reCAPTCHA Founder’s Duolingo Teaches Languages While Crowdsourcing Translation
Luis von Ahn, the crowdsourcing pioneer who developed reCAPTCHA, has come up with the perfect crowdsourcing project.
His new company, Duolingo, teaches people how to speak languages by reading, writing, speaking and listening to translations of Web content. The byproduct of that activity, if all goes well, will be a fast and accurate human-powered translation engine.
By contrast, reCAPTCHA digitizes books word by word when Web users are asked to prove they are humans and not bots. Participants get value out of unlocking Web sites, but they’re not learning anything useful besides how to squint at squiggly letters really hard.
ReCAPTCHA has been incredibly successful; it was bought by Google and now digitizes 100 million words per day, and more than one billion people have used it.
Duolingo is “much more ambitious,” von Ahn said in an interview Monday. “Here we have to create a very compelling Web site for people to come back to. In the case of reCAPTCHA, people were forced to use it.”
But the opportunity for translation is much larger than book transcription, von Ahn said — something like $30 billion annually worldwide, he estimated.
And, meanwhile, online language learning is expensive and generally teaches in a textbook environment lacking real-world context.
Duolingo learns as it teaches. For language learners, it personalizes: The system tracks words that users don’t know and finds other instances of them. For translations, it sends the same sentences to multiple students until they start to agree.
The site isn’t just a random pile of text to translate. Each language has a basic curriculum, with lessons on grammatical topics and vocabulary, as well as gamification and social features to reward progress and daily participation.
In beta testing since late last year, 125,000 active Duolingo users have translated 75 million English, Spanish, French and German sentences. Von Ahn said the single most active student has spent 650 hours on the site — the equivalent of a full-time job — while many have spent more than 100 hours.
Duolingo is free, and it doesn’t have ads. That’s because language learners are von Ahn’s free translation labor. The company will sell translations, and is working on a system for content owners to input raw translation materials.
Von Ahn — who is still a teaching professor at Carnegie Mellon — said he knows that translation can be more art than science, but he’s not concerned. His goal is to pick off the best automated systems, like Google Translate.
“What we want, at the end of the day, is something that’s correct and reads well,” he said.
Von Ahn also acknowledged that true fluency is best achieved by moving to a place where people speak a language every day.
But at scale, Duolingo could help take down language and geographic barriers. “If everybody’s using this, I think it could really bring the whole world together a little better,” von Ahn said.
The company raised $3.3 million in Series A funding last fall from Union Square Ventures, Ashton Kutcher’s A-Grade Investments and Tim Ferriss. It is based in Pittsburgh and has 13 employees.
In beta testing, the most common user request was for mobile versions, von Ahn said. So expect those sometime soon, along with expansions to Portuguese, Chinese and other languages.