It should be noted that the first person to chose AI over rules systems was George Hotz from comma.ai. I cannot find the video but he made a pretty powerful argument for AI being superior over rules systems years before the arrival of ChatGPT and it was enough to convince me this approach would be the eventual winner.

Now I wonder who could have given him that idea. I don't think I've ever seen him put together a line of reasoning by himself.

It says that this does for driving what Stable Diffusion does for images, ChatGPT for chat, AlphaZero for chess. And that they're finding it to be a profound improvement as in the other domains. From Walter Isaacson that evaluation makes me bullish on competent general self driving real soon now.