AI Safety via Debate

Link post

New pa­per and blog post by Ge­offrey Irv­ing, Paul Chris­ti­ano, and Dario Amodei (the OpenAI safety team).