[人人能懂] AI内卷启示录：效率、稳定与创造力的再平衡

00:00:35 AI进化论：当机器开始自己设计自己00:04:59 AI训练场上的新规则：别纠结错别字，看的是整篇文章 00:09:53 AI当医生？先等等，咱们换个活法儿 00:15:29 给AI做体检：我们能不花钱就看出模型好坏吗？ 00:20:35 AI没书读了怎么办？一个“笨”方法里的新智慧 本期节目介绍的五篇论文：[LG] AlphaGo Moment for Model Architecture Discovery  [Shanghai Jiao Tong University & SII]  https://arxiv.org/abs/2507.18074 ---[LG] Group Sequence Policy Optimization  [Qwen Team]  https://arxiv.org/abs/2507.18071 ---[LG] Towards physician-centered oversight of conversational diagnostic AI  [Google DeepMind & Google Research]  https://arxiv.org/abs/2507.15743 ---[LG] SETOL: A Semi-Empirical Theory of (Deep) Learning  [Calculation Consulting & Onyx Point Systems]  https://arxiv.org/abs/2507.179 ---[LG] Diffusion Beats Autoregressive in Data-Constrained Settings  [CMU & Lambda]  https://arxiv.org/abs/2507.15857 在小宇宙查看该单集文稿