The Pentagon has deployed machine learning algorithms to identify targets in more than 85 airstrikes against targets in Iraq and Syria this year.
The Pentagon has been working on this kind of effort since at least 2017, when it launched Project Maven, which sought suppliers that could develop object recognition software for drone footage. Google withdrew from the project after its own employees objected to the idea of using AI in warfare, but other tech companies have been willing to cooperate.
In 2017, Marine Corps Col. Drew Cukor said the Pentagon wanted to integrate the software into government platforms “by the end of the year” for intelligence gathering.
U.S. Central Command, which currently operates in the Middle East, Central Asia, and parts of South Asia, used this algorithm to conduct more than 85 airstrikes in seven locations in Iraq and Syria on February 2. .
Schuyler Moore, chief technology officer for U.S. Central Command, said the military began deploying Project Maven's computer vision system in actual operations after Hamas' surprise attack on Israel last year.
“Everything changed on October 7th,” Moore told Bloomberg. “We quickly shifted into high gear and achieved a much higher operating tempo than before.
Object recognition algorithms are used to identify potential targets. Humans then operate the weapon system. The US reportedly used the software to identify enemy rockets, missiles, drones, and militia facilities.
“Over the last 60 to 90 days, we've definitely seen an increase in targeting opportunities,” Moore said. U.S. Central Command is also running an AI recommendation engine to see if it can suggest the best weapons to use in operations and create attack plans. However, this technology was “often inadequate.
“There's never an algorithm that you just run, come to a conclusion, and move on to the next step,” she says. “For every step that involves AI, there is a human check-in at the end.”
Concerns that the United States could fall behind more capable adversaries has led the Pentagon to ramp up efforts to integrate and test AI combat capabilities. Craig Martell, the agency's chief digital and artificial intelligence officer, envisioned a large-scale language model to guide combat decisions last week at the Advantage Department of Defense 2024: Defense Data and AI Symposium.
“The Department of Defense has a mandate to responsibly generate AI models while identifying appropriate safeguards and mitigating national security risks that may arise from issues such as improperly managed training data. “We have an obligation to promote adoption,” he said. “We also need to consider the extent to which adversaries may exploit this technology and seek to disrupt the use of our own AI-based solutions.” ®
