Pentagon’s AI ambitions require high-quality data, CDAO’s Martell says

SAN ANTONIO – To advance artificial intelligence that the U.S. Department of Defense can deploy and rely on, a foundation of “really high-quality data” must first be laid, according to the Pentagon’s AI czar, Craig Martell.

Such a task, he said on December 13 at the DODIIS Global Conference in Texas, became his main task for about eight months to lead the main office for digital and artificial intelligence.

“What is our job at the CDAO? We originally thought our job at the CDAO was to produce tools for those in government to do modeling. We no longer think that is the case,” he said. “We say we want to build the scaffolding that goes around the model. We want to help people build and use that model.”

The CDAO, established in December 2021, took its first full steps in June. It brought together what was the joint artificial intelligence centre, the defense digital service, the Advana audit platform and the chief data officer’s role.

RELATED

In short, the office is seen as an overseer and expeditor of data analytics and AI at the Pentagon, where they are increasingly the focus of spending, experimentation and implementation.

“I’m a firm believer that what most people are asking for when they say they want AI is actually a really good dashboard that just tells them where their stuff is,” Martell said. “And so we believe our job is to get the data right, and then provide a very strong data analytics layer.”

More than 685 AI projects, including several tied to major weapons systems, were underway at the Defense Department as of early 2021, according to a Government Accountability Office report published in February.

Mass amounts of reliable data and training are what fuel AI capabilities; the digital lifeblood allows AI to assist with navigation and target recognition — as expected aboard the Army’s optionally manned combat vehicle — as well as with maintenance forecasts and supply logistics. However, not all data is created equal, which complicates the work of Martell and his colleagues.

“First, we have enormous amounts of data. We have massive amounts of data spread all over the world,” he told the conference. “Some of that data is going to be really effective for decision-making at scale. And other parts of that data are going to be completely ineffective for decision-making. So, a big part of our job is figuring out the ways to manage that data, so that the data that is effective for decision making is front and center for decision makers when they need it.”

Martell was named chief digital and AI officer in April, succeeding John Sherman, who took on the role in an acting capacity and remains the Pentagon’s chief information officer. Martell previously worked at Dropbox, where he was head of machine learning, and LinkedIn, where he led a number of AI teams and initiatives.

On a previous occasion, Martell said that he took the government’s action because of the gravity of the situation.

“There aren’t many people who have the intersection of AI and a government background,” Martell said in June. “So when the deputy secretary of defense calls you and says: ‘We’d like you to take this job,’ you have to think very carefully about why you won’t take the job, and not the other way around. And I think it’s extremely important to get this mission right.”

Courtney Albon contributed to this report.