LLMs seem to have a pretty fundamental problem: they can't learn beyond their training data.

Also, can you really say Google isn't AI? Pretty sure there's a whole expert system lurking in there...