China's Top Air Force Pilots Are Training AI in Fighter Combat

If you can't beat them, join them.

So seems to be the reasoning behind the Chinese People's Liberation Army (PLA) Air Force, which has deployed artificial intelligence (AI) in simulated aerial combat training against its flesh-and-blood aces, enabling human pilots to adapt to the rapid-reaction moves of AI fighter jets, while also training the AI on how to win every engagement, according to an initial report from the Global Times.

And, as was the case with recent battles against AI in the United States, the human pilots from China were often bested by their computer-generated counterparts.

China's Air Force is integrating AI into daily routines

While this exercise successfully trains human pilots to be more efficient in combat, AIs are slated to become an integral part of China's future air fleet, assigned to join and assist human pilots to optimize and streamline in-situ combat decisions, observers explained on Monday, according to the Global Times report. The training simulation went down early this summer, under the supervision of an aviation brigade linked to the PLA's Central Theater Command Air Force, in which a group leader in the brigade named Fang Guoyu — who has a high reputation from earlier combat exercises — was shot down during a mock aerial battle against the new AI-run aircraft in the simulator, according to a Sunday PLA Daily report cited in the Global Times report.

Notably, AI has shot Feng down before, and he's not the only fighter pilot ace in China taken out by a computer program. "The AI has shown adept flight control skills and errorless tactical decisions, making it a valuable opponent to hone our capabilities," said Commander of the Brigade Du Jianfeng, according to the report. He added that simulated and AI-assisted training has been increasingly integrated into daily training regimens of China's Air Force.

The advantages to simulated training exercises are numerous. Beyond enhanced efficiency, they lower flight risks, save big on cost, and, as the technology continues to unfold in militaries around the world, it will become a key tactical standard in fast response times to potential threats with unprecedented effectiveness. And this is why the U.S., too, has pitted its top guns in the air against advanced AI pilots.

Last August, an AI pilot shot won five consecutive rounds of simulated air combat against flesh-and-blood Air Force pilots in virtual F-16 Vipers, as part of a broader effort from the Defense Advanced Research Projects Agency (DARPA) to accelerate its Air Combat Evolution (ACE) program, in which the role of AI is central.

Nations of the world enhancing air forces with AI upgrades

The United States' aerial battle with AI "pilots" happened during a three-day exercise called AlphaDogfight, from August 18 to August 20. The winner was Heron Systems, which beat three other teams, including Physics AI, Aurora Flight Sciences, and Lockheed Martin. Heron Systems defeated an F-16 fighter pilot who goes by the callsign Banger. Since then, DARPA's aim was to "energize and expand a base of AI developers," according to a report from The Drive. "As we are kind of pushing the roles and responsibilities of pilots into this battle manager category, then what we're essentially doing in this program is enabling the autonomy to be even more capable to handle the aircraft maneuver and these rapid, high-tempo decisions in a dynamic environment," added Air Force Colonel Javorsek in an Air Force Magazine report.

As with any technological boom, the continuing advance of AI will see much of its early growth in military applications. When integrated into warplanes, it can assist pilots in combat by rapidly calculating the best options with which pilots may choose to engage a combat situation, optimizing real-time intelligence at speeds and levels of comprehensiveness unachievable by a human brain, or even several. Human pilots are not going obsolete anytime soon, but we may be nearing an age when human pilots without AI aid are simply no match for their pseudo-cybernetic counterparts.