[人人能懂] AI进化下一站：学会犯错、遗忘与谦虚？

00:01:32 想让AI变聪明？关键不是喂饱它，而是教它如何“犯错” 00:07:12 你的“电量焦虑”，AI来想办法了？00:11:21 给AI装上一个“选择性失忆”的大脑00:15:41 AI“现学现卖”的秘密，可能被揭开了   00:21:06 AI也需要“自知之明”：我们如何教会机器谦虚？本期介绍的几篇论文：[LG] From Reasoning to Super-Intelligence: A Search-Theoretic Perspective  S Shalev-Shwartz, A Shashua  https://arxiv.org/abs/2507.15865 ---[LG] Expert-Guided LLM Reasoning for Battery Discovery: From AI-Driven Hypothesis to Synthesis and Characterization  S Liu, H Xu, Y Ai, H Li...  Université de Montréal & University of Oxford  https://arxiv.org/abs/2507.16110 ---[CL] Beyond Context Limits: Subconscious Threads for Long-Horizon Reasoning  H Luo, N Morgan, T Li, D Zhao...  MIT CSAIL  https://arxiv.org/abs/2507.16784 ---[CL] Learning without training: The implicit dynamics of in-context learning  B Dherin, M Munn, H Mazzawi, M Wunder...  Google Research  https://arxiv.org/abs/2507.16003 ---[LG] Beyond Binary Rewards: Training LMs to Reason About Their Uncertainty  M Damani, I Puri, S Slocum, I Shenfeld...  MIT  https://arxiv.org/abs/2507.16806  在小宇宙查看该单集文稿