An AI-controlled virtual fighter jet just beat a human pilot in a dogfight.
Victory came in the finals of the U.S. Army’s AlphaDogfight competition , hosted by the Defense Advanced Research Project Agency (DARPA), in which big names in the military, including Lockheed Martin, competed to win the title.
The victorious company, Heron Systems, emerged victorious with its AI fighter jet by defeating a human fighter pilot – call sign “Banger” – who was seated inside a simulator wearing a virtual reality headset with five rounds to zero, as reported by Defense One .
A turning point
For some, victory means a turning point in what machine learning systems are capable of doing on the battlefield.
“The role of man is noble, but it symbolically points to a future where more and more machines will play more and more roles,” PW Singer, an analyst at the New America Foundation in Washington, told The Daily Beast .
Flatten against the ground
Teams had to start from scratch to teach their AIs to fly a fighter jet.
“You don’t have to teach a human not to crash into the ground… They have basic instincts that the algorithm doesn’t,” said Lee Ritholtz, director and architect in head of AI, from Lockheed Martin, to Defense One. “It means dying a lot. To hit the ground, many times. »
For Lockheed Martin, it took multiple servers to engage in round-the-clock trial and error to come up with the final AI, software that could run on a single graphics card.
The winning team’s AI had been put through more than 4 billion simulations, or about 12 years of flight time, as Ben Bell, principal machine learning engineer at Heron Systems, told Defense One.