Why the Pentagon can’t go it alone on AI

Artificial intelligence — with its requirements for vast sets of data and the potential to disrupt government and military operations — can’t be created in a diplomatic vacuum.

This message was recently reiterated by AI policy experts during a panel about data sharing. It’s also a core tenet of the U.S. national defense strategy. And military strategists on both sides of the Atlantic are championing the idea with the hopes of seeing more collaboration and a diplomacy refresh for the digital age.

For those working on AI policy in the Department of Defense‘s Joint AI Center, it’s a call they are working to heed, but one that remains in the early stages.

“We know when the U.S. goes into combat operations…rarely do we do [it] alone,” Mark Beall, head of strategy and policy at the JAIC, told FedScoop in an interview. The DOD must work to ensure that AI’s applications in the military are built from the ground up with allies in mind, he said.

“At the highest level, JAIC is very much interested in how it is we upgrade our alliances for the digital era,” Beall added.

For experts like Cameron Kerry, former general counsel to the Department of Commerce and visiting fellow at the Brookings Institution, state-led solutions are “not effective” in the development of AI systems at the technical and policy levels. AI has an insatiable thirst for data, powerful computers and engineers who can build world-class algorithms. It’s an ingredients list that Kerry and others on both sides of the Atlantic say needs international cooperation to be truly effective.

“NATO is taking these issues very seriously,” Jan Havránek, a policy adviser to the secretary general of NATO, said during the virtual event sponsored by the Center for Data Innovation.

For the JAIC, work on international cooperation is in the “initial stages.” Conversations with NATO allies are primarily focused on “ethics, principles and regulations,” Stephanie Culberson, director of international AI policy at JAIC, told FedScoop. Technical conversations and data sharing comes later, JAIC officials hope.

The center is working on three pillars of diplomacy for AI: shaping norms around democratic values, ensuring data interoperability and working to create pipelines to enable the secure transfer of technology.

“Given the importance of the NATO alliance, we desire a future that enables digital-age cooperation and interoperability between the U.S. and NATO while respecting and honoring the strong commitment to safe, responsible, and ethical uses of technology,” Lt. Gen. Jack Shanahan, the JAIC’s soon-to-be departing leader, said in January following a meeting with NATO allies.

Still, DOD’s AI ecosystem remains relatively immature — far short of the science-fiction vision some have of the military’s development of the technology. The technologies for now mostly focus on the “less sexy” back-office tasks. But as they mature, the need for technical working groups will be stronger, said Culberson, who also attended NATO meetings in January.

AI partnerships to deter Russia, China

The JAIC is also forming bilateral relationships with other militaries, like Singapore’s. In June the center held a “multi-day technology exchange” with Singapore’s Defence Science and Technology Agency to work on disaster relief projects.

The DOD is working to hire “exchange officers” to facilitate the procurement of foreign AI technology and vice-versa. Some limited export controls on national security technology powered by AI have already been put in place and are seen as a means to deter adversary advancement in AI. Expanding data sharing and technology transfer are also measures taken to deter China and Russia’s use of AI.

“I don’t think you can have that conversation about data spaces without taking into account China,” Kerry said. “China has clearly created a Chinese data space.”

And with that, the country the U.S. describes as a strategic competitor holds some advantage in AI’s implementation and potential weaponization. The best response is sharing data, several experts said.

“Data sharing is at the core of developing artificial intelligence,” said Jose-Marie Griffiths, president of Dakota State University and former government official working on science and technology development.