AI解决问题，一定要更强更快吗？今天我们换个角度，看看AI如何学得更“巧”。我们将探讨，如何让AI从一个“全能设计师”变成一支“专业施工队”，甚至学会从你选择时的犹豫中读懂你的心思。我们还会发现，AI制药不再是闭门造车，而被赋予了空间直觉；而处理音频的AI，也终于懂得直接从MP3里学习，不再做“返工”的傻事。准备好，让我们一起进入AI的“巧思”世界！00:00:36 AI制药的新玩法：如何把每个原子放到正确的位置？00:06:12 让AI当好“施工队”，而不是“总设计师” 00:11:42 自动驾驶的新思路：扔掉老师，你也能当状元？00:17:20 AI读心术：你选得有多快，它学得有多好00:22:34 AI新范式：为什么你的MP3比原始音频更“聪明”？本期介绍的几篇论文：[LG] Pearl: A Foundation Model for Placing Every Atom in the Right Location[Genesis Molecular AI]https://arxiv.org/abs/2510.24670---[LG] TDFlow: Agentic Workflows for Test Driven Software Engineering[CMU & UC San Diego]https://arxiv.org/abs/2510.23761---[RO] ZTRS: Zero-Imitation End-to-end Autonomous Driving with Trajectory Scoring[Fudan University & NVIDIA]https://arxiv.org/abs/2510.24108---[LG] Preference Learning with Response Time: Robust Losses and Guarantees[Stanford University]https://arxiv.org/abs/2505.22820---[LG] Transformers from Compressed Representations[King Abdullah University of Science and Technology]https://arxiv.org/abs/2510.23665

Ask this episode

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

Transcript

Related