Earlier this year, Leidos and Domino Data Lab conducted a digital wargame applying machine learning to a combat scenario. What made the game different is that the software they used was for Reinforcement Learning, an approach to machine learning that mimics the trial-and-error interaction with the environment that characterizes much of human learning. RL, as it is called, could transform the way in which artificial intelligence is applied to military operations, because it does not require supervising the learning process or anticipating in advance all of the specific circumstances that might occur in a dynamic combat environment. The machine teaches itself, and as it accumulates experience its responses become more subtle and precise. I have written a commentary for Forbes here.
Find Archived Articles: