Nathan Lambert: belief

Transcript context

…I guess we should state explicitly, so we’ve been talking about these Chinese models and open weight models. Oftentimes, the way they’re run is locally. So it’s not like you’re sending your data to China or to whoever developed the model in Silicon Valley. A lot of American startups make money by hosting these models from China and selling them. It’s called selling tokens, which means somebody will call the model to do some piece of work. I think the other reason is for US companies like OpenAI. OpenAI is so GPU deprived; they’re at the limits of the GPUs. Whenever they make a release, they’re always talking about how their GPUs are hurting. And I think in one of these gpt-oss-120b release sessions, Sam Altman said, “Oh, we’re releasing this because we can use your GPUs. We don’t have to use our GPUs and OpenAI can still get distribution out of this,” which is another very real thing, because it doesn’t cost them anything. And for the user, I think also, I mean, there are users who just use the model locally how they would use ChatGPT. But also for companies, I think it’s a huge unlock to have these models because you can customize them, you can train them, you can add more data post-training, like specialize them into, let’s say, law, medical models, whatever you have. And you mentioned Llama; the appeal of the open weight models from China is that the licenses are even friendlier. I think they are just unrestricted open source licenses, whereas if we use something like Llama or Gemma, there are some strings attached. I think it’s like an upper limit in terms of how many users you have. And then if you exceed so many million users, you have to report your financial situation to, let’s say, Meta or something like that. And I think while it is a free model, there are strings attached, and people like things where strings are not attached. So I think that’s also one of the reasons besides performance why the open weight models from China are so popular, because you can just use them. There’s no catch in that sense.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.