"I don’t think we're attracting enough people. Whether they're the right people or not, I think that's a separate question. I'm not sure that we'll be able to answer that until we're working with a broader set of the industry base that's working AI,” Will Roper, the Air Force’s undersecretary for acquisition, technology and logstics, told reporters at Farnborough Airshow.
“I contend that the companies driving AI are a different breed of company than those who drive evolution in hardware, especially companies that drove hardware that have gotten us to today’s military. The paces are faster, turnovers are quicker. Software is done in month cycles not year cycles.”
Over the past year, the Air Force has charted some successes and some failures in its attempts to integrate tech like AI and big data analytics with legacy hardware systems like fighter jets or air operations center.
It has established the Kessel Run Experimentation Lab, a group of industry and airmen in Boston that are iterating new capabilities for air operations centers. Instead of rolling out a large software package, the coders focus on app-like updates that can more rapidly insert new functionality into the AOC.
But it’s also suffered setbacks — most notably, Google’s stated intention to withdraw from future Defense Department projects after some employees objected to the company’s work on Project Maven, a program would allow the Pentagon to use AI to review footage from drones. Some have worried that could have a chilling effect on future efforts.
Roper said that a big focus of his job is changing how the Air Force approaches software. In the past, software was a product that could be bought in cycles, just like a physical product like a missile or aircraft. Now, it’s a service that must be reworked constantly, he said.
“You get a good set of coders in, they can push out so much code per month. You put them with the user that's going to use the code and together they're able to collaborate to make sure that the developer is creating something that the operator is using,” he said.
“That’s working very well for us in Boston, and we’re looking to expand that. That’s the basic mechanism to move towards AI. We’re going to need people that are working with us that are software people that are working, tweaking algorithms with the users that use them, and it’s probably a different company than have worked with us over the past 10 years.”
The Air Force has to get those companies under contract faster, in weeks instead of months, Roper said. It’s looking for opportunities to use contract vehicles specifically delegated for small businesses and to use AFWERX — its outreach arm to nontraditional contractors who are creating promising commercial technologies — to introduce startups to the service.
But Roper acknowledged there was no easy answer to the problem.
One possible way to inject AI into Air Force programs — although a mundane one — is to use it for predictive maintenance technologies that use sensors to forecast when a component will break, said Air Force Under Secretary Matt Donovan.
“It’s very exciting for us and I think it holds a lot of potential to reduce our sustainment costs,” he said, noting that sustainment makes up a whopping 70 percent of the life-cycle cost of any given product.
Roper agreed that sustainment was a great area to begin employing AI, and that experience could help the Air Force begin to figure out how to use the technology for classified applications.