Inside the Air Force Training Program that Will Pit Human Pilots Against AI


The work could lead to a new era of human and technology teaming.

Air Force fighter pilots will soon face new opponents in their training: artificial intelligence-based enemy pilots that can match humans based on their personal learning needs.

After steering the production of numerous AI-enabled pilot agents for years, Aptima, Inc. confirmed it landed a four-year contract with the Air Force Research Laboratory to build an “automated librarian” that will categorize those AI pilots and pair them with military trainees in scenarios that are right to advance their skillsets.

“The best case outcome is that AFRL determines that the products of this research are so promising that they create a library into which AI training technologies are shelved like books are shelved and they refine the sort of librarian that we're trying to build here so that it can sweep through that enormous library of AI, sweep through a library of scenarios—and for each individual student—pick out just the right pairing to advance them to expertise reliably and more quickly than we can do today,” Aptima’s Chief Scientist Jared Freeman told Nextgov during an interview on Tuesday.

Freeman joined the company in 1999, four years after its launch. Aptima’s project portfolio has grown increasingly diverse since then, he noted. Now, much of it concerns AI support for human teams, like forming and measuring them, and helping people and AI to manage those groups. The company also does a bit of work developing sensors and wearables to monitor and improve military members’ physical and cognitive fitness. 

Efforts under Freeman’s purview as chief scientist deal heavily with big data and AI. He’s currently a principal investigator overseeing Aptima’s role in a Defense Advanced Research Projects Agency-led initiative to help imbue AI with social intelligence to pave the way for next-level human/AI teaming down the line. He’ll also be very involved in this newly unveiled contract with AFRL, which is worth potentially more than $5.2 million.

“There's a wonderful line drawn, a thread from our work we did at our founding to this project that we are talking about now,” Freeman noted.

The company’s formation dates back to “a contract from the Navy to use computational methods to design more effective command and control teams, which we then tested with human beings,” he explained, adding that that work lasted about a dozen years. The latest pursuit with the Air Force lab also involves people and technology teaming up, as well.

“This time, the teams are human pilots and their adversaries—who are in a sort of a dance together—one trying to beat the other. And those adversaries are AI,” Freeman explained. “So here we're on, I think, the edge of really the next generation of AI research, which is human teams interacting with AI. AI as advisors. AI as trainers. AI as adversaries who exercise these skills.”

The roots of this now-formal contract trace in part back to collaboration Aptima had been engaging in with the lab’s 711th Human Performance Wing and its Senior Principal Research Psychologist and Readiness Product Line Lead for the Warfighter Readiness Research Division, Dr. Winston “Wink” Bennett. This “builds on an ambition of Dr. Bennett's and the wing to ensure that America has robust architectures for building artificial intelligence,” Freeman noted.

He and his team previously had been overseeing and assessing  the development of AI agents that were meant to become enemies in Air Force pilot training. 

Often in simulated combats or “dogfights” for educational purposes, trainees will fly against human experts, to learn how to operate and survive in combat. But those expert trainers are costly and highly in-demand, so the services are pivoting to explore the potential for AI to fill that gap. 

Aptima facilitated the making of numerous AI pilots across a variety of commercial developers. In the prior work, which received recognition and an award from NATO in 2020, the researchers found that the performance of each of the computer-based pilots the various entities’ deployed varied significantly between scenarios—as if each were an expert in some types of encounters and a novice in others.

“This tells us that [the Defense Department] needs to be quite careful about investing all of its funds in a single AI solution,” Freeman said.

Under this new contract, he and his team will direct the development of AI agents by those other companies and form a scalable cloud-based testbed that allows many agents to generate flight data in parallel that can be used to assess the technology. Further, Aptima also aims to capitalize on AI-based pilots’ varying performances, by making a “smart” librarian—underpinned by AI—that will organize and match the computerized pilots to the human ones, depending on their experience levels. A “unique feature” of the system is that it will use training data in real-time to personalize learning based on each individual.

Eventually, the pilots of all types will interact in simulators.

“We don't have human pilot trainees in this loop. We're running AI through simulations against fairly scripted but smart—those other AIs are like the students, right?” Freeman noted. “It’s a very long term program. We have several more years of work to do and we hope to be working with human beings on the trainee end soon.”