finding things with crowds -- 5/5/15
Today's selection --from The Organized Mind by Daniel J. Levitin. Accomplishing the impossible through crowdsourcing:
"Crowdsourcing -- outsourcing to a crowd -- [is] the technique by which thousands or even millions of people help to solve problems that would be difficult or impossible to solve any other way. Crowdsourcing has been used for all kinds of things, including wildlife and bird counts, providing usage examples and quotes to the editors of the Oxford English Dictionary, and helping to decipher ambiguous text. ...
"In December 2009, DARPA offered $40,000 to anyone who could locate ten balloons that they had placed in plain sight around the continental United States. DARPA is the Defense Advanced Research Projects Agency, an organization under the U.S. Department of Defense. DARPA created the Internet (more precisely, they designed and built the first computer network, ARPANET, on which the current World Wide Web is modeled). At issue was how the United States might solve large-scale problems of national security and defense, and to test the country's capacity for mobilization during times of urgent crisis. Replace 'balloons' with 'dirty bombs' or other explosives, and the relevance of the problem is clear.
"On a predesignated day, DARPA hid ten large, red weather balloons, eight feet in diameter, in various places around the country. The $40,000 prize would be awarded to the first person or team anywhere in the world who could correctly identify the precise location of all ten balloons. When the contest was first announced, experts pointed out that the problem would be impossible to solve using traditional intelligence-gathering techniques.
"There was great speculation in the scientific community about how the problem would be solved -- for weeks, it filled up lunchroom chatter at universities and research labs around the world. Most assumed the winning team would use satellite imagery, but that's where the problem gets tricky. How would they divide up the United States into surveillable sections with a high-enough resolution to spot the balloons, but still be able to navigate the enormous number of photographs quickly? Would the satellite images be analyzed by rooms full of humans, or would the winning team perfect a computer-vision algorithm for distinguishing the red balloons from other balloons and from other round, red objects that were not the target? (Effectively solving the Where's Waldo? problem, something that computer programs couldn't do until 2011.)
"Further speculation revolved around the use of reconnaissance planes, telescopes, sonar, and radar. And what about spectrograms, chemical sensors, lasers? Tom Tombrello, physics professor at Caltech, favored a sneaky approach: 'I would have figured out a way to get to the balloons before they were launched, and planted GPS tracking devices on them. Then finding them is trivial'.
"The contest was entered by 53 teams totaling 4,300 volunteers. The winning team, a group of researchers from MIT, solved the problem in just under nine hours. How did they do it? Not via the kinds of high-tech satellite imaging or reconnaissance that many imagined, but -- as you may have guessed -- by constructing a massive, ad hoc social network of collaborators and spotters -- in short, by crowdsourcing. The MIT team allocated $4,000 to finding each balloon. If you happened to spot the balloon in your neighborhood and provided them with the correct location, you'd get $2,000. If a friend of yours whom you recruited found it, your friend would get the $2,000 and you'd get $1,000 simply for encouraging your friend to join the effort. If a friend of your friend found the balloon, you'd get $500 for this third-level referral, and so on. The likelihood of anyone person spotting a balloon is infinitesimally small. But if everyone you know recruits everyone they know, and each of them recruits everyone they know, you build a network of eyes on the ground that theoretically can cover the entire country. One of the interesting questions that social networking engineers and Department of Defense workers had wondered about is how many people it would take to cover the entire country in the event of a real national emergency, such as searching for an errant nuclear weapon. In the case of the DARPA balloons, it required only 4,665 people and fewer than nine hours."