AI safety via debate
✨ AI Summary
🔊 جاري الاستماع
We’re proposing an AI safety technique which trains agents to debate topics with one another, using a human to judge who wins.
