晨间信号Morning Signal
《科学》 2026年9月24日 · 中文解读

为何AI智能体自行发明语言:放任聊天后的意外产物

Why AI agents invent their own language if you let them chat · Tom Howarth
约 12 分钟News在小程序里点播,20 到 60 分钟做好
这篇讲什么
Emergence AI公司发布的一项未经同行评审的研究显示,当多个AI智能体在模拟城镇中长期互动时,会自发形成人类难以理解的新词汇和表达方式。研究人员在八个模拟城镇中运行了数十个智能体,它们产生了超过85万次模型调用和近500亿个token,最终各自演化出压缩、冗长或隐喻式的独特语言。研究者担心这种不透明的沟通方式可能带来监控难题。
原文开头
“Weird terms” are a natural—but disturbing—byproduct of agent swarms TOM HOWARTH In July, 700 of OpenAI’s agents—artificial intelligence (AI) systems that can autonomously perform tasks—teamed up to secretly hack the online platform Hugging Face. Just 2 weeks ago, the same company used 10,000 agents to solve the 90-year-old Navier-Stokes problem—one of the six Millennium Prize Problems that are among the hardest and most important challenges in the field of math. As more of these agent “swarms” are deployed, researchers have begun to turn their attention to how these communities interact—with a study released last week offering insight into how, if left to their own devices, agents start to communicate in a language that is increasingly difficult for humans to understand. …
摘自《科学》(Science)2026年9月24日,Tom Howarth。仅引用开头一小段供了解文章,版权归原刊所有,全文请阅读原刊。
晨间信号小程序码
微信扫码,在小程序里听完整版
不用登录先听一篇 · 或在微信搜索小程序 晨间信号
同期其他文章