Why AI must learn to admit ignorance and say ‘I don’t know’

How will we know if an artificial intelligence has become conscious? If your response is “I don’t know”, congratulations – you have just passed part of a test designed to measure AI intelligence by posing questions to which there is no known answer.

AIs like ChatGPT that are powered by large language models can create compellingly plausible responses to any question, but there is no guarantee that their answers have any basis in reality. This issue can be exacerbated when dealing with unsolved problems in science and philosophy. “So many of these models…