> But as LLMs lack a way to reason, you get nonsense instead.

I suppose we should conclude humans don't reason either, since they do this all the time. Reason over blindspots without realizing it, is endemic in humans.

LLMs reason differently because their architecture is different, the way they learn is different, and their mode of inference is different. They will be bad at some things we are good at.

And yet their reasoning is far wider than ours (able to weave together virtually any disparate topics sensibly), and almost always better than humans at moderate depth one-shot responses.

Human's are truly awful at one-shot reasoning.