> ChatGPT does not know more than you. The fallacy is always that you compare AI to a human without literature references and a database.

If the human needs a literature references and a database to answer a question, can they be said to "know" the answer?

ChatGPT doesn't have an endgame database for chess. Despite having "read" all the literature about chess, it will hallucinate the board state if you try to play chess with it directly. But it "knows" how to write a chess engine that would beat me… and more than that, one which would beat a competent player.

It is a very weird and spiky form of intelligence, but it's also definitely not just a database.