The Defense Department’s Joint Artificial Intelligence Center is now out of its most nascent stages and off and running to make changes in the Pentagon; what that means for the next year is the organization will tackle big projects that have the most potential impact to spread AI throughout the military.
Just last fall, JAIC entered into a new phase of life by moving away from small test projects and turning toward issues that will have serious impact on the force and its readiness.
Now, for 2021, JAIC acting deputy director Jacqueline Tame, says the center is picking projects that will do the most good for the department due to their ability to scale up and touch multiple areas of the military.
That may not give JAIC what many gung-ho, sci-fi-loving AI advocates imagine when they think of the military and computers that think for themselves, like weapons that can detect enemies without human help — though JAIC is working on those things too. Instead, Tame said it means a lot of work with DoD’s back office and business process transformation space.
“We have had real interest in the last several months in particular in the joint logistics enterprise,” Tame told Federal News Network.
That means working with entities like U.S. Transportation Command and the Defense Logistics Agency — two organizations that provide the invaluable service of delivering people, weapons and necessities. In other words, they use a lot of business practices like supply chain management and inventory management to do their jobs.
“You can easily see where the intersections of the biggest return on investment are going to be with their work with logistics and moving materiel,” Tame said. “The technology is actually pretty mature in these areas where there’s an abundance of data in some of these back office areas. That’s one of the ways in which we’re really mapping the landscape of DoD from a missions and mission needs perspective and targeting those areas where we think AI applications and helping to instantiate AI is going to scale the most quickly or the most robustly.”
Another area JAIC is working is in the test and evaluation space and how the military services can do testing specific to AI.
“There needs to be an entirely different framework for how you think about and instantiate test and evaluation for AI applications than for other technologies, especially when you layer on AI to something that is incredibly near and dear to our hearts,” Tame said.
DoD’s test and evaluation leadership is working closely with JAIC to build that framework and promote AI adoptions.
“Those initiatives are actually then lending themselves to further learnings and further things that we end up undertaking,” Tame said. “That test and evaluation framework for AI has actually showed us that there are some sort of gaps right now in the way in which we procure those services, whether they be, data labeling services, or the actual customer evaluation services themselves.”
JAIC is also getting to the tactical edge as well, but getting AI to that position means building the plumbing and cleaning the data pipes that get to the edge.
Tame said there needs to be a solid infrastructure for the data to freely flow so that information can go from sensors to AI applications and then to humans.
“It is not easy to bring those capabilities to the tactical edge, there’s a lot of work that has to go into cleaning and curating data and ensuring that we are identifying and removing barriers to entry, whether they be technical or policy or even cultural in some cases,” Tame said.
One example of how JAIC is working on the tactical edge is its partnership with U.S. Special Operation Command’s 160th aviation regiment to improve the quality of its helicopter maintenance records.
“Now fully operational, the AI-enabled capability has corrected more than 400,000 historical work unit code entries,” a press release from last December states. “These corrected maintenance records provide improved data analytics and reporting for maintenance and supply, and enable AI-ready data for future predictive readiness capabilities.”
The regiment is basically able to fix things before they break and save time and money.
“The program enables both proactive and reactive data maturation to create AI-ready maintenance information,” said Chris Shumeyko, of the JAIC’s Joint Logistics mission initiative. “By suggesting correct codes on data entry, dirty data never makes it into the maintenance systems, which is a problem that has deterred the maintenance community for years. For the first time, failure and corrective action information will be AI-ready from the start, enabling more rapid deployment of advanced analytical capabilities to improve operational readiness.”
JAIC is currently building out its Joint Common Foundation, which will allow different areas of DoD to work on AI programs like the predictive maintenance program and share it with other military components.
Tame said JAIC is thinking of the foundation as more of a fabric because the military services already have some programs that share applications on a smaller scale. JAIC wants to incorporate all of those together.