[人人能懂] 当AI不再需要人类“喂”数据，会发生什么？

00:00:23 AI的自我进化：当左手成了右手的老师00:04:02 教AI“左右互搏”：成为高手的秘密00:07:59 AI界的“四两拨千斤”00:11:40 AI当学徒：一个“解题大师”是如何炼成的？00:16:30 AI的“后悔药”：让机器学会“重新理解”本期介绍的几篇论文：[LG] Language Self-Play For Data-Free Training  [Meta Superintelligence Labs]  https://arxiv.org/abs/2509.07414  ---[CL] Parallel-R1: Towards Parallel Thinking via Reinforcement Learning  [Tencent AI Lab Seattle]  https://arxiv.org/abs/2509.07980  ---[LG] K2-Think: A Parameter-Efficient Reasoning System  [Mohamed bin Zayed University of Artificial Intelligence]  https://arxiv.org/abs/2509.07604  ---[LG] Autonomous Code Evolution Meets NP-Completeness  [NVIDIA Research]  https://arxiv.org/abs/2509.07367  ---[CL] Causal Attention with Lookahead Keys  [ByteDance Seed]  https://arxiv.org/abs/2509.07301  在小宇宙查看该单集文稿