Increasing integration of artificial intelligence (AI) into military systems has the potential to influence nuclear escalation even when that integration occurs outside nuclear weapon systems. Non-nuclear applications of military AI may compress decision-making timelines, potentially increasing miscalculation risks during a crisis.
Opaque recommendations from an AI-powered decision-support system can bias a decision maker towards acting, while autonomy in a system with counterforce potential may undermine strategic stability by threatening the integrity of second-strike capabilities. Such uses of AI raise the fundamental question of whether they introduce new risks,
exacerbate existing ones or fundamentally alter the nature of nuclear escalation. Contextual and socio-technical factors that might affect nuclear escalation pathways can help to answer this question. Understanding these dynamics is essential for using current risk-reduction measures or developing new strategies to address nuclear escalation risks posed by military AI.
No comments:
Post a Comment