Comment by jdietrich

The question I gave is a literal textbook example of abstract reasoning. LLMs are just very advanced text prediction, but they are also provably capable of abstract reasoning. If you think that those statements are contradictory, I would encourage you to read up on the Bayesian hypotheses in cognitive science - it is highly plausible that our brains are also just very advanced prediction models.