Rapid AI Innovation is a Pentagon Priority

Artificial intelligence leaders at the Pentagon believe AI will play an integral role in the Joint All-Domain Command-and-Control (JADC2) concept, and they want industry to innovate in AI as fast as possible.

“The pace of innovation taking pace [with AI, RPA] — it’s not slowing down, it’s accelerating,” said Douglas Bush, assistant secretary for logistics and technology at the Army, during the National Defense Industry Association’s JADC2 Symposium Monday. “How do we make sure JADC2 elements are continuously open to state-of-the-art technologies?”

Bush said industry partners and the military services should “default” to speedy, iterative innovation processes.

“When you go faster you create more risk, but right now the demand signal from Congress and leadership, the default is speed,” he said. “Getting comfortable with that should be normal.”

Bush’s comments come after the House Armed Services Cyber, Innovative Technologies and Information Systems Subcommittee questioned in June how quickly the Defense Department will be able to implement JADC2.

The biggest challenge is not acquisition authority, but coordinating and integrating the services’ different capabilities.

“JADC2 can sound like an impossible problem, but once you focus on discreet tasks it’s a lot more plausible,” Bush said.

Jaime Fitzgibbon, program manager for the Defense Innovation Unit’s AI/ML portfolio, said she wants to see DOD integrate AI into existing processes rather than try to overhaul whole IT systems.

“How do we see the future of AI, but also how do we make it fit today?” she asked during an AI panel at the symposium Monday. “DOD likes to think their problems are super special and only specific to them. But lots of companies have the same issues.”

In June, the DOD Responsible AI Working Council released a Responsible AI Strategy and Implementation Pathway, which aims to guide ethical, responsible, scalable and efficient AI innovation across the department. DIU’s Responsible AI Guidelines released in November 2021 provided a baseline for the Responsible AI Strategy and Implementation Pathway and relied on significant input from industry.

“[It was] something DIU developed with commercial customers,” Fitzgibbon said. “And we asked, is this going to squelch innovation, is this going to crush you with unnecessary reporting? And with that input we created a framework and now all these other groups (like CDAO) are using [it] to create a new baseline of what [AI] quality is, so people [can] innovate to that level.”

AI can also improve workforce training, she added. One of her AI prototypes involves senior intelligence analysts encoding their expertise with AI to help teach junior analysts what to look for so they “spin up” quicker and make fewer mistakes as they learn.

“It’s augmenting the speed, especially since we have a people problem, one person can do a lot more than what 10 people used to do because we just don’t have those 10 people and we’re drowning in information to find those signals,” Fitzgibbon said.