Breakthrough in Machine Learning as AI Pilots F-16 in Dogfight Simulation
The US Air Force Test Pilot School and DARPA have reportedly achieved a breakthrough in machine learning by demonstrating that AI software can pilot a modified F-16 fighter jet in a dogfight against human pilots. The achievement was made possible through the integration of machine learning agents into an X-62A VISTA testbed aircraft. This accomplishment was recognized as one of the finalists for the National Aeronautic Association’s 2023 Robert J. Collier Trophy, an annual award for exceptional feats in aeronautics or astronautics in America.
The DARPA ACE program has been testing AI agent software for piloting simulated planes since 2020, with the AI winning in AlphaDogfight trials. The X-62A’s successful flight in a mock dogfight against a human-piloted F-16 in September 2023 marks a significant milestone in autonomous aviation. Despite not winning the Collier Trophy, the achievement is seen as a validation of autonomous aviation for both military and commercial applications.