In A Historic First, US Air Force Tests Dogfight Simulation Between AI And Human Pilot: Here's All About It

Details have emerged about what is believed to be the first in-flight aerial dogfight simulation between an artificial intelligence (AI) and a human pilot.

Conducted by the United States Air Force (USAF) last year, the simulation involved a modified F-16 (known as the X-62A) engaging in a battle of wits with a human pilot.

While a human "failsafe" pilot was present in the cockpit, the exercise was otherwise entirely AI-controlled, marking a significant milestone in AI development.

The test is seen as a potential game-changer for the future of unmanned aerial vehicles (UAVs), including initiatives like the US Air Force's Collaborative Combat Aircraft (CCA) programme.

The dogfight footage has been released as part of a video showcasing the Defense Advanced Research Projects Agency's (DARPA) Air Combat Evolution (ACE) programme, which has made remarkable progress in 2023.

Industry and academic partners, including Shield AI, played crucial roles in developing the AI "pilot" that won DARPA's AlphaDogfight Trials in 2020.

Unlike previous AI victories, this dogfight required adherence to flying norms, making it more realistic. The X-62A, also known as the Variable-stability In-flight Simulator Test Aircraft (VISTA), played a pivotal role in these tests, with its flight systems able to replicate those of various aircraft.

The ACE programme aims to make AI-enabled aircraft more "trustable" for human crews.

The X-62A completed 21 test flights from Edwards Air Force Base in California, with the AI "agents" requiring reprogramming almost daily. AFRL has highlighted the ability to rapidly train and retrain algorithms in entirely digital environments to support such flight testing.

While the US military asserts that human involvement will always be necessary to operate autonomous weapon systems, the extent of their participation in decision-making processes is expected to evolve over time.