Paul Christiano

4 posts

Learning Complex Goals with Iterated Amplification

Gathering Human Feedback

Learning from Human Preferences

Concrete AI Safety Problems