Harvard Psychologist Urges Pragmatic AI Safety
Harvard psychologist Steven Pinker has called for a shift toward pragmatic safety engineering in artificial intelligence, arguing that fears of human extinction are overblown. According to The Decoder, Pinker made his case in an open letter published on Quillette in response to a debate challenge from psychiatrist and blogger Scott Alexander.
Pinker stated that apocalyptic scenarios, such as an artificial intelligence prioritizing a single objective to the detriment of humanity, mistakenly equate raw intelligence with a drive to dominate. While Pinker acknowledged that he previously underestimated the rapid capability gains of large language models and criticized the reckless deployment schedules of tech companies, he maintains that catastrophic extinction narratives are unhelpful.
Instead, Pinker emphasized that the industry should focus on tangible and immediate hazards, including bioterrorism, advanced cyberattacks, and autonomous software agents. He argued that doomsday rhetoric fails to address these pressing challenges, advocating instead for sober safety engineering supported by independent oversight, legal liability, and strict human control mechanisms.
Alexander previously estimated a notable probability of human extinction resulting from artificial intelligence, advocating for enhanced safety research and international treaties to moderate development speeds. Pinker declined the public debate proposal, characterizing such events as unconstructive spectator sports in favor of addressing concrete policy and technical safeguards.
Based on reporting by the-decoder.com.
