DARPA Teaches AI How To Dogfight

The Air Force has pitted a human fighter pilot against artificial intelligence but it won't say who won. The Defense Advanced Research Projects Agency and the Air Force's Test Pilot School at Edwards Air Force Base poured AI into a modified F-16 called the X-62A VISTA for a head-to-chip battle over the high desert last fall that "broke one of the most significant barriers in combat aviation" by employing machine learning to put man against machine in an old-fashioned dogfight, said Secretary of the Air Force Frank Kendall. The X-62A was first taught defensive maneuvers and it built on the knowledge gained to move on to hair-on-fire 1,200-MPH exchanges with the crewed aircraft less than half a mile away. There were 21 test flights.

There were safety pilots with an on/off switch in the X-62A but they never had to take over from the algorithms. And as cool as robot dogfighting sounds, it was just a means to a variety of ends in integrating AI into Air Force flight operations. “It's very easy to look at the X-62A ACE program and see it as under autonomous control, it can dogfight, but that misses the point," said Bill Gray, the school’s chief test pilot. "Dogfighting was the problem to solve so we could start testing autonomous artificial intelligence systems in the air. Every lesson we're learning applies to every task you could give to an autonomous system.”