News for hackers. Stuff that matters.
Re:Does Reasoning Emerge? Probabilities of Causation in Larg... by slashdave on Friday August 16, 2024 @08:41PM (#41271545)
Given that LLMs are trained on tokens and not symbols, that is, they have no context or world model, I would say that LLMs cannot symbolically abstract, given that they have neither a mechanism nor the data to do so.

There are no comments on this comment yet.