视频加载失败？在 YouTube 打开 ↗
通过奖励建模训练AI，无需编写奖励函数
Training AI Without Writing A Reward Function, with Reward Modelling
AI 翻译，未经编辑审核
讲什么
Robert Miles 介绍让 AI 从人类偏好比较中学习奖励，并讲解相关强化学习论文。
HN 网友推荐理由
HN 网友推荐理由：被视为 RLHF 思想的清楚入门，也有人借它类比人类动机。
摘要与推荐理由由 HackerTube 用 AI 生成，大榕 AI 翻译整理，未经编辑审核。
视频数据来源：HackerTube（https://hackertube.org），基于 Hacker News 社区推荐整理，依 ODC-By 1.0（https://opendatacommons.org/licenses/by/1-0/）许可使用。筛选与中文整理：大榕 darong.ai。
展开英文原文字段
Title
Training AI Without Writing A Reward Function, with Reward Modelling
Summary
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.
Why HN
Seen as the clearest introduction to the idea behind RLHF, and also borrowed by commenters as a metaphor for human motivation.
Primary topic
reward modelling and rlhf