Discuss as:

DARPA wants smarter machines

US Military

The Pentagon is drowning a "data deluge" from drones such as the one pictured here and other intelligence gathering activities. It has put out a call for a mathematical language to make sense of it all.

The Pentagon's Defense Advanced Research Projects Agency has sharpened its focus on a future where machines do most of the dirty work — and a good deal of the thinking, too.

Last week, the military think tank kicked off a program called Mind's Eye, which is aimed at developing "a visual intelligence capability for unmanned systems." This week, it sent out an announcement that hints at the agency's plans for a mathematical language that would give data-collecting sensors the ability to speak to each other, think for themselves, and take action with scant human interference.


The Mind's Eye program essentially ups the ante for camera-equipped unmanned vehicles by giving them not just a processor for collecting visual information, but a "visual intelligence, enabling these platforms to detect operationally significant activity and report on that activity so warfighters can focus on important events in a timely manner." That would take boots off the ground and free data analysts from their chairs.

DARPA watchers have also pointed out an announcement calling for proposals to participate in the Mathematics of Sensing, Exploitation, and Execution program, or MSEE. The announcement says MSEE's mission is to find new ways to handle a "data deluge."

"The amount of data collected by Department of Defense sensor systems far outstrips the ability of both human analysts and current automated decision systems to extract actionable information," it reads. In other words, all the data coming in from drones and satellites taking pictures and making videos, plus tapped phone lines and who knows what else, is simply too much. Instead of being helpful in stopping the bad guys, it's a hindrance.

MSEE seeks a unified mathematical language that can teach sensors what data to collect as well as how to interpret and act on it. The agency's call for proposals says the goal of the program is "to capture the economy and efficiency that derives from an intrinsic, objective-driven unification of sensing and exploitation." To get there, "all proposed research must describe a unifying mathematical formalism that incorporates stochasticity fundamentally."

Wired.com's Spencer Ackerman notes that in about three and a half years, the agency wants prototypes to "furnish sensor output products" from imagery and video, communications intercepts and the tracking of a moving target.

"If your algorithm can train those very distinct sensors how to determine for themselves what relevant data is, you'll have gone a long way to draining oceans of data into a customizable kiddie pool for military analysts," he writes.


Skynet, anyone? Feel free to weigh in with your comments below.

John Roach is a contributing writer for msnbc.com. Connect with the Cosmic Log community by hitting the "like" button on the Cosmic Log Facebook page or following msnbc.com's science editor, Alan Boyle, on Twitter (@b0yle).