I know people like that - the amount of inane, obsessive and just strange conversations they have with AI is concerning - there's never any actually useful result or information that they get out of these chats.

Furthermore, there are some clearly wrong questions where person asks AI to make some kind of numerical evaluation of some data. And evaluation is done entirely through inference - essentially a hallucination, instead of some one-off python script which can actually give deterministic calculated evaluation. Yet they accept the answer AI gives them.