The Defense Department’s Chief Digital and Artificial Intelligence Office (CDAO) is supporting autonomy efforts throughout the enterprise, including the nascent Replicator Initiative, by creating the infrastructure to better leverage artificial intelligence and machine learning across programs, an office official said on Tuesday.
Autonomy is one of three lines of effort within the CDAO’s Alpha One portfolio, which is a set of capabilities and services related to AI and ML “scaffolding” the office will bring to the larger enterprise, Navy Capt. M. Xavier Lugo, the CDAO’s Algorithmic Warfare Division Chief For AI Scaffolding and Integration, said at an event hosted by the Center for Strategic and International Studies. Autonomy scaffolding refers to the infrastructure elements such as modeling and simulation and “other synthetic areas” separate from traditional AI capabilities such as data labeling and containerization, he said.
The CDAO is working to cross-pollinate sensor data across the program offices that have autonomy efforts, he said.
“So, the program offices have been…very stove-piped in the way they think of autonomy,” Lugo said. “As with AI/ML, in general, we are looking at how can we leverage the similarities of autonomy across weapons systems and more specifically the data.”
For example, data from one aircraft will likely benefit another aircraft, and the same is true between ground vehicles, and between surface vehicles, Lugo said.
“A lot more similarity than not,” he said.
Regarding Replicator, which the Pentagon initiated last August to quickly begin fielding uncrewed autonomous, attributable, all-domain systems geared for a potential war with China in the Western Pacific, Lugo said his office is enabling “perception autonomy.”
Perception autonomy refers to using AI to what a sensor is seeing or interpreting. Command and control autonomy for a vehicle remains the domain of the program offices, he said.
In addition to autonomy scaffolding, Alpha One consists of two other lines of effort, which Lugo described as “traditional AI” and would include computer vision, and generative AI. There is a lot of overlap between traditional AI and autonomy scaffolding, such as data labeling, model repositories, control systems, containerization, and instrumentation, he said.