Annual Meeting Reports

Plenary Address: Human Computation

It can be annoying. You want to log onto a Web site or get ready to purchase an item online, but before you can proceed, you are instructed to type specified funny “words”. The scientist behind these words is Luis von Ahn, an assistant professor of computer science at Carnegie Mellon University, whose interests include using humans “cheaply” to address problems that computers cannot solve.

“The inscription you type before you proceed to a desired Web page is called a CAPTCHA,” said von Ahn, who invented CAPTCHAs. “Computer programs cannot identify CAPTCHAs, so they are used to verify that only humans are using certain Web sites.” von Ahn said that people usually type CAPTCHAs as part of security measures on Web sites to allow only humans access, but he noted that spam companies have employed humans to type CAPTCHAs and thus circumvent this measure. “CAPTCHAs are generating jobs for some people,” he joked. “Such people earn $2.50 per hour.”

von Ahn said that many people spend hours playing computer and video games and that more than 200 million CAPTCHAs are typed every day by people around the world. To transform this typing into useful work, he has introduced the program reCAPTCHA (www.recaptcha.net) to digitize printed materials. von Ahn said that his project gives CAPTCHAs free to Yahoo, Twitter, Facebook, and many other online companies. The CAPTCHAs originate from print sources to be digitized. He explained that words in older publications often cannot be recognized accurately by computerized optical character-recognition systems. Such words are presented as CAPTCHAs. When a CAPTCHA is consistently translated into a particular word, that word is assumed to be the original word. von Ahn said that through the use of CAPTCHAs, the digitization of The New York Times published from 1851 through 1980 would be completed in 2009.

von Ahn also discussed computer projects that use humans to label images on the Web. He termed them “games with a purpose” (GWAPs). GWAPs have been applied in such fields as security, computer vision, Internet accessibility, adult-content filtering, and Internet searching. One such game, the ESP Game (www.espgame.org), shows how humans, as they play, can solve problems that computers cannot yet solve. The ESP Game randomly pairs players. Players do not know each other’s identity. The partners see the same image at the same time. Each partner guesses what label the other would give to the image. When both partners type the same words, they earn points, and another image is presented. Some images have “taboo words” that players must avoid. When an image has more taboo words, a series of matches leads to more points. During the game, partners must learn to think the same. It has been found that the series of words on which the two players agree are usually good labels for an image. von Ahn said that more than 200,000 players have made 50 million agreements in the ESP Game. He noted that many people play more than 20 hours a week.

Although people may be playing the game for fun, its importance is huge. “5,000 people playing simultaneously can label all the images on Google in 2 months!” von Ahn said. He also described a similar game called Audio ESP, which encourages partners to listen to sounds and type what they hear.

von Ahn said his interests in computing include treating human brains as processors in a distributed system, each performing a small part of a massive computation. He therefore uses online games to encourage participation in the process. The plenary address indeed showcased a new era of human computing and its use in Internet security.