Arthur Kroeber: belief

Transcript context

…A previous guest I just had on, Victor Shih, said that the Chinese government might be reluctant to let AI development go full speed because of the destabilizing effects it might have on the political system or the inability potentially to control it. What's your sense on what will happen as AI becomes more economically valuable, a focal point in discussions about technological development and so forth? I have a lot of respect for Victor. I think he has a very good grasp of the Chinese bureaucratic mindset and how political calculations get made. That's his territory more than mine. So I'm very respectful of that view. I however tend to think that the Chinese are going to be pretty in favor of high-speed AI development. The reason is essentially the same reason that I was talking about earlier, about informatization and electrification. This is essentially a basic enabling technology that is useful for all of the things that can flow from it. My hunch would be that they would make the same bet with AI that they made with the Internet back in the 90s. With the Internet they said, “This technology could be dangerous, it could be out of our control, but we think actually it's going to be really beneficial and that we can control it. ” That bet paid off for them very, very well. My hunch is that they will look at AI as a similar sort of thing. Yes, it's possible that this could get out of our control, but we think unbalanced, this is really important technology. It's going to enable a lot of other things and we'll probably be able to figure out how to control it because our control systems are so good. It is interesting because if you look at the big development in Chinese AI recently, DeepSeek, it was not the result of some state program. It was the result of basically a quant hedge fund guy in Hangzhou sitting around and saying, “Oh, I could actually build a pretty good LLM model flowing out of my quant activities. Let's give it a go, see how it happens.” It worked great for him. The reports that I have seen suggest that this created a certain amount of disquiet in Beijing. They said, “Well, where did this guy come from and why is he not coming out of the official plan?” So there is some risk. I think that they would be a little bit nervous about freelancing on this. But on the whole I think they'll see this as another basic enabling technology that they should promote and figure out how to control over time. To the extent that China's advantages lie in its ability—in situations where centralization is important—to do it… Ironically, this might be a case where centralization might be helpful and they might not do it because there are these tremendous economies of scale in AI training. You have to keep increasing the training cost by 4x every single year. Especially given that the availability of these advanced chips is lower in China, it would be even more helpful to have one person who can use them all for training. This might be the situation where it would make more sense for the government to say, Huawei and ByteDance, you have to give all your chips to High-Flyer.…

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