The Pentagon’s New Algorithmic Warfare Cell Gets Its First Mission: Hunt ISIS

By year’s end, the Pentagon wants computers to be leading the hunt for Islamic State militants in Iraq and Syria, through turning countless hours of aerial surveillance video into actionable intelligence.

It’s part of Project Maven, a fast-moving effort launched last month by Deputy Defense Secretary Bob Work to accelerate, improve, and put to wider use the military’s use of machine learning.  

“We have to tackle the problem a different way,” said Air Force Lt. Gen. John N.T. “Jack” Shanahan, director for defense intelligence for warfighter support, and the man tasked with finding the new technology. “We’re not going to solve it by throwing more people at the problem...That’s the last thing that we actually want to do. We want to be smarter about what we’re doing.”

Thousands of military and civilian intelligence analysts are “overwhelmed” by the amount of video being recorded over the battlefield. These analysts watch the video, looking for abnormal activities. Right now, about 95 percent of the video shot by drone aircraft is from the campaign against ISIS in Iraq and Syria.

The Pentagon has raced to buy and deploy drones that carry high-resolution cameras over the past decade and a half of war in Afghanistan and Iraq. But on the back end, stateside analysts are overwhelmed. Pentagon leaders hope technology can ease the burden on the workforce while producing better results on the battlefield.

“How do we actually begin to automate that in a way that gives time back to analysts who otherwise spend 80 percent of their time doing…mundane, administrative tasks associated with staring at full-motion video,” Shanahan said.

If an analyst sees something now, he or she typically types the data manually into a spreadsheet. Pentagon leaders do not believe that’s a good use of these analysts’ time.

So last month, Work — perhaps the Pentagon’s lead champion for marrying people and technology — established a special cell called the Algorithmic Warfare Cross Functional Team. The group will look to integrate big data and machine learning across the military.

Shanahan — whom Work put in charge of the team — has nearly 33 years in uniform. He spent most of his first 15 years of his career as a weapons systems officer in the backseats of F-4 and F-15E combat jets. Since then, he’s been deeply rooted in intelligence roles in the Air Force and joint positions.

The Algorithmic Warfare team is working hand in hand with the Pentagon's Strategic Capabilities Office, a group working to modify existing weapons and technology to make them more versatile and lethal.

"The integration between our respective initiatives and organizations will be immensely beneficial in moving the ball forward," Shanahan said.

Officials see the effort to automate intelligence processing as the starting point for using new technology in other military functions.

“There are a thousand things we want to do with artificial intelligence, machine learning, deep learning, computer vision, but everybody has cautioned us, don’t take on too much the first time you do this,” Shanahan said. “You have to go after a manageable problem, solve it, show early wins and then start to open Pandora’s box and go after all of these other challenges across the department.”

Right now, there are fragmented, one-off efforts across the military services to use this type of technology. Shanahan’s team is looking to change that.

“The Algorithmic Warfare team “will bring much-needed unity of effort and energy to bring artificial intelligence and machine learning to DoD,” he said.

When Shanahan talked to Defense One, the general was in Silicon Valley scouting for new technologies and searching for assistance in the high-end engineering community.

“There’s only so much of the high-end AI, machine-learning software engineering talent in the world,” Shanahan said. “Most of it happens to be [in Silicon Valley]. We’re looking for the very best. We need world-class.”

This type of technology ”is being unveiled almost in real time as these companies start going deeper and deeper into it,” he said. “But industry is light years ahead of the Department of Defense in these areas. Defense needs to catch up so we can give our analysts the tools they need to go after this.”

The Pentagon is looking to get the technology in place quickly so it can make an impact on the battlefield. Then it wants to look for ways to scale technology into other areas besides intelligence.

“It’s no good if it’s just shown and demonstrated in a research lab somewhere,” Shanahan said. “We want to deliver — in this calendar year — some capabilities out to the warfighters that make a difference. That’s probably the most aggressive schedule you’ll hear out of DoD in terms of delivering capabilities from concept to fielding something.”

Work laid out a three-phase effort, each of which will last 90 days. The first is to acquire data labeling algorithms. The second will acquire the hardware and software necessary to make it happen. And the third will put the technology into existing intelligence projects.

Shanahan sees different companies playing a role in the endeavor.

“There will be lots of different opportunities for companies to be involved in this,” Shanahan said.

“We see all sorts of things for intelligence, for targeting, for collection management, for sensor fusion. For the department … logistics, command and control, communications,” he added. “Everything that industry is working on has some applicability throughout the entire department.”

For more than a year, the Pentagon has been working a similar project that looks to use automation to find and track mobile missile launchers, like the ones used by North Korea. But that effort is largely using images from satellites, while the current one is centered around full-motion video from drones and aircraft.

“There’s some hard work behind the scenes that has to be done to first of all get that [video] cleaned up,” Shanahan said.

About 60 percent of the full-motion video gathered over the battlefield is benign while the drone or aircraft flies to and from a point of interest. Then there could be bad weather where the camera cannot see through the clouds.

The goal is to find technology that can clean up that video, “finding the juicy parts where there’s activity and then labeling the data,” Shanahan said. That work today is done by three-person teams of analysts.

“The feedback we’re getting from the analysts on the worker level is they’re excited about an opportunity,” he said. “These analysts are incredibly talented and they want to use their talents to perform high-end analysis, not administrative or mundane tasks.”

Right now, the Algorithmic Warfare cell needs to get its finances in order. Because the group has been stood up so quickly, without its own budget line, it needs Congress to approve a $70 million transfer of funds.

“The return on investment for this is so high compared to other weapon systems, we should have paid these bills a long time ago,” Shanahan said.

And the possibilities are endless.

“Once we show success, people are going to say what else can we apply this to,” he said. “To me that breaks things wide open and we’re going to figure out how we really, at scale, bring in some of these capabilities into the department.”