[ericmay]

Any country that developed sufficiently advanced models will pursue the same path. The EU as a regulatory body would pursue the same path, as will/is China.

It’s rather straightforward to think through. If China (they’re practically the only competitor) built a sufficiently advanced AI system would they allow it to propagate on the free market? Of course not. The fact of the matter is that they are behind, even if it’s just a little bit, so the best strategy they have is to try and compete with “good enough” models with lower/subsidized cost - but that is a losing strategy because AI is and always has been a winner-take-all strategy.

Likewise if, idk, France someone built an AI system that was valuable do you think they’d just hand it over for sworn enemy Donald J. Trump to utilized? Of course not.

The American strategy in the context of the current geopolitical landscape is the only valid one and the obvious one. If you find yourself criticizing the American strategy it’s because you aren’t in the arena where you would, inevitably, make the same decision to restrict access.

[somenameforme]

This isn't how things typically work. For instance the US is increasingly adversarial towards China yet China continues to largely power the US economy both through manufacturing and market access, which makes up an increasingly large share of all revenue for many US companies. Why? Because that position as a dependency is not only directly economically beneficial to themselves, but also provides leverage which can be utilized in extreme circumstances.

This isn't really going to achieve much besides incentivizing the growth of non-US models and minimizing market access for US-based models. I was expecting the US government to try to ban foreign models, which is also a self-own but orders of magnitude less than this. All this will do is greatly diminish the influence of the US in the future, and minimize the benefits they might reap from a global LLM explosion. It'd be like if 30 years ago, China decided that their manufacturing could only be used by white-listed individuals. Their economy and influence wouldn't be even a fraction of what it is today.

[ericmay]

> This isn't really going to achieve much besides incentivizing the growth of non-US models and minimizing market access for US-based models.

The Catch-22 here is that the growth of non-US models is something Americans can take advantage of if they are successful while simultaneously denying access as deemed necessary by the US government to advanced American models.

If open source or models made outside the US surpass US controlled models, then the US would just switch to those and then American companies can leverage those models for their own development or for consumer sales or whatever.

If they don’t surpass US models (as I expect they won’t, though they will remain perhaps marginally useful) then the US maintains the lead in a positive feedback loop development.

If others start controlling advanced models and denying access to superior models developed outside of the United States (ex: China) then my assumption was correct.

I don’t think comparing China’s manufacturing capacity to the American manufacturing base makes a lot of sense in this context and as we know China has in fact weaponized that manufacturing capacity. If nothing else, the economic arrangement isn’t comparable partly because if Chinese factories stopped making iPhones or whatever they would simply not make anything and workers would lose jobs and such hurting the Chinese. Today if the United States government prevents Chinese users from accessing advanced American models there isn’t really a loss on the American side. Quite the opposite - these models will accelerate productivity and industrial capability for the country as a whole and make it more competitive economically.

[janalsncm]

> AI is and always has been a winner-take-all strategy

People claim that AGI is. AI is turning out to be a fairly competitive but “normal” product. Companies carving out niches on cost, quality, and speed.

If it was a winner take all OpenAI’s head start would have been decisive. For years ChatGPT was far ahead of everyone, then Anthropic released Opus 3, then OpenAI released 4o, then in mid 2025 it seemed like everyone had strong reasoning models including Google with Gemini 2.5, and now Claude is probably the best coding model. So taking the top spot is not a guarantee you can hold it.

Also the top model becomes a prime target for distillation, making it easier for competitors to keep up.

[Razengan]

> and now Claude is probably the best coding model

Do people who keep saying this actually used Claude side by side with other models?

Because in my experience it turned out to be hilariously crap: https://i.imgur.com/jYawPDY.png

[_3u10]

I don’t think they have. I used to think that Claude sucked because of the harness so I tried it on opencode and it was still trash.

I’m sure if the primary problem I was solving was multi file complex exploits it would be better but even that I am not sure of. I’ve always found the it’s too dangerous claims to be a bit underwhelming.

[janalsncm]

The point is that if AI was really a winner-takes-all technology it would have meant OpenAI’s models would have been unequivocally better, permanently.

[prmoustache]

Can a model turn crap? shouldn't it be consistent once realized? Isn't it probably Anthropic facing capacity problems and comitting fraud by silently routing you to smaller models for the price of the current Claude?

[d5lt5]

I think you view this situation from the US point of view and assume that China has the same guiding principles and values in their foreign policy, for which it doesn't. They might do what you said, of course. But they very well might also treat LLMs as another goodwill investment like the Belt and Road Initiative (https://en.wikipedia.org/wiki/Belt_and_Road_Initiative) and export the capability to partner countries, for example, in Africa, to strengthen relationships.

[ElFitz]

Belt and Road wasn’t goodwill.

A lot of it was financed through large (sometimes unsustainable) loans to recipient countries, sometimes leading to unsustainable debt burdens, irrespective of the potential ROI for the recipient (ie Sri Lanka’s port).

In many cases, much of that debt paid for Chinese companies, contractors, suppliers, and imported workers who built or operated the projects.

And the infrastructure didn’t necessarily line up with the recipient’s actual needs, mostly with China’s (ie the Laos–China railway, in large part financed by Laotian debt, which may someday bring some benefits to Laos, but mostly serves China’s regional trade ambitions).

Not to say other countries do it better or have purer ambitions or whatever. It’s just the "goodwill" part that made me twitch.

[d5lt5]

Can you argue that the principle of the BRI is humanitarian and it should benefit both partners, but not equally? Imho, that policy is far better for humanity than blockading Cuba, bombing Venezuela and Iran.

> A lot of it was financed through large (sometimes unsustainable) loans to recipient countries, sometimes leading to unsustainable debt burdens, irrespective of the potential ROI for the recipient (ie Sri Lanka’s port).

I see that you blame China for Sri Lanka, while China wasn't the only creditor there.

> And the infrastructure didn’t necessarily line up with the recipient’s actual needs

Easy to say in hindsight.

[ElFitz]

> Can you argue that the principle of the BRI is humanitarian

No. You can argue some projects, if done well, benefit both sides. That doesn’t make it humanitarian. It makes it basic foreign policy.

> China wasn’t the only creditor there.

I didn’t say it was. I said Hambantota was a costly development failure for Sri Lanka, and Chinese lending was part of that specific project and problem. Basically, that unlike your "goodwill" claim, China isn’t just giving away infrastructure for free out of the goodness of its government’s heart.

Don’t make me say what I did not.

> Easy to say in hindsight.

Yes. That’s why development and debt are hard problems. Also why calling it “goodwill” is, at best, too generous.

> Better than blockading Cuba / bombing Iran / etc.

“The US also does very bad stuff” doesn’t make BRI goodwill. Plus, there are more than two countries in the world. Some even try viable (if self-interested) development policy without bombing people.

[d5lt5]

> Yes. That’s why development and debt are hard problems. Also why calling it “goodwill” is, at best, too generous.

One can call the intent 'goodwill'. It doesn't mean that the outcome is satisfactory for your economic expectations. Judging from exceptions is not a valid approach and is a weird take.

> “The US also does very bad stuff” doesn’t make BRI goodwill.

True. I used that as an example of an alternative approach. The reader can decide which one is more 'goodwill'.

> Some even try viable (if self-interested) development policy without bombing people.

What countries are you referring to here: France (Douala and Abidjan ports, North–South railway in Vietnam), Japan (also ports in Sri Lanka, Thilawa), something else?

> Don’t make me say what I did not.

That conclusion says more about your reading than about what I actually wrote.

> Basically, that unlike your "goodwill" claim, China isn’t just giving away infrastructure for free out of the goodness of its government’s heart.

I shoot back with "Don’t make me say what I did not.", and 'goodwill' doesn't mean 'free stuff', you may want to check the dictionary ;)

[ElFitz]

> One can call the intent “goodwill”.

One can call anything anything. And intents can only be guessed at, while outcomes can actually be evaluated. The question is whether the label explains the policy.

BRI is state-backed finance tied to Chinese strategic, commercial, and diplomatic interests. Some projects may benefit recipients too. Great. That still doesn’t make "goodwill" a useful description of the principle.

> Judging from exceptions is not valid.

I’m not judging from one exception. I backed my point of view with examples of a broader pattern: debt-financed infrastructure, Chinese contractors/suppliers/operators capturing much of the spending, and projects that often also serve China’s trade and influence goals.

Feel free to provide substantive counter-examples instead of just waving the word "goodwill" around.

> I used that as an example of an alternative approach.

No, you used US violence as a contrast to make BRI look benevolent. That’s whataboutism. "Less bad than bombing" is not the same thing as goodwill.

> What countries are you referring to here: France (Douala and Abidjan ports, North–South railway in Vietnam), Japan (also ports in Sri Lanka, Thilawa), something else?

Given its own propensity to rely on bombing, I would not use France as an example. The EU-financed port and airport in Gaza, for all the good they were allowed to do, come to mind. Japanese development and aid efforts too, even if they, like most if not all state-sponsored efforts, come with strings attached.

> goodwill doesn’t mean free stuff

Obviously. But you’re the one explicitly framing it as humanitarian. If your "goodwill" consists of lending countries money for infrastructure that often serves the lender’s own strategic interests and pays its own companies, then I’m going to question the framing.

Keep the dictionary for Scrabble.

[fragmede]

Tales of an Economic Hitman was an instruction manual for the Belt and Road Initiative.

[d5lt5]

It's a long stretch in mental gymnastics with no factual proofs.

[bostik]

More likely the PRC sees the open-weight models' progress as a way to prevent an existing dominant player from cementing their (finicky) lead and pulling up the ladder.

That strategy happens to have beneficial side effects to the global Hoi Polloi, but to attach any kind of benevolence to it would be naive.

[d5lt5]

How would open-weight models benefit PRC better than their own closed-weight models, but still available at lower prices? If anything, open-weights can be distilled far easier.

[bostik]

Thinking like a business vs. thinking like a state.

If you see a given technology as fundamental[tm], you want to ensure that you will retain access to it AND its ongoing development. China may well foresee a possible future where US imposes export controls and global sanctions to block PRC from having access to the necessary equipment to either train or use the most advanced models - let alone its alternate parallel universe where US might go as far as prevent anyone else than US themselves having the most advanced forms of the technology at all.[ß]

To ward off such a scenario, China doesn't need to become the sole leading supplier. They only need to guarantee that nobody else can even try to block them off, and that the technology itself can never be yanked.

ß: What could possibly give them such ideas?

[fc417fc802]

Because lower prices with closed weights would be severely compute constrained which would tightly cap the damage to american firms. As it is there's a plethora of providers (many of them american) serving up the cheap open weight models. Even tightly regulated industries with security concerns can use the latest deepseek.

It also enables further R&D using the open models as a starting point. That doesn't benefit china directly but it does serve to further undermine the lead that the american frontier labs have which limits their future ability to cut geopolitical adversaries off. In that sense it provides a long term hedge by minimizing the damage in the worst case scenario where china ends up suffering a crushing defeat in the AI race for whatever reason.

[d5lt5]

> Even tightly regulated industries with security concerns can use the latest deepseek.

That is not necessarily true, as "tightly regulated industries with security concerns" are also afraid of deepseek models generating vulnerable code. Even a possibility of that prevents those industries from deployments.

> That doesn't benefit china directly but it does serve to further undermine the lead that the american frontier labs have which limits their future ability to cut geopolitical adversaries off.

So, basically, competition is bad for US models? That argument doesn't address open-weights. And it doesn't work the other way around, because in that case China should be releasing close-weights instead.

[fc417fc802]

> So, basically, competition is bad for US models?

Having less of a lead is bad for US frontier labs.

> That argument doesn't address open-weights.

Open weights does far more to further that goal than closed weights ever could. The bit right before the sentence you quoted was about enabling further R&D using open models as a starting point. In the same way that open weights serves as a force multiplier when it comes to flooding the market with cheap inference it also serves as a force multiplier if your goal is the raise the R&D floor. In this scenario a reasonable move for a capable opponent that's lagging behind is to attempt to raise the floor of the whole market so as to erode the position of whoever is in the lead.

Also consider that (as I previously mentioned) raising the capabilities of the global market as a whole serves to minimize damage in a worst case scenario. If you find yourself needing to depend on a foreign supplier for a key technology then it's better to have as many choices as possible from as many different jurisdictions as possible.

[holoduke]

They have the same values. Domination it is. People are people. Really no difference between the US and China. None at all.

[d5lt5]

Well, you are wrong. Maybe you should visit and learn more about China to understand it. For starters, China's society is high-individualistic with a strong sense of community and with high respect to their elders. On the contrary, US's society is hyper-individualistic with a strong sense of family and basic respect to their elders.

[KPGv2]

> AI is and always has been a winner-take-all strategy

AGI sure. But I don't think we're going to get there in our lifetimes, if ever. There are too many structural and physical limitations. One everyone seems to be catching onto now is that they're running out of human data and are incestuously feeing AI output to itself as input.

Current state AI is barely an improvement over where it was fifty years ago. We just have stronger hardware and more content to train on. We need a new paradigm. One that hasn't come in half a century.

[59nadir]

> Any country that developed sufficiently advanced models will pursue the same path.

Looking at most of the available evidence, Mythos is an incremental upgrade over other models and nowhere near the implied advancement that this seems to point to. I guess you could be right in that a sufficient advancement would cause this type of withholding of it, but I think it's kind of silly to think that the US has reached that level.

[_3u10]

I don’t think the ai market has been around long enough to prove winner take all, I see it more like a commodity which is why the winner quickly switched from OpenAI to Anthropic at least on coding.

I generally find the Chinese models to be superior. If you look at the token pricing or hosting cost it’s still about 80% margin on Chinese models. Deep cuts should still be possible.