4. The Masses 4.大众
In the late 1760s, a Hungarian nobleman named Wolfgang von Kempelen built the first machine capable of beating a human at chess. Called the Turk, von Kempelen’s automaton consisted of a small wooden cabinet, a chessboard, and the torso of a turbaned mannequin. The Turk toured Europe to great acclaim, even besting such luminaries as Benjamin Franklin and Napoleon. It was, of course, a hoax. The cabinet hid a flesh-and-blood chess master. The Turk was a fancy-looking piece of technology that was really powered by human intelligence. Which explains why Amazon.com has named its new crowdsourcing engine after von Kempelen’s contraption. Amazon Mechanical Turk is a Web-based marketplace that helps companies find people to perform tasks computers are generally lousy at – identifying items in a photograph, skimming real estate documents to find identifying information, writing short product descriptions, transcribing podcasts. Amazon calls the tasks HITs (human intelligence tasks); they’re designed to require very little time, and consequently they offer very little compensation – most from a few cents to a few dollars. 在18世纪60年代的晚期,一名叫做Wolfgang von Kempelen的匈牙利贵族发明了第一台能够在象棋上击败人类的机器。这台被称为Turk的von Kempelen的机器人，由一个木盒子，一块棋版，以及一个假人的部分驱赶所组成。Turk在欧洲广受追捧，甚至与本杰明.弗兰克林和拿破仑齐名。当然，这是一场闹剧。这个木盒子里面藏着一名活生生的象棋高手。Turk是一项真正由人类智慧所控制的还不错的技术。这解释了为什么Amazon.com将其众包引擎以von Kempelen的奇妙装置命名。Amazon的机器人Turk是一个基于Web的市场，它能够帮助企业找到人群后执行计算机一般不怎么在行的任务，如：在一张照片中发现细节，浏览房地产文档以发现与众不同的信息，写作精悍的产品描述，抄写播客。Amazon把这些任务称为“人类智能任务”（HIT，human intelligence tasks）；它们的设计只需很少的时间，因此花费很少 - 大多数只需几分钱到几美元不等。
InnoCentive and iStockphoto are labor markets for specialized talents, but just about anyone possessing basic literacy can find something to do on Mechanical Turk. It’s crowdsourcing for the masses. So far, the program has a mixed track record: After an initial burst of activity, the amount of work available from requesters – companies offering work on the site – has dropped significantly. “It’s gotten a little gimpy,” says Alan Hatcher, founder of Turker Nation, a community forum. “No one’s come up with the killer app yet.” And not all of the Turkers are human: Some would-be workers use software as a shortcut to complete the tasks, but the quality suffers. “I think half of the people signed up are trying to pull a scam,” says one requester who asked not to be identified. “There really needs to be a way to kick people off the island.”
Peter Cohen, the program’s director, acknowledges that Mechanical Turk, launched in beta in November, is a work in progress. (Amazon refuses to give a date for its official launch.) “This is a very new idea, and it’s going to take some time for people to wrap their heads around it,” Cohen says. “We’re at the tippy-top of the iceberg.”
A few companies, however, are already taking full advantage of the Turkers. Sunny Gupta runs a software company called iConclude just outside Seattle. The firm creates programs that streamline tech support tasks for large companies, like Alaska Airlines. The basic unit of iConclude’s product is the repair flow, a set of steps a tech support worker should take to resolve a problem.
Most problems that iConclude’s software addresses aren’t complicated or time-consuming, Gupta explains. But only people with experience in Java and Microsoft systems have the knowledge required to write these repair flows. Finding and hiring them is a big and expensive challenge. “We had been outsourcing the writing of our repair flows to a firm in Boise, Idaho,” he says from a small office overlooking a Tully’s Coffee. “We were paying $2,000 for each one.”
As soon as Gupta heard about Mechanical Turk, he suspected he could use it to find people with the sort of tech support background he needed. After a couple of test runs, iConclude was able to identify about 80 qualified Turkers, all of whom were eager to work on iConclude’s HITs. “Two of them had quit their jobs to raise their kids,” Gupta says. “They might have been making six figures in their previous lives, but now they were happy just to put their skills to some use.”
Gupta turns his laptop around to show me a flowchart on his screen. “This is what we were paying $2,000 for. But this one,” he says, “was authored by one of our Turkers.” I ask how much he paid. His answer: “Five dollars.”
Contributing editor Jeff Howe (jeff_howe@wiredmag.com) wrote about MySpace in issue 13.11. To read more about crowdsourcing, please visit Jeff Howe’s blog on the subject.
