Dwarkesh Patel: belief

Transcript context

…I do think the tide is changing there for whatever reason. Neel Nanda has had a ton of success promoting interpretability in a way where Chris Olah hasn't been as active recently in pushing things. Maybe because Neel's just doing quite a lot of the work, I don't know. Four or five years ago, Chris was really pushing and talking at all sorts of places and these sorts of things and people weren't anywhere near as receptive. Maybe they've just woken up to the fact that deep learning matters and is clearly useful post-ChatGPT. It’s kind of striking. Okay. I'm trying to think of a good last question. One thing I’m thinking of is, do you think models enjoy next token prediction? We have this sense of things that were rewarded in our assessor environment. There's this deep sense of fulfillment that we think we're supposed to get from things like community, or sugar, or whatever we wanted on the African savannah. Do you think in the future, models that trained with RL and a lot of post-training on top, they'll like predicting the next token again in the way we just really like ice cream. Like in the good old days. So there's this ongoing discussion of “are models sentient or not” and “do you thank the model when it helps you?” But I think if you want to thank it, you actually shouldn't say thank you. You should just give it a sequence that's very easy. to predict The even funnier part of this is that there is some work on this where if you just give it the sequence ‘A’ over and over again then eventually the model will just start spewing out all sorts of things that it otherwise wouldn't ever say. So I won't say anything more about that but you should just give your model something very easy to predict as a nice little treat.…

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