This is How ‘Cultural Artifacts’ Impede the Defense Department's Ability to Go Big on AI

The Defense Department has many predicaments to untangle when it comes to adopting emerging, and existing, technologies from workforce to acquisition. But when it comes to artificial intelligence, cultural and budget issues are the largest obstacles to widespread implementation, officials say.

“Implementation is the key to successful transformation” but is also “extraordinarily challenging” because it cuts across nearly everything the Defense Department does such as educating the workforce, acquisition,” Marine Corps Lt. Gen. Michael Groen, the outgoing director of the Defense Department’s Joint Artificial Intelligence Center, said during a keynote at the Atlantic Council event on May 26.

Groen continued saying that “cultural artifacts” from policy to budget execution accounted for 85% of the Defense Department’s broad implementation of artificial intelligence. 

“Those cultural artifacts, whether that’s policy, whether that’s dealing with service budgets and service budget execution, getting authorization to operate on networks, ATO, networks and network policy, network security. The role of the department in all of those things, the relationship of the department with the services becomes a key conversation,” Groen said. 

Moving past those entrenched cultural dynamics means being willing to share capabilities, he said.

Margarita Konaev, a security and defense fellow at the Atlantic Council and Georgetown University’s Center for Security and Emerging Technology, described the Defense Department’s implementation of AI was like “working on a common goal but, unfortunately, on parallel tracks” and that improving alignment means investing in safe, reliable, secure, trusted AI that is battle-tested. 

“You will hear a lot about the urgency of implementing AI, magnified by the strategic competition with China, the changing nature of warfare, and the massive promise that AI has. That is all absolutely true and correct. But rushing towards the operational deployment of untested technologies that have not been effectively verified or assured…will end up being counterproductive,” Konaev said. 

Jaime Fitzgibbon, program manager for the Defense Innovation Unit’s AI and machine learning portfolio, said pushing AI to the battlefield to help commanders make more informed decisions also means confronting the Defense Department’s worst enemy: the budget process. 

“The capability is there...even DIU has pushed out prototypes that could help get that information that’s critical for decision-making up to the senior leaders, but…our budgetary cycles, you have to plan it five years ago, when the technology or even the data feed wasn’t around two years ago,” she said. 

Fitzgibbon said that conundrum is DOD’s “own worst enemy” and “everyone has a role in the problem and the solution,” which could include making funds available further down the command chain, perhaps even to the unit level, as some military needs are not communicated up to senior leaders. 

“And putting in an unfunded request that doesn’t get fulfilled for two years still doesn’t answer that capability challenge,” Fitzgibbon said.