In this video segment adapted from NOVA scienceNOW, meet Luis von Ahn, a computer scientist and a professor at Carnegie Mellon who is already at the top of his field at age thirty. Learn about one of his most successful ideas—CAPTCHA—a test that humans can pass but computers cannot, which has been used to improve the security of Internet sites. Explore how he comes up with his innovative ideas, and how CAPTCHAs have been reinvented to help digitize old books.
Visit the program page here.
You have probably solved at least one CAPTCHA—a distorted word image that users are asked to decode on Web sites, such as when registering a new email address or posting to a site. Solving a CAPTCHA is a relatively easy task for humans, who have the ability to recognize distorted letters and numbers. However, it is nearly impossible for a computer to figure out the answer. As a result, this innovative test allows valid users access to Web services, while protecting the services from being abused by automated computer programs designed by spammers with malicious intent.
Although perhaps best known for developing CAPTCHAs, the computer scientist Luis von Ahn has many other interesting ideas. For example, he has also developed "Games With A Purpose" (GWAP)—computer games that humans play for fun but that also accomplish useful computing tasks. Since people tend to spend quite a bit of time playing video games, Luis von Ahn wanted to try to harness all that "wasted" time to produce something beneficial for society. The games that he develops serve the purpose of training computers and improving information content on the Internet. For example, in his ESP game, two users are randomly paired together and shown the same image. The goal of the game is to describe the picture using words that the other player also chooses. When both players type the same word, they earn points and get a new image. This game helps label pictures with appropriate descriptive words for use in search engines. The game has been licensed to Google (as Google Image Labeler) and is currently being used to improve the quality of Google's image search results.
There are a number of GWAPs that are similar in purpose to the ESP game. In Tag a Tune, players describe what they hear in an audio recording; in Popvideo, players describe what they see and hear in video clips; and in Verbosity, players work together to collect common sense descriptions of words. Other games are meant to train computers in a more abstract sense. For example, in Matchin, players rate images according to attractiveness, to teach computers what humans view as beautiful. In Squigl, players outline objects in images—this program is being used to improve computer vision.
When thinking of ways to make use of the time millions of people spend online, Luis von Ahn has certainly conceived of some exciting projects. Of course, he fully admits that not all of his ideas are good. In fact, he claims that the vast majority of his ideas are bad, especially since he comes up with so many ideas every day. He recognizes good ideas by letting them stew for a while. If he's still thinking about an idea a few months later, then maybe it actually is a good one. And if it's a good idea, maybe it's worth working on it some more to make it a great idea.
© 2014 PBS & WGBH Educational Foundation. All rights reserved.