Shield AI buys company whose AI beat a fighter pilot in a dogfight

Shield bought Heron Systems, which won the much-hyped AlphaDogfight trials that pitted a fighter pilot against AI in a simulated dogfight.
Two L-39 Jets fly in tandem during a mock dogfight performance at the Air & Space Expo, Joint Base Charleston, S.C., April 28, 2018. (U.S. Air Force photo by Senior Airman Christian Sullivan)

The company that built an artificial intelligence system advanced enough to beat a fighter pilot in a simulated dogfight has been acquired by defense tech startup Shield AI.

Heron Systems is a small team of researchers based around the Beltway that develops multi-agent deep reinforcement learning AI for defense applications. Shield focuses on “AI for maneuver,” selling AI-enabled drones and robots aimed at helping to keep troops out of harm’s way.

Purchasing Heron Systems will expand Shield’s business portfolio into the cockpits of the military’s airplanes.

“Truly special AI companies are incredibly rare assets in the defense market,” Shield AI co-founder and CEO Ryan Tseng said in a statement. “Heron has developed the most advanced AI-pilot for fighter aircraft in the United States.”


Shield did not disclose how much it paid for Heron, but the company will continue to operate as a wholly owned subsidiary of Shield.

“Shield AI enables us the opportunity and scale to accelerate the integration of our AI-pilot on a next generation fighter and UAS,” Brett Darcey, Heron Systems general manager, said in a statement. “What stood out about Shield AI for us – is that they’re really the only ones who have an operational AI pilot that can operate on the edge without GPS or comms, and this has been proven on combat operations.”

Heron put its AI to the test in the Defense Advanced Research Project Agency’s  AlphaDogfight, which pitted a trained F-16 pilot against several companies’ AI systems in a simulated dogfight. The event was set up as a contest, with Heron coming out on top beating the human pilot 5-0. While some were impressed by the results, others saw the tests as “AI theater,” showing off interesting technical achievements that won’t necessarily translate to real world applications.

Latest Podcasts