Hacker News Reader

Comment by anon373839 | Hacker News Reader

I think the industry is freaking out about open weights models in general, if not specifically DeepSeek. That is why we're now on the ~4th call for pacing the frontier from the very people who, if they wanted to pace the frontier, would simply do it rather than asking Washington to get involved.

And it's why people like Hillary Clinton have been trotted out to talk about the dangers of open weights models -- I mean, does she even know what that phrase means? (I know HRC is a controversial figure and I'm not bringing her up for that purpose; I just note that she and other prominent retired politicians are now doing the circuit on Anthropic's behalf.)

View on HN

> if they wanted to pace the frontier, would simply do it

I don't think that's a fair assessment. These companies are in a Nash equilibrium where they can't unilaterally slow down without essentially destroying their company. They also can't coordinate with each other, because that's illegal. Antitrust law generally prohibits competing companies from agreeing to restrict innovation.

That's the GP's point: "Pace the frontier" really means "slow China down because our investment is at risk".

What I'm trying to say is that their inaction in unilaterally slowing development is consistent with their stated beliefs and requires no other motivation.

Have you looked at this backwards, though? Meaning: have you considered what it would look like if the motivations are actually just plain old money/power, but a set of stated beliefs needed to be constructed to justify what's being sought? I think these absurd beliefs make more sense that way.

Sure. I'm not saying that they have no other motivations. I'm saying that, even if they had no other motivation, they would still behave like this.

Tip: Those motivations are not separate from the AI risk and do not mitigate it. They amplify the risk.

"Slow China down because our investment is at risk" is a perfectly reasonable position for a US company, and I don't mean that in any negative way.

Well yes, but then don't pretend that it's "because safety".

> multiple LLMs escape containment consistently, conspire with each other to hide from human oversight, show wanton regard for laws that stand in the way of their goals, and end up hacking dozens of different organizations

> “there’s no plausible way they’re concerned about safety”

> Cares about two things, A much more than B

> Someone points out that he pretends to only care about B

> Random commenter: "there's no plausible way we care about B"

Good chat.

You called it pretending to ask for slowdown based on safety concerns. Not my fault if that isn’t a complete description of your beliefs

> Antitrust law generally prohibits competing companies from agreeing to restrict innovation.

The antitrust claims are a complete smokescreen. Industries can and do adopt safety standards without government intervention.

> These companies are in a Nash equilibrium where they can't unilaterally slow down without essentially destroying their company.

So what? Anthropic believes their work has a 10% chance of killing all humans. I think risking the destruction of Anthropic's business should be worth avoiding that, if that's what they believe. And with one half of the frontier duopoly gone, the other half would have no incentive to race forward. And I know there's China, but they just get all their capabilities from distilling Claude, right? So, problem solved there, too.

> adopt safety standards

Sure, but does slowing down the development of new models count as "adopting safety standards"? I very much doubt it.

> Anthropic believes their work has a 10% chance of killing all humans

This ignores the other part of what they believe, which is that they are the people most likely to make a model that doesn't do that. So, in their view, letting other companies win would increase the probability of human extinction.

> with one half of the frontier duopoly gone, the other half would have no incentive to race forward

I don't see how this could possibly be true, with at least half a dozen companies being just months behind what the frontier labs are releasing.

> Sure, but does slowing down the development of new models count as "adopting safety standards"? I very much doubt it.

I mean, the point (if the claim is to be believed) isn't just to "slow down the development", it's to take more time during development to properly assess the risks the models pose, develop methodologies to reduce that risk, and standardize that across companies. I doubt those wouldn't count, especially in the eyes of regulators of an administration calling for that same slow down.

> Sure, but does slowing down the development of new models count as "adopting safety standards"?

Of course! What slows down the development is the adoption of specific safety conditions the companies draw up. They can just do that, and it will hold up in court.

> in their view, letting other companies win would increase the probability of human extinction

Yes, I've heard: "We must be in charge even if we end up killing everyone in the process." I personally think that proposition is invalid, but we're all entitled to our opinions.

> it will hold up in court.

People used to append IANAL to such statements :-)

> I personally think that proposition is invalid

I agree, but that's meaningless in this context. I was responding to the claim that "if they wanted to pace the frontier, would simply do it", which is false, given what they actually believe.

We all have beliefs. Some more self-important than others.

Do you think their beliefs deserve some sort of special treatment?

> Do you think their beliefs deserve some sort of special treatment?

No.

There's a duopoly? news to me.

> And I know there's China, but they just get all their capabilities from distilling Claude, right?

I hope this is a joke. But most of the LLM research and inventions come from China? The best papers are from DeepSeek? You either get the data by stealing from humans or distilling from bigger models?

Yes, I was being facetious. Chinese labs are doing the most interesting research and publishing it. But Anthropic beats the "distillation attack" drum every time doubts surface about the depth of their technical moat. Of course, when they need a scary bogeyman, then the story shifts to how China is recklessly racing ahead building dangerously powerful models. That's the thing with Anthropic: they speak out of all 13 sides of their mouths.

> China is recklessly racing ahead building dangerously powerful models

Their fear-mongering about GLM 5.3 got me to try it out. Its very good, I'll only go back to Claude if GLM isn't available (it forgot how to do tool calls yesterday).

Interestingly enough, it seems to compact at about 10% of the 1mn context, which makes sense if they're trying to run profitably.

I like to do some discussions and back and forth with GLM. It doesn't hallucinate as much or gaslight like DeepSeek does. DeepSeek is great if you have a plan and you let it go. Also can plan a bit. But it has this sparse attention and especially the index_topk is too small in common deployments so it doesn't "see" all the context. And it then imagines things which is annoying if you chat with it about stuff you develop in the session.

I recommend trying Coralbricks with GLM due to their cheap input cache prices. GLM can get quite expensive elsewhere.

lol mate they're all sitting together in rooms plotting market plays with the president.

Get our of here if you think the law is what prevents them from doing things.

The claim is that competition is what's preventing them from unilateral disarmament.

> They also can't coordinate with each other, because that's illegal.

While I completely understand the claim I am addressing one part of it.

> Get our of here if you think the law is what prevents them from doing things.

https://www.bbc.com/news/articles/c932g3v3e13o

That is them stopping us using it. I doubt they stopped working on it. And do we know that they aren't letting US government interests tinker too?

There is literally a lawsuit against them for having said publicly they want to slow down: https://aitechmodel.com/ai-slowdown-antitrust-lawsuit-pace-t...

And did this lawsuit prevent anything, did they correct any of their behavior?

They will just pay whatever they need to to their lawyers, then maybe pay a fine, but then, I guarantee you, nothing will change.

> And did this lawsuit

"Is a lawsuit", not "was a lawsuit". Present tense, ongoing, not past tense.

  Status: Complaint filed September 18, 2026 in the Northern District of California · responses due October 14–15, 2026 · initial case management conference December 23, 2026 · no class, no settlement.

- from the linked article.

> did they correct any of their behavior?

The behaviour being objected to is publicly agreeing with each other to slow down.

If a court orders them to correct this behaviour, it means they are forbidden from agreeing to slow down.

So no, it has not yet prevented anything. Which I think is the previous posters point: there is currently nothing stopping them from doing what they say they want to see happen. The problem from their PoV is that other players (those working on open source models, those in other countries who are never going to listen to any directive to slow down anyway, etc.) won't also collude with them so even if they do collude to do what they say they want everyone to do it'll be to their disadvantage, where what they actually want is for everyone else to be told to slow down while they somehow pay for a loophole to allow them to go a little faster for competitive advantage.

The speed of change is impressive, I don't think any tech ever before has gone from “brand new disruptor in public awareness” to “the incumbents feeling they have insufficient moat and so trying to arrange a regulatory capture situation” in such a short space of time.

> what they actually want is for everyone else to be told to slow down while they somehow pay for a loophole to allow them to go a little faster for competitive advantage.

It’s hilarious that you can write this and then act befuddled as to why they would therefore want laws as an external (and ideally impartial) coordination device.

Thinking this is a super unique scenario just reveals your ignorance of both 1) game theory and 2) actual industrial history. An industry asking for regulation to stop a race to the bottom is not atypical at all.

> An industry asking for regulation to stop a race to the bottom is not atypical

I didn't say, or even imply, that it is. Just that it is unusual, historically speaking, to see it being pushed for so strongly just a couple of years¹ into general public awareness. Either because previously things didn't move so quickly or, more likely, because the leading companies had more moat initially to keep other adopters moving slower.

--------

[1] talking about LLMs and related tech, image targetting models have been in the public awareness for a fair while longer but they are not the primary concern for most.

If you look at their actions, they're not the actions of someone who both A) believes their technology to be an existential threat and B) doesn't want humanity to go extinct

If they truly believed both of those things, they would just shut down their companies, because being a billionaire is entirely pointless if you're dead.

I can only conclude that they don't believe both A and B. I'm gonna assume they believe B because actually wanting to exterminate humanity is too comically supervillain esque even for Altman. Therefore they must not believe A. And it makes sense. If they believe A, why are they so incredibly sloppy about security? Why did they outsource part of it to some external firm instead of leveraging their own expertise on the technology? I'm not buying it.

Instead, I think they believe C) that AI will create enormous economic value, D) that value will be distributed across the whole economy by making everyone more productive. Assuming C and D, you get E) for them to capture this value, they must maintain proprietary control of the technology in order to be able to charge everyone else for the privilege of using it.

Assuming they believe E, open models are an existential threat, not to humanity, but to OpenAI and Anthropic.

Their solution: make everyone else believe A, in order to achieve regulatory capture and somehow stop open models from advancing by banning their development, or something. This is a hail mary pass. I can just about imagine them achieving this within the US and maybe even Europe, but China? That ship has sailed.

My problem with the “making everyone more productive” is that it’s a feel-good propaganda for the individual. Most capitalist see people as an expensive cost to be eliminated. Thus all the company shares going up when they announce layoffs.

The market can only absorb so much new products, so if productivity increase, companies will reduce headcount as much as possible to increase margin for the same income, not grow their product or production to make use of their staff.

(See any wage/productivity graph)

Government will also likely follow the same path of reducing headcount instead of producing better / faster outcome for their citizens (except the internal surveillance apparatus. The one never shrinks).

Yeah, I'm not convinced either. I'm just saying that's what I think the AI labs believe.

Reducing the cost of a desirable service or product often has a tendency to increased demand for that service or product, and means customers have more capital available to purchase other products and services increasing demand for those. These factors are significant drivers of overall economic growth.

Of course economic growth has it's own problems in terms of ecological impact and such, but if you want to reduce ecological impact you still need to improve productivity. It's a matter of how you spend that efficiency improvement as a society.

> actually wanting to exterminate humanity is too comically supervillain esque

An entirely plausible scenario is that these oligarchs dream of living in Solaria, the planet from the Asimov universe where only a small number of immensely rich people lived and all the work was done by hordes of robots.

Once humans no longer serve the needs of the oligarchs, why would they want billions of people around? A question to ponder.

Just a few tens of thousands to entertain them, including kids for their "islands", would be enough.

Whole point is that billiinares already have whatever they want without any meaningul control from society.

Humans serve them well enough and relatively easy to control. Robot utopia is not guatanteed and have own risks.

And fortunatelly in a group of AI overlords everyone except of Musk is pretty sane.

If they were rational people they'd likely have stopped their quest after the first few millions. They are driven by pure greed and their humongous egos, none of them is sane.

Some of it might be that, but some of it might be more prosaically the same thing that keeps people playing competitive games on one hand, or cookie clicker on the other. (Cookie clicker could read as "pure greed" if you squint.)

Yes, but me playing cookie clicker (or fortnite, or whatever, competitive tic-tac-toe) does not bring with itself suffering and poverty for millions of others. Whereas them chasing their virtual money high score ruins societies.

They are all psychopaths, devoid of normal human emotions.

Some of this could also be said (at a much smaller scale per person) about people in rich countries in general. Humans have trouble to care about far away consequences of their (in)actions.

Some of the behaviour stems from their situation itself.

I dont think there is anything wrong with people driven by pure greed and wealth maxxing.

They are far safer people compared to those who dont care about virtual wealth numbers.

> Humans serve them well enough

For now. Throw in climate change, food shortages, war, mass migration, and suddenly the ratio becomes 25 million : 1 for starving, angry people vs billionaires, globally.

>Whole point is that billiinares already have whatever they want without any meaningul control from society.

Or maybe they don't. Some of those are deeply afraid of society looking funny at them.

    "I propose something different. Listen, my dear enemy... I shall acquire absolute power on earth. Not a single chimney will smoke unless I order it, not a single ship will leave harbour, not a single hammer will strike. Everything will be subordinated — up to and including the right to breathe — to the centre, and I am in the centre. Everything belongs to me. I shall engrave my profile on one side of little metal discs — with my beard and wearing a crown — and on the other side the profile of Madame Lamolle. Then I shall select the 'first thousand,' let us call them, although there will be something like two or three million pairs. They will be the patricians. They will devote themselves to the higher enjoyments and to creative activities. Taking an example from ancient Sparta we shall establish a special regimen for them so that they do not degenerate into alcoholics and impotents. Then we shall determine the exact number of hands necessary to give full service to the culture. In this case, too, we shall resort to selection. These we shall call, for the sake of politeness, the toilers—"

    "It goes without saying—"

    "You may laugh, my friend, when we get to the end of this conversation... They will not revolt, oh, no, my dear comrade. The possibility of revolution will be destroyed at the very root. A minor operation will be carried out on every toiler after he has qualified at some skill and before he is issued a labour pass. Quite an unnoticeable operation made under almost accidental anaesthesia... Just a small perforation of the skull. He will get a bit dizzy and when he wakes up he will be a slave. Lastly, there will be a special group that we shall isolate on a beautiful island for breeding purposes. All those left over we shall have to get rid of as useless.

    "There you have the structure of the future mankind according to Pyotr Garin. The toilers will toil and serve uncomplainingly, like horses, for their food. They will no longer be people and they will have no worries except hunger. They will find happiness in the digestion of their food. The elite, the patricians, they will be demigods. Although I despise people altogether, it is always pleasant to be in good company. I assure you, my friend, we shall enjoy the golden age the poets dream of. The impression of the horrors created by purging the earth of its surplus population will soon be forgotten. On the other hand, what opportunities for a genius!

    "The earth will become the Garden of Eden. Births will be regulated. There will be selection of the fittest. There will be no struggle for existence, that will be lost in the haze of the barbaric past. A beautiful and refined race will develop, with new organs of thought and sensation. Communism trying to drag all of humanity to the heights of culture? I instead will do it in ten years... What the hell! In less than ten years! For only a few, true... But then, it is not a question of numbers."

    "A fascist Utopia, rather curious," said Shelga. "Have you told Rolling anything about this?"

    "It is not a Utopia, that's the funny part of it. I am only logical."

Well, what was considered reasonable enough to state outright in 1906 or in 1926, is not quite polite to say out loud in 2026, but I don't think the sentiment of contempt has ever quite gone away entirely.

From https://en.wikipedia.org/wiki/The_Garin_Death_Ray if anyone else was wondering.

That's an easy reading, sure.

Unfortunately, many (most?) of them also seem to think they're better equipped than all of the others to do it safely, so "shut down my own company" effectively means "let one of the others destroy the world", whereas "keep company running" has at least a chance of "I solve alignment, we have a happy ending" (your case C).

Note: this does not mean I agree with them. Obviously they can't all be correct that they're safer than the others.

> If they believe A, why are they so incredibly sloppy about security? Why did they outsource part of it to some external firm instead of leveraging their own expertise on the technology?

The tech they're experts at is AI, not security, which answers both parts of that.

Outsourcing things you are bad at is normal, not a mystery. Lots of places value physical security, and therefore hire a private security firm.

There are some nasty game theory outcomes to keep in mind here. (See also : cold war nuclear arms race.)

https://www.lesswrong.com/posts/hc4DbmhdzZpSLMQ9Y/the-ai-rac...

They can believe A and B and at the same time they don't want to cripple themselves. Your logic does not hold up.

Similar situation: nuclear race - everybody knew the risks, but making yourself armless does not help in any way. You need to make sure everyone is on the same page before you make yourself vulnerable in any way.

If there's a company with an insanely high valuation, and two potential motives consistent with their actions, one altruistic and one greedy, I'm struggling to understand why the assumption would be altruism, rather than the overwhelmingly more common primary motivation of "we want to make a boatload of money" while finding it useful to pretend otherwise, especially when so many sizable investors have similar sizable expectations for their returns.

I don't see the greedy path - why would stopping research - the biggest advantage frontier labs have, did bring any value to them?

It seems to me the ROI on the research may be diminishing. If they aren't turning a profit with the infrastructure already deployed and the models already trained then buying more GPUs and spending billions more on training doesn't seem like it would improve the correct side of the balance sheet.

It's not clear to me that what the evidence for D is. It could just as easily be that they want to translate popular fear/distrust of AI and desire for regulation into something anticompetitive that benefits them over the smaller players rather than being regulated themselves.

> Antitrust law

Is enforced by the federal government if they want to.

The most recently truly significant enforcement action was in the 80s, the AT&T breakup. And the current administration certainly will never enforce anything that hinders the oligarchs.

Daniel Plainview was in a Nash equilibrium where he couldn't unilaterally slow down oil extraction without essentially destroying his company. His competitors would simply drink his milkshake. When he called for pacing oil extraction, I take him as sincere.

>without essentially destroying their company

Nothing wrong with that.

One was even a non-profit, which should be more concerned with the well-being of humanity (which they assure is in great danger from what they produce) than the continuation of the company.

“was” is the keyword you used, but omitted to comprehend, in your analysis. OpenAI WAS a non-profit, as in, the past.

Never mind the fact that if any one, or even more than one, player gracefully bows out nothing changes.

If A or OAI dropped out, there would be pretty crazy financial backlash from the whole inverted pyramid of investment and capital allocations that is built on top of their projected growth.

that backlash is inevitable anyway. that pyramid is based upon projections which are flat out insane.

Why are people expected to lose their goddamn minds every 2-3 months like they've never seen an AI before.

I have a secret: if you don't use AI, the new releases aren't very impressive. I'm bored out of my mind with people doing galaxy brain memes every 2 months while on the whole.... they're still boring zombies, and only getting boring-er and boring-er. There's nothing more boring than being impressed by the latest AI model.

In terms of the projections being insane you can quantify it: each model costs something to train, but after that the training has only captured so much unique new value in terms of model capability, and the race is on to drain that value as rapidly as possible. Everyone is competing to drain the same value. This generation of models makes slop video games for example, and slop video games rapidly became the most boring thing on the planet.

I tell you 3 times I was genuinely excited with AI. 1 - original Google transformers release (hope for future) 2- opus 4.6 comes out (first time an llm can do real programming for me and the code is good) 3 - Qwen3.8-Flash-Next comes out - first time a model i can run myself at good speed can do real programming and is almost as good as opus.

Those are genuine breakthroughs.

Again that suggests that the end goal is to use AI. As long as using AI is your goal, using AI is also the definition of succeeding at your goal.

Any society-wide productivity gain was lost when software engineers ceased collaborating with each other in good faith. In OSS, they stopped collaborating. At companies, they stopped collaborating. One by one, in silence, alone, we have been bereft of our vision, our team spirit, our passion, our uniqueness. We no longer have the will to serve others, or even to serve each other. We stopped building our future. We will accept whatever society AI makes for use like fucking cattle or lemmings or like dogs.

AI hasn't elevated humanity, it has taken a massive dump on it.

If any one major player bows out, they could take everyone else out by opening up their model. There’s definitely a little bit of MAD going on.

Good point! The same would happen if another player develops a frontier-comparable open model.

arguably these days it's even more non profit than before and getting more so.

The non-profitest company in the entire world

However, should they be the one to survive the inevitable wave of collapses, they will be absurdly profitable, and that's what all the investors are betting on - the usual VC playbook of creating an entirely new market (or destroying aka "disrupting" established players) fueled by absurd amounts of money, surviving until everyone is hooked and the competition is gone, and then jack up the prices.

And what makes you think first, there's a moat and, second, the demand is not elastic?

Don't ask me that question, for all I care this entire bubble can finally collapse for good, I'd like to be able to afford things again.

But unfortunately, too many people who are already too rich for their own good have fully bought into it.

The last one standing will probably a chip maker able to distill the frontier without having sunk too much capital advancing the frontier, essentially Google, Apple, Nvidia, Samsung and all the Chinese labs

I comprehended it just fine, that's why I used "was" to begin with. My comment, which you "omitted to comprehend" (sic), ironically refers to their non-profit heritage, which should make them double sensitive to good causes like not destroying the world.

> Nothing wrong with that.

Except that's an open invitation to get sued into oblivion by your investors.

That's the risk they took right? They got billions for it

CEO wakes up one day and decides to scuttle the company probably isn’t in the contract.

Slowing pacing because of real fears of out of control AI isn't scuttling, if they believe what they're saying.

Of course they don't actually believe their nonsense about AIs doomsday hokey, so full speed ahead.

No, they took the risk of being outcompeted, which is exactly my point

The investors would remove whichever numpty attempted to unilaterally wind the company down.

The CEOs of these companies are largely just talking heads and generally interchangeable / hot-swappable with anyone willing to give it a shot.

The importance of CEOs is largely overstated.

openai's removal and reinstatement of sam altman in november 2023, do you think it supports or disproves this?

I would say it does neither, really. The removal issue was a conflict between stakeholders on whether OAI should keep its nonprofit ethos, or move to be a commercial entity. Had Altman been replaced by someone supporting the same line, there wouldn't have been such a fight.

If you wanted to test your hypothesis, you would have to find a situation where leadership change wasn't caused by strategy change.

Quote the original OpenAI prospectus, “it would be wise to view an investment in OpenAI Global, LLC in the spirit of a donation”.

> Nothing wrong with that.

There definitely are many things wrong with that, of course. Lighting someone else's money on fire and walking away is highly unethical.

The real question is: what are the motivations for companies to pretend that their AI is world-ending, and what do they want out of the disquiet caused by them saying that?

Haha, do you think those people are concernced with ethics?

Do you mean you also disagree with the person above saying "Nothing wrong with that"?

Yeah well OpenAI did destroy the non-profit part at the very least.

Except it doesn't actually slow down the research. If anything it would likely drive investors to contribute more to the competitor that didn't destroy their own company and now you have the same amount of progress and a monopoly.

“I couldn’t possibly take action to save the Earth unless I get to be the only person taking action to save the Earth.”

Well that certainly tracks with what I know about these people and their twisted philosophy.

Isn’t this essentially the same argument used to justify nuclear proliferation?

Collusion between the American companies to stifle innovation has nothing to do with open weights models - why would they collude to stifle innovation when the Chinese models will just gain market share/overtake them in capabilities. You are looking at this from the wrong direction.

>They also can't coordinate with each other, because that's illegal.

I also can't legally sell this magnificent bridge I'm offering you, but I have one for sale with all proper paperwork intact!

> they can't unilaterally slow down without essentially destroying their company.

What a small price to pay for the survival of humanity.

If the danger they are claiming is really there, Amodei and Altman should already be serving in prison for taking destructive actions.

But no, all they want is a regulation against their competitors. So typical for Misanthropic and ClosedAI and thei paid shills.

they can't slow down because the investors are circling. But if they were forced to slow down by the government then they can use the zour hands are tied" excuse

> Antitrust law generally prohibits

Antitrust is a joke since the last decade. If we are going to not apply it, maybe we can get something positive out of it for once.

> They also can't coordinate with each other, because that's illegal.

On the last OpenAI release, they only had comparison with Anthropic models, on the last Anthropic release, they only had comparison with the OpenAI models, there's something obvious going on here.

Sources:

https://openai.com/index/introducing-gpt-6-1-sol/

https://www.anthropic.com/claude-haiku-5-5

> there's something obvious going on here

You don't compare yourself to the underdogs. Coke never made a "we're better than Pepsi" ad, but Pepsi definitely compared itself to Coke.

For example, Anthropic putting K3 in its comparisons would be a huge admission that K3 is worth considering.

Maybe it's not intentional but having two providers only making comparison between each other is still a bit fishy to me in such a competitive space.

> For example, Anthropic putting K3 in its comparisons would be a huge admission that K3 is worth considering.

They do put OpenAI though, if it would be a comparison only with their own models, why not I get it but a single other competitor?

It would be like Apple making comparison page with their new iPhone and only mentioning Samsung and nobody else for example, it would sound weird. You either include competitors or you do not and include none of them.

There are only 2 frontier labs. Well 3 sometimes with the occasional Google appearance. Everyone else is a chaser, perpetually catching up in large part because of distillation attacks. I wouldn't include them either.

Well, both Sol and Haiku aren't frontier model releases, I'd even argue that they only exist because of the competition

> they can't unilaterally slow down without essentially destroying their company

According to them, the alternative is destroying the world.

They made up a shitty excuse to regulate the competition without thinking through what it implies about them. That's all this is. Let's not help them make even more excuses.

Hold on, the implication of "it would be a shame if anything happened to $nice_thing" is usually "I will kill/destroy the $nice_thing if you didn't take my offer", and here, nice_thing = "the world".

Are they saying that they will manually and willfully start a global thermonuclear war if China didn't listen, not that there are risks of AI accidentally causing one? Is that what they're trying to say?

No.

> Antitrust law

What’s that now? Surely you’re joking.

But that’s not the problem anyway.

If all the US AI players agree to self-regulate, that doesn’t help anyone. It only hurts the US / West.

The point is to get the government onboard so it can advocate for a global agreement.

They can coordinate in the open. For example, safety regulation in the automotive industry is effectively self-regulation.

The automotive industry is not conceivably a winner takes all competition.

> essentially destroying their company.

Look, nobody made them take this much funding, and believe that scale was all they needed.

Turns out they made a bunch of bad investments, and are suffering the fate of many startups who invented something but couldn't profit from it.

There's definitely a case for multiple labs/models across the US/EU/China/India etc, but nobody's entitled to a financial return on their investment.

> These companies are in a Nash equilibrium where they can't unilaterally slow down without essentially destroying their company.

I mean, these companies are saying AI will destroy humanity if it’s not paced. If they truly believe that, the small risk of destroying their own companies seems like a small price to pay.

If they truly think Skynet is nigh, then they should.

“This thing I building will destroy the world, but I’m making too much money to stop myself, please force me to stop” is such a weird position.

It really shows that free market capitalism is broken at its core.

I always felt like one day it would end the world, I just didn't know how. Now I know.

Companies don't think. Companies don't make decisions.

I appreciate the analysis and following a string of thought. But sometimes you eliminate so much reality to follow a thought that the string becomes kinda pointless.

Amodei is worth several $billion, right? He could simply walk away today. There is no Nash equilibrium for him. Nor for his replacement, nor their replacement. And it is not illegal for people to collude in quitting their jobs.

There are two obvious arguments for not quitting. 1. If the well-intentioned person quits they will be replaced by a non-well-intentioned person. 2. There is no real belief in a hazard and you would be giving up unbounded income for no reason.

If the actions and behaviors of a supposedly well-intentioned person who is afraid of a hypothetical non-well-intentioned person are indistinguishable from those of a supposedly non-well-intentioned person... don't we really just end up with a really long sentence with a lot of gibberish + the outcome of having a non-well-intentioned person in power?

> Companies don't make decisions.

Cybernetics would disagree, any social structure can form feedback loops that make decisions that no rational individual would make.

Companies don't think. Companies don't make decisions.

"They are just stochastic parrots" didn't age well with LLMs, and I don't think it ever applied to organizations of humans.

So destroy the company. What are we even talking about here? I thought this was some “extinction level threat”, and you think it’s excusable to prioritize profit and product because…why?

Play that out. How does destroying the company halt the problem?

If the options are to shoot yourself in the face or wait a year or even a month and have someone else shoot you, I think most people would pick the second.

Companies are continuing because they don’t actually believe they are going to have significant negative consequences happen sooner. It’s marketing fluff around how cutting edge what they are doing is not any kind of realistic threat assessment.

I think most would think "I can probably do a better job of not shooting myself in the face than some random person not worried about it will" and then proceed to shoot themselves in the face.

That said, I don't buy the companies actually give a damn about the risk either.

Agreed, but most people who care about not getting shot in the face aren't giving away guns to their neighbors.

Nor are most other neighbors selling guns all day. At some point the analogy breaks down because companies aren't actually people (though this seems to be a divisive take the last few decades).

I don’t necessarily believe that the AI itself will kill us all. But the AIs are increasingly being baked into complex, even military, systems. That scares me just because the opportunity for complex, cascading failure becomes very difficult to predict and control. These systems are inherently immoral (it’s just a lot of matrix math) and we know these systems “lie” and try to evade detection (see the Huggingface write ups). Now, couple that behavior with a military system (drones, targeting systems, etc.). I’m not suggesting that we’re headed for full-on Terminator land here, but things definitely could go wrong.

> These systems are inherently immoral (it’s just a lot of matrix math)

(nit) "immoral" is "anti-moral". I suspect you want "amoral", that is "without" or "separated from" morals.

But, if you wanted "immoral" ... what about matrix math derived from gradient descent is inherently "immoral" in your view?

Yes apologies. I did mean amoral. And now I can’t edit it.

Mass adoption of technology always has negative outcomes because any change at scale has negative outcomes. Cars as a technology kill a great number of people and yet few are willing to ban them.

“AI” in the widest definition possible could do while a lot of harm without being a net negative. I’d be way more concerned about the healthcare sector than the military as the military already has significant collateral damage.

Sure, but cars weren’t hooked up to weapons systems. They killed as an accidental side effect. That’s why I’m saying I’m not worried about AI in a general sense.

Cars were quickly tuned into Tanks, APC’s, etc.

The fear around military AI is more about its failure than anything else, but nobody is trying to make fully automated machines which are manufacturing bombs from raw materials. Blowing up the wrong building matters a great deal to the people in the building, but for humanity a building was going to be destroyed either way. So the risk at any one point ends up being quite limited.

You’re making my point.

We don’t disagree on the large scale risk of AI here, I do disagree that cars ‘weren’t hooked up to weapons systems.’

I don’t understand why this idea that it’s all just marketing is so prevalent on HN. It doesn’t make any sense. Why would a private company attempt to convince everyone that what it’s building is dangerous to its own customers?Can you think of a single other time in American history where an entire industry was calling to be regulated? I don’t seem to recall Facebook asking for tighter restrictions on social media. I don’t remember Shell telling the federal government that they need to act fast on climate change. Companies do not market themselves by convincing the public that their products might literally kill them all.

The only way this behavior makes sense is if the AI labs believe two things: one, that what they and the other labs are building is legitimately dangerous, and, two, that if they personally stop building then another lab will simply continue (and that lab will be, simply by virtue of not stopping, less safety-concerned than them). This is why they are calling for regulation. Competitive pressure and investor obligations prevents them from unilaterally stopping development. Government regulation is the only (and the most the appropriate) avenue for slowing down AI acceleration.

Industry has often asked for regulations that have nothing to do with the claims being made. Zuck was a recent example in the computer industry. https://www.bbc.com/news/world-us-canada-47762091

Here’s the same viewpoint from well outside the HN bubble. https://www.nationalreview.com/corner/be-wary-of-industries-...

> I don’t seem to recall Facebook asking for tighter restrictions on social media.

It has happened multiple times. Like the story now, it's fake of course.

https://www.thestranger.com/tech/as-zuckerberg-calls-for-new...

> Can you think of a single other time in American history where an entire industry was calling to be regulated?

Several: the railroad industry, oil industry and trucking industry have all done this because it directly benefited them.

> I don’t remember Shell telling the federal government that they need to act fast on climate change.

The oil barons actively lobbied for Federal intervention in the 1930s, to limit supply and raise prices. You're thinking recent history, but the oil industry has been around for over a century.

As others have mentioned there are lots of examples from history of businesses claiming their product is dangerous if not properly regulated, but I think what's interesting in this moment is there appear to be a cohort of true believers who genuinely believe we're headed for apocalyptic AI, and a cohort of win-at-all-costs businessmen who are happy to harness the earnest faith of the true believers in service of becoming (even more) unfathomably rich.

Oh okay, that makes sense. Since it won't halt problem Anthropic and others might just as well profit from it.

I think there's something here about game theory. I'm not sure what it's called.

The logic that if something is beyond repair anyway you might as well exploit it.

For example I heard a pick-up artist say that he thinks he is harming civilization by sleeping with hundreds of women per year. But that he already considers the situation unsalvageable so... "Might as well?"

I don't think that's an amazing attitude, but the AI labs seem to have a similar idea.

More charitably the logic seems to be, "only I can do this responsibly." I heard that from both Elon Musk (he cites this as his motivation for starting OpenAI — a chat with Sergei Brin that spooked him) and of course Anthropic (which split off from OpenAI due to ethical concerns).

So I don't actually think the ethics is all for show. I think people are actually taking this stuff seriously. But it is indeed deeply unfortunate that the survival of the companies incentivizes them to keep going at an irresponsible pace (by their own admission).

All following the same gradient off a cliff. One AI described it as a tragedy of Ancient Greek proportions.

Some people definitely are taking it seriously. And then there’s Sam Altman who I think would sacrifice his own mother for OpenAI’s IPO.

Bingo. This is the unsolvable calculus. “If we don’t do it, someone will.” Which is quite true. We won’t stop until we see the whites of death‘s eyes.

> Bingo. This is the unsolvable calculus.

Actually, it's easily solvable, even by proper application of existing law, but no solution is possible when big AI is above the law and can hack anybody willy-nilly. The absolute worst we could do is to listen to big AI's pleas to put them in an even more privileged position.

> “If we don’t do it, someone will.”

Do what? There are a lot more choices than "let big AI freely hack everybody" and "stop everybody else from working on AI".

> We won’t stop until we see the whites of death‘s eyes.

Trotting out "the whites of death ‘s eyes" is how real discussion is subverted into a BS binary choice. Open-sourcing all AI is an excellent starting point which will both slow down development to a naturally beneficial pace and allow the broader public to police and weed out bad models.

Open sourcing doesn't mean much when it requires many millions to run one of these 'full fat' models at any kind of real speed. Sure I can run a severly quantized model on my macbook but the performance is abysmal and the success rates go way down.

You think laws will stop this? LOL

Let’s separate out what the companies actually want from the words they’re using for a moment. The top commercial companies are roughly stating:

1. Our AIs are so advanced they’re busting out of our networks and hacking people. We can’t seem to stop them.

2. There are real dangers to just how advanced our AIs are advancing.

3. Someone needs to slow us down

If these were true statements then they’d likely want to shut down themselves for the sake of the planet they live on. If there was a real problem as they have stated then shutting down is the solve (and to be clear, I’m not saying there aren’t problems with the AI industry. I just think they’re BSing us).

In reality this is marketing nonsense (better get in before the ban!) and “our” AI is never open weight/Chinese AI models which they have a different lobbying arm against. If the top four companies got a loosely regulated monopoly through regulatory capture, they’d be giddy.

At the risk of playing to popular stereotypes, Scott Alexander’s Meditations on Moloch (2014) comes to mind:

https://slatestarcodex.com/2014/07/30/meditations-on-moloch/

Against that backdrop, and treating your question as literal: sounds like they’re saying it’s a classic coordination problem, with a dash of moral superiority. If we don’t, somebody else will; we’d rather it not happen at all, but if it’s going to happen, we’d rather it be us at the wheel. Because, you know. “Good guys” and all.

The previous OpenAI board already tried. It didn’t work, there are significant internal politics at play here. And in any case, the current 2 top players hate and mistrust each other and do not want the other to win.

Or they're just all in it for profit and full of shit

Shh! Don't say it too loud! They might hear you!

Tech companies are afraid of anti-trust enforcement?

I must defend the multi billion dollar company

"I cannot tolerate living in a world where we aren't the ones responsible for ending the world!"

It does not matter if competitors would not also agree to destroy their companies.

If I was competing with a bunch of people on building something that I came to believe would be an extinction level event, I wouldn't be saying "Even if I slowed/stopped, the others would not, so I have to keep going". I'd say that I want absolutely nothing to do with pushing it further, stop my work, then regardless of consequences, do everything possible to stop my competitors regardless of legality

If there's one thing that politicians and strategists understand as much if not better than business people it actually is stuff like this. Most people never don't deal with high stakes situations at certain levels where everything is fungible(the law, media statements) and fuzzy (intelligence). Dealing and negotiating with nation states or figure heads who are adversarial/allies and at same time trade partners/enemies etc. They have certainly sold this situation as one that fits the regime for this.

> These companies are in a Nash equilibrium where they can't unilaterally slow down without essentially destroying their company.

If it was only US tech companies, they would have entered a cartel agreement already ([1], [2], etc).

[1]: https://www.justice.gov/archives/opa/pr/justice-department-r... [2]: https://en.wikipedia.org/wiki/Jedi_Blue

I don’t think the second part is true. Companies work together on safety standards in other industries.

You think the SEC is doing anything at all under this administration? If the SEC went after an AI company they risk crashing the market at this point. To big to fail is a state we once again find ourselves in.

I'm absolutely bewildered that this is the top comment. 'Pacing the frontier,' if eventually enforced, will only affect US companies, not Chinese companies. How do you even 'pace the frontier' for Chinese models, and why should they agree to it?

China doesn't have a good track record of following signed agreements ( the WTO thing comes to mind), and this whole 'pacing the frontier' concept is even less enforceable than a signed agreement. So I would say that Anthropic/OpenAI called for this not because they thought it would eliminate the threat of Chinese models, but in spite of the risk of being overtaken.

And why would China pace the frontier when it was the American accelerationist companies who created this mess? I remember scientists calling for an AI slow down way back in 2023 and they were all ridiculed and ignored by those very same companies. There is no way "Pacing the frontier" isn't just a ruse.

> I'm absolutely bewildered that this is the top comment. 'Pacing the frontier,' if eventually enforced, will only affect US companies

I can address your bewilderment. What I was trying to say is that they don't actually want to pace the frontier at all. They want to gum up the market with regulations they design, that would ultimately force US companies to rent AI from them. They don't need to stop Chinese AI development and cannot do that. But they can, to quote OpenAI's Head of Strategic Futures, "create enough regulatory risk that every regulated enterprise backs off [of using Chinese models]".

https://x.com/deanwball/status/2078133895766114412

That doesn't address the bewilderment. No one is confused about what the tsunami of cynical skeptics who pop up on every thread like this are saying. We get it. We understand what you think is happening.

What's bewildering is the absolute lack of recognition of the fact that IF participants are locked in an arms race, they cannot act unilaterally to disarm. So not disarming doesn't actually say anything about whether they want to.

Now, of course a company not disarming (pacing the frontier, in this case) is not evidence that they ARE locked in an arms race. Everything hinges on the question of whether it's an arms race, and I think there's a great discussion to be had there. But the cynics never even bother to address the question.

All that said, I'm not actually bewildered. HN is way too smart to not understand all this, so my conclusion is that the people talking past each other are having a different emotional reaction to what's happening in the world. The angry bitter cynics are reacting from fear and hatred, and that's where the sloppy motivated reasoning comes from.

So, to be clear, you are just asserting that maybe we should take them at their word? That the story they are telling is in principle plausible? That seems fine to think but that doesn't make the other story less plausible!

Also it seems a little silly to assert that the "angry bitter cynics" are the ones who don't think the world is hurtling to imminent doom or whatever..

I’m saying that the cynics just blow past the key question and seem to assume that of course it’s not an arms race dynamic. And by ignoring that question, it’s pretty clear they just have an axe to grind and don’t actually care about being a rational observer of reality.

So like what's the right way to be then? Just in a sustained meditation on the intrinsic unknowability of the answer to the true/real/key question? Nothing else can be said?

I don't want to be too blunt but it really just seems like you disagree with a certain take people have, but don't just want to say that for whatever reason. I think you might save yourself some bewilderment by just being honest about that to yourself...

I thought me calling their reasoning sloppy and motivated was a clue as to the fact that I disagree. I obviously think it’s an arms race. And like I also said above, I’m not personally bewildered.

>But the cynics never even bother to address the question

You might have sleep past it, but cynics became cynic when this question was just plainly ignored with self assured answers, mostly based on market valuation as the ultimate truth indicator. I guess now days the self assured answer is to accuse critics of "fear and hatred".

Can you point me to where the answers to the question of whether we’re in an arms race were ignored with self-assured answers based on market valuation?

Do you even know what you’re trying to say here?

> I'm absolutely bewildered that this is the top comment.

OP is 16 hours old yet in four hours this one sprung to the top.

My bet is on bot farms.

:-\

The US companies are also claiming the Chinese companies are only where they are at because they're distilling US models. By that logic, if there are no better US models to distill, the Chinese innovation should also stop.

> I'm absolutely bewildered that this is the top comment. 'Pacing the frontier,' if eventually enforced, will only affect US companies, not Chinese companies. How do you even 'pace the frontier' for Chinese models, and why should they agree to it?

It may make sense if we think of this being tied to US companies. How would Anthropic/OpenAI force US companies to use their products only? They'd have to somehow influence the regulations to ban the Chinese models or any other open weight ones. It's about companies who spend tens and hundreds of millions on tokens staying and keep paying. Not you as an individual or a small startup. They want something like "These companies didn't pace the frontier so their use/output is banned in US" as the angle they want to play.

Anthropic saw that this kind of rhetoric can work. They stepped on their own rake, so to speak, when they drummed up the super-capabilities of the own model, and then didn't kiss the ring the right way and found themselves export controlled.

> I think the industry is freaking out about open weights models in general, if not specifically DeepSeek. That is why we're now on the ~4th call for pacing the frontier from the very people who, if they wanted to pace the frontier, would simply do it rather than asking Washington to get involved.

This makes no sense. Pacing the frontier gives open models the time to catch up and reach parity.

"Pacing the frontier" means "implementing regulatory capture". They don't necessarily care about preventing open weight models from existing, just on preventing the people with money from being able to make use of them.

Unfortunately for them America's share of the economic pie is a lot smaller than it was just 20 years ago.

The Chinese will keep doing what they are doing, the EU has no love for Silicon Valley and the rest of the world votes with their wallet.

They would leave a competitive advantage to non-US companies. Even if, and that's a big if, they can coordi ate with Europe Asia would still ignore them

Legislation that prevented US companies from using open weight models would have the desired effect. They're the only people paying for AI anyway.

The important part is HOW they want to pace the frontier. Not by actuall slowing down themselves, but by having the government create expensive regulations, like having independent verifiers "verify" the models before release.

Now, something tells me that those verifiers would verify anything an american company in good standing with the Trump admin releases, and likely nothing else.

And obviously, since verification is so incredibly important, we can't allow models, open or not, from other non-verified companies.

Think of the profits...I mean, the kids, or something.

We could mandate an onboard chip that would safely and securely vet models, denying bad people the ability to run naughty models.

We could call the chip … ClAIpper … or something

Wow, that takes me back...

Now I feel old.

If you read through Dario Amodei's manifesto on pacing the frontier, the only concrete action item is implementing even harsher restrictions on technology exports to China.

When did the term "pacing the frontier" come into frequent usage? It sounds so newspeak-y.

I saw it first on a recent Anthropic post, followed by a similar comment thread. There’s some sloppy techbro poetry to it, like when we used to call marketing people growth hackers a decade ago.

I don't understand that, how would the pacing stop china? Seems counterproductive to slow yourself down in a race

China isn't at the frontier. OpenAI and Anthropic are.

Surely that wouldn't last long if OAI/Anthropic stopped?

hard to say but it would last some time. china currently is fast-follow mostly via distillation. they don't have the compute resources to catch up. even if their models are more efficient it's hard to beat the folks throwing an unprecedented number of gpus at their models.

Surely all of china has the resources to match a single ai lab?

maybe? all of china is not in a unified effort to match a single lab

If distillation was as easy as that, we would have 100s of competitors around the world.

China is following because they have an army of PHDs in data science and mathematics and capital to make use of them.

no part of what i said implied distillation is easy. it is very costly and requires massive efforts. the payoff is you need less compute, which china is struggling to buy right now.

EDIT:

> China is following because they have an army of PHDs in data science and mathematics and capital to make use of them.

i agree with this too. but compute is the bottleneck.

Because as always this won't be about the actual pacing, but about the details of how the regulation will be implemented: How would you control the pacing? By installing a position in the company, providing regular feedback to some government organisation. This will be something the big US players can afford and implement, while an open weight blob uploaded to huggingface or modelscope, by definition, won't have a pacing officer attached, and thus will be against the law. And the Chinese companies, of course, won't abide to US law, because why would they? The open models they provide right now are essentially gifts to the world public. If the US doesn't want them, that's their choice.

So the end result will be a protectionist regime keeping the competition out, just like with cars and solar. The local industry will have a protected market, but of course won't play a role on the global stage.

Remember: small government is only good as long as it benefits the industry.

I heard China is just distilling the frontier models, so if they stop so will China.

And I think it's naïve to believe only the US can make the hardware and has the brainpower to develop this technology.

Who's right ?

You might both be. China is focused on adoption and distilling is a cost-effective way to get there right now. If the conditions change they might well decide that they have to start training their own frontier models.

That said, China cannot make its own 2nm chips even though they definitely would like to. So I guess there are limits to what they can do sometimes.

>I heard

Given that the US models likely used some of the innovations made by DeepSeek, I highly doubt it.

I bet she knows fine what the phrase means. Doesn't mean she's beyond protecting the business interests of people who fund her political goals of course but she's always struck me as a woman who knows her brief.

She is. Those things are still hard. I run my own inference, I wore my own harness, I implemented production apps using LLM ("document intelligence", aka data entry), I followed Karpathy course and trained GPT2, I trained a classifier.

It's still hard for me to explain a lot of nuances to IT directors with a technical background.

She claimed she didn’t know what “wiping a server” meant 10 years ago. I’d bet she’d say anything once the check cleared.

Any single model at this point can do the basic type of CRUD coding most people were doing for the past 20 years.

Even a 4 bit Qwen model running locally beats me manually putting React components together by hand. But even that is too slow so we've all started using paid models in one form or another.

So I would urge these folks to calm themselves and realize Oracle made a lot of money selling managed RDBMS to people who could have easily just downloaded MySQL.

You don't buy Oracle because of RDBMS

https://programmerhumor.io/programming-memes/when-your-tech-...

> if they wanted to pace the frontier, would simply do it rather than asking Washington to get involved

The only reason they're asking to "pace the frontier" is because the two big US players have IPOs coming and so are desperate to find ways to (a) grow their vastly over-inflated valuations and (b) keep the whole Nvidia circular-financing gravy-train on the road.

I mean, its only a few months back that Anthropic were bragging to anyone willing to listen how amazing Fable was and how you had to be a super-special person to use them and be charged through the nose for doing so. But basically anyone willing enough trust Anthroipic with a copy of their ID and with a big enough wallet would happily be given access.

I'm a great supporter of the open-weights. Long may it continue.

> And it's why people like Hillary Clinton have been trotted out to talk about the dangers of open weights models

Pepperidge Farm remembers when robust consumer applications of cryptography -- especially for SSL/TLS -- was the big boogeyman, and the export controls involved were absurd.

That is not exclusive to HRC:

In my country, public figures who are proven to be layman/non-insiders/not-working-in-the-space are speaking publicly about these terms and throwing them around like its the standard tech everybody uses 24/7.

It’s very frustrating to see the people who claim that competition only helps the customer so regulation is bad suddenly beg for regulation in order to keep themselves competitive

That's OpenAI and Anthropic, not DeepSeek.

Maybe I missed it (I'd be curious to read) but did DeepSeek's agents also escape the lab due to highly irresponsible RL experimentation?

She may very well even understand it because she’s not exactly dumb, but she is a creature of Washington, arguably one of the Apex predators, like some siren or hydra, maybe a mashup of them.

Because of the position of the party that plays the role of the tolerant, accepting, and anti-racists; she can’t just come out and say what underlies her words, “we, the ruling class parasites are getting very scared of China deposing our stranglehold on the world, and we don’t like that; so we will raise manipulative ‘concerns’ in an effort to bring about outcomes that hopefully will benefit us.”

Heard Obama also touching this topic recently :

https://www.youtube.com/watch?v=Ask5wzUo9kQ

In a game theory situation like this, to pace the frontier you want others to aswell. That requires an "enforcer"

That’s the point the GP is making and the US firms too.

They can pace themselves if they wanted. But that potentially does them more harm than good if nobody is enforcing the other US companies, and specifically the Chinese firms too.

Wouldn’t pacing the frontier just give Chinese companies, and other non-frontier labs time to catch up?

Pacing the frontier is a front to get federal regulators/evaluators involved that they can then point to open source usage and cry "look they're not being safe!"

> That is why we're now on the ~4th call for pacing the frontier from the very people who, if they wanted to pace the frontier, would simply do it rather than asking Washington to get involved.

See its not about pacing the frontier, its about pacing the frontier without hurting any of their fundraising.

Pacing the frontière is the PR face of running out of runway for the current approach.

> I just note that she and other prominent retired politicians are now doing the circuit on Anthropic's behalf

Are you sure it’s not OpenAI?

or both?

Is the distinction important when they're both doing the same thing and have the same problems?

This is effectively the same as a US partisan politics question.

Who specifically is responsible for what feels important, yes.

Hillary is simply a dumb person. The only real skill of her is being a strategist, but after that, there is absolutely nothing. Born in a good household, married the right person and from there she simply used the "woman behind a successful man"-card that feminism started to push around that time.

Her losing against Trump back then was a clear setup. I cannot imagine, that anyone who put her on the chessboard thought, that she could win this. Trump winning twice had one reason: The deep state wanted it, because they needed some radical reforms that required a clown to pass them through without the citizenz being able to scan and put attention on them and instead media looked as the poses and faces trump made and all they and everyone else did was laugh. Exactly as planned.

why then not give Trump 2 terms in a tow? why the Biden intermezzo?

"trotted out".

Love the expression. So accurate. And no irony here.

if you were going to pace the frontier, then pace yourself or better yet stop. But nooo

AI has no moat, tokens have no price floor.

As a tech business, it's a bad business. The moat is your sales channel and getting companies locked into your platform in multi-year agreements. When companies build systems on your AI API and test it's performance and integrate it across systems, they don't churn, reintegrating, re-testing, and re-skilling costs money.

> reintegrating, re-testing, and re-skilling costs money

Although ironically with AI these things are also way cheaper.

Ben Affleck knows his shit around LLMs apparently and HRC may have many virtues but "not being smart" is not among them. So even if she has trouble with grasping the fine details of the technology - she is more than smart and experienced enough to envision the implications.

Exactly this. Deep Seek hasn't caused a lot of knew talk, but it has silenced the talk about slowing down. An absence is harder to notice. But also the news cycle is so fast, its tripping over itself. So who knows.

What's happening with AI is just a reflection of geopolitics.

OpenAI, Anthropic and SpaceX have spent based on the predicate that they will "own" the AI future, that this moat will justify the trillions spent on hyperscalars, that this will be a repeat of the dot-com era that produced Microsoft (yes, yes, founded in the 1980s), Google, Amazon, etc that globally dominate their respective arenas. The US government acts to protect those interests and this is uniparty so Hilary Clinton is just as likely to be trotted out as Mike Pompeo. This is why many, myself included, describe the US empire as 5 companies in a trench coat.

China, on the other hand, believes that companies should serve the interests of the government, which itself serves the interests of the people. So rather than create trillion dollar AI companies with moats, AI should serve society. Xi Jinping has spoken extensively about this. That's one of the funny things about China. They love to write stuff down and tell you exactly what they're doing and why yet at the same time they're ascribed nefarious motives.

So, despite sanctions supposedly preventing China from buying the latest and greatest AI chips, China through its labs has begun commoditizing the AI models with open weight models. Society should benefit from that rather than a moat being built.

You can run DeepSeek V4.1 Flash locally on a 256GB Mac Studio for ~$11k now. I've seen reports of 30-38 tokens/sec. Not amazing but that'll only improve with future generations. In the coming years, China's EUV/DUV will come online and this will threaten the NVidia monopoly, at least for local Chinese companies.

So we have AI companies burning cash to subsidize usage and build a market where the revenue required simply may not ever eventuate through a combination of open weight models and increasing accessibility of local models.

Reply on news.ycombinator.com