By Sandra I. Erwin
Hyper-intelligence machines have permeated every layer of modern society — from smartphones to self-driving cars. Silicon Valley continues to pour billions of dollars into efforts to keep pushing the boundaries of artificial intelligence. So it might come as a surprise that in the military and intelligence agencies, the technology is not moving as fast as it could.
Even as their needs for data analysis continue to grow, the U.S. military and intelligence communities are not harnessing machine learning technology to its full potential, says Brian Wilson, senior vice president of Zenoss, an information technology company that does business in both the commercial and government sectors.
“It’s a common thread I’m seeing,” Wilson says in an interview. “There has been continued advancement of artificial intelligence and machine learning for analyzing big data. Palantir [a Silicon Valley company] is an absolute leader in that space.”
The government has been slow to tap into the commercial revolution and take that technology to a lower price point, he says. “Today’s problem boils down to ‘What do I do with data? How do I interpret it? How do I analyze it? How do I draw knowledge out of information?’” Wilson adds. “I would say that is still a problem area where the intelligence community will need help from commercial vendors.”
The potential is obvious in the defense and security fields. Border patrol and homeland security agencies use facial recognition but they need better capabilities to interpret data. “Can you get down to the point that you can infer emotions in images? Can I start to predict what actions might occur so I can be proactive based on the posture of a particular person? This would require deep machine learning and artificial intelligence. It has to happen in real time,” Wilson says. The data exists to make new applications possible, he says. “It’s a matter of merging data from scattered databases around the world.”
Outside advisers have urged the Defense Department to become more aggressive in the use of artificial intelligence not only to increase its capabilities but also to save money. Many labor-intensive functions that today demand a large workforce would be performed by computers, suggested the Defense Innovation Board. The 20-member panel of tech industry and government officials, established in April by Defense Secretary Ash Carter, last month offered a new set of recommendations, one of which is to “catalyze innovations in artificial intelligence and machine learning.”
The board characterized artificial intelligence and machine learning as “important, defining technologies.” The military “needs to use massive amounts of data rapidly. Autonomous vehicles use massive amounts of data. And people are being tasked to do things that machines are fundamentally better at, such as image recognition,” Defense Innovation Board Chairman Eric Schmidt, executive chairman of Google’s parent company, Alphabet, says during a public meeting in January.
Board members noted during visits to military installations in the United States and overseas that “many of our soldiers and sailors are tasked with just watching.” This is one example of areas where computers could pick up the load.
“We’ve all seen artificial technology at play,” says astrophysicist Neil DeGrasse Tyson, a member of the Defense Innovation Board. “We’ve seen it beat the world’s best chess player. We’ve seen it beat the world’s best Jeopardy player.” Computers are capable of more complex tasks, he says. “This is going on in the commercial sector. I’d be disappointed to learn that the world’s greatest military is not ahead of all other activities in this regard.”
The risk to the United States is that technological advances will continue to “occur at our own peril if other nations or non-states embrace this in ways that we do not,” says DeGrasse Tyson. “There’s no end to how much damage it could cause us if we’re not ahead of this curve.”
The history of computers has been about “replacing tasks,” he says. “With artificial intelligence you not only have machines that do things better than we can, you’ll get machines that make cross-tasked decisions in areas you would not have thought of. That is the real future here.” Computers can watch screens, but artificial intelligence does much more,” DeGrasse Tyson says. “It connects to a command post, it advises, it researches the history of what went on, suggests or comments. The future of this is much more integrated that just being a ‘task master.’”
The Defense Innovation Board would like to see the Pentagon make artificial intelligence a “core competency.”
Werner J.A. Dahm, chairman of the U.S. Air Force Scientific Advisory Board, also is a proponent of greater use of machine learning and artificial intelligence in the Defense Department.
“Machine learning and data science are getting a lot of attention for good reason,” he tells National Defense.
Why so much interest? “There is an enormous payoff coming from the development of technology to allow computers to predict and anticipate,” says Dahm. Machine learning is used by retailers to predict what you want. Most of that research is happening in the commercial industry and academia. The Defense Department is “leveraging” technology but it is a “small player in the big picture,” he says. “There is a big push to continue and expand it. The amount of money going into machine learning commercially is huge” compared to what the government invests.
The Air Force Scientific Advisory Board recently completed a study on data analytics that looked at what the commercial sector has been doing, says Dahm. “When you can automate, you can process intelligence totally differently.”