Humans, computers and the quest for Sensing as a Service

The Air Force Research Laboratory is working to break the “linear relationship” that makes analyzing ISR data so time-consuming and manpower-intensive, though it’s going to require a significant leap in computing technology.

That was part of the message Monday from Dr. Steve “Cap” Rogers, a senior scientist at AFRL and the Air Force’s principle scientific authority for automatic target recognition and sensor fusion. Rogers spoke during the opening day of the GEOINT 2015 Symposium in Washington, discussing the lab’s efforts toward what’s called Sensors as a Service, or SaaS.

The Air Force, like all of the military services, collects mountains of ISR—intelligence, surveillance and reconnaissance—data from radars, cameras and other sensors aboard unmanned vehicles and other platforms. But when it comes to most of the analysis of that data, “the state of the art is people—throwing people at the problem,” he said during a keynote.

AFRL is working toward sensor fusion, integrated ISR and SaaS through programs such as Qualia Exploitation of Sensor Technology, or QuEST, which is researching the possibilities of an artificially conscious computer, Rogers said. As new technologies come along, AFRL will test them with its development and demonstration program called Planning & Direction, Collection, Processing & Exploitation, Analysis & Production, and Dissemination Experimentation, or PCPADx.

“And we’re assuming major technology breakthroughs by you guys,” he said the roomful of attendees.

One key to an integrated ISR system will be matching what computers are capable of doing with what humans can do better and coming up with the right processes for working together, he said. For example, after IBM’s Deep Blue beat world chess champion Garry Kasparov in chess in 1997, many people assumed that a milestone had been passed and that question of whether computers were now officially “smarter” than humans was settled. Then the website Playchess.com held a tournament, inviting people to play against computers and, surprisingly, several mid-level players, rather than grand masters, emerged victorious. It’s not always a matter of the best technology or the best person, but whether the right process is in place, Rogers said. “If you don’t have the right process, you lose.”

Rogers applied that principle when the company he started after he retired from the Air Force took on a problem with breast cancer screenings—specifically, that many of the signs of breast cancer were easy to miss when examining readouts with the naked eye. His company developed and patented what’s now considered the most accurate detection systems by focusing that, while humans are better at recognizing some indicators, computers are much better at detecting others.

Automation is one thing—just about any repeatable process can be automated. But autonomy is something else, and the gap between the two is still large. Rogers said that right now researchers are good at automating products and are gradually getting there on tracking. Target recognition, however—in which a program could take sensor data and reliably identify potential targets—is still a long way off, said Rogers, who described that kind of research as “job security.”