Training AI Without Writing A Reward Function, with Reward Modelling
Robert Miles AI Safety · 3 則 HN 推薦
影片問答
可以請 AI 解釋目前片段，或總結整部影片。如果回答附有時間戳記，點選即可跳到對應片段。
試著問：這裡講的是什麼意思？
需使用 Google 登入。確認前可查看最多會扣除多少點數。
Robert Miles explains reward modelling, in which an AI learns what humans want from their preference comparisons instead of a hand-written reward function. He walks through the 2017 paper on deep reinforcement learning from human preferences.
HN 推薦的原因： Seen as the clearest introduction to the idea behind RLHF, and also borrowed by commenters as a metaphor for human motivation.
“I find this video by Rob Miles still the best presentation of the ingenious original 2017(!) paper” 閱讀 HN 討論 ↗