原文整理页

Demis Hassabis 通过与 Gemini 下棋来追踪其思维链,发现模型有时明知是错招却依然执行,展现了“参差不齐”的智能

来源作者:Garry Tan (@garrytan)原始来源:https://x.com/garrytan/status/2051055687510438075

中文导读

Demis Hassabis 通过与 Gemini 下棋来追踪其思维链,发现模型有时明知是错招却依然执行,展现了“参差不齐”的智能。

正文 Markdown

Demis Hassabis says he plays chess with Gemini to trace the model's chain-of-thought. As a former chess prodigy, he can tell when the model starts reasoning itself into trouble. Sometimes it sees a blunder, searches for something better, then plays the blunder anyway. That’s what jagged intelligence looks like.