The U.S. military this year plans to deploy artificial-intelligence algorithms, including object-recognition technology, to sift through its intelligence database to gather insights about the Islamic State (ISIS) terrorist group.
The military will leverage existing commercial AI technologies to accelerate the deployment of these algorithms, as reported by Breaking Defense. Such algorithms will take on the challenging task of examining images and identifying whether they depict ISIS activity.
For example, the object-recognition technology will need to be able to distinguish a civilian pickup truck from an armed ISIS vehicle, or a hospital from an enemy stronghold. However, because the object-recognition software is dealing with a limited number of enemy weapons and structures, it will be possible to train and utilize it quickly.
The software will need to recognize and distinguish only “about 38 classes of objects,” according to Col. Drew Cukor, in comments reported by Breaking Defense.
“We’re not talking about three million lines of code,” Cukor said. “We’re talking about 75 lines of code… placed inside of a larger software (architecture)” for intelligence-gathering.
Cukor heads the Algorithmic Warfare Cross Function Team, a group specifically formed to manage the flood of data being gathered by various electronic intelligence systems. The establishment of the team represents a recognition that the military’s capability to gather intelligence has outstripped its capacity to analyze that information. Like many private-sector companies, the military is finding that traditional, manual monitoring techniques are unequal to the task of processing today’s massive quantities of data.
“Our work force is frankly overwhelmed by the amount of data,” Cukor said, adding that, “staring at things for long periods of time is clearly not what humans were designed for.” Because of this, human reviewers often miss key information.
The Algorithmic Warfare Cross Function Team is in the process of training its machine-learning systems by labeling images in its massive database. Cukor said the team is working with a data-labeling company on this task.
The first project for the team’s object-recognition AI will be sorting through the video data collected by the U.S. military’s vast fleet of unmanned surveillance drones.
The team’s work represents an element of a Pentagon initiative called the Third Offset Strategy, which is focused on using AI to gain military supremacy.
The effort is placing an emphasis on the collaboration between humans and AI. For example, object recognition can sort through massive quantities of data, identify patterns and report them back to human decision-makers.
John Newsom is executive vice president and general manager for Veritone Government. He is a software executive with an evangelical passion for AI technology who aligns the Veritone Platform with customer and market needs.