DoD Creating Standards For AI Programs

WASHINGTON: DoD’s Research and Engineering (R&E) office has launched a new initiative to develop best practices for the many programs to design and build artificial intelligence (AI) applications, says Mark Lewis, director of modernization.

AI is one of DoD’s top research and development priorities, charged to the Director of R&E Mike Griffin.

The standards initiative is the brain child of newly appointed AI technical director Jill Chrisman, Lewis told the virtual “Critical Issues in C4I Conference 2020,” sponsored by AFCEA and George Mason University.

“When Jill first joined us just a couple of weeks ago, I asked her to give me a site view of all the efforts underway in AI across the department, and kind of give me an evaluation of where we stood,” Lewis explained today. However, he said, because DoD has “so many hundreds of programs that we really couldn’t do a fair evaluation of each individual activity.”

So, instead R&E has decided “to establish a series of standards, if you will, principles and practices that we consider to be good practices for artificial intelligence engineering,” he said. “I liken it to systems engineering.”

A key goal of the new effort is to break down stovepipes in order to allow the various DoD AI efforts to share databases and applications.

In addition, Lewis said, R&E is aiming to “figure out what are the artificial intelligence applications that will have the biggest impact on the warfighter.” This could involve moving out prototypes rapidly so that warfighters have an opportunity to “play with them, experiment with them, and figure out what makes their job more effective,” he added. At the same time, it would enable warfighters to quickly reject things that are not useful or overly complicated.

Lewis said that developing autonomous systems is another top priority. That portfolio of effort is handled by assistant director Wayne Nickols, and is focused on development autonomous systems that can team seamlessly with humans.

“We want autonomy systems that will operate in ways that put human life at lower risk,” he explained. “If we can if we can have a robotic system as a target, instead of a human being as a target, that’s that’s our preferred approach.”

In his wide-ranging discussion, Lewis also expounded on DoD’s research goals for quantum science — a focus area that he said is somewhat less well developed than others on DoD’s high priority list.

“There is a lot of hype associated with quantum science,” he said bluntly. “People are talking about quantum computers that will, in a few years, replace our fastest supercomputers, quantum communication technology, quantum key encryption techniques. And frankly, a lot of it is promising —  but it’s also very very far term.”

That said, Lewis noted that there are two near-term opportunities for DoD in the field: enabling back-up positioning, navigation and timing capability in case GPS satellites are degraded in anyway; and future “exquisite sensors for a variety of applications.”