Artificial Intelligence and Warfare

Abstract

New developments in Artificial Intelligence may overshadow the traditional foundation of nuclear deterrence by reducing the survivability of missile launch platforms. AI has the capability to gather and quickly synthesize information from thousands of sources to produce highly accurate estimates of locations for missile-submarines, or land-based mobile launchers. Thus, mobile nuclear assets for deterrence may become easier targets, which reduces the threat of successful retaliation after a deliberate first-strike. Other traditional forms of defense may also be rendered obsolete. While traditional defenses are designed to protect against larger, stronger weapons, AI has the new capability to operate autonomous weapons at the miniaturized level. Miniature devices can fly toward a target undetected and then swarm together at the last minute to deliver a powerful, coordinated blow through autonomous synchronization. Many critics encourage policy makers to prevent AI weapons from ever killing humans as part of a military mission without first receiving explicit permission from a human operator. However, other experts warn that AI may someday evolve beyond submission to their human controllers, and may independently and dynamically set up their own new missions and goals which may conflict with the original designers. AI is currently an area of intense research for China and Russia. Moral objections to AI by contractor-technicians in the US may slow new development by the DOD. However, US commercial development of AI is moving forward rapidly, and can be seen in new products such as self-driving cars. AI may be considered a dual-use technology, similar to strong encryption, which makes legal controls to limit uses and prevent export somewhat difficult.