AI safety via debate 事件

REGULATION2018-05-03影响: MEDIUM

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