The founder of ReCAPTCHA, Prof Luis voh Ahn demonstrates how computing time can be harnessed effectively
It may come as a surprise to some that whenever you decipher scrambled words – known as captchas – before accessing sites, you are doing much more than merely authenticating that you’re a human being and not a computer bot.
When you make sense of a “Completely Automated Public Turing Test To Tell Computers and Humans Apart” or captcha, you are also helping online search giant Google digitise books and newspapers.
Books are digitised using the optical character recognition (OCR) software. But OCR cannot understand all words. These words appear gibberish and the software fails to understand them, but we as humans can decipher these words. While humans can read distorted text, computers are still far away from doing so.
The word captcha was coined in 2000 by Prof Luis von Ahn and his colleagues from Carnegie Mellon University (CMU).
“About 200 million captchas are solved by humans around the world every day. In each case, roughly 10 seconds of human time are being spent. Individually, that’s not a lot of time, but in aggregate these little puzzles consume more than 150,000 work hours every day. reCAPTCHA channels the effort spent solving captcha online into “reading” books. We showed how this time could be used for something else,” said Ahn.
Born in Guatemala, Ahn is both a professor in the Computer Science department at CMU and an entrepreneur. He continues to lead the team inside Google after the online search giant acquired his company reCAPTCHA last September.
“I first got the idea (for captchas) some nine years back when I saw people solving crosswords on a plane. At that time, computers were unable to solve crosswords. I wondered, then, if I could create something which could make use of this human skill and time. I now build systems that combine human and computer intelligence to solve large-scale problems that neither can solve alone. I call this Human Computation. Others call it crowdsourcing. For the last eight years, I have been working to develop this area,” explained Ahn — one of the pioneers of the idea of crowdsourcing, wherein large number of people are enticed to perform work over the Internet.
After this field was developed, Amazon created Mechanical Turk, a marketplace for human computation tasks (or “human intelligence tasks” as they call them), which is now used and studied by hundreds of researchers worldwide. Since then, similar services have emerged where workers are paid to perform micro-tasks that are hard for computers.
Ahn’s work on “human computation” has earned him international recognition and numerous honours. He was awarded the MacArthur Fellowship in 2006; the David and Lucile Packard Foundation Fellowship in 2009; a Sloan Fellowship in 2009; and a Microsoft New Faculty Fellowship in 2007.
Ahn believes in making this entire exercise “fun-packed”. He and his team created the ESP Game (licensed to Google and called the Image Labeler) and Peekaboom — known as Games with a Purpose (GWAP).
“When you play a game at Gwap, you aren’t just having fun. You’re helping the world becoming a better place. By playing our games, you’re training computers to solve problems for humans all over the world,” said Ahn.
Thousands of Internet users, till date, have been lured into tagging photos online — doing it so effectively that Google bought Ahn’s idea (ESP has been licensed to Google) last year to improve its Image Search engine. The Image Labeler is a feature of Google Search that allows you to label images and help improve the quality of Google’s image search results. Players are randomly paired with a partner who’s online and using the feature.
Over a two-minute period, you and your partner will be able to see the same set of images; provide as many labels as possible to describe each image you see; and receive points when your label matches your partner’s label. The number of points would depend on how specific your label is. You will see more images until time runs out. After the time expires, you can explore the images you’ve seen and the websites where those images were found. And you will see the points you’ve earned through the session.
But computers are fast learners too and soon captchas will be solved by bots. In fact, there have been some reported breaches. Ahn agreed, but believed the scope of human computation will extend for many years to come. “The human brain is very complex. Computers will take at least another 50 years to match the human brain. Till then, we can put that time to good use.”
Going forward, he is “trying to do something on the language-translation front”.
(The author, on a sabbatical from Business Standard, is an MIT Knight Science Journalism Research Fellow 2010-11)
