Wading through drone data is a 5,000-person job. Darpa wants you to help them change that. Illo by Arikia Millikan.
If living in a college dorm for two months while you’re prodded by government officials in a “short-fuse, crucible-style environment” is your idea of summer fun, the Defense Advanced Research Projects Agency would like a word with you. Oh, and there’s $50,000 to be earned if you can help them algorithmically crack the problems they’ve run into while trying to interpret the massive amounts of surveillance data the Pentagon is now capable of collecting.
Darpa, partnered with George Mason University, announced Tuesday that it is now accepting proposals for the Innovation House Study, a challenge that aims to attract top civilian geeks to attack the problem of efficiently wading through and extracting useful information (such as people, places, things and activities) from massive piles of visual and geospatial data.
If your typical hackathon is the equivalent of a sprint around the track, this challenge is like one of those hundred-mile death races. “This is an experiment in rapid, break-through development,” said Tod Levitt, principal investigator for Innovation House. “The goal is to speed radically innovative solutions obtained by radically decreasing the timeline allowed but opening the gate wide on the technology target.”
Levitt claims the topic of the Innovation House challenge doesn’t have anything to do with the Pentagon’s struggle to find relevant information in the loads of street-level video feeds, cellphone data, and drone data. But the military sure wouldn’t mind an elegant algorithmic solution to those issues. Currently, the Air Force alone employs more than 5,000 people just to monitor feeds from its surveillance drones.
The Innovation House challenge is the latest in a series of Darpa challenges that beckon civilians to lend their genius to the Pentagon. Previous challenges have pushed participants to solve problems with self-driving cars, create humanoid robots, re-assemble shredded paper into the original document, locate red weather balloons using social networking tools, and create the spy drone of the future.
The contests have had mixed success. Not one of the 140 teams that competed in the recent drone experiment were able to meet all the technical requirements of the competition. Then again, it took Darpa a couple of “Grand Challenges” before it got truly self-driving cars. Now, those autonomous automobiles are poised to become a roadside reality.
“A lot can happen when you put seriously intelligent, seriously motivated people in a room with a mission and a deadline,” Darpa program manager Michael Geertsen was quoted saying in the Innovation House press release. “We are inviting a new generation of innovators to try out ideas in an environment that encourages diverse solutions and far-out thinking. If this model proves to be as successful as we believe it could be, it represents a new means for participating in Government-sponsored research projects.”
For this challenge, Darpa will lodge a selected six to eight teams at George Mason University and provide them with an initial $10,000 for equipment and access to unclassified data sets including “ground-level video of human activity in both urban and rural environments; high-resolution wide-area LiDAR of urban and mountainous terrain, wide-area airborne full motion video; and unstructured amateur photos and videos, such as would be taken from an adversary’s cell phone.” However, participants are encouraged to use any open sourced, legal data sets they want. (In the hackathon spirit, we would encourage the consumption of massive quantities of pizza and Red Bull, too.)
There are two month-long phases of the challenge. After the first month, the teams will show off their design and software prototypes, and will be evaluated based on novelty, potential impact on revolutionizing technology and military operations in the field, and feasibility. If they’re up to par, they get an additional $20,000 and will enter the second phase where they will solidify their software designs. After the final phase, the teams will collect an additional $20,000 — if they agree to hand over all of their source code and documentation about how to use it and issue a Government Use Rights license. However, the teams will get to keep the commercial rights.
So if you want to help the Pentagon robotically plow through their loads of surveillance data, possibly to take the load off telecom employees, you have three weeks to submit a proposal. But you may not want to push this deadline, as they only anticipate looking at the first 100 proposals.
