Harvard psychologist calls for sober AI safety engineering over doomsday rhetoric
decoder ·
Harvard psychologist Steven Pinker has issued an open letter to Scott Alexander, arguing that fears regarding AI's existential risks are exaggerated. Pinker declined a public debate on the topic, characterizing such events as mere spectator sports rather than productive discourse. He contends that doomsday scenarios, such as the paperclip maximizer thought experiment, mistakenly conflate intelligence with a desire for domination and assume unrealistic single-goal fixation. While acknowledging he underestimated the rapid capabilities of large language models and the reckless release strategies of tech companies, Pinker remains skeptical of apocalyptic narratives. Instead of focusing on extinction-level threats, he identifies bioterrorism, cyberattacks, and unchecked AI agents as the primary dangers requiring attention. Pinker advocates for sober safety engineering grounded in independent oversight, liability frameworks, and human control mechanisms. In response, Alexander maintains a 20 percent probability of humanity being wiped out by AI and calls for improved safety research alongside international regulatory agreements.
decoder