Jensen Huang on China open source AI
by Vojtech: Jensen Huang
So the question of where open models come from might be missing the actual mechanics here, and Jensen Huang laid it out pretty plainly. His read on it: the overwhelming share of open source work happening worldwide traces back to China at this point. Not some grand coordinated strategy, just raw volume. More engineers, and universities like Tsinghua turning out science and math grads at a scale Huang openly calls a disadvantage for the US side. But the interesting bit is what happens once you hit download. Linux, Kubernetes, a huge chunk of everyday software has Chinese fingerprints all over it, and the second it lands on your machine it's yours. Fork it, improve it, claim it. Open weights models work the same way. The moment one is sitting on your local drive, its national origin is basically gone. It's just whatever you build with it. So governments can keep arguing about bans while the guy whose chips sit under nearly every AI lab points somewhere else entirely: these models cost nothing, they keep getting sharper, and no restriction really changes who can grab them. Which one are you actually running on your own setup though? Full cost rundown for each one below.
Transcript (en)
Let me just ask, does it matter if the open models come from China or the US? Well, we're doing everything we can to make a contribution in open models. However, the moment you download, for example, probably the vast majority of the world's contribution to open source today is coming from China. They just have a lot more engineers. They produce everything in large scale because it's a larger country. And so they produce science and math students in volume. That's one of our disadvantages. They're manufacturing them through amazing universities like Tsinghua University in high volume. Well, they contribute open source today. We download Linux, we download Kubernetes, we download all the software. A lot of it has been touched by Chinese. Once you download it, it's yours. We fork it, we improve it, we make it ours. When you download one of these Chinese models, it just happens to be made by some really great researchers in China. Thank you.
