For the first time, a fighter jet controlled by artificial intelligence (AI) has recently been pitted against a human-controlled jet in a practice dogfight scenario.
The boundary-breaking test flight took place in September 2023, but by DARPA’s Air Combat Evolution program last month.
The video shows the two aircraft – one human-controlled, one AI-controlled – flying with each other at speeds of 1,931 kilometers (1,200 miles) an hour, completing several “dynamic combat maneuvers” including nose-to-nose dogfighting positions.
The AI-guided plane is a one-of-a-kind model developed by Lockheed Martin Skunk Works in collaboration with Calspan Corporation for the US Air Force Test Pilot School.
Known as the X-62A, or VISTA (Variable In-flight Simulator Test Aircraft), the project was first developed as a simulator program to train pilots. The initial software involved AI technology, although it was limited to the virtual world – effectively like a super-realistic video game.
That changed in December 2022 when the AI agents were put in charge of a physical plane in the real world, allowing it to autonomously cruise along a flight path. Since then, it has carried out and , preparing it for its ultimate task: within-visual-range combat scenarios (aka “dogfighting”).
“When DARPA went to explore this problem, it looked as it does for the hardest kind of challenge it can find, and dogfighting is a perfect case for the application of machine learning. Dogfighting is extremely dangerous, so if machine learning can operate effectively in an environment as dangerous as air-to-air combat it has great potential to earn the trust of humans as we look to applications that are less dangerous, but equally complex,” Colonel James Valpiani, Air Force Test Pilot School commandant, said in the video.
In another flex of the program’s progress, US Secretary of the Air Force Frank Kendall on an AI-guided flight on May 2, 2024.
“It’s a security risk not to have it. At this point, we have to have it,” Kendall told the upon landing.
AI-assisted weapons have the potential to be more accurate than human-guided ones, potentially reducing collateral damage, civilian casualties, and friendly fire. However, this uncharted territory also raises a bunch of One of the is whether it is safe or appropriate to put life-and-death decisions in the hands of sensors and software – effectively killer robots – with little human oversight. Furthermore, can AI systems be held accountable if something were to go wrong?
Questions like this remain unanswered and there have been before the genie is let out of the bottle. However, some believe that the AI arms race is already upon us – with little sign of slowing down.