The positive power of the Internet

Want to learn a language for free? Find out how the Internet is allowing projects like DuoLingo to thrive...
01 December 2014

Interview with 

Luis von Ahn, DuoLingo, captcha




The Internet has allowed more people to connect and communicate with each other than ever before. This means that collective people-power can be harnessed in a way that was previously unimaginable.

A pioneer in this field is Carnegie Mellon computer scientist Luis von Ahn. Luis first became famous for creating CAPTCHA, which is a system for checking that visitors to a website are genuine humans and not just computers. The system involves typing in a word hidden inside a picture; a task that computers struggle with.

But Luis realised that this was wasting the time of a lot of people's time for little benefit. So, instead of users being asked to enter a random word, he realised that they could instead be asked to enter a word photographed from an old book. And with so many people filling in these CAPTCHA tests every day, in no time it's possible to digitally translate an entire book and then make it freely available for everyone to read. Using this process, over 2 million books a year are now being translated in this way.

More recently, Luis has embarked on a new project that works in a similar way but aims to help people learn a new language - again, for free, as he explained to Chris Smith...

Luis - I started a new project which is Duolingo which is a free language learning platform and it uses a very similar concept.  So, when I started Duolingo, we started thinking, "Can we teach languages for free?"  there is 1.2 billion people in the world wanting to learn another language, but the majority of them, about 800 million of them aren't satisfied properly.  First of all, they're learning English.  Second, the reason they're learning English is to get a job and third, they have low socio-economic conditions.  So, most people wanting to learn a language are poor people that are learning English more to get out of poverty.  Ironically, most of the ways to learn languages usually, it's pretty expensive.  So, if you think about a software for learning a language, it usually costs hundreds of dollars.  So, we wanted to have a completely free way to learn a language.  But if it's going to be completely free, how are we going to finance it?  And then an idea very similar to this idea with book digitisation came up which is, can we in some way generate something at a value while people are learning a language?  The answer is yes and this is what we're doing now with Duolingo where instead of charging people to learn a language with us, what we do is that while they're learning, some of the exercises that we give them in order to practice are exercises to translate things that have never been translated before and that we get paid in order to translate.  So for example, CNN is one of our clients where the idea is that they write, all of their news in English then they send it to us and then we give the article in English to our users. And then the ones that are learning English, in order to practice English, they're translating this CNN article from English into their native language.  And then a few different people translate the same article and correct each other.  And then one translation emerges of the whole article and then we'll return it to CNN and CNN pays us for that translation.

Chris - Wow!  How many people are signing up for this?  How many subscribers do you know have learning languages with you?

Luis - Right now, it's already the largest platform for learning languages in the world.  We have about 50 million people that have signed up for this.

Chris - And how many different languages can you boast to CNN that you can turn their news into?

Luis - We have about 12 different languages that we've been translating to.

Chris - How much news are you turning out a day?

Luis - We can do a single CNN article in about 4 hours, but we can do multiple at the same time.  Right now, we're doing a few hundred articles.  Not just CNN, we're also translating other things.  For example, Buzz Feed is one of our clients.  We're doing a few hundred articles a day, but in terms of how much we could be doing, it's tens of thousands of articles a day.

Chris - What do the end-users say of the experience?  Do they find it is helping them to learn language or do they find they end up speaking like a journalist?

Luis - No, because the translation part is only part of the language learning experience.  You're right, that you can't learn a language just from translating news articles.  So, what happens is that Duolingo is a full language learning service.  We teach you how to speak, how to write.  We do the whole thing because we want to attract as many users as possible.  But some of the exercises are basically things where in order to practice, you get to translate things.

Chris -   It's an amazing thing, the internet isn't it, in terms of giving us this way to harness human brain power and ingenuity and do it massively in parallel which we've never been able to do before.

Luis - Yeah.  I mean, just having everybody connected, literally, billions of people connected allows you to do things we would never been able to do before.  There's a lot of projects, but just as an example, some of the things that we do at Duolingo.  We are now able to do something that was essentially impossible before which is, we have so many millions of users that we can run tests on them to try to really figure out what is the best way to learn a language.  So for example, if we want to figure out whether we should teach you plurals before adjectives or adjectives before plurals, we want to figure out which is the best order.  We can just run an experiment with the next 50,000 people that sign up which with Duolingo, it takes about 12 hours to get 50,000 new people.  With the next 50,000 people that sign up, we just - for half of them at random, we teach them plurals before adjectives, and for the other half, we teach them adjectives before plurals and then we measure which ones learn better.  We can know once and for all in the system, is it better to teach plurals before adjectives or the other way around.  This is something that we could figure out in a matter of hours and figuring this out in a standard classroom is really difficult.

Chris - And it's also going to differ from one language to another so you can actually do the study and do the experiment now to work out what works best for each different language rather than say, one-size-fits-all which is probably what we used to do, isn't it?

Luis - It's not just for one language.  It also matters what your native language is.  So in fact, teaching English to native Chinese speakers is very different than teaching English to native Spanish speakers.  They have just different things that they find difficult.  We are now teaching English differently to Spanish speakers than we do to Chinese speakers.  We can even get way more personalised.  We can even start looking at, it is different to teach English to Chinese speakers that are younger than age 20 than to Chinese speakers that are older than age 20.  We can start really finding the best way to do things in a pretty personalised way, something that is essentially impossible when you're dealing with a classroom.

Chris - Now, you've come up with two world-changing ideas for the public good already.  Every entrepreneur has a whole suite of ideas in their minds.  So, what also have you got your eye on?  Where do you think this is going to go next?

Luis - Personally, I would like to continue working on education.  Education is a huge problem.  I personally don't think that the world's educational system is really getting the job done.  I mean, it's not like it's bad, but there are many things that could be done better.  I mean, just as a good example, there are 1 billion adults in the world who don't know how to read and write.  That is a failure of the current system and I think we can do better.  It's not just that.  I think there are a lot of other things.  I mean, for example in my country, I was born in Guatemala.  The high school graduation rate is about 50%.  So, about 50% of the people graduate from high school.  Of those that do graduate, only about 10% of them have the required mathematics level and only about 25% of them have the required reading level.  So really, the vast majority of people are just not learning what they should be learning.  This is pretty common in a lot of developing countries, but even when you looked at developed countries, I mean, I don't know how it is in the UK, but here in the US, people learn math for 12 years and at the end, barely know how to add fractions.  And this is 12 years of every single day, of education every single day.  I would consider that extremely inefficient because I can teach you how to add fractions in probably, a week.  So, it's extremely inefficient and I think there's a lot to do in terms of improving the educational system and that would be what I would like to concentrate on.


Add a comment