Pentagon Built AI Program to Navigate Its Bloated Budget

As tech luminaries like Elon Musk issue solemn warnings about artificial intelligence’s threat of “civilizational destruction,” the U.S. military is using it for a decidedly more mundane purpose: understanding its sprawling $816.7 billion budget and figuring out its own policies.

Thanks to its bloat and political wrangling, the annual Department of Defense budget legislation includes hundreds of revisions and limitations telling the Pentagon what it can and cannot do. To make sense of all those provisions, the Pentagon created an AI program, codenamed GAMECHANGER.

“In my comptroller role, I am, of course, the most excited about applying GAMECHANGER to gain better visibility and understanding across our various budget exhibits,” said Gregory Little, the deputy comptroller of the Pentagon, shortly after the program’s creation last year.

“The fact that they have to go to such extraordinary measures to understand what their own policies are is an indictment of how they operate.”

“The fact that they have to go to such extraordinary measures to understand what their own policies are is an indictment of how they operate,” said William Hartung, a senior research fellow at the Quincy Institute for Responsible Statecraft and expert on the defense budget. “It’s kind of similar to the problem with the budget as a whole: They don’t make tough decisions, they just layer on more policies, more weapons systems, more spending. Between the Pentagon and Congress, they’re not really getting rid of old stuff, they’re just adding more.”

House Republicans reportedly aim to pass their defense budget later this week. They had planned to vote on an $826 billion proposal last week before the far-right Freedom Caucus blocked the proposal, demanding cuts to non-defense spending.

“The fact that the Pentagon developed an AI program to navigate its own policies should be a stark wake-up call for lawmakers who throw more money at the department than it even asks for nearly every year,” said Julia Gledhill, an analyst at the Project on Government Oversight’s Center for Defense Information. “It’s unsurprising, though: The DOD couldn’t adequately account for 61 percent of its $3.5 trillion in assets in the most recent audit, and those are physical!”

The Pentagon did not respond to a request for comment.

Military brass use GAMECHANGER to help them navigate what the Defense Department itself points to as an absurd amount of “tedious” policies. The program contains over 15,000 policy documents governing how the Pentagon operates, according to its GitHub entry.

“Did you know that if you read all the Department of Defense’s policies, it would be the equivalent of reading through ‘War and Peace’ more than 100 times?” a press release about GAMECHANGER from the Defense Intelligence Agency, the military’s spy wing, says. “For most people, policy is a tedious and [elusive] concept, making the idea of understanding and synthesizing tens of thousands of policy requirements a daunting task. But in the midst of the chaos that is the policy world, one DIA officer and a team at the Office of the Undersecretary of Defense for Intelligence & Security saw an opportunity.”

The press release went on to decry the Pentagon’s “mountain of policies and requirements.”

As unusual as it is for the military to publicly air its contempt for its own sprawling bureaucracy, members of Congress have been similarly harsh. In its portrayal of U.S. military policy — which it also had a hand in creating — the Senate Armed Services Committee called rules governing the department “byzantine” and “labyrinthine.”

“The committee notes that the Joint Artificial Intelligence Center developed an artificial intelligence-enabled tool, GAMECHANGER, to make sense of the byzantine and labyrinthine ecosystem of Department guidance,” the committee said in a report for National Defense Authorization Act — the law that appropriates cash for the Pentagon budget — for fiscal year 2023. (Amid the critique of the Pentagon’s bloated bureaucracy, the NDAA would later become law, authorizing $802.4 billion in funding for the defense budget.)

Though announced in February of last year, GAMECHANGER has received scant media attention. The military’s Joint Artificial Intelligence Center, a subdivision of the U.S. Air Force created in 2018, developed the program. Upon its completion, the Joint Artificial Intelligence Center transferred ownership of it to the office of the Defense Department comptroller, which handles budgetary and fiscal matters for the Pentagon.

Shortly after its release, GAMECHANGER was already used by over 6,000 Defense Department users conducting over 100,000 queries, according to the Defense Intelligence Agency.

Described as a natural language processing application — a broad term in computer science generally referring the use of machine learning to allow computers to interpret human speech and writing — GAMECHANGER is just one of a vast suite of AI programs bankrolled by the Pentagon in recent months.

The Pentagon is currently funding 686 such AI projects, according to the National Academy of Sciences, a nonprofit that frequently conducts research into the government. The figure does not include the Department of Defense’s classified efforts.

Before it was formally released, GAMECHANGER was granted an award by the Office of Personnel Management, the federal government’s human resources agency for civil servants.

“GAMECHANGER is an ironic name: They’re patting themselves on the back for, in the best case, figuring out what they’ve said in the past, which is pretty modest,” said Hartung, the Quincy Institute defense budget expert. “It’s more a problem of how they make policy and not a problem of how to surf through it.”