Sholto Douglas: evaluation

Transcript context

…Okay, today I have the pleasure to talk with two of my good friends, Sholto and Trenton. Noam Brown, who wrote the Diplomacy paper, said this about Sholto: “he's only been in the field for 1.5 years, but people in AI know that he was one of the most important people behind Gemini's success.” And Trenton, who's at Anthropic, works on mechanistic interpretability and it was widely reported that he has solved alignment. So this will be a capabilities only podcast. Alignment is already solved, no need to discuss further. Let's start by talking about context lengths. It seemed to be underhyped, given how important it seems to me, that you can just put a million tokens into context. There's apparently some other news that got pushed to the front for some reason, but tell me about how you see the future of long context lengths and what that implies for these models. So I think it's really underhyped. Until I started working on it, I didn't really appreciate how much of a step up in intelligence it was for the model to have the onboarding problem basically instantly solved. You can see that a bit in the perplexity graphs in the paper where just throwing millions of tokens worth of context about a code base allows it to become dramatically better at predicting the next token in a way that you'd normally associate with huge increments in model scale. But you don't need that. All you need is a new context. So underhyped and buried by some other news. In context, are they as sample efficient and smart as humans?…

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