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.
replySo 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.
China isn't at the frontier. OpenAI and Anthropic are.
replySurely that wouldn't last long if OAI/Anthropic stopped?
replyhard 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.
replyIf distillation was as easy as that, we would have 100s of competitors around the world.
replyChina 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.
replyEDIT:
> 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.
Surely all of china has the resources to match a single ai lab?
replymaybe? all of china is not in a unified effort to match a single lab
reply[dead]
replyI heard China is just distilling the frontier models, so if they stop so will China.
replyAnd I think it's naïve to believe only the US can make the hardware and has the brainpower to develop this technology.
replyWho'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.
replyThat 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
replyGiven that the US models likely used some of the innovations made by DeepSeek, I highly doubt it.
reply