How LLMs Might Think
Joseph Gottlieb, Ethan Kemp, Matthew Trager
Why It Matters
What makes this one worth your time
Understanding the nature of LLM 'thinking' could influence how we interpret their outputs and guide future AI development.
The paper suggests LLMs might think in an arational, associative way if they think at all.
Summary
The paper critiques an existing argument against the idea that large language models (LLMs) think, proposing instead that if LLMs do think, they do so in an arational, associative manner.
Key contributions
- Critique of the argument from rationality against LLM thinking.
- Proposal of an alternative view on LLM thinking as associative.
Notable insights
- The notion that LLMs might engage in arational, associative thinking is an unconventional perspective.
Possible limitations
- Not stated in the abstract
Abstract
arXiv:2604.09674v2 Announce Type: replace Abstract: Do large language models (LLMs) think? Daniel Stoljar and Zhihe Vincent Zhang have recently developed an argument from rationality for the claim that LLMs do not think. We contend, however, that the argument from rationality not only falters, but leaves open an intriguing possibility: that LLMs engage only in arational, associative forms of thinking, and have purely associative minds. Our positive claim is that if LLMs think at all, they likely think precisely in this manner.