AI coming to the US Military [Link]
[Summary]
This objective expose into the US military’s plans for further advancing AI in the battlefield tells a somewhat surprising story. It is not going to be an all-automated system as one might think - rather the military is mainly using this technology to help human operators be better warriors in an increasingly competitive landscape for military supremacy. The article breaks machine learning in military systems into two camps, defensive and offensive. The defensive strategy has less qualms with using automated systems for targeting attacking adversaries or locating and diffusing mines since the risks are lower and offer resource allocation efficiency. Offensive systems, on the other hand, must take into account impacts to civilian lives and are not expected to take autonomous lethal action, but rather to assist human operators in making decisions. For example, an AI system may prioritize the order in which certain targets in an area should be attacked, or combine sensor information from multiple units to provide a more holistic view of a hostile situation. Systems like these are not entirely new though, the Navy has been using auto-targeting systems like the Phalanx on ships for 30 years, offering the ability to target, shoot, and track performance in real time.
[Commentary]
The age of AI in the battlefield has arrived whether anyone likes it or not. The competitive landscape for global military dominance will continue to push ethical boundaries. In the case the US is attacked by fully autonomous offensive systems originating from a country of differing values, the stance of prioritizing fully-automated systems for defensive systems only may shift in order to keep that dominance. At least the US DoD is trying to find a clear path to this, as indicated by their search for an ethicist in this area
Another interesting challenge introduced by autonomous systems in military systems is akin to a self-driving car issue: As human operators rely on automated systems, will they lose their “edge” in operating these systems in a manual-mode? Airline pilots rely on automated flight systems, but must regularly undergo training to keep their skills up in times of need. The same argument could apply for controllers of defensive systems, it’s not unlikely that future adversarial attacks will attempt to render autonomous systems useless, requiring human intervention.
#GlobalAIandDataScience#GlobalDataScience