The department, we think, must ensure it has in place the resources processes and organizations to enable AI innovation. The department needs to establish a common digital ecosystem — it’s called the joint common foundation — in the department of defense. That’s the technical foundation for all AI development and fielding. So it’s going to include access to a secure cloud AI software trained models, data, and algorithms, as well as high performance computing power and a development environment that allows the entire AI stack to be put together in a way that is secure and will do what we expect it to do on the battlefield.
We think JAIC should be designated as the department’s AI accelerator. The JAIC in our view should focus on applications, not on hard research. Essentially, the JAIC’s role is to try to get as many applications into the field as possible, and they provide the resources through the digital ecosystem or the joint common foundation they also can provide subject matter expertise to support AI efforts across the department without becoming a central clearinghouse.
The department has to expand the use of specialized acquisition pathways and contracting authorities to source and deliver the best AI. As ranking member [Elise] Stefanik mentioned, software and algorithms are just a different kettle of fish than ships, airplanes, missiles etc., We have to come up with ways that are specific to get those algorithms and models developed, and into the hands of our war fighters.
We have to come together also to reform the Planning Programming Budgeting and Execution process [emphasis added]. Congress has provided the department of defense with an expanded toolkit of acquisition and contracting approaches, and the department’s effort to adopt AI will be impeded by processes that are unsuited to digital technologies and the pace of development of AI right now.
The department should also increase its overall S&T spending, increase AI R&D to 8 billion annually by 2025. We think that’s totally appropriate within the size of the DoD R&D budget, which is the largest in its history.
Third we think AI adoption has to be accelerated. We think one of the ways to get these algorithms and models across the valley of death and into the hands of warfighter would be if congress could create a dedicated AI fund specifically designed to speed operational prototyping and transition overseen by the under secretary defense for R&E.
That was NSCAI vice chair Bob Work testifying to the HASC on March 12, 2021, “Final Recommendations of the National Security Commission on AI.” He had to budget recommendations there, reform the PPBE and create an AI innovation fund to bridge the valley of death. The latter is a short-term fix, the former the long term fix. He doesn’t provide more details on that, but readers of this blog know that I have plenty of meat to put on them bones.
I haven’t read the Final Report, so it might be there, but it’s not clear to me whether NSCAI envisions the JAIC being an accelerator for applications analogous to DIU or AFWERX, or as delivering critical elements of the AI stack in the joint common foundations analogous to a Platform One or Black Pearl, or some combination.
Leave a Reply