30-03-2010, 07:14 AM
Pentagon: Replace Human Intel With High-Tech ‘Guard Dog’
![[Image: armymil-2007-10-17-090657-589x395-custom.jpg]](http://www.wired.com/images_blogs/dangerroom/2010/03/armymil-2007-10-17-090657-589x395-custom.jpg)
U.S troops operating overseas face insurgent threats and affiliations that are constantly changing. Not to mention the language barriers and cultural differences that can make even minor interactions — let alone intelligence and interrogation — more difficult.
Now Darpa, the Pentagon’s blue-sky research arm, wants to develop a foolproof system that analyzes social networks and cultural tendencies using graphs, complex algorithms and new advances in computing, to interpret and predict human actions.
The agency is hosting a proposal workshop for Graph Understanding and Analysis for Rapid Detection - Deployed on the Ground (priceless acronym: GUARD-DOG). Ideally, Darpa wants a replacement for current war-zone human intelligence, called HUMINT, which involves putting trained interrogators on the ground, identifying and tracking sources, and compiling data on relevant social networks. HUMINT is effective, but it can be dogged by slow turnaround: As Darpa notes, the lag between data collection and analysis can be 48 hours. And that means more than 80 percent of information may be irrelevant by the time troops take action.
A computerized intel analysis system, however, could rapidly grasp the size and complexity of the “human terrain,” and create new scenarios based on constantly-updated inputs. The real-world social networks in which troops operate have thousands of variables: people, locations, social affiliations, and organizations, to name a few. Spotting one small, hard-to-detect change in that landscape can be significant.
And Darpa wants more than just mind-bendingly fast analysis: the new programs should also be able to fill in the blanks. “Real-world social networks are likely to contain conflicting information and have missing data,” the agency’s proposal reads. “Patrols are also likely to be given false or misleading information.” So where human intel collectors might not pick up on inconsistencies, algorithmic interpreters somehow will. (Good luck with that — Ed.)
Having real-time access to the ins-and-outs of communities, whether friendly or hostile — not to mention accurate predictions of how those communities are apt to evolve — would be invaluable to troops operating among foreign cultures. Not to mention that it might teach them how to win friends and influence people.
“GUARD DOG will provid[e] dismounted soldiers with real-time assessments of the human networks relevant to their local battlespace, including threats, vulnerabilities, and uncertainties; and cues on engaging the people they encounter.”
[Photo: U.S Army]
Tags: DarpaWatch
Read More http://www.wired.com/dangerroom/2010/03/...z0jdbYWOTY
- By Katie Drummond
![[Image: envelope.gif]](http://www.wired.com/dangerroom/wp-content/themes/wired/images/envelope.gif)
- March 29, 2010 |
- 2:15 pm |
- Categories: DarpaWatch
U.S troops operating overseas face insurgent threats and affiliations that are constantly changing. Not to mention the language barriers and cultural differences that can make even minor interactions — let alone intelligence and interrogation — more difficult.
Now Darpa, the Pentagon’s blue-sky research arm, wants to develop a foolproof system that analyzes social networks and cultural tendencies using graphs, complex algorithms and new advances in computing, to interpret and predict human actions.
The agency is hosting a proposal workshop for Graph Understanding and Analysis for Rapid Detection - Deployed on the Ground (priceless acronym: GUARD-DOG). Ideally, Darpa wants a replacement for current war-zone human intelligence, called HUMINT, which involves putting trained interrogators on the ground, identifying and tracking sources, and compiling data on relevant social networks. HUMINT is effective, but it can be dogged by slow turnaround: As Darpa notes, the lag between data collection and analysis can be 48 hours. And that means more than 80 percent of information may be irrelevant by the time troops take action.
A computerized intel analysis system, however, could rapidly grasp the size and complexity of the “human terrain,” and create new scenarios based on constantly-updated inputs. The real-world social networks in which troops operate have thousands of variables: people, locations, social affiliations, and organizations, to name a few. Spotting one small, hard-to-detect change in that landscape can be significant.
And Darpa wants more than just mind-bendingly fast analysis: the new programs should also be able to fill in the blanks. “Real-world social networks are likely to contain conflicting information and have missing data,” the agency’s proposal reads. “Patrols are also likely to be given false or misleading information.” So where human intel collectors might not pick up on inconsistencies, algorithmic interpreters somehow will. (Good luck with that — Ed.)
Having real-time access to the ins-and-outs of communities, whether friendly or hostile — not to mention accurate predictions of how those communities are apt to evolve — would be invaluable to troops operating among foreign cultures. Not to mention that it might teach them how to win friends and influence people.
“GUARD DOG will provid[e] dismounted soldiers with real-time assessments of the human networks relevant to their local battlespace, including threats, vulnerabilities, and uncertainties; and cues on engaging the people they encounter.”
[Photo: U.S Army]
Tags: DarpaWatch
Read More http://www.wired.com/dangerroom/2010/03/...z0jdbYWOTY
"Where is the intersection between the world's deep hunger and your deep gladness?"
