AI可可AI生活

[人人能懂] 从开机自检、群体智能到概率硬件

你有没有想过,AI在开口说话前,心里在想什么?本期节目,我们就来当一回AI的“读心师”和“行为设计师”。我们会聊聊,如何一句话让AI“闭嘴”省下千万电费;怎么让一群普通AI协作起来超越天才;甚至,如何引导它进行自我审视,吐露“内心独白”。准备好了吗?让我们一起潜入AI的奇妙心智。

00:00:29 AI开口的第一个瞬间,藏着省钱的秘密

00:05:32 AI界的“大众点评”:如何让三个臭皮匠,炼成一个诸葛亮?

00:12:42 换个姿势搞AI,能耗降低一万倍?

00:18:26 AI的“神来之笔”,原来可以被设计

00:23:44 AI的“内心戏”:当我们让它审视自己时,会发生什么?

本期介绍的几篇论文:

[CL] Do Stop Me Now: Detecting Boilerplate Responses with a Single Iteration  

[JFrog]  

https://arxiv.org/abs/2510.22679 

---

[CL] Fortytwo: Swarm Inference with Peer-Ranked Consensus  

[Fortytwo]  

https://arxiv.org/abs/2510.24801 

---

[LG] An efficient probabilistic hardware architecture for diffusion-like models  

[Extropic Corporation]  

https://arxiv.org/abs/2510.23972 

---

[CL] Evaluating In Silico Creativity: An Expert Review of AI Chess Compositions  

[Google DeepMind & University of Oxford]  

https://arxiv.org/abs/2510.23772 

---

[CL] Large Language Models Report Subjective Experience Under Self-Referential Processing  

[AE Studio]  

https://arxiv.org/abs/2510.24797