Free beta access. 5 free AI tokens each week. Optional top-ups available. See what is included
Opening Word document…

AI early warning systems undermine nuclear deterrence and stability – automation bias and false alarms ruin reliability

Horowitz and Scharre 2021 - Director of the Emerging Capabilities Policy Office in the Office of the Under Secretary of Defense for Policy and Vice President and Director of Studies at CNAS [Michael, Paul, January 12 2021, “AI and International Stability: Risks and Confidence- Building Measures”, https://www.cnas.org/publications/reports/ai-and-international-stability-risks-and-confidence-building-measures, Acc 6/18/22. M.A.]

The introduction of AI into military operations could also pose risks in certain circumstances due to the nature of the military mission, even if the AI system performs correctly and consistent with human intentions. Some existing research already focuses on the intersection of AI with specific military mission areas, most notably nuclear stability. Nuclear stability is an obvious area of concern given the potential consequences of an intentional or unintentional nuclear detonation. Lethal autonomous weapon systems (LAWS), a particular use of AI in which lethal decision-making is delegated from humans to machines, also represents a focus area of existing research. Other areas may deserve special attention from scholars concerned about AI risks. The intersections of AI with cybersecurity and biosecurity are areas worthy of exploration where there has been relatively less work at present. Potentially risky applications of AI extend beyond the battlefield to the use of AI to aid in decision-making in areas such as early warning and forecasting adversary behavior. For example, AI tools to monitor, track, and analyze vast amounts of data on adversary behavior for early indications and warning of potential aggression have clear value. However, algorithms also have known limitations and potentially problematic characteristics, such as a lack of transparency or explainability, brittleness in the face of distributional shifts in data, and automation bias. AI systems frequently perform poorly under conditions of novelty, suggesting a continued role for human judgment. The human tendency toward automation bias, coupled with the history of false alarms generated by non-AI early warning and forecasting systems, suggests policymakers should approach the adoption of AI in early warning and forecasting with caution, despite the potential value of using AI in intelligent decision aids. Education and training to ensure the responsible use of AI in early warning and forecasting scenarios will be critical.

The Impact - this increases the risk of accidental nuclear war because the speed of AI causes countries to adopt Lunch On Warning postures.

Shah, 2019 - Research Assistant at the Center for International Strategic Studies [Syed Sadam CISS Insight Vol.VII, No.2 “The Perils of AI for Nuclear Deterrence” https://journal.ciss.org.pk/index.php/ciss-insight/article/download/10/9 Acc 5/25/22 TA]

The impact of AI on nuclear doctrines The use of autonomous delivery platforms and autonomous detection systems will increase the time urgency and pose several risks to the second strike capability of the states with less AI capabilities. This will compel states to rely on comparatively more ready nuclear arsenal where weapons are placed close to the delivery systems. The less AI capable states may also delegate the authority to use these weapons to lower commanders owing to the time urgent decision making needs. These states may decide to increase the number of their nuclear warheads to ensure survival of a sufficient number of their weapons in an attack situation. Time urgent nuclear doctrines will also be more prone to accidental and unauthorized nuclear launch. However, on the other hand, states with superior AI based detection systems may resort to first strike doctrines, though for different reasons. They might become confident to destroy their enemy in a comprehensive first strike and intercept the remaining retaliatory enemy strike by the timely identification of their AI based detection systems and leaving their missile defences to do the remaining job.

Sign in to recut, update, or save this card.