[人人能懂] 抛弃“大力出奇迹”后，AI的下一站在哪？

00:02:21 AI巨头们的“乐高”心法：应对复杂世界的终极武器

00:07:24 AI小助理的养成记：笨徒弟如何变高徒？ 

00:12:47 你家AI聪不聪明，得看它“跑”得稳不稳 

00:18:43 AI也会“看走眼”？一招教它“看真切”

00:22:03 AI的“笨”办法：像婴儿一样学习世界 

00:26:54 AI也“开小差”？不说话的思考，可能更强大 

今天介绍的六篇论文：

[LG] AXLearn: Modular Large Model Training on Heterogeneous Infrastructure  

[Apple]  

https://arxiv.org/abs/2507.05411 

---

[LG] MobileGUI-RL: Advancing Mobile GUI Agent through Reinforcement Learning in Online Environment  

[Tencent AI Seattle Lab]  

https://arxiv.org/abs/2507.05720 

---

[LG] A Dynamical Systems Perspective on the Analysis of Neural Networks  

[Freie Universitat Berlin & Universiteit van Amsterdam]  

https://arxiv.org/abs/2507.05164 

---

[CL] Perception-Aware Policy Optimization for Multimodal Reasoning  

[University of Illinois Urbana-Champaign]  

https://arxiv.org/abs/2507.064 

---

[LG] Thousand-Brains Systems: Sensorimotor Intelligence for Rapid, Robust Learning and Inference  

[Thousand Brains Project]  

https://arxiv.org/abs/2507.04494 

---

[CL] A Survey on Latent Reasoning  

https://arxiv.org/abs/2507.06203