LLM 意图性：多层视角
分析
本文探讨了大型语言模型（LLM）在多层面上拥有意图性的可能性，挑战了它们的行为仅仅由“聊天游戏”心态驱动的观点。作者认为，即使底层的LLM仅仅是在预测下一个token，它也可能像Searle的中文房间一样，召唤出具有自己独特意图的“模拟体”。
要点
引用 / 来源
查看原文"I think the mistake Keith is making is to restrict his analysis to the level of the LLM itself and to fail to consider that there may be a distinct agent supervening on top of it in much the same way as The Chinese Room supervenes on top of Searle."