The tests we are using to assess the intelligence of AI are missing an essential aspect of human inquiry — the query itself.