U.S. Air Force Secretary Frank Kendall is putting his money where his mouth is.
Last week, Kendall got in the cockpit of a self-flying fighter plane during a historic flight at Edwards Air Force Base (KEDW) in California. The aircraft—called the X-62A Variable In-flight Simulator Test Aircraft, or VISTA for short—is a modified F-16 testbed and represents the Air Force’s first foray into aircraft flown entirely by machine learning AI models.
As Kendall and a safety pilot observed, the X-62A completed “a variety of tactical maneuvers utilizing live agents” during a series of test runs. Incredibly, the aircraft was able to simulate aerial dogfighting in real time, without Kendall or the safety pilot ever touching the controls. According to the Associated Press, VISTA flew at more than 550 mph and within 1,000 feet of its opponent—a crewed F-16—during the hourlong simulated battle.
“Before the flight, there was no shortage of questions from teammates and family about flying in this aircraft,” Kendall said. “For me, there was no apprehension, even as the X-62 began to maneuver aggressively against the hostile fighter aircraft.”
It wasn’t VISTA’s first rodeo. In September, the Air Force for the first time flew the uncrewed aircraft in a simulated dogfight versus a piloted F-16 at the Air Force Test Pilot School at Edwards. The department said autonomous demonstrations are continuing at the base through 2024. But Kendall’s decision to get into the cockpit himself represents a new vote of confidence from Air Force leadership.
“The potential for autonomous air-to-air combat has been imaginable for decades, but the reality has remained a distant dream up until now,” said Kendall. “In 2023, the X-62A broke one of the most significant barriers in combat aviation. This is a transformational moment, all made possible by breakthrough accomplishments of the ACE team.”
ACE stands for Air Combat Evolution, a Defense Advanced Research Projects Agency (DARPA) program that seeks to team human pilots with AI and machine-learning systems. The Air Force, an ACE participant, believes the technology could complement or supplement pilots even in complex and potentially dangerous scenarios—such as close-quarters dogfighting.
“AI is really taking the most capable technology you have, putting it together, and using it on problems that previously had to be solved through human decision-making,” said Kendall. “It’s automation of those decisions and it’s very specific.”
ACE developed VISTA in 2020, imbuing it with the unique ability to simulate another aircraft’s flying characteristics. The aircraft received an upgrade in 2022, turning it into a test vehicle for the Air Force’s AI experiments.
VISTA uses machine learning-based AI agents to test maneuvers and capabilities in real time. These contrast with the heuristic or rules-based AI systems seen on many commercial and military aircraft, which are designed to be predictable and repeatable. Machine learning AI systems, despite being less predictable, are more adept at analyzing complex scenarios on the fly.
“Think of a simulator laboratory that you would have at a research facility,” said Bill Gray, chief test pilot at the Test Pilot School, which leads program management for VISTA. “We have taken that entire simulator laboratory and crammed it into an F-16, and that is VISTA.”
Using machine learning, VISTA picks up on maneuvers in a simulator before applying them to the real world, repeating the process to train itself. DARPA called the aircraft’s first human-AI dogfight in September “a fundamental paradigm shift,” likening it to the inception of AI computers that can defeat human opponents in a game of chess.
Since that maiden voyage, VISTA has completed a few dozen similar demonstrations, advancing to the point that it can actually defeat human pilots in air combat. The technology is not quite ready for actual battle. But the Air Force-led Collaborative Combat Aircraft (CCA) and Next Generation Air Dominance programs are developing thousands of uncrewed aircraft for that purpose, the first of which may be operational by 2028.
The goal of these initiatives is to reduce costs and take humans out of situations where AI could perform equally as well. Some aircraft may even be commanded by crewed fighter jets. The self-flying systems could serve hundreds of different purposes, according to Kendall.
Even within ACE, dogfighting is viewed as only one use case. The idea is that if AI can successfully operate in one of the most dangerous settings in combat, human pilots could trust it to handle other, less dangerous maneuvers. Related U.S. military projects, such as the recently announced Replicator initiative, are exploring AI applications in other aircraft, like drones.
However, autonomous weapons, such as AI-controlled combat aircraft, have raised concerns from various nations, scientists, and humanitarian groups. Even the U.S. Army itself acknowledged the risks of the technology in a 2017 report published in the Army University Press.
“Autonomous weapons systems will find it very hard to determine who is a civilian and who is a combatant, which is difficult even for humans,” researchers wrote. “Allowing AI to make decisions about targeting will most likely result in civilian casualties and unacceptable collateral damage.”
The report further raised concerns about accountability for AI-determined strikes, pointing out that it would be difficult for observers to assign blame to a single human.
The Air Force has countered that AI-controlled aircraft will always have at least some level of human oversight. It also argues that developing the technology is necessary to keep pace with rival militaries designing similar systems, which could be devastating to U.S. airmen.
Notably, China too is developing AI-controlled fighter jets. In March 2023, Chinese military researchers reportedly conducted their own human-AI dogfight, but the human-controlled aircraft was piloted remotely from the ground.
Leading U.S. defense officials in recent years have sounded the alarm on China’s People’s Liberation Army’s growing capabilities, characterizing it as the U.S. military’s biggest “pacing challenge.” The country’s AI flight capabilities are thought to be behind those of the U.S. But fears persist that it may soon catch up.
“In the not too distant future, there will be two types of Air Forces—those who incorporate this technology into their aircraft and those who do not and fall victim to those who do,” said Kendall. “We are in a race—we must keep running, and I am confident we will do so.”
Like this story? We think you’ll also like the Future of FLYING newsletter sent every Thursday afternoon. Sign up now.
