| Advertisement |
Aug. 21 (UPI) -- An artificial intelligence algorithm soundly defeated an experienced U.S. F-16 fighter pilot in a simulation of an aerial dogfight, the Pentagon said. The event, called Alpha Dogfight Trials, was part of the Air Combat Evaluation program of the Pentagon's Defense Advanced Research Projects Agency.
Advertisement
Designed to accelerate the development of artificial intelligence in combat situations, DARPA invited teams from eight defense contractors to fly F-16 fighter planes in competitive, simulated combat.
The weeklong competition at the Applied Physics Lab of Johns Hopkins University in Baltimore led to the winner facing off against a live, experienced pilot.
Heron Systems, a Maryland-based AI developer, was the last company standing, and on Thursday its algorithm quickly defeated a U.S. Air Force pilot identified only as "Banger" in five rounds of combat, all in less than two minutes.
The dogfights involved simulated speeds of up to 500 mph, 9-G maneuvers and laser beams instead of machine-gun fire.
"There are a lot caveats and disclaimers to add in here," Col. Dan Javorsek, DARPA program manager, said after the event. Heron's AI system had "perfect information," a situational condition fighter pilots never face, and Heron could aim its guns more accurately than the human pilot.
Between rounds, Banger admitted he faced a tough opponent.
"Standard things we do as fighter pilots are not working," he said, noting that at one point he attempted to duck out of the way as the AI plane approached. The enemy plane was able to maneuver around his plane more easily than a human pilot could.
The DARPA program seeks develop AI technology to enhance, and not replace, human combat capabilities, Javorsek said.
It could react more quickly than a human to enemy aircraft actions, for example, and could be used in drone technology. Although drones fly without pilots, they generally still require an on-the-ground operator.
