WASHINGTON: The Pentagon’s Joint Artificial Intelligence Center has hit a key milestone in the rollout of its Joint Common Foundation, a cloud-based platform that enables users to access Defense Department data and develop AI solutions in a secure environment.

JAIC Director Lt. Gen. Michael Groen said the JCF has achieved Initial Operating Capability (IOC). His remarks came during the opening keynote at the National Defense Industrial Association AI National Security Conference.

“In IOC, the JCF can host data and algorithms,” Groen told me in a follow-up interview, and he noted that JCF is already being used by customers from across the services and DoD. “The JCF also hosts some data science and data engineering tools that provide AI developers with some introductory capabilities to test and provide feedback to our infrastructure team.”

The JCF is a set of cloud-based tools that enable JAIC customers to develop AI, share data to train machine learning algorithms, and to work in a common environment guided by secure development operations (DevSecOps). DevSecOps merges security-focused software development and IT operations to speed development and enable continuous improvement. The JCF’s cloud environment will serve any Defense Department customer, from Pentagon teams to forward-deployed military.

“In the coming months,” Groen told me, “the JAIC’s infrastructure team will add DevSecOps capabilities that will provide an agile and secure AI development environment for DoD customers and users. Moving forward, the goal is to add new tools and resources to the platform, while expanding access to DoD data.”

JAIC’s plan to rapidly grow capabilities, which it will do via a series of monthly updates, reflects the US’s current AI competition with adversaries, including China which has said it plans to dominate AI by 2030. “To a Marine, this is dangerous close,” Groen said in the NDIA keynote.

Embracing a commercial sector ethos, JAIC is already focusing on customer needs. “JAIC recently completed an extensive survey of DoD users and their requirements in February 2021 to capture the tools, processes, and potential accelerators for AI development,” Groen told me. “Moving forward, the JAIC will iteratively update the JCF to incorporate responses from that survey and future surveys.”

While the JCF is now live as a “minimum viable product” — not an uncommon phase in rolling out large-scale, complex tech from scratch — the initial success reflects Groen’s view of the JAIC as a “do tank” in a town of think tanks. JAIC began its initial JCF “big push” last August by awarding a contract to enable IOC capabilities and implementation tools.

The JCF is part of JAIC’s broader initiative to get DoD “AI-ready” by 2025.

In the NDIA keynote, Groen urged the military, Defense Department, and industry to recognize that the AI race with adversaries is not some future event, but something that’s already here and happening now. “[The US] cannot separate AI from national security,” Groen said. “We need to start building [AI] now. Getting started now is critical.”

Major themes throughout the keynote included the “scale and scope” of DoD’s AI challenge, breaking down DoD “stovepipes” to facilitate resource sharing, “core” public-private partnerships, ethical AI, US competitiveness in the AI race, and — perhaps most prominently — the urgency of implementing AI now.

Groen highlighted four themes around what he called the “artifacts” of AI. Groen prefaced the four themes by referencing prior examples of warfighting paradigm shifts — “things that were imminently foreseeable [yet] not foreseen.”

Groen then turned to current day and asked what’s foreseeable in today’s AI environment that could impact future warfare. What does AI mean to the execution, tempo, and other factors in warfare?

The first is how close AI is. Groen reiterated how China has been “very vocal about dominating AI by 2030.” He stressed that, given the scope and scale of DoD’s AI challenge, 2030 is not as far away as it sounds.

Second, Groen highlighted how dangerous the situation is. Here, he emphasized a sub-theme throughout his speech: “If our systems are not prepared, then we’re not prepared.” The AI race is “truly becoming a systems competition. This fight is on from a systems perspective. Today, with the threat of systems warfare, there’s no time” to merely think about AI development, he said. To these points, he urged DoD to “break down stovepipes” and become a more integrated organization.

The third point is how different the AI challenge is. Groen said JAIC’s vision for DoD AI requires rethinking a lot of sequential processes, which are limited in scale and scope, and to instead build more dynamic “data-driven” processes. Groen’s specific example reiterated his statements throughout on the “scale and scope” of the AI challenge, which must be addressed from the highest organizational levels down to myriad specifics, including how AI fits into functional processes such as artillery. He also noted the importance of US academic and private sector partnerships to DoD’s success in meeting this challenge.

Finally, Groen pointed to the question of how to react. Given the “criticality” of the challenge, Groen said, it’s imperative DoD act. Transformation must be “wholesale” to be effective. “If one data process can affect another, then [we’re] getting there,” he observed.

He also mentioned the JAIC’s vision for a “fabric of platforms” to efficiently share data from multiple sources and in disparate formats across warfighting and support services. Groen noted that support services are the “gears the Department rides on for effective warfighting” and called them a “natural target for AI” to create more efficiencies across the DoD.

“I’m optimistic,” Groen said. “If I’m passionate today, it’s because of the scale and scope of the challenge — and what’s at stake.”