Yeah, I think making a decent, nuanced analysis of ChatGPT and related systems (which the gp starts trying to do) requires one to avoid terms like smart/intelligent/etc. The term intelligence is naturally not fully defined and primarily used by humans to rank each other. So our human intuition about intelligence comes from the ranking we do of people. IE, we have the intuition that "if they can do X, they can easily do Y" about a given person and such an intuition clearly shouldn't carry over to even a very complex machine.