find a local soup kitchen, or animal shelter, or library, and volunteer some of your time there. This will help provide for people and animals without other support.

Trying to do something bigger than that is going to require more people, and may require resources you don't have; I don't know your situation. You haven't really given anyone much to work with in giving you examples, and the post you're replying to a) isn't really giving "empty platitudes", and b) is talking very broadly about problems in the abstract, so there's not much guidance there as to what kind of "something you could do" you're looking for.

The only way out for climate change is through. While we're doing crazy things like data centers in space, we "just" have to capture CO2 from the air and store it. While that will take some revolutionary breakthroughs in materials science and engineering, all we can do is dump money into those areas and hope they pan out. And despair a little less.

We can reduce climte change-causing emissions. We could do it dramatically if we wanted. The only thing stopping us is that it's hopeless.

AI for his own pundit profession is not allowed but "open source hackers do amazing things with AI"?

He evidently has no clue about open source (people just plagiarize with AI and don't do amazing things).

If he wants to build resistance as stated in the last paragraphs maybe he should be a bit more careful.

Plenty of people are doing novel hard work with AI in open source: https://news.ycombinator.com/item?id=48867961

So novel the person you link to says he's mimicking an existing CAD software

"I reimplemented a state-of-the-art closed-source thing in OSS" is extremely valuable and novel work.

It may be valuable.

It seems like a stretch to call a reimplementation novel.

Darn, I meant to link to https://doctorow.medium.com/https-pluralistic-net-2025-09-11... which is easier to read because of the typeface.

I am reading on mobile and actually preferred the pluralistic link, glad you posted that one.

> what crypto will leave behind is bad Austrian economics and worse monkey JPEGs

its what the crypto industry will leave behind, and you could say the main bubble already burst when ftx went bankrupt.

the defining property of public blockchains is they are designed to never go down, and the biggest ones like ethereum probably never will until its replaced by something objectively better. if you take off the scams (aka most of the industry) what youre left with is one of the missing pieces for a real fully decentralized internet.

first off, ethereum is a real alternative to iana and icann. todays open web, atproto and fediverse all depend on dns for name resolution. that system is managed by a single us based corporation thats technically controlled by a non profit but always submits to big tech demands with things like branded tlds and trademark disputes. if you replace that you get a single global authority thats fully transparent and runs on consensus instead of arbitrary decisions.

the other side is anonymous payments. the good thing crypto did in finance is not pumpfun type schemes or centralized stablecoins that run on their issuers word. its systems like railgun or tornado cash (rip) that give you the privacy of cash without risk of counterfeit money or getting set up at a sale. it doesnt matter if its illegal (and some uses of it probably should be illegal like a business doing tax evasion), just that its out there as a fallback in case digital payment systems turn against the public.

crypto is not the world saving end all technology all the scammers want you to believe it is. its way more limited in scope but what it actually gives us is one more tool to get around big tech and oppressive governments. imo thats worth all the damage it caused.

I agree that the infrastructure is useful, though not as useful as it said on the tin, and having it around for other purposes is a good thing. I'm not sure it'll persist as "one more tool to get around big tech and oppressive governments."

There's still a lot of work to be done before big tech and governments don't have an outsized influence, even in blockchain-mediated spaces.

Ethereum actually contains a time-bomb designed to make it stop working a few months after the latest protocol revision, which of course can only be published by Vitalik.

On a side note, I wish Doctorow would stop listing URLs in the middle of his articles, instead of using descriptive text with a URL anchor.

My screen reader always reads them in full ("H T T P S colon slash slash..."), which is very disruptive when listening to an article, especially with 14 interruptions.

So you are a cog, you are a watchmaker, or you just want to read the time hoping that all these cogs will tell it. Software engineers are becoming line workers, it's okay, we had a good run. But realise we needed the press to give way for mass literacy and scientific revolution. Isn't anybody at least curious how our children will leapfrog us and what they'll do once agentic thinking is engrained in their thinking?

Probably at least half of the people reading this shudder at the above. But I thank the Lord my job has been never to typeset a press.

Great analogy and really good points. I'm hopeful for a "slow pop" on the bubble though but we'll see! Very nice and so much easier to actually read a non-ai generated article too.

I feel so much less immiserated and precaratized.

https://tvtropes.org/pmwiki/pmwiki.php/Fanfic/LeftBeyond From 2015. Could a LLM be used to help coordinate a resistance movement?

Heinlein's The Moon is a Harsh Mistress is probably a better-known example of that.

Libertarian fever dream.

Semi-related: Boko Haram using LLMs for plotting, planning, and preparing for attacks http://archive.today/2026.07.10-132100/https://www.nytimes.c...

I have been fortunate enough to be in the centaur setting but I find my issue is simply that it’s often fools gold. The output is not correct, but in a way that looks correct, and I have to spend hours backtracking and hunting down the issue(es).

> A reverse centaur is a machine that uses a human being as its assistant

So like a political, military or religious machine does. We're already bred for it. We're Konrad Lorenz's geese, desperate to be led. It's the path of least resistance.

The machine may happen to be the Nietzsche's https://en.wikipedia.org/wiki/Übermensch . Nietzsche analytically extrapolated and foreseen the shape of the things to come, while the state of the tech at the time didn't allow to foresee the actual implementation.

>We're Konrad Lorenz's geese, desperate to be led. It's the path of least resistance.

Us being that way may also be considered a result of evolution driven by the natural selection enforced by the society (similar to how we breed cattle for weight and milk production or turkey for wide breast).

> while the climate costs of creating those models is terrible, they're here now

> It doesn't matter what happens to Openai; Whisper is here to stay.

Yes and No. They may be here now, but they will be outdated tommorrow.

From the article

His job was to be the AI's "accountability sink" (in the memorable phrasing of Dan Davies): he was being paid to take the blame for the AI's mistakes.

Doctorow has consistently one of the most sober takes on AI in my opinion. He acknowledges that we are in a bubble right now, but also understands that underlying tech has actual merit to it. And he does a good job explaining how the issues people project onto this technology isn't inherent to it, but rather stems from capitalist relations in our society.

Aren't those completely obviously points?

I see a lot of people struggling with them every day. Some people can't see that there's a bubble and think we're on a cusp of AGI. Others can't acknowledge that this tech has merit outside the hype. A lot of people seem to be at either one extreme of the spectrum or the other.

(2025)

“ AI is a bubble. If it wasn't a bubble – if it was just a bunch of computer scientists and product teams tinkering with possible uses for advancements in back-propagation, generative adversarial networks and machine learning – there wouldn't be any controversy here.”

Hard disagree with this. There’s plenty of AI controversy driven by the open source AI scene. The vast majority of AI art controversy, for example, seems to be between artists and guys training SD:XL LoRAs to make porn on CivitAI rather than between big frontier AI companies and artists. While CivitAI is definitely profiting off that, the controversy wouldn’t go away if the guys using data science techniques and their own hardware to train fetish porn models went to HuggingFace or their own volunteer-hosted forums instead of relying on their current distribution platforms.

centaur / reverse centaur seems to me like glass half full / empty.

tough for Cory to get one wrong, though.

I'm not sure his argument holds that foundational models will no longer function after the bubble pops. There's plenty of open weight models that are competitive which are more likely to exist in a world with abundant cheap GPUs.

In addition to what wrs said, it's true that the foundation models won't necessarily stop working after the bubble pops...but if new training isn't happening, then they're not getting updated.

Day 1, that's not going to matter much.

Day 30, it's barely going to be noticeable.

Day 365, you're going to be able to tell that you're using an outdated model—whether you're using it for general world-knowledge stuff, or for coding.

Day 1826 (ie, Year 5), they're going to still have some uses, but they'll be so out of date that they'll be actively detrimental in many areas.

https://qntm.org/mmacevedo

That is what he said.

>When the AI bubble bursts, there will be stellar bargains on GPUs…

>these standalone models can do amazing things

>The things these open source standalone models can do will only expand, and they will become a given for our computing applications.

I think of the “big” foundation models as the “fossil fuel” of AI. Once the bubble pops and we can’t afford to train any more of them, we’ll be distilling and remixing the ones we managed to make during this weird period where they were feasible.

Quaaludes are maybe a better analogy since the original, strong ones are mostly gone and the pills being manufactured today are significantly weaker. And their effect on Wall Street's decision making is pretty comparable to AI's

Based on my experience, even the models we have now are a huge benefit when properly used. And we probably have a decade or two of significant gains we can make just with harnesses, skills, heuristics, etc. even if no further progress were made on models.

They'll have stale knowledge though. It's like someone stopped interacting with society five years ago. Are they employable?

I am a developmental autism researcher using multimodal neuroimaging, standardized and experimental behavioral/cognitive assessments, with a focus on longitudinal modeling to study condevelopment of brain and behavior. I also manage lab servers, databases, I program pipelines for neuroimaging, and I also write grant applications. I also tend towards slow and rigorous over fast and loose in publications. This is both the reason why my publication land solidly mid-high tier, never publishing in low quality journals, but also not flashy enough to into the highest tiers (Biological Psychiatry is my top tier so far, a very good journal).

This preamble is to provide context: I wear many hats, and have to be familiar and competent over a wide range of things, and I must be comfortable enough to trust that I can do things I don't know, but scared enough, that I obsess in making sure I didn't get it wrong (especially if the result is suspiciously strong).

How I spend my day has completely changed since powerful AI. While I still read papers, AI often helps me find relevant papers quickly. Interested in surveying multivariate longitudinal statistical models, AI can provide generally competent overviews to help me decide on directions.

At times, I feel like it is a better neuroscientist than I am. It definitely has a larger breadth of relevant information at its immediate disposal. I'll run ideas against it, working hard to get it to not be syncophatic (usually by telling it that a local LLM suggested it but that I was unsure), and it can often raise points that I had not considered.

At times it can even seem like it is running the show, and I'm just getting in the way: Reverse Centaur.

But other times, I feel like it is helping me leverage my time to explore MY goals much more quickly, and it is empowering.

Also, Opus and ChatGPT have a real functional fixedness issue, and a hard time weighing the importance if various aspects of the research question/space. It will get stuck in solving minutea while missing the obvious large problem with an approach. I would be lying if I said that I did not get you in telling it this. One does want to feel useful after all.

AI today feels like a partner,and it is rewarding. I have yet to find a person that I can work this closely with. Usually other scientists are too busy pursuing another paper, etc.

Once AI surpasses me, this joy will go away. I am fine with a machine being better than me in almost everything, but if it's the thing I like doing, and if it's value is only judged by the extent in which it pushes the frontiers of science, I don't want to live in that world.

Most people do jobs that they are not in love with. I forsook higher paid careers in order to have a career where I can choose what to do, then work to make it happen. So I don't want to say that my experience and fears generalize, but I want AI to stay at roughly this level. A level where we aren't the meat tool, or worse, unneeded, but we are empowered for self fulfillment.

I feel like AI is set to surpass us on most left brain tasks, yes. I know left/right is a myth, please forgive me the shorthand. I see it in my own work where the AI is incapable of brute forcing a solution because it can't do the fuzzy wet pattern matching we do. The kinds of problems I can't solve in front of the keyboard either, but that come to me when I step away and take a walk.

CTF + F "bubble" - 15 results. This guy was manically predicting the bubble in 2025

> AI is a bubble.

> AI is a bubble. Bubbles burst. We're in for a near-total collapse of the AI investment mania

> AI is a bubble, and when bubbles burst, they sometimes leave behind a productive residue

> When the AI bubble bursts, there will be stellar bargains on GPUs

> The bad news is all the damage the bubble is doing now and all the further damage that will come from its collapse.

> After the bubble bursts, there will be the mass incineration of everyday people's retirement savings and the knock-on effects as the whole market craters

> Every day the bubble persists, the harms of today and tomorrow increase. We need to burst that bubble as soon as possible.

Holy. Shit. This guy is insane. He's been pulling the same schtick since 2014 calling many other companies a bubble. Why why is this guy taken seriously.

I don’t agree with Doctorow on this point, but there’s nothing inherently unreasonable about predicting an economic phenomenon is a bubble and having to wait several years to be proven correct. Michael Burry first purchased credit default swaps on mortgage backed securities in May of 2005.

A stopped clock is, after all, correct twice a day.

People say this about everyone who calls something a bubble, right up until it pops.

Or doesn't.

"The stock market has predicted nine out of the last five recessions." And people who don't put their money where their mouth is? The track record gets so much worse.

So far every single one has popped. The record isn't particularly good for "this time is different".

Not really. Not everything is a bubble. Cloud computing is not a bubble. And screaming “it’s a bubble” 10 times is clinically insane.

there's no prize for predicting 12 of the past 2 bubbles

Your whole argument for not taking someone seriously is that they use a specific word 15 times? That's your standard for insanity?

Reply on news.ycombinator.com