Dr. Will Roper, Assistant Secretary of the Air Force for Acquisition, Technology and Logistics, discusses ÒThe Future Air Force, Faster, Smarter: The Next GearÓ during the Air Force Association Air, Space and Cyber Conference in National Harbor, Md., Sept. 16, 2019. The ASC Conference is a professional development forum that offers the opportunity for Department of Defense personnel to participate in forums, speeches, seminars and workshops. (U.S. Air Force photo by Staff Sgt. Chad Trujillo)

Will Roper

WASHINGTON: The Air Force’s sixth-generation fighter jet that emerges from the Next-Generation Air Dominance (NGAD) program is almost certain to come with an AI copilot, says service acquisition czar Will Roper. The bigger question is what tasks the human pilot can, and most importantly should, cede to that artificial intelligence algorithm in what circumstances.

Whereas low-cost aircraft, such as the Skyborg drone, could very easily be flown by solo AI pilots in the near future, Roper told reporters during a Defense Writers Group briefing, the stakes are much higher for crewed aircraft.

“For many missions, we’re ready today. Skyborg, the attritable airplane, that’s going to be flown by ARTUµ or another sci-fi named equivalent,” Roper said. “In fact, I wouldn’t be surprised to see ARTUµ make it into a Skyborg attritable in the near future,” he said, referring to the AI “copilot” — pronounced R2 and named after the Star Wars droid R2D2 — flight tested in the U-2 spy plane on Tuesday.

But “all of those AI pilots that are flying solo will be vulnerable to counter-AI techniques” brought to bear by adversaries, Roper said. This is because current AI systems are based on gaming that involves know rule sets — such as chess or Go — but warfare is run by humans and so isn’t necessarily logical. “To gamify warfare at that level, in an algorithmic sense, will be exceptionally challenging, which means there will be so many opportunities to exploit AI’s need to extrapolate rules,” he explained.

This means that AI-piloted aircraft are going to fall prey to countermeasures, Roper said. But with low-cost drones, losing them at relatively high rates is one of the precepts behind their development in the first place — to send them into the fight where the risks of being shot out of the sky are extremely high. The entire point is to use attritable aircraft to protect pilots and high-end planes.

Nonetheless, Roper said he is “confident” NGAD will “have an AI assisted copilot, maybe even ARTUµ,  inside of it.” But in that case, the AI “should have more of a support function, and the human is there to help augment” when the AI is being targeted. “What I expect will happen in the pilot, copilot role — the Luke Skywalker, R2D2 role — is that pilots will gain an instinct, just like they have an instinct for stealth today, about when their AI crew pilot is performing well, or could perform well, and will turn over more of the reins to it. And [the pilot] will have a similar instinct of when it won’t be performing well, and will pull the reins back to the human.”

Northrop Grumman depiction of Next Generation Air Dominance fighter

The next step for brining AI copilots and pilots into the real world, Roper said, is to figure out how to certify that they are flight ready — in the same manner that the Air Force needs to judge when a pilot is sufficiently trained up to take the stick.

“We have to have a process to certify AI operators,” Roper said. “We have a process to train and equip people today, and determine that they are ready to go into operations, and now we need to do that for AI. That’s the task that we’re beginning now. I’m meeting with the ARTUµ team today to talk through what it will take to get ARTUµ ready to go into real world ops, and to do valuable missions supporting the pilot.”

In addition, he said, the Air Force will keep working on how to improve AI resiliency to countermeasures, with the service considering giving hackers a stab at ARTUµ during next-year’s DefCon 29 conference, currently slated for this coming August.

Other Cool Things

Roper’s roundtable — as usual — touched on a myriad different issues and new capabilities his shop is trying to push forward, including:

Scramjets

Roper said that research and development on scramjets needed to power hypersonic cruise missiles “is moving faster than I expected. … The acceleration period is compelling us to go ahead and start thinking through future programs of record. I would not be surprised at all to see a hypersonic cruise missile program enter into our future Air Force set of programs,” he said, citing a potential four year development timeframe. Roper said scramjet development will allow “cheaper and smaller” hypersonic cruise missiles to be fitted to fighters, such as the F-15EX that can carry “quite a lot of weapons.” The pace of the research is the reason why the DoD and Australia launched the bilateral Southern Cross Integrated Flight Research Experiment program (SCIFiRE) effort last month, designed to jointly develop air-breathing hypersonic weapons prototypes, Roper noted.

ABMS

The Air Force hopes by January to have a full-up acquisition strategy for “releaseONE,” a data relay pod for the KC-46 tanker that carries machine-to-machine software (called gatewayONE) linking it to F-35 and F-22 fighters (and translating the different machine languages used by the two fighters), Roper said. ReleaseONE “will have the data gateways that are acting as a cellphone tower connecting big cloud-based analytics to forward-edge fighters, and doing it very similarly to the way the Internet works,” Roper explained. The Rapid Capabilities Office, which has been tapped to shepherd ABMS through the formal acquisition process, is responsible for delivering releaseONE, which will be the first ABMS program of record to hit the street. As such, Roper said, he hopes it will go a long way to help mitigate criticism about the lack of specificity about the acquisition strategy and budgetary plans for the ABMS program so far levied by Congress and the watchdog Government Accountability Office in an April report.

Digitizing satellites

One of the big things Space Force acquisition “will have to crack” is “bringing digital acquisition, digital engineering into satellites. That hasn’t been done yet; to our knowledge hasn’t been done commercially.” Roper reiterated that there is already one satellite program attempting to do this, but it is classified. (He told the Defense Writers Group that he would try to get the name released by noon today, but …) “I think it can be done,” he stressed. “I’ve done a thorough deep dive on the program, and I see the same path to simplifying acquisition, simplifying touch labor and increasing interoperability through digitizing interfaces that I have seen in aircraft, and I’m really excited about that.”