OpenAI & Anthropic have both made calls for Chinese AI models to be banned in the US on national security grounds. While it is true countries have reason to distrust other countries’ tech, I doubt this is the real reason they are upset.
Their big problem is that Open-Source AI annihilates their chances of succeeding as businesses. Silicon Valley’s model of VC funding is to bet on many small start-ups, hoping one becomes a ‘unicorn’ - a multi-billion dollar company (like Google, Meta, etc) able to dominate an industry and rake in hundreds of billions of dollars.
Even if they succeed in banning Chinese Open-Source - does this mean they’ll become unicorns? I doubt it. The Chinese Open-Source AI models are superior to theirs. Most of the rest of the world will use them, and the real AI innovation will happen in the rest of the world. Meanwhile Americans will make do with the second-best AI, that can only survive when it gets the best banned.
I agree with some of the overall premise, open weight models will put immense financial pressure on the ai-only companies doing the foundation training like openai and anthropic. But lmao at the notion that ANY open source models are better than the existing cutting edge ones from the big players. The difference just doesn’t matter for many applications.
Chinese or not doesn’t matter for this, releasing open weight models is a strategy for those whose cutting edge models are not the absolute best. Metas entire strategy has been open weight since day 1. It’s not benevolence. They’re propaganda/ad machines and they want people to use them. If you’ve got to pay someone for that little extra reliability it’s still probably going to be anthropic, openai or Google.