works
Christian Ruhl Autonomous weapon systems and military artificial intelligence (AI) applications report online The likely future proliferation of autonomous weapon systems and military artificial intelligence applications presents under-studied strategic risks. These technologies can act as threat multipliers, exacerbating pathways to global catastrophic risks such as great power conflict and nuclear war. Specific dangers arise from accelerated decision-making, automation bias, increased system complexity leading to accidents and escalation, and the potential for destabilizing military AI competition. While public discourse and existing efforts often concentrate on humanitarian concerns and the pursuit of a formal, multilateral ban, this focus leaves the more fundamental strategic threats relatively neglected. A more tractable and impactful approach involves prioritizing funding for research into these strategic risks and developing confidence-building measures focused on the key state actors most likely to develop and deploy such systems. – AI-generated abstract.

Abstract

The likely future proliferation of autonomous weapon systems and military artificial intelligence applications presents under-studied strategic risks. These technologies can act as threat multipliers, exacerbating pathways to global catastrophic risks such as great power conflict and nuclear war. Specific dangers arise from accelerated decision-making, automation bias, increased system complexity leading to accidents and escalation, and the potential for destabilizing military AI competition. While public discourse and existing efforts often concentrate on humanitarian concerns and the pursuit of a formal, multilateral ban, this focus leaves the more fundamental strategic threats relatively neglected. A more tractable and impactful approach involves prioritizing funding for research into these strategic risks and developing confidence-building measures focused on the key state actors most likely to develop and deploy such systems. – AI-generated abstract.

PDF

First page of PDF