It’s unclear if anyone really knows just how many pilot projects in the Defense Department are using artificial intelligence, machine learning or intelligent automation.
Some say it’s around 300, while others say it’s closer to 600, and then there are those who believe the number could be more than 1,000.
Insight by Confluent: Learn about how agencies are benefitting from that concept of data-in-motion to improve mission outcomes in this exclusive e-book.
But unlike so many technology innovations that came before it, the Pentagon, through its Joint Artificial Intelligence Center (JAIC), is taking aggressive action to stop, or at least limit, AI-sprawl.
“There’s a lot of efforts that are out there that are not very well tied together and there’s a whole bunch of them that are dealing with exactly the same thing. So one of them is talent. Do they have talent? Or do they have to grow their talent or do they have to acquire the talent? The other big one, of course, is data and it’s almost invariably when anybody in the Department of Defense talks about doing work, they get to the data saying, ‘Okay, my data hasn’t been cleansed so is it usable?’” said Anthony Robbins, the vice president of the North American public sector business for NVIDIA, in an interview with Federal News Network. “They try to assess use cases, and then they’re trying to figure out how to get started. The JAIC wants to help them figure out this out.”
DoD launched the JAIC in June 2018 with a much different vision than where it stands today. Whereas the Pentagon saw JAIC nearly three years ago as pushing AI to the military services and defense agencies through pathfinder projects, it’s now focused on providing services and setting the foundational elements for mission areas to take advantage of the technologies.
In November, DoD announced JAIC 2.0 detailing its new vision and mission. As part of that new approach, the JAIC awarded a $106 million contract in September to build the Joint Common Foundation Artificial Intelligence (JCF), and plans to create three new other transaction agreements (OTA) vehicles in the coming year under the Tradewinds moniker to further build out its services catalog.
Jacqueline Tame, the acting deputy director of JAIC, said the move to 2.0 is a recognition that the services and defense agencies need a different kind of help to ensure AI tools improve and measure mission readiness.
The JAIC doesn’t need to be a doer, but a trainer, educator and supporter because the adoption of AI and AI-like capabilities —think robotics process automation (RPA) and predictive analytics — are spreading across the department like wildfire.
“What we have been able to do over the last two-and-a-half years is really test what the department actually needs, what the department is actually ready for and what the foundational building blocks of AI-readiness actually are. JAIC 2.0 is a recognition and learnings that we’ve undertaken that there are some key building blocks we have to put in place departmentwide to be AI ready,” Tame said during AFCEA NOVA IC IT day. “Where we are today, having developed a lot of capabilities, deployed a lot of prototypes and implemented a lot of solutions across the department is that we’ve learned that what the department actually needs is enabling services.”
Tame said while some like the Army Futures Command, the Special Operations Command and in the Air Force have matured their AI capabilities, the efforts too often are rolling out in siloes.
“What is still not happening, and this is the underpinning of JAIC 2.0, is the connective tissues between all of those capabilities that is being researched or deployed. What is still lacking in our assessment is the aggregate of the components of AI-readiness,” she said. “That includes removing some of the barriers to entry that present themselves in terms of both education and awareness about what AI is and what AI is not, what things actually lend themselves to AI and AI-enabled applications. Really understanding what the data need to looks like, the status of AI readiness in order to leverage it, test it appropriately and an understanding of the ethical underpinnings in terms of what that needs to look like as we consider some of the more advanced capabilities that we are trying to deploy across the force. Having a really foundational understanding of the types of infrastructure and architectures that need to be able to be interoperable in order to achieve the goals we are trying to achieve here. And really trying to understand the culture barriers to entry that still exist.”
Like with any new technology, the culture barriers to AI aren’t unusual. But Tame, Robbins and other experts say trust, confidence and usability are at the heart of AI-readiness.
“This is a technology that is and will affect every person, every country and every industry around the world,” Robbins said. “It is a technology that can go into every industry from transportation to healthcare to defense. Technology transformation is as much about leading change in transformation as it is the technology. The technology is ready.”
Robbins said a predictive and preventive maintenance program, as well as its use to help with humanitarian assistance, are two examples of how DoD already is using AI.
One example is the Army’s Aviation and Missile Command G-3’s work with the JAIC since 2019 on the predictive and preventive maintenance for the UH-60 Blackhawk helicopter.
“When it comes to logistics and maintenance, there is an overwhelming amount of data available — anything from aircraft sensor data to maintenance forms and part records,” Chris Shumeyko, JAIC product manager, said in an Army release. “Ordinarily, subject matter experts play a huge role in understanding this data and identifying trends that may affect the readiness of the Army’s vehicle fleet. However, as the amount of data grows, you either need more experts to comb through that data or possible warning signs of problems may get missed. By injecting AI/ML, we’re not replacing these experts, but rather providing them with tools that can find hard-to-spot trends, anomalies or warning signs in a fraction of the time. Our goal is to increase the efficiency of the experts.”
It’s this type of service that the JAIC is providing under its latest iteration.
Tame said the new services include or will include:
Robbins said these services and other recent actions by JAIC is part of how DoD is moving AI out of the testing phase and into the operations phase.
Tame added part of the way to address that operational need is not to develop, test and deploy in the siloes of yesterday, but through a common framework that creates a starting point for all AI technology.
“These critical building blocks will enable us to get to the point of implementation of AI across the force in a really cohesive way are not there yet,” she said. “The JAIC’s role really needs to be driving that advocacy and education of our senior executive leadership all the way down to line analysts and intelligence agencies about institutionalizing the ethical underpinnings that need to be talked about every time we are thinking about AI, about ensuring there is a departmentwide test and evaluation framework that is specific to AI, which is different than everything else the test and evaluation community has been saying before, and ensuring we have a really foundational understanding across the board of those data standards, many of which do not exist yet or haven’t been agreed upon, and the level of infrastructure interoperability that we need to both put in place in terms of new systems and reimagine in terms of our legacy systems.”
The end goal of JAIC 2.0 isn’t just about offering new services or changing its mission focus, but addressing the AI-sprawl that seems to be quickly happening by giving the military services and Defense agencies a common baseline to build on top of and ensure the necessary trust, confidence, security and ethical foundations are in place. This is something that was missing with cloud, mobile devices and many other technologies that led to unabated sprawl.