AIHT

内容

精选全部 AI 动态AI 日报热点漫剧榜主题收藏

接入

Agent 接入

更多

关于公开接入条款隐私模型榜更新日志反馈
← 返回原文 ↗
OpenAI Blog47

AI safety via debate

2018/05/03 15:00· 3086天前
AI 摘要

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

Agent 智能体论文论文研究
查看原文(OpenAI Blog)
精选全部日报更多