今天，我们来聊聊如何让AI不再只靠“大力出奇迹”，而是学会更聪明地工作。我们会看到，AI如何学会“继承”自己的思考，不再用后即焚；又如何像个聪明的导演，把算力“增援”到最关键的地方。我们还会发现，机器人如何掌握了快慢有度的“节奏感”，以及一个好的系统为何要懂得“聪明的懒惰”。这几篇最新论文，将带我们一窥AI从“野蛮生长”到“精耕细作”的进化之路。00:00:32 让AI告别“用后即焚”的思考模式00:05:29 AI解题新思路，如何把一份算力，掰成八瓣花？00:10:45 机器人也懂的“快慢之道”00:15:42 成大事者，为什么都懂得“懒惰”的艺术？00:21:29 给AI一盒乐高，让它自己搭出新世界本期介绍的几篇论文：[AI] Full-bandwidth transformer[Johns Hopkins University & Princeton University & Microsoft]https://arxiv.org/abs/2608.08888---[AI] Thought-Level Beam Search for Reasoning[Princeton University & MIT & Meta AI]https://arxiv.org/abs/2608.08020---[RO] SpeedTuning: Speeding Up Policy Execution with Lightweight Reinforcement Learning[Stanford University]https://arxiv.org/abs/2608.09138---[LG] Beyond Binary: Continuous State Optimization with Graph-Structured Objectives[Google Research & Tel Aviv University]https://arxiv.org/abs/2608.09366---[LG] Idea Search: Guiding Tree Search with Ideas to Explore Diverse Scientific Methods[California Institute of Technology & Google Research]https://arxiv.org/abs/2608.08958在小宇宙查看该单集文稿

Ask this episode

“What did the council decide on item 42?” — answered with the speaker and the moment.

Transcript

Related