IISS | Nathan Orlando
Incorporating artificial intelligence (AI) into nuclear protocols risks disastrously suppressing human reason, which historically prevented nuclear war despite rapid technological innovations. While no government publicly endorses AI making launch decisions, its complete exclusion is unrealistic given the number of nuclear actors, innovation incentives, and verification difficulties. AI offers operational advantages like speed, potentially reducing decision windows to three minutes with advanced hypersonic systems, and could perfect a "Dead Hand" system, especially tempting for nations like China with smaller, more vulnerable arsenals. However, historical incidents, including the 1960 NORAD false alarm, the 1962 Cuban Missile Crisis near-misses (e.g., Vasili Arkhipov's veto on B-59, the Duluth bear incident, U-2 error, Okinawa launch order), 1979-1980 NORAD computer errors, Stanislav Petrov's 1983 judgment, and the 1995 Norwegian rocket incident, demonstrate human fallibility but also the critical role of prudential judgment in averting nuclear catastrophe. These events underscore the danger of fully automating nuclear responses.
No comments:
Post a Comment