CAPTCHA Creator Explains His Next Project, Duolingo: Translate the Web, Learn a Foreign Language

My fellow tech blogger MG Siegler is known for making some pretty bold predictions. And his headline to the story introducing Duolingo, a startup that hasn't even launched yet, seems like no exception: "Meet Duolingo, Google's Next Acquisition Target."

What Duolingo does is still not 100% clear, but even so, MG's statement here doesn't feel too far out. After all, Duolingo is the latest project by Carnegie Mellon University professor Luis von Ahn, with two acquisitions of his research by Google already under his belt: the ESP Game and reCAPTCHA.

Duolingo is very much in the same spirit as both these projects in so far as it utilizes human microtasks "for good" -- in order to help identify and verify information (images in the first case, digitized text in the second) that computers cannot yet process.

We creatures of the Web fill out a lot of CAPTCHAs, and some 750 million users have helped Google's book digitization efforts by using reCAPTCHA to help correct OCR misreads. Now, with Duolingo, von Ahn hopes to be able to use a similar sort labor in the wonderful and mammoth goal of translating the Web into all major languages.

Unlike reCAPTCHAs, which we tend to approach with a certain annoyance, people are unlikely to scowl at Duolingo. That's because Duolingo is also a language learning platform, one that von Ahn boasts seems to show results as good as expensive software packages. Learn a language as you translate the Web.

Duolingo is taking aim at those expensive language learning services -- indeed, it's a huge market. But von Ahn says that he wants to offer a "fair business model for education" -- so Duolingo will be free.

More details are sketchy, and Duolingo hasn't even opened to beta yet. It's not clear what the language learning will look like, and it's not clear if publishers like The New York Times or Wikipedia will have to pony up to have their sites translated. Clearly paying for professionals to translate the Web would be cost-prohibitive. von Ahn estimates it would cost about $50 million to translate the English-language portions of Wikipedia into Spanish, for example.

But with 1 million people using Duolingo, he contends that translation would only take about 80 hours. That's definitely worth something. Right, Google?

von Ahn's recent TEDx talk is now available. It's definitely worth a watch.

Image credits/TEDxCMU: News Grange


Tags: , , , , , , , , , , ,