New research finds AI can pass as human until it remembers ‘too well’, with simple memory tests exposing chatbots by their lack of normal human errors. Researchers from Princeton have developed a ...