Harvard Psychologist Pinker Refuses AI Doomsday Debate, Calls Extinction Risk Overstated

On September 28, Harvard psychologist Steven Pinker published an open letter in Quillette addressed to psychiatrist and tech blogger Scott Alexander, stating that the risk of AI wiping out humanity is overstated and declining to take part in a public debate on existential AI risk. He described such debates as “spectator sports” and argued doomsday scenarios like the “paperclip maximizer” confuse intelligence with dominance.

In the letter, Pinker said he has been relatively optimistic about AI, but acknowledged he underestimated how capable large language models could eventually become and did not expect AI companies to release products so recklessly. He noted the irony that many people driving the release race come from the same community that has been warning about these risks.

Pinker said the risks he actually takes seriously are bioterrorism, cyberattacks, and unconstrained AI agents. He wrote that dwelling on doomsday scenarios does not help address these dangers, and that the effective response is careful safety engineering based on independent oversight, accountability mechanisms, and human control.

Alexander had previously argued on his blog that there is a 20 percent probability AI will cause human extinction, and proposed improving AI safety research while slowing AI development through international agreements.