今天，我们来一场深入AI大脑的探秘之旅，看看它是如何像一个聪明的徒步者一样，高效打包海量知识的。接着，我们会揭开一个流行“省钱”捷径背后的意外代价，并拷问那些华丽的商业模型，我们看到的“聪明”究竟有多少是障眼法。我们还会戳破量子AI的“皇帝新衣”，看看真正的“量子优势”何时才能走出实验室。最后，我们将见证AI如何化身基因侦探，不仅找出答案，更能画出罪犯间的“社交网络”，真正理解“为什么”。00:00:38 你的大脑如何打包信息？AI训练给了个新答案00:05:39 你用的大模型，是个“盲盒”？00:12:23 人工智能的“省钱”智慧，一个你不知道的代价00:18:38 量子AI的“皇帝新衣”？00:24:55 AI当侦探，如何破译基因里的“社交网络”？本期介绍的几篇论文：[LG] Sharp Capacity Scaling of Spectral Optimizers in Learning Associative Memory [UC Berkeley & Princeton University & New York University] https://arxiv.org/abs/2603.26554 ---[CL] How Open Must Language Models be to Enable Reliable Scientific Inference? [MIT & EleutherAI & University of California San Diego] https://arxiv.org/abs/2603.26539 ---[CL] Weight Tying Biases Token Embeddings Towards the Output Space [EleutherAI & UC Berkeley] https://arxiv.org/abs/2603.26663 ---[CL] Entanglement as Memory: Mechanistic Interpretability of Quantum Language Models [Stanford University] https://arxiv.org/abs/2603.26494 ---[LG] A Boltzmann-machine-enhanced Transformer For DNA Sequence Classification [Tsinghua University & UC Berkeley] https://arxiv.org/abs/2603.26465

Ask this episode

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

Transcript

Related