The Army’s 100-day investigation of risks tied to artificial intelligence implementation is coming to a close, and now, the service branch is exploring what the findings mean for the next phase of its AI journey.
Dan Andrew, software systems architect with the Office of the Deputy Assistant Secretary of the Army for Data, Engineering and Software, said the most surprising finding from the program is that AI is already pervasive within the Army, but was not fully understood and sometimes misidentified.
This discovery, he said in a panel discussion at the Potomac Officers Club’s 2024 Army Summit last week, brought to light the need for an assessment of how existing tools – such as large language models and deep learning technologies – are managed.
“I think the undercurrents of applying AI are already there,” added Ki Lee, senior vice president at Booz Allen Hamilton. “We just have to make sure that we’re answering the question,” which he explained is about how models are engineered and trained.
Though the program officially took place over 100 days, Andrew “would say it started two years ago” due to the foundational role of gathering data in the AI adoption process.
“Our AI implementation is 100 percent dependent upon quality data, and the [Unified Data Reference Architecture] is tuned and intentionally there to ensure data quality to support our AI,” he said.
If the data is not right, he said, the quality of the AI will suffer.
Reiterating Andrew’s points, Lee noted that AI is “the tip of the iceberg” and the Army “is very intentional about focusing on all the fundamental building blocks underneath that water line.”
Another question that emerged from the 100-day project is how the Army can accelerate AI adoption.
“How do we get to AI faster? Do we get to AI faster by using existing open industry models, taking the time and resources to train ourselves, all of those things?” Andrew said.
When planning courses of action, he continued, the Army must also balance risk and make sure that test, evaluation, verification and validation processes can “give us the appropriate amount of confidence that we aren’t exposing ourselves to unintended consequences by being fast and adopting already-out-there” technologies from industry.
Offering his industry perspective, Lee said that providing feedback to the Army as the strategy rolls out is a major way businesses can help. He commended the Army’s transparency, which allows commercial entities to weigh in on projects so the service branch can “take that feedback and iterate upon it.”
Once this phase of the project concludes, the Army will enter a 500-day sprint that will leverage the information collected so far. Andrew sees the next phase not as a “500 days until AI” journey, but a way to start addressing all the questions from phase one as quickly as possible while evaluating any gaps.
Learn more about Department of Defense modernization initiatives at the Potomac Officers Club’s 2024 Air Defense Summit on July 23, which will feature Air Force Under Secretary Melissa Dalton as the opening keynote speaker. To learn more and register to attend the event, click here.
