Nathan Lambert: belief

Transcript context

…Yeah, it has to learn a lot about you specifically. And this goes to continual learning—about the general mistakes that are made throughout, and then mistakes that are made through you. All the AI interfaces are getting set up to ask humans for input. I think Claude Code we talked about a lot. It asks feedback and questions. If it doesn’t have enough specification on your plan or your desire, it starts to ask questions, “Would you rather?” We talked about Memory, which saves across chats. Its first implementation is kind of odd, where it’ll mention my dog’s name or something in a chat. I’m like, “You don’t need to be subtle about this. I don’t care.” But things are emerging, like ChatGPT has the Pulse feature. Which is like a curated couple paragraphs with links to something to look at or to talk about, and people talk about how the language models are going to ask you questions. It’s probably going to work. The language model knows you had a doctor appointment or something, and it’s like, “Hey, how are you feeling after that?” Which again, goes into the territory where humans are very susceptible to this and there’s a lot of social change to come. But also, they’re experimenting with having the models engage. Some people really like this Pulse feature, which processes your chats and automatically searches for information and puts it in the ChatGPT app. So there’s a lot of things coming. I used that feature before, and I always feel bad because it does that every day, and I rarely check it out. How much compute is burned on something I don’t even look at, you know?…

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