@clegra @Gargron @erictopol Anybody who thinks LLMs work the same way as humans, or any animal, is confusing the simplified explanations of LLMs and intelligence with reality.

Also a model built to understand something complex is by no necessity actually behaving internally like that more complex model.
I get the feeling Hinton is stuck equating external with internal behaviour, like psychologists who insist people with ASD "don't feel empathy" just b/c they don't express it like neurotypicals.