The Department of Defense’s flagship artificial intelligence program, Project Maven, has delivered half a dozen capabilities since launching last April with officials looking for new industry and open system opportunities to expand deployment of its algorithms.
Lt. Col. Garry Floyd, deputy chief of the Algorithmic Warfare Cross Functional Team, told attendees at a May 1 event held by sister publication Defense Daily that Project Maven’s AI algorithms for full-motion video data analysis have successfully been deployed to commands in Africa and the Middle East, and further partnerships are needed to rapidly push out future machine learning capabilities.
“We’re very, very early with Project Maven. We’re at five or six locations now in USAFRICOM and USCENTCOM with more on the way,” Floyd said at sister publication Defense Daily’s Modular Open Systems Summit. “We’re basically developing what I think of as ‘digital wingmen’ that we’re deploying.
Project Maven dropped its first set of algorithms in December, and then another round in March and April, according to Floyd.
The AI initiative focuses on freeing up DoD analysts from having to manually make sense of the massive amount of ISR data collected from tactical UAVs and video sensors, with flexible algorithms.
Deployed capabilities include a “Train AI” button that allows the system to be retrained and react to changing operational environments, according to Floyd.
“When we made that first deployment to USAFRICOM, I think we trained that algorithm six times in five days,” Floyd said. “We think of these algorithms as augmenting an assistant analyst in our range, we don’t see them as replacing the analyst. We trying to free up time for them to focus on tougher tasks.”
With several successful algorithm drops in just a year since the program’s incarnation, officials are looking to open architectures systems for opportunities to improve data collaboration and user interface opportunities.
“It needs to be open. It needs to be available, not just to vendors but to all our partners,” Floyd said. “Open needs to be the standard and the expectation, both in the actual deployment of the algorithm and the assembly of user interfaces. And then we need to make sure that we have relatively open access to the department’s labeled data sets.”
Floyd said industry partners should have a key role in helping finalize Project Maven’s capability ‘pipeline’ for future AI capability development to function in an open system environment.
“There are numerous areas all across the pipeline, from the data wrangling piece to the data labeling piece, where we know that we will need help. Some of those things we can do internally, but we know that we’ll need partners to help us. Particularly, when there’s operational need to field rapidly.”
Floyd addressed previously reported concerns that employees of Google [GOOG], a main partner with Project Maven, have expressed related to their company’s participation in a program that assists with warfare technology.
“I think what you’ll see is there will be those who will partner with us and help us do the things that we need to do to be successful, because success revolves around completing the mission successfully so we can all come home safe. And can keep collateral damage to a minimum. I think AI will help us do all those things,” Floyd said, citing the technology’s ability to reduce human error in data analysis.
Project Maven’s AI capabilities could eventually extend beyond sensors for full-motion video to other use cases in the future, according to Floyd.
“You’ll see us continue to expand upon the initial work that we’ve done. There are additional capabilities that we want to develop in that sensor area. We’ll see where we go from there. There are some other concepts and use cases you might see us pursue,” Floyd said.
Future success of the AI program could be tied to DoD’s upcoming multi-billion dollar cloud project, Joint Enterprise Defense Infrastructure (JEDI), according to Floyd. JEDI would migrate the department’s operations to the cloud and eventually extend to tactical operations.
“We’ll see how this thing plays out. But we know from the Maven experience that the cloud will be a key enabler of the things we’re trying to do with AI and machine learning,” Floyd said.