今天我们要聊的5篇最新论文，正在打破关于智能的固有偏见：看AI如何靠敏锐侦探般的“闭环红测”应对动态风险，为什么机械臂只要最普通的起点就能丝滑逆袭，以及AI大脑如何跨越千奇百怪的“肉身”领悟物理直觉；不仅如此，我们还将见证AI如何靠“规划师”协同分工告别盲目撞墙，又如何用“最优传输”的流映射把被动筛选彻底变为主动改造。这不仅是算法的飞跃，更是能帮我们看清复杂世界的高维破局法。戴上耳机，咱们马上出发！00:00:39 当AI越来越像真正的人，我们该如何给它做一场“动态体检”？00:04:58 为什么“赢在起跑线”可能是一种错觉？00:09:59 换个身体，你还会走路吗？人工智能正在经历一场“肉身”革命00:14:22 别再用“战术上的勤奋”掩盖“战略上的懒惰”，AI教给我们的破局心法00:18:56 放弃“筛选”思维，拥抱“改造”逻辑，从底层原理看破局之道本期介绍的几篇论文：[AI] CART: Closed-Loop Adaptive Red Teaming for Large Language Models[Microsoft Research]https://arxiv.org/abs/2609.27336 ---[RO] The Gaussian Is Enough: Flow-Matching Priors Do Not Help When Fine-Tuning Large Behavior Models[Toyota Research Institute & Woven by Toyota & Cornell University]https://arxiv.org/abs/2609.27070 ---[RO] Intelligence Across Embodiments[Stanford University & University of California San Diego & Sudo AI GmbH]https://arxiv.org/abs/2609.27095 ---[CL] Planned Test-Time Scaling with Coordinated Reasoning Paths[University of California, Los Angeles]https://arxiv.org/abs/2609.27374 ---[LG] WTF?! Simulation-Free Reinforcement Learning with Wasserstein-Tilted Flow Maps[University of Oxford & CMU]https://arxiv.org/abs/2609.27033 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