The Defense Department’s Joint Artificial Intelligence Center has launched a DoD-wide development platform meant to accelerate testing and adoption of AI tools across the department.
JAIC Director Marine Corps Lt. Gen. Michael Groen said this week that the Joint Common Foundation is in its “initial operating capability,” but will build it up to full operational capability through doing block upgrades.
“Every month, we want to add more services, more capacity, more capability to the JCF, but it’s up and running,” Groen said Tuesday at an AI conference hosted by the National Defense Industrial Association.
The JCF is a DevSecOps platform led by the JAIC that looks to lower the barriers to AI adoption by providing a secure testbed where users can share data sets, training data and algorithms across DoD components. Groen said the platform is similar to a “lending library,” where one service can pull training data from another service’s AI tools and use it to their own algorithm.
The Air Force and the Army have their service platforms, but Groen said the JAIC has several services signed up to use the JCF, and wants to integrate these development platforms into a “fabric of platforms.”
“As we build our Joint Common Foundation, we’re thinking about this joint common fabric. How do you stitch these development and operational environments together so that you can actually share data readily from an Army sensor into an Air Force system, or vice versa?” Groen said.
The JCF will help the JAIC become an “accelerator” for AI adoption in the Defense Department, as outlined in the final report from the National Security Commission on AI.
That same report also calls on the Defense Department to be AI-ready by 2025, which will require building out training across the department.
The NSCAI sets a long-term recommendation for building out a service academy focused entirely on emerging technology, but in the near-term, Groen said the JAIC is working with some of its vendors to identify a pipeline of training with several AI tracks.
“If you are an AI operator, perhaps you only need a certain level of AI coursework that we can really give you. If you are a data scientist, or if you’re a coder, maybe you need something a little bit more sophisticated. If you are in a leadership role, and you really don’t do AI, but you manage it or you oversee it, then we have a training package that we’re putting together for that archetype as well,” Groen said.
Groen said the JAIC is leveraging best practices from services that have already done some of this work and is sharing best practice across DoD. But to build AI competency across DoD, Groen said the services need to work together to identify common competencies and skillsets needed for this work.
“If Service A says, ‘No, I got it, I’m going train my people, and I don’t want any help, and I don’t want anybody to look over my shoulder,’ then we’re going to wind up with at least five different approaches, at least five different levels of certification. If you go into a joint job, what level are you actually certified as an AI expert? We won’t know if we just stick with a stovepipe approach,” Groen said.