AI Safety Fundamentals: Alignment cover image

AI Safety via Debate

AI Safety Fundamentals: Alignment

CHAPTER

The Properties of Debate as an Approach to AI Alignment

The goal of this paper is to lay out theoretical and practical properties of debate as an approach to AI alignment. We can approximate optimal play by training ML systems via self-play, which has shown impressive performance in games such as Go, Chess, Showgi, and Dota 2. Where the debate works with humans and machine learning is more subtle and requires extensive testing and analysis before the model can be trusted.

00:00
Transcript
Play full episode

Remember Everything You Learn from Podcasts

Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
App store bannerPlay store banner