The Inside View cover image

Collin Burns On Discovering Latent Knowledge In Language Models Without Supervision

The Inside View

00:00

Is a Model Like RL Breaking the Law or Not?

I think that sort of thing I have in mind is okay we maybe get to human level language models and maybe we hook that up to other modalities as well like vision. We start with you know massive self-supervised pre-training something like that but then maybe we Hook that up to RL and we sort of fine tune this model using RL to do all sorts of tasks. So it seems like a really natural thing that there will be very strong economic incentives to do is have in RL agent maximize like the amount of money in my bank account something like this. This is a sort of system I think is scarylike if if the model is especially capable  it would have these long

Transcript
Play full episode

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!
App store bannerPlay store banner
 Get the app