David Krueger–Coordination, Alignment, Academia
The Inside View
00:00
The Importance of Language Models for Safety
There was this simulator post on less wrong about how like the language, any language model can just like simulate an image in inside them. So kind of the language model is like living in some quantum physics world and when he's like, observe some, something like, Oh yeah. I mean, this world now this is what follows. And then as soon as you like actually plug it into a robot and like train it for a little bit on that's kind of tasks, it's just like, Oh, oh, I get it. Yeah. Some people have these stories where, you know, just through this carefully controlled like text interaction, like maybe we just ask this, uh, thing
Transcript
Play full episode
