Can AI police itself? Experts say chatbots can detect each other's gaffes.AI chatbots sometimes give inaccurate or nonsensical answers, termed 'hallucinations.' Researchers propose using chatbots to detect errors made by other chatbots.