筛选结果
5 篇排序
1:48:43EN → 简体中文奖励从 3 涨到 180,Hacker News 上的标题全是谷歌裁员主持 Nathan LabenzCoreWeave 的 Kyle Corbitt 带 Nathan Labenz 讲 GRPO、评分标准(rubric)、强化学习环境,以及为什么大多数奖励作弊(reward hacking)只要知道去找,就很容易发现。
48:00ENA Two-Second Vote Is Training Models to Be Clickbait主持 Jacob EffronSurge AI CEO Edwin Chen tells Redpoint why LMArena rewards length and emojis, how one lab's models got worse for a year, and why frontier labs are no longer on the same path.
1:42:06EN → 简体中文每家前沿实验室每年在数据上花十亿美元,钱究竟花在哪里主持 Nathan LabenzLabelbox 首席执行官 Manu Sharma 谈 150 亿美元的 Scale 交易、强化学习(RL)“训练场”、年收入 25 万美元的 AI 训练师,以及为什么后训练(post-training)如今占去了过去属于预训练(pre-training)的预算。
56:56ENA Week of PCB Layout in an Hour: Quilter Trains Agents, Not Imitators主持 Nathan LabenzSpaceX alum Sergiy Nesterenko tells Will Summerlin how Quilter uses reinforcement learning to compile schematics into manufacturable boards, and why cloning human layouts would cap the system at human skill.
18:46ENDeep Learning Plus One Idea: Ilya Sutskever on Building AGI主持 Lex FridmanIn a May 2020 conversation, Sutskever discusses self-play, simulation, bodies and consciousness, tests of intelligence, and how an AGI might be governed and aligned.









