6 comments

[ 2.1 ms ] story [ 7.7 ms ] thread
The export control on architectures for any open-weight model with 2T or more parameters OR in the top 10 of a global leaderboard is huge.
> APIs, apps and cloud services would be exempt, preserving overseas access to Chinese models

Makes sense. A lot of people I have seen that use chinese models, just download it and host it on US or their country's infra, so no money goes to alibaba. If they are OK with API access, they just use anthropic or openai. This move will basically cut that tap and make it so if you want the cheaper chinese models, you have to pay deepseek or alibaba themselves.

Yep. This was also why some of the more idealistic open model types at Alibaba were axed.

I said this was happening months ago but no one listened.

What's the chance of it actually working though. The distribution of chinese software is not that good (e.g theyre pretty much never gonna be the first model in whatever frankenstein salesforce is going to sell).
The point was never about exports or soft power - that was an HN spin and idealism.

They just want to monetize and scale out their hyperscalers and the associated shared services, which is what AI has become.

So it'll work to meet their needs - a decoupling of the US and the China ecosystems, along with reducing the risk that India, Vietnam, Korea, or another country tries to conduct distillation attempts on Chinese models as well.

Most ExChina organizational buyers were they to use a Chinese open-weight model would create very strict sandboxes (none of the VPC or monitoring fuck-ups that OAI and Anthropic made) and not use the Chinese cloud.

> They just want to monetize and scale out their hyperscalers and the associated shared services, which is what AI has become.

Yeah, for that, distribution via SaaS-ish products is important right? My thinking for now is that 10 years down the line, the responses API is not going to be what any of these guys will be selling. Rather it will be like whatever salesforce is doing, or that recently announced GPT-Synopsys product. Or stripe, AWS, type companies will sell it integrated here and there. Chinese software doesn't have that same reach right.

Within china of course they can scale out if thats what you meant