How to Train Language Models to Act Assist Humans

The idea here was just like, can we get language models to act as assistants to humans and in particular to ask clarifying questions from the human about what they want from the system. So far it's just an entirely abstract model of what it do like, but then it's another a concrete task. And I think you can relate it to LMM because it's like kind of close to, uh, cooperative IRL as well. Yeah. Do I, I think it's very good because I tried Googling on, on our archive and I think it is not out yet.

Transcript

Play full episode

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!

Get the app