[人人能懂AI前沿] 从具身认知、探索式建模到并行解码

本期我们来聊聊AI如何突破成长的瓶颈：最新的几篇论文告诉我们，聪明的AI正努力摆脱那个看不见自己身体的“幽灵”状态，并学会了用“笨办法”来激发创造力。它甚至开始成为自己精明的“预算会计”和高效的“项目经理”，最终踏上了“自我进化”的道路，试图自己教会自己如何变得更强。

00:00:26 你的AI为什么像个“幽灵”？

00:05:40 AI绘画的“笨办法”，如何成了进化的新方向？

00:11:12 AI画画的下一关，不是更有才，而是更会算计

00:17:14 AI作画，如何从“精雕细琢”到“一挥而就”？

00:22:51 如何让AI自己进化成一个更强的AI？

本期介绍的几篇论文：

[CV] HumanCLAW: Can Vision-Language Models Act Through a Body?

[Meta & University of Washington & Nanyang Technological University]

https://arxiv.org/abs/2607.27180

---

[LG] Explorative Modeling: Unlocking a Third Pretraining Axis and End-to-End Generation

[UIUC & Harvard]

https://arxiv.org/abs/2607.27372

---

[CV] Chimera: Designing and Chinchilla-Scaling Hybrid Visual Diffusion Transformers

[Adobe Research]

https://arxiv.org/abs/2607.28611

---

[CV] Parallel Decoding Distillation for Fast Image and Video Generation

[NVIDIA & Weizmann Institute of Science]

https://arxiv.org/abs/2607.26004

---

[CL] Frontis-MA1: Training an AI4AI Model towards Recursive Self-Improvement in Machine Learning Engineering

[Horizon Research & Tsinghua University]

https://arxiv.org/abs/2607.28568