Advertisement

Six months into JAIC 2.0, DOD still needs to move faster on AI

The Joint AI Center's transformation into 'JAIC 2.0' has brought added speed to AI adoption, but not enough, its director said.
Lt. Gen. Michael Groen, Joint Artificial Intelligence Center director, conducts a press briefing about the DOD's efforts to adopt and scale artificial intelligence capabilities, from the Pentagon, Washington D.C., Nov. 24, 2020. (DoD photo by U.S. Air Force Staff Sgt. Jack Sanders)

Despite revising its strategy six months ago to speed up artificial intelligence adoption across the Department of Defense, the Joint AI Center is still not moving fast enough, its director said.

The JAIC made the significant change from being an organization focused on product delivery to being more of an “enabling force” that supports AI offices across the military in November. That change brought more unity of effort across the department’s AI projects, but it’s still not moving as fast as it should be, Lt. Gen. Michael Groen said during a press briefing Friday.

“Is JAIC 2.0 enough? Are we moving fast enough?” Groen said. “I lay awake at night and say the answer is no.”

Groen’s message of raising the bar for the JAIC’s performance was echoed by Bob Work, former deputy secretary of defense, who appeared alongside Groen in his capacity as vice chair of the National Security Commission on AI. Work backed the JAIC’s new strategy of being an enabling force designed to help others, pointing to his commission’s 700-page report on how to achieve more of the speed Groen is after.

Advertisement

New development networks

Groen says he is satisfied with much of the work the JAIC has done, but he measures success in the achievements of other AI- and data-focused organizations through DOD. That means getting program offices more tools like the Joint Common Foundation, the JAIC’s AI coding environment, and the data-know how to achieve AI at scale.

Groen also detailed his hopes of stitching together a cloud-based “operating layer” of databases and development platforms to turn disparate efforts into unified ones.

“You need a network of operating platforms,” he said. “This will give us the capability.”

When civilians and service members want to work with data or code new solutions they still turn to a patchwork of different environments, like the Air Force’s Platform One. There are also disparate cloud systems with different authorities to operate, a problem officials hope the still-stymied Joint Enterprise Defense Infrastructure (JEDI) cloud program will fix.

Advertisement

Beyond purely technical work, Groen said members of the JAIC have fanned out across the department to share best practices. He said the center has less of a “teacher-student” relationship and more of a partnership and supporting role with other AI initiatives. He said the JAIC is also finding ways to use its new acquisition authority.

“We can do our own acquisition,” Groen said. “Now we can start a much broader array of support services.”

A new set of recommendations

The NSCAI report Work helped lead has more than 100 recommendations for the DOD to speed up the work the center is already engaged in.

“We thought about this as a blueprint,” he said. “You should not look at the recommendations individually, you have to do them all together to get the effect that the commission feel is important.”

Advertisement

The recommendations push the DOD to be “AI-ready” by 2025 through widespread education, leadership support for  AI, computer hardware development to run AI on, and greater investment in cutting-edge and basic AI research.

Groen said the JAIC is fully behind all of the recommendations but taking a “hard look” at some— which he didn’t enumerate — on how exactly they could be implemented.

Latest Podcasts