Pentagon launches new generative AI ‘cell’ with $100M for pilots, experiments

WASHINGTON — The Pentagon today announced it’s setting up a new AI Rapid Capabilities Cell meant to accelerate the military’s adoption of Large Language Models and other forms of generative artificial intelligence.

The new AIRCC (pronounced “Arc”) will implement the recommendations [PDF] of Task Force Lima, which was created in 2023 to get a handle on the unexpected explosion of “GenAI” and is now officially disbanded — although some as-yet unnamed members of the Task Force will move over to the accelerator cell, said outgoing Chief Digital & AI Officer (CDAO) Radha Plumb.

“Over the course of 12 months, Task Force Lima analyzed hundreds of AI workflows and tasks, [from] warfighting functions like command and control decision support [to] enterprise management functions like financial management and healthcare information management,” Plumb said.

“AI adoption by adversaries like China, Russia, Iran and North Korea is accelerating and poses significant national security risks,” she warned. “We are taking an all hands on deck approach to ensuring the US continues to lead the way and accelerate DoD adoption of these tools.”

AIRCC will be run by CDAO but work closely with the Defense Innovation Unit, the Pentagon’s outreach arm to Silicon Valley and other centers of private-sector innovation nationwide, Plumb told reporters today. It’ll have $100 million in funding to launch efforts on four fronts:

• $40 million in Small Business Innovation Research (SBIR) grants to small and non-traditional companies working on GenAI. “We’ve got about 400 initial applications,” Plumb told reporters this morning. “We’re still in the selection process, [but] those awards will be made in mid-January.”
• $35 million for four pilot projects to test “frontier” GenAI. Two of these are meant to aid frontline military operations, while the other two will help streamline back-office business processes. These pilots will get underway “immediately,” Plumb said this morning.
• $20 million to build digital “sandboxes” with the computing power required to test the expected wave of new GenAI tools. Plumb plans four of these testing-and-experimentation systems, one run by each of the four private-sector giants — Amazon, Google, Microsoft, and Oracle — on contract with CDAO’s Joint Warfighting Cloud Capability (JWCC). Two of the sandboxes should come online in January, Plumb said, and the other two by summer.
• $5 million to beef up ongoing GenAI testing under the CDAO’s quarterly Global Information Dominance Experiments (GIDE), which let the military’s Combatant Commands field-test and give feedback on new AI tools, with the most successful going into immediate use.

(Technically and confusingly, this funding comes from the budgets for federal fiscal years 2024 and 2025. However, Plumb said, all $100 million come from Research, Development, Test & Evaluation funds, aka RDTE, which can be used up to three years after the end of the year for which it was appropriated).

GenAI applications that prove successful in testing and get positive feedback from would-be users — whether military officers or civilian officials — will get further funding to scale up to widespread deployment.

When will that happen? “I can’t give you a specific timeline, because it really depends on the performance of the technology,” Plumb told reporters. That said, she noted proudly, when Deputy Secretary of Defense Kathleen Hicks challenged CDAO and its partners to come up with a global data-sharing system for the Combatant Commands, known as Combined Joint Global Command & Control (CJADC2), they got a “Minimum Viable Capability” operational within a year.

“That was about a nine to 12-month cycle to get all the way to the MVC for a pretty complex technology,” Plumb said. The new GenAI effort will take the same accelerated approach, she said.

Radha Iyengar Plumb, seen in a 2022 photo, has been tapped as the next Chief Data and Artificial Intelligence Officer. (DoD photo by Lisa Ferdinando)

Indeed, she argued, the Pentagon is already moving quickly on generative AI, at least by its own standards. Today’s creation of the new accelerator cell comes two years after OpenAI unveiled ChatGPT and 16 months after the Pentagon stood up its Task Force Lima to figure out how to harness the explosion in “GenAI.”

“Today the Department is officially sunsetting Task Force Lima and, to concretely implement their recommendations, we’re establishing the AI Rapid Capability Cell,” Plumb said. That’s ahead of schedule, she emphasized.

“Task Force Lima did an incredible job,” she said. “We had slated it for 18 months, and within 12 months, they had made so much progress and had such concrete recommendations that we moved to stand up the cell for actual execution.” (That implies the Task Force submitted its recommendations in August and it’s taken the last four months to get the AI accelerator cell organized and ready for today’s official launch).

Those recommendations go beyond the GenAI accelerator cell officially launched today. According to a four-page summary of Task Force Lima’s report [PDF]CDAO and other Pentagon agencies must work to safeguard the new technology. That includes researching potential security issues unique to GenAI, adapting government classification protocols and security standards, and potentially using GenAI itself to accelerate the notoriously cumbersome Authority To Operate (ATO) procedures. The Task Force also calls for DoD to “develop and resource a comprehensive acquisition strategy for GenAI” — which would be difficult to figure out at any time, but especially on the cusp of a presidential transition.

It’s worth noting TF Lima’s original 18-month deadline would have had it delivering its recommendations in early 2025 — after the Biden Administration and Biden appointees like Plumb had all left office. Once-and-future President Donald Trump and his advisors have publicly blasted the Pentagon for being too slow and bureaucratic to adopt new tech, while his inner circle includes tech enthusiasts and AI entrepreneurs who have called for a “Manhattan Project”-style push on AI.

So while specific Biden-era AI efforts may go under the axe, AI as a whole is expected to keep accelerating.

“I will transition at the end of this administration, but the rest of CDAO will still be here, and of course the mission continues,” Plumb said this morning. “We’ve seen broad, widespread support in this administration [and] from the incoming team. We see it … across the aisle, bipartisan, in Congress; we see it bicamerally in the House and Senate. We see it here in DC and in Silicon Valley and internationally.”