Army Works to Safeguard & Secure AI for Future War

By Kris Osborn, President, Center for Military Modernization

Here is a key portion of Warrior’s discussion with Mr. William Nelson, Deputy, Army Futures Command, about efforts to improve reliability of AI in warfighting operations. 

************** Related Essay on How AI is Already Changing War Tactics**********

(Washington D.C.) Comparing unknown data against a seemingly limitless database of known information and variables to arrive at new conclusions represents the heart of Machine Learning, as it can enable an AI system to accommodate, organize and integrate new, previously unknown information.

What happens if Russia or China builds a new secret tank or heavy armored vehicle that even the most advanced U.S. databases are not able to recognize? What if a weapon attacks U.S. forces that is simply not in any known threat library? Does the U.S. military have any recourse with which to make a fast, informed, combat-sensitive decision? What kind of munition should be used to counterattack? What kind of ammunition does the new threat fire? What is its range and scope? Are there AI-enabled computer programs now equipped to confront some of these challenges likely to present problems for U.S. commanders operating long-range sensors?

The answer is: maybe. If not now, not too far away, according to Army drone and robotics requirements writers now tracking threats and technical trends in autonomy and Artificial Intelligence (AI).

“Using AI, a small unit UAS (drone) identifies an enemy tank, asks other sensors to confirm and then reports back to a platoon leader, giving him various courses of action with which he can make a decision,” Col. Sam Edwards, Director of Robotics Requirements, Capability Development Integration Directorate, Ft. Benning, Ga., told Warrior in an interview several years ago

Also, what if, as Edwards also posited, the enemy tank is not recognized by an AI-capable database? This is where analytics comes in; a complex series of AI-informed algorithms would assess a range of additional variables to make a determination, to include analysis of the various configurations and components of known tanks, surrounding context, heat signature or even previous circumstances presenting similar dynamics.