How AI Will Predict Chinese and Russian Moves in the Pacific

HONOLULU—On the site of the most infamous sneak attack in American history, U.S. Pacific Air Forces is collating tens of millions of radar contacts and other data in a bid to stave off a latter-day surprise — and even reveal the adversary’s weaknesses.

Airmen and researchers at PACAF’s Pearl Harbor headquarters are using the data — as old as a year and as new as real-time — to draw up a portrait of normal air traffic in the vast Pacific region. Ultimately, that should make it easier to spot abnormal events, such as an impending attack, the deputy chief of PACAF’s C3 Integration Division said at the Defense One-Nextgov Genius Machines event here last Tuesday.

“If you’ve got six months, eight months, a year's worth of data, you start to understand what the pattern looks like,” Lt. Col. Ryan Raber said. “Here’s what I know is ‘normal.’ Then we start to pick out the data points that are abnormal. What does abnormal look like? And then we start to focus on those and figure out what they mean to us. Is that adversary aircraft preparing for something? Are they just off their air route? What’s going on with that specific data?”

The goal is to use artificial intelligence to compress a process that now takes days into just minutes. 

“So, if someone is going to take a potshot at us, we want to be in front of their decision cycle to know where we need to be to best use every capability that we have to stop that from happening before it happens.”

Raber didn’t mention who that someone might be. But in recent years, the Chinese and Russian militaries have been expanding their presence in PACAF’s area of operation and growing closer to each other as well. 

China has also become more aggressive in its air activities in the region as it builds and arms artificial islands in the disputed South China Sea that could threaten U.S. pilots and ships. 

PACAF says their efforts will eventually help individual pilots and operators to make minute-by-minute decisions about whether the behavior they are seeing from adversary aircraft implies a potential attack.

“We are also looking to develop the capability for the ‘AI wingman’ to look back months and years into the past to show the operator what happened the last time that the anomalous behavior was detected. Based on past data, current intelligence, and other factors, the AI will make recommendations or predict behavior to enable superior decision-making,” Raber said in a later statement. “Overall, we want the AI wingman to provide advice based on experience and long-term looks to enable operational art by the warfighter/decision maker.”

Robert Mohan, an automated-decision-support analyst at U.S. Indo-Pacific Command, is working with DARPA on a slightly different project. Called Causal Exploration, it aims to build “a tool that uses artificial intelligence to enhance what machines do best in terms of researching large amounts of data and pattern recognition…and combining that with the human and allowing the human to do what he does best in terms of coming up with new ideas,” Mohan said.

The program seeks to build a “modeling platform to aid military planners in understanding and addressing underlying causal factors that drive complex conflict situations,” according to a writeup on the DARPA website.

Indo-Pacific Command is also using artificial intelligence to give commanders a better sense of what’s happening across the theatre based on all the publicly available information that can be gleaned, structured, and analyzed. The operation is similar to the 2012 IARPA program called Open Source Indicators, but larger in scale, Mohan said. It’s the sort of data analysis and observation that would have been very difficult to do on a continual basis a few years ago, but is now possible because of the availability of massive amounts of compute and storage power through enterprise cloud capabilities, such as Microsoft’s Azure program, Google’s Cloud, or Amazon’s AWS services, which is what Mohan is using for his program. He says they currently have an unclassified and classified version of the program and are looking to set up a top-secret version soon.

 “We’ve got an intelligence feed that’s publicly available information. They bring in over 2,000 separate sources of information right now. It will grow to 10,000 feeds of data within a year. One of those feeds of data is Twitter, with 500 million tweets a day. That’s a lot of information,” he said. “It’s difficult to discern truth from fiction; that’s where the AI comes in.”

Indo-Pacific Command's and PACAF’s efforts may also reveal weaknesses in the enemy’s defenses — or Raber put it, find the adversary's “catastrophic endpoints.” Those could be vulnerabilities in radar coverage, software vulnerabilities in critical defenses, gaps in maneuvering, or other holes that the enemy might not be aware of but that might reveal themselves in the mountains of radar and other data that the military is working with.

Both Mohan and Raber said it’s hard to woo experts in the field to Hawaii. “The problem I cannot solve with money is the workforce,” Mohan said. “I have three openings now open more than 60 days I cannot fill.”

Pentagon Accelerating AI Efforts

Back in Washington, D.C., the Pentagon’s weeks-old Joint Artificial Intelligence Center, or JAIC, will soon launch three projects, including one for maneuvering and fires.

“The project will focus on individual lines of effort or product lines oriented on warfighting operations like operations/intelligence fusion, joint all-domain command and control, accelerated sensor to shooter timelines, autonomous and swarming systems, target development and operations center workflows,” Lt. Gen. Jack Shanahan, JAIC’s director, told reporters  

Read that to mean applying AI to virtually everything that the military does in combat, from identifying what targets to hit to figuring out the best weapon to use and how to put that weapon in position. 

The Center will also look to apply AI to t better predict when soldiers might face medical problems, including PTSD, a project that involves the Department’s Defense Innovation Unit. A third project will seek to build a common library of tools and data that military operators and units can use to build their own AI tools. Shanahan said this Joint Common Foundation already exists, though in a primordial, not fully-usable form. 

“It will be a platform that will provide access to data, tools, environment, libraries and other certified platforms to enable software and AI engineers to rapidly develop, evaluate, test and deploy AI-enabled solutions to warfighters,” he said. “It is designed to lower the barriers to entry, democratize access to data, eliminate duplicative efforts and increase value added to the department. This platform will reside on top of an enterprise cloud structure.” 

The Pentagon is requesting $268 million for the JAIC in 2020, more than double the $93 million it got last year.