685
0人工智能已经能下赢世界冠军、写诗、写代码，但让它把碗放进洗碗机却依然是一场灾难。这期节目从一个看起来很离谱的问题出发：为什么洗碗这种小事，对 AI 来说反而比下棋更难？答案不在算力，而在“智能到底是什么”。
结合《智能简史》（A Brief History of Intelligence）这本书，节目会从进化的角度拆解智能的来源：从最早能分清好坏的虫子，到会学习时间规律的鱼，再到能在脑中“预演未来”的哺乳动物。你会发现，人类的智能并不是一次性出现的能力，而是一层层叠加出来的结果。也正因为缺少这些底层能力，AI 才会在常识、物理世界和日常判断上频频翻车。听完这一期，也许你会对“智能”这件事，以及当下的 AI 热潮，多一分冷静的理解。
收听指南🧭
00:00 The Rosie paradox shows that machines can master games and language but fail at simple physical tasks because they lack basic common-sense understanding of the world.
03:54 The first evolutionary breakthrough appears in simple worms, which can label things as good or bad and make trade-offs instead of acting on pure reflex.
06:22 Fish introduce reinforcement learning by linking events over time, using prediction errors to learn patterns rather than only reacting to the present moment.
09:15 Mammals gain the ability to simulate the future in their minds, allowing them to imagine consequences before acting, which highlights what current AI systems are still missing.
Episode Transcript: pan.baidu.com 提取码:yyds
Original Book (epub): pan.baidu.com 提取码: yyds
