OpinionMIT Tech Review

Don’t be fooled—LLMs don’t reason

#llms#reasoning#alphago#ai#science

English

The article argues that large language models (LLMs) lack genuine reasoning capabilities, contrasting their operation with AlphaGo's ability to combine intuition and deliberation. While LLMs can generate intermediate reasoning steps, they do not maintain a persistent epistemic state or separate knowledge from reasoning, which limits their effectiveness in high-stakes fields like science and medicine.

中文

文章认为大型语言模型（LLMs）缺乏真正的推理能力，与AlphaGo将直觉和深思熟虑结合的能力形成对比。虽然LLMs可以生成中间推理步骤，但它们不维护持久的认知状态或将知识与推理分开，这限制了它们在科学和医学等高风险领域的有效性。