Duolingo founder and CEO Luis von Ahn is an associate professor in the Carnegie Mellon University Computer Science Department, and was one of the original developers behind reCAPTCHA. Google acquired ReCAPTCHA in 2009 for "an undisclosed sum," a bit of history that led TechCrunch to speculate back in 2011 that Google would buy Duolingo within six months -- which didn't happen. But don't despair. It's still possible that Google (or another big company) might absorb Duolingo. We'll just have to wait and see -- and possibly improve our foreign language skills while we wait. (Alternate Video Link)
What is Duolingo? We are a sort of massive crowdsourced approach to language learning. Importantly, we try to be free and accessible to as large an audience as possible. We offer languages – we offer learning between many popular languages, English to many European languages—Spanish, Italian, French, German, Portuguese, with more on the way. Conversely, we do a lot of languages back to English. So the same languages can now go back to English. We have Russian to English, Chinese to English, Japanese to English, Korean to English; we have many other things that go back to English.
Tim: What are the most popular languages and what direction is it mostly?
Franklin: Spanish, English to Spanish and Spanish to English seems to be the most popular. Part of that is probably because our founder Luis grew up in Guatemala. He had a lot of influence there and he knows whom to contact I guess. I don’t know – but we do have a lot of Spanish users. I guess, next is probably Chinese to English, it is very big. I mean that’s sort of to be expected when it’s just such a big market in China.
Tim: More than a billion people.
Franklin: Yeah. More than – yeah.
Tim: When you say you want the language learning here to be accessible to a lot of people
Tim: Explain that a little bit—in what way?
Franklin: So first we try to be free. Unlike some other alternatives which charge hundreds of dollars, our whole goal is to try and be free. Again our founder Luis, who grew up in Guatemala, he knows about the hard life people have to live and how painful it can be when you don’t have access to language education. Especially because once you do have access a lot of doors start opening up. A lot more opportunities appear. So we really want to try and just increase the amount of people we can reach. And part of the ways we do that is by adding support from our languages. We started with some of the bigger popular languages, but we’re rapidly working on adding support for other languages that people are requesting. And the other way we do that is to try and be on many different platforms and support any different way people have with learning.
Tim: And the translations themselves are crowdsourced—is that correct?
Franklin: Yes. So as part of the learning experience you go through and you do we have a course that we make for each of these directions. They consist of little lessons, bite-sized chunks of knowledge, that are usually focused on something. So we have one that’s focused on plurals, we have one for food, animals and that sort of thing. Originally, we had a language team dedicated to making these things. But then we quickly realized that this doesn’t approach, this scale does not approach, this approach does not scale, once you start adding more and more languages.
Tim: Why don’t you back up and just say, “We realize”?
Franklin: Okay. Where should I start?
Tim: I can just ask the same question about crowdsourcing you got that right? I’ll probably rephrase it myself: Franklin, one of the ways that this is made open and free is in part through outsourcing of the exact translations, is that right?
Franklin: Yes. Our courses are made up of little skills that you go through and the skills focus on different subjects or content or just sort of sentence structure that sort of thing. So we have ones for verbs, we have some for plurals, animals, food, also sort of different things. And in those what happens is we originally had specific people, language experts who would come in and they would think about what skills we want people to learn. And then they would craft the sentences themselves. But we realized that approach doesn’t scale well, once you start adding dozens of languages. Because suddenly with every new language, it’s like an N-squared thing—that gets pretty big pretty fast. So we started developing what we call the Incubator. You can reach it at incubator.duolingo.com and there you can apply to be a moderator. Basically, it’s a volunteer driven system, where you can add new sentences, you can tweak existing sentences, and you can even work on building a new course if you want.
Tim: Very crowdsourced.
Franklin: Very crowdsourced. So you can add any language you can imagine. I think we have applicants for Dothraki, Klingon and Elvish—I can’t speak for the progress on those. You have to check it out.
Tim: For the accuracy?
Franklin: I can’t speak for that either. But we do get applications for all sorts of things. We try to add them as fast as people are interested.
Tim: Okay. I had another question that I have to pause here to remember what it was: Here we go so Duolingo started out on the web as a web app, and then it went from there to iOS and now to Android. So you must have an extensive background as an Android developer?
Franklin: Originally, we started on the web because originally there were just a couple of people in a small room somewhere. As we grew, we realized that there is a lot of potential if you can reach people in their everyday lives. Like on their commutes, just whenever they have a couple of minutes. And we also realized that mobile devices were increasingly prevalent everywhere in the world, I guess. So then we got an iOS developer, and he made an iOS app. Originally, we actually had a little betting pool in the beginning, about how many users we thought it would have. And we were completely off. It was crazy! Then we realized how important it was to reach these users. So we decided to also focus on Android as well. Unfortunately, we were in a little bit of a hurry so we didn’t have time to find an Android developer. So one of our existing developers, Vicki, made the first version of the Android app that launched last year May. I showed up around that time and a couple of weeks later I was doing the Android app.
Tim: Also not as an Android developer at that time?
Franklin: Not as an Android developer. So I can’t say that my code was particularly nice. I’ve learned a lot of things—it’s a learning experience. But we make do with what we can. And recently in January we hired another developer, who has never done Android before but is also now working on the app.
Tim: Three in a row?
Franklin: Yeah, I mean it worked the first time, it seemed to work the second time and hopefully will work the third time.
Tim: What kind of constraints does it impose on you to go from a web app where you have a full browser screen to a display that may be 3” x 5”, maybe a different operating system really even if it’s in the same family?
Franklin: So one of the big challenges is reducing the amount of stuff we have to show the users. It gets tricky. It’s very easy to try and squeeze too much content on a single small screen but then the user feels overwhelmed—they don’t know what to click on—it’s just a bad experience. So first we had to figure out what was the core experience we wanted our users to see, and then we tried to design around how we can make that feel the best.
In the original version, we didn’t have what we now call the tap exercise. Where you build the translation by tapping out individual words. But we quickly realized that people typing on their keyboard on a mobile device weren’t very happy—they made more mistakes, and they were less likely to do things like type out umlauts and accents and that sort of thing. So then we tried letting them tap from a word bank. Of course, there are fake words in there and incorrect options to try and throw them off. So they still have to know the correct answer. We found that throwing that in there helped a lot.
So definitely, I think the biggest thing is just how much content you can show to the user at a single point in time, and trying to reduce the influence of the keyboard—they were probably the two biggest challenges.
Tim: Has it actually changed your own language learning, outside a computer language, that is?
Franklin: My language learning? I’ve learned a little bit of German through it. Not as much as I would have liked—I had a little bit of a motivation thing. I started back in high school, I took some German classes. I think I am about up to where I was back then. But it’s also a little tricky. German has a lot of different things from English. Like now I am on genitive case—and I am going to be honest—I don’t fully understand it. But every day I try to learn a little more and one day I’ll get it, hopefully.