By Marty Graham, Contributor
In February, when the United States Department of Defense’s (DoD) Futures Command released an unclassified plan for its artificial intelligence (AI) strategy, it also created its first, large-scale taskmaster: the Joint Artificial Intelligence Center (JAIC, pronounced “Jake”). The AI center is focused on the speed and agility the military hopes to attain through understanding information and managing processes. Funded this year with $90 million, the budget will grow to $414 million next year as the AI center launches its first projects.
These projects are more than ideas and tasks: They’re a pragmatic move to get the technology up and running, and demonstrate its usefulness for a wide range of projects, since AI is a tool the Defense Innovation Board says will change the way the Pentagon does business. That’s no small matter: The DoD’s fiscal 2019 budget is $750 billion.
The AI center is the first broad-scale vote of confidence in algorithms. Though the Defense Advanced Research Projects Agency (DARPA) is widely seen as a key pioneer of AI, little of DARPA’s work crossed the hall to military application. DARPA‘s latest investment is $2 billion toward AI 3.0: teaching machines to adapt to unexpected change in circumstances.
The JAIC grew from a smaller effort, Project Maven, which started—like so many image-based machine learning projects —with the painstaking task of identifying and labeling thousands of images to help the machine learn what it’s supposed to look for. The military’s goal was to develop AI capable of using drone video in its counterterrorism efforts.
Google won the $28 million contract to build an algorithm that can interpret video to advance the drone video effort and provide cloud services to Project Maven, quickly sparking controversy among Google employees who balked at the idea of the company entering the war-fighting arena. In response to a petition signed by several thousand people from the Tech Workers Coalition and its own employees, Google opted not to renew the contract.
The Pentagon’s lead on Project Maven, U.S. Air Force Lt. Gen. Jack Shanahan, is heading the new initiatives with the JAIC. In a March interview, Shanahan said the past controversy stemmed from “grave misperceptions about what the DoD is actually working on.”
Working With Brains and Brawn
Using the Maven model, the JAIC looks to academia and the high-tech industry for the expertise and bandwidth to dive quickly into large-scale AI. The AI center teamed with Carnegie Mellon University‘s National Robotics Engineering Center, bringing $72 million to the table.
With the far-reaching mission of rapidly introducing AI throughout the military, the JAIC is looking for partners in the private sector to supply the needed tools and expertise, including cloud computing and algorithm development. Working on a short list of demonstration projects creates a real-time framework for implementing these tools.
“The benefits to the department will continue to accrue over time, increasing the level of understanding of AI across the force while accelerating the delivery and adoption of AI throughout the DoD.”—Elissa Smith, JAIC spokesperson
“It is important to show early wins that provide practical results and demonstrate the art of [what’s] possible, followed by scaling capabilities across the enterprise,” Elissa Smith, spokesperson for the JAIC, says. “As we build and scale each project, our ability to harness the full operational potential of AI increases. The benefits to the department will continue to accrue over time, increasing the level of understanding of AI across the force while accelerating the delivery and adoption of AI throughout the DoD.”
The center is focusing its early efforts on projects in several different areas: applying machine learning to maintenance of its fleet of Sikorsky UH-60 helicopters, and to humanitarian and disaster relief missions. A third, recently added project called cyber sense-making will develop an algorithm to detect and deter cyber attackers whose skills are aimed at DoD information systems.
The predictive maintenance project is expected to result in better forecasting, diagnosis, and management of maintenance of the military’s Black Hawk helicopters that will reduce costs, increase safety, and improve efficiency, Smith explains. It will incorporate data that includes staffing and supply issues, location and use of helicopters, and maintenance records to analyze patterns of work being done.
Using the flood of image data from the ever-increasing video sensors and video gathered on fly-overs by military-manned and unmanned aircraft, the humanitarian assistance and disaster relief project aims to reduce the time it takes to understand conditions on the ground, such as in disasters like the recent California wildfires. This will allow for better decision-making about how to allocate resources more effectively, and will enable smarter, quicker, lifesaving-efforts and relief assistance—from providing food and clean water to beginning to clear debris and rebuild.
“As the JAIC’s capacity increases, the team will explore how to broaden the use of AI capabilities.”—Elissa Smith, JAIC spokesperson
“This mission initiative is using AI to automatically map active fire lines from full-motion video, and will reduce the time and cost of mapping active wild fires,” says Smith. “The mission initiative has plans to extend this to other types of imagery, as well as to expand the work to assess infrastructure damage during other events, such as flooding. As the JAIC’s capacity increases, the team will explore how to broaden the use of AI capabilities.”
There are other projects—from small ones that achieve real efficiencies in existing processes to larger ones with broader goals that haven’t previously been considered or seen as possible. “We are in early discussions on the applicability of AI to help with solutions in areas as diverse as talent management, suicide prevention, preventive medicine, installation and force protection, and other solutions,” she says.
But the JAIC has even bigger goals.
By focusing on limited scale projects, the JAIC aims to prepare the rest of the Pentagon to think about where AI can make things sleeker, smarter, and faster. Anticipating and managing maintenance for a few thousand helicopters, for example, can lay the groundwork for different ways to staff and supply what’s needed to track and maintain the DoD‘s 97,000 trucks and 66,000 passenger vehicles, 2,100 UH-60 helicopters, and thousands of other aircraft and hundreds of ships.
“The JAIC will develop a common foundation that is essential for scaling AI’s impact across the DoD,” Smith says. “This foundation includes shared data, reusable tools, frameworks, libraries, standards, and cloud and edge services.”