Harvard psychologist warns against AI doomsday narratives, advocates for prudent safety engineering

On September 28, 2026, Harvard psychologist Steven Pinker published an open letter to psychiatrist and tech blogger Scott Alexander, arguing that fears of AI extinction are exaggerated. Pinker, who previously expressed optimism about AI, acknowledged underestimating the capabilities of large language models and the recklessness of AI companies in product releases. He criticized the 'paperclip maximizer' concept, which assumes an AI would sacrifice humanity to produce more paperclips, as a misleading portrayal of intelligence and control. Pinker emphasized risks like bioterrorism, cyberattacks, and uncontrolled AI, urging a focus on safety engineering with independent oversight and human control. Alexander, in a June blog post, estimated a 20% chance of AI causing human extinction and called for improved safety research and international agreements to slow AI development.