The scoop on reCAPTCHA founder's new startup Duolingo. | Hacker News Reader

The scoop on reCAPTCHA founder's new startup Duolingo.

tedxcmu.com

84 pointsgregg198234 comments

Interesting, but I'm skeptical.

I studied translation, and noticed to my surprise that the foreign language part of translating is fairly easy, unless it's highly technical, or contains dialect or slang.

The hard part was the command of my native language. I was pretty literate, but the translations I produced weren't good enough - they were accurate, but wooden, and they stayed too close to the original. What I had to learn was my own native language.

So I have doubts about the quality of the translations, and I have doubts about how they're planning to merge translations from different users, which sounds like a very tricky problem.

And it looks like the only exercise they use is to translate the foreign language (e.g. Spanish) into the native language, which will teach you to read, but not to write, speak or understand Spanish.

Nevertheless, a fascinating experiment, and an extremely ingenious idea.

Speaking as a professional translator, I gotta say it's time to transition back out of professional translation. Another five years, maybe, and I'm going to be out of a job.

It was lucrative while it lasted: the Internet made it possible for me to do well as a freelancer, but now the Internet is going to supersede me. Which I knew. It's just scary to be right in this particular instance.

This idea is immense in its brilliance and in the impact I expect it to have.

You'll be just fine. As someone who has employed a number of translators in my professional career, the biggest issue is trust. Translation is scary because it's one thing you have no ability to self-validate. While duolingo is incredibly smart, they're going to have to go a LONG ways towards building trust that what they're producing is truly what my app, brand, whatever wants to convey.

That's going to take awhile.

Nah, it's not just Duolingo. Google is making frightening inroads where it counts - they just won a hugemongous contract from the WIPO that a lot of people were hoping for (I was asked to be part of the team by at least four agencies placing bids). Statistical methods work pretty well for a lot of applications, and patents are probably one of those applications.

Actually, I think Duolingo is going to help the translation market for a while - anything that brings people closer together between languages is going to encourage more people to need quality translation - but they intend to train massive numbers of people to do what I do for a living. And once that system is in place and honed on Wikipedia, they'd be insane not to employ it for placing contract translations. The notion of employing multiple human translators and combining their input to produce a better overall output is fantastic! It really is! The only reason nobody's done it yet is the expense of paying for the translation multiple times - and that's just about to go away.

No, seriously. I've been predicting five to ten years for the limit for the industry already. This just tells me how right I am: things are going to start getting freaky weird for me.

It's about damn time, too. I'm bored sick with translation. I want to get back into programming anyway. If I can figure out how to combine them, I'm going to be a happy man.

I'm blown away by the simplicity of this. Also, there's something that feels weird about taking two things people pay for (translation and learning a new language) and somehow simultaneously making them both free.

This is both exciting and ingenious.

> there's something that feels weird about taking two things people pay for (translation and learning a new language) and somehow simultaneously making them both free.

The exact same principle that is behind reCAPTCHA, except in that case it's CAPTCHA generation & OCR.

Hmm, I guess you're right...

A bit older, but way better presentation by Luis von Ahn on duolingo:

The Royal Society - "Augmented intelligence: the Web and human computation" (Louis von Ahn) http://royalsociety.tv/dpx_royalsociety/dpx.php?cmd=autoplay...

+ has excellent discussion at the end.

tl; dr summary please?

The content of the presentation is much the same, albeit pitched at a different audience. To get straight to the discussion at the end (which is impossible effectively to summarise for you and well worth the time) skip to about 26:00 in the video.

I saw his tech talk at google a few years back where he shares his ideas for image tagging. Absolutely amazing work. Watch if you haven't yet: http://video.google.com/videoplay?docid=-8246463980976635143...

Too bad the video doesn't go into details on how the application really works from a user perspective and how merging is done.

This makes so much more sense. I heard he was working on duolingo but could not figure out how a translation application could work as a captcha. The video explains that it isn't a captcha technology, instead it's an app about translating, wholly disconnected from his previous two "monumental" inventions.

I look forward to working with it. It seems like it's another excellent step forward for independent study along with the Khan academy.

This is amazing. I wonder how they combine the beginner stuff though. Also, captchas have the benefit of being adopted by lots of websites, while this is a standalone destination. They may not get 10 million active users, but if they do, it will be because so many people want to learn a language. How will they market the site?

As a free way to learn a language, I guess.

Sure, but the economics of the marketing are not the same. Websites needed captchas, so it made sense. They saw other websites using them, and the link to the recaptcha was embedded. Are they going to embed language learning modules in language learning programs?

You shouldn't compare it to reCAPTCHA just because it's made by the same guy...

I agree with your point, yet as the parent points out. The uptake will be less, it has to be. Websites and people learning languages are vastly different size wise.

I also think the numbers he quoted about buying learning material are a little off. Purely because they are quoting numbers which cover a broad range of media. Whereas duolingo is the web and maybe mobile apps.

I personally think the market is huge, I just felt the numbers were sexed up for effect.

The "avg" of user sentences to create a better version sounds very exciting from a tech point of view.

I haven't heard anything about grammar in the videos. People are expected to absorb the grammar magically somehow?

The software he mentions (rosettastone.com) doesn't work with grammar either if I remember correctly. Instead, for one spoken sentence, they display various images of which you have to pick the right one (i.e. the image the sentence corresponds to).

It's the child approach: When you were a kid, you probably didn't have any clue about english grammar either, yet you were able to produce perfectly valid english sentences.

Why do you think such an approach would work with adults?

I didn't say that it does. And I didn't say that it doesn't either.

The actual duolingo part in this video starts at about 12 minutes mark. Before that it's about reCAPTCHA and such.

This was the best of the talks at TEDxCMU.

Duolingo could have the unintended effect of making English even more dominant by teaching everyone English.

I'm not sure whether you're intending to pose that as a problem -- why should language be a barrier to anything? It could open the possibilities to a lot of people in less well off countries.

What would be a problem (and is kind of assumed in what you're saying) is if people stopped learning their native language in favour of English.

I'm sure many people will feel that their culture is being threatened if English becomes even more dominant.

Yes, and I meant to say that would be a bad thing. For example, Welsh is compulsory in schools in Wales where most will know English fluently, there is quite a political effort to keep it going.

Could you clarify the english? I have assumed British english. (the real one ;))

Honestly your comment sounds troll like.

Languages bring richness and expression. Im going to guess your from an English speaking country and only have english speaking parents.

Sorry, I didn't mean to come off as a troll!

The OP originally struck me as a negative comment but reading it again makes me think it was more of just an observation.

I wasn't trying to say having one language is a good/bad thing, but everyone having say, a common language, could benefit all as it opens up more possibilities with people.

I'm not sure I came off right in the second sentence. I meant as if people only chose English and leaving behind their heritage / culture (as amichail suggested below) that would be bad.

Yes, I'm an English only speaker (I believe the majority of the world are bilingual now too), and so are my parents :)

Im living at persent in Oslo, young and old talk about how english has altered Norwegian. Words have krept in, some phrases are now said in english. I think a common language would accelerate the decline of many languages with small populations.

Glad you responded.

It seems more likely to make English less dominant by (for example) letting people read all of Wikipedia in their native language. Same goes for countless other resources that are currently published only in English.

Either way, if someone cares enough about learning English to put in the huge amount of time required to do it, I would bet they have a pretty good reason already. Giving them a free tool that helps is a clear win for everyone.

I'm so excited by this technology. Can't wait to learn Spanish.

I submitted this right before :)

Reply on news.ycombinator.com