So, does every day is at least 10x an 8 minute workout, so you don't need maths to really know that that's true.
因此,每天至少要进行10倍的8分钟锻炼,因此您不需要计算就能知道这是真的。
Ladies and gentlemen, please give a warm, warm welcome to Lewis Van Anh.
女士们先生们,请热烈欢迎 Lewis Van Anh 的到来。
Hi, can you hear me? Good. Okay, let's see if this works. Good. Or not. Good.
你好,能听到我说话吗?很好。好的,让我们看看这个能否工作。很好。或者不行。很好。
Okay, so I am from Guatemala. This is a public service announcement that is where Guatemala is. And very importantly, this is not where they keep the prisoners that is called Guantanamo, not the same.
Now, a few years ago I was in a very fortunate position in my life. I had just sold my second company to Google. And I decided that I want to work on a new project, and I wanted to work on a new project. And I decided that I wanted to work on a project related to education. Education has always been my passion.
So I wanted to do something related to education, but my views on education are very related to where I'm from. I'm from Guatemala. And if you don't know anything about Guatemala, it's a very poor country. It's very representative of developing countries throughout the world.
Now, a lot of people talk about education as something that brings equality to different social classes. But I always saw it as the complete opposite, as something that brings inequality to social classes. Because what happens in especially in developing countries is that people who have a lot of money can buy themselves the best education in the world. They can go study at Harvard, for example. And because they have such good education, they continue having a lot of money. Whereas people who don't have very much money, they barely learn how to read and write. And because of that, they can never make a lot of money.
So education was this thing that created a divide between social classes. And I wanted to do something that would give equal access to education to everybody, regardless of how much money they had.
教育是在社会阶层之间创造出的一道鸿沟,而我想做的是让每个人都能平等获得教育机会,无论他们有多少钱。
Now, education is very general. And I can't think that generally. So I decided to concentrate on one type of education, which happens to be very big throughout the world, except for the United States. It's not very big here. And it is learning a foreign language. So it turns out there's 1.2 billion people in the world learning a foreign language.
Now, this is a very strange market. 800 million of these satisfy three properties. The first is that they're learning English. The second is that the reason they're learning English is in order to get a job or a better job. And the third is that they are of low socioeconomic conditions. So most people learning a foreign language are basically people trying to learn English in order to get out of poverty.
But of course, before Dueling, most of the ways they were to learn a language were very expensive, especially with software. For example, in the United States, there was this thing called Rosetta Stone, which cost between $500 and $1,000. In other countries, there was this thing called Open English, which cost about $1,000. So the irony of it all is most people are trying to learn English to get out of poverty. But it seems you needed about $1,000 in order to get out of poverty. This made no sense.
So what I decided to do with my co-founder was to create Duelingo. And we launched it about four years ago. And the idea is that it's a completely 100% free way to learn a language. You can do it on the web or on apps, or basically anywhere electronically. But it is 100% free.
Now, since we've launched, Dueling has now become the most popular way to learn languages in the world. It's also the most popular education app in iTunes and also Android. And there's a lot of really cool numbers. I mean, for example, there are more people learning languages on Dueling when the US than there are people learning languages in the whole US public school system. Or here's another one.
We teach a lot of languages. We teach the standard languages you would imagine, like German, French, Italian, Spanish, et cetera. But we also teach some smaller languages like Irish. I actually didn't know Irish was a language. I thought they spoke English there, but it turns out Irish is a language. There's about 94,000 people who are native speakers of Irish. On Dueling, we have over a million people learning Irish. So we can actually multiply the number of people that speak Irish by a factor of 10. There's a funny tidbit about this on St. Patrick's Day, that number doubles every day. And then the next day it goes back to normal. So it's very funny.
Now, what makes me the proudest about Dueling was, there are entire countries where every public school that has access to the internet is using Dueling to teach English to their students. And many of these are developing countries. For example, in Costa Rica and Colombia, this is the case.
So on one side of the spectrum, students in public schools in very poor developing countries are using Dueling to learn a language. On the other side of the spectrum, Bill Gates uses Dueling to learn French. So to me, this is what makes me the proudest. The fact that the richest man in the world uses the same system as kids in public schools in developing countries.
That means that, at least for this particular thing, more money cannot buy you a better educational system, which is exactly what I wanted to do.
这意味着至少就这个问题而言,更多的钱并不能为您带来更好的教育系统,而这正是我想做的。
Now, why do people like Dueling so much? So let me show you some of the things that people say about Dueling. So there's this person, for example, who said in the past couple of days, they learn more from Dueling than in four years of high school. And I think that says more about high school than about Dueling.
I don't think Dueling goes all that good. Here's another good one. This was my mother. She likes it a lot. Now, one of the reasons that people like Dueling so much is because we've spent a lot of effort.
When we were starting Dueling, we realized the hardest thing about learning a language or learning anything by yourself is keeping yourself motivated. So what we've done is we spent a lot of time making Dueling go feel as much as a game as possible.
It's not exactly a game. We don't call it a game, but it really does feel like a game while you're learning language. We made it as easy as possible as we can. So for example, this is what the home screen for Dueling looks like on iPhones.
If you notice, the idea is that the language is split up into multiple units that we call skills. For example, food is a unit. That's where we teach you all kinds of things of food or plurals is a unit. That's where we teach you how to pluralize.
And the idea is that at the very beginning, only one of these skills is unlocked. And you have to complete it to unlock other skills. There's this unlocking mechanism that is very motivating. Now, completing a skill, what it involves, is basically doing a bunch of exercises, a very short exercise, each of which takes about 20 seconds, where the idea is that you get something and you have to do something.
And you can always get it either right or wrong. So at no point on Dueling, or you're reading anything about theory or grammar, anything at all points, you're just doing stuff. So you're learning by doing.
Now, each of these exercises, they have multiple modalities. Some of them, you have to translate between your native language and the language you're learning.
In others, you have to kind of click on the picture that is related to the word. In other exercises, you have to speak to the app and it tells you whether you set it correctly or not. And the idea is there's this progress bar at the top.
And whenever you get an exercise correct, it goes up, whenever you get an exercise wrong, it goes down. And at the very beginning, that progress bar is empty. And the whole goal of the lesson is to complete this, to fill up that progress bar.
Now, this is, it looks very simple. But it's actually in the background, it's very sophisticated. For example, the amount that the progress bar goes up and down varies depending on the person and also the exercise.
So we actually have a pretty good idea of how much each person knows while they've been using Duolingo, because we record everything that they've done. So we know every exercise that you've done on Duolingo and every time you got something wrong, every time you got something right, we even record things like if you've, every time you see the word for elephant, you take, let's say, a second longer to respond, that means that the word for elephant is a little harder for you, so you don't know it that well.
So we have a very good model for how much each person knows, and for each exercise, we give you, we have a pretty good idea of whether you're going to get it right or wrong. And if we give you an exercise that we think is too hard for you, we sometimes give you exercises that we think are too hard for you, just kind of to test you and also to motivate you.
If we give you something that we think is a little too hard for you, and we think you're going to get it wrong, but you somehow manage to get it right, that progress bar goes up more because you're breaking our expectations. But conversely, if you get it wrong, this is an exercise that we anyways expected you to get wrong, the progress bar's not going to go down very much because we kind of expected you to get it wrong anyways.
So we do a lot of things like that in the background where we keep a pretty sophisticated model of how much you know. Now, when we started Duolingo, I didn't know anything about how to teach languages. English is not my native language. I learned English.
Then in high school, I unsuccessfully tried to learn French for a year. And so that was my experience of language teaching up until I started Duolingo. It was a computer scientist.
My co-founder was also a computer scientist. So we basically two computer scientists who didn't know anything about how to teach languages. So what we did is we started reading books about how to best teach languages. Literally, we read French for dummies. And we read a bunch of books.
At some point, we found a book that was called Something Like The Best Method to Teach a Language. And it was a pretty thorough book. It had a whole method for how to teach a language, and it had a lot of scientific evidence proving that it really was the best method to learn a language. At that point, this was not a very popular book. We found it in some university library, and we thought, oh, we struck gold. Because this is really the best method to learn a language.
All we have to do is make an app related to this book. And that's it. That'll be the best app to teach a language. Unfortunately, after we kept on looking, we found other books that also claimed to be the best method to learn a language, which were completely different from the first one. And at that point, we thought, well, these methods to learn a language are a lot like diets, where they all claim are the best diet in the world, but they are completely contradictory to each other. And so we didn't know what to do, and we got confused.
And we had very simple questions. We wanted to know things like, because we were engineers, we wanted to know things like, should you teach plurals before adjectives or adjectives before plurals? That's a very simple question. And it didn't seem like there was a right answer to this, or at least nobody knew it.
So what we did is we launched what we could, and fortunately, the bilingual started growing quite a bit. And after it grew enough, after we had enough uses, we realized we were in a very unique position. We realized we could actually figure out the answer to many of these questions using our own users, or using our own data.
So for example, nowadays, if we want to know whether we should teach plurals before adjectives or adjectives before plurals, we just do an experiment. So for the next 50,000 people that sign up to do a lingo, which takes about six hours to get 50,000 new users, in that time, what we do is, to half of them, at random, we teach them plurals before adjectives. To the other half, we teach them adjectives before plurals. And then we measure which ones learn better.
And then once and for all, we will know, in our system, it turns out it is better to teach this before that. And we're doing that all the time. At any point in time, we're running about 30 to 50 of these AV tests. And literally, the lingo is getting better every week at teaching languages, because we're experimenting with our users.
That didn't sound so good. That's what we're doing. Now, this happens to be, it worked very well. Somebody at the City University of New York did a study to figure out how good lingo is at teaching languages. And what they found is that if you use the lingo for 34 hours, you learn the equivalent of one semester course in that language, which usually takes a lot more than 34 hours if you're actually doing the homework. So it's very successful.
Now, a lot of people ask us about the lingo. If you've ever used the lingo, there's a lot of exercises you get to speak to it. But a lot of people ask us about conversational practice. This is one of the most common requests we get is something like the following.
I would like to be paired up with somebody else who speaks, who teach me the language, maybe with some sort of video or something like that. And there's a version of this that is a very elegant idea that a lot of people have suggested.
I am an English speaker learning Spanish. I would like to be paired up with a Spanish speaker learning English, and then we magically will be able to teach each other the language. And this is a very elegant idea. And this has been suggested a lot. And over the last four years, we've really tried to add this to the lingo. We've had many different attempts at adding this. And every time we try to add it, we realize this really sounds like a good idea, but it turns out it doesn't work.
And the biggest problem is that most people don't actually want to do this. The activation energy, the social anxiety that this causes, being paired up with another person to practice a conversation, and it's usually a stranger, to practice a conversation in a language where you're not very good. Most people just don't want to do this. And we've been scratching our heads about what to do with that, because we really want to provide people with conversational practice.
And as we're doing this, somebody from the company actually found a study that the US Army did. So there's this funny thing. So the US Army is trying to teach people languages. It turns out the language they really want to teach is Arabic. I don't actually know why that would be, just kidding. But they're trying to teach Arabic. And because it's the US Army and because all the prices are inflated, it costs them something like $50,000 to teach a soldier Arabic. Now, the problem they were having is that for some people, they do spend the $50,000 and they learn Arabic. For other people, they were spending the $50,000 and they weren't learning Arabic at all. And so they were scratching their heads and trying to figure out why is that some people can learn Arabic versus others can't.
And what they decided to do is they decided to develop a test to figure out how good somebody is at learning a language. And they were successful at it. They actually developed a test that somebody can take. And afterwards, they'll tell you whether you are actually good at learning a language or not. And this test has a lot of things. I mean, the master's general intelligence, unsurprisingly, you have to be at least a certain amount of intelligence to learn a language. But that's not very high.
But the biggest predictor about whether you're good at learning a language or not is how much you are OK with sounding stupid. Turns out that if you're OK sounding stupid, you're very good at learning a language. Whereas if you're not, you're not good at learning a language. And it's because the people who are OK sounding stupid have no trouble practicing. At the beginning, if you take somebody who is OK sounding stupid versus somebody who's not OK sounding stupid, the person who's OK sounding stupid will just start talking. Whereas the other shier person like me just won't. And they won't practice. And whereas the other person, at the beginning, both sound equally stupid. But the person who's getting the practice is just getting better and better and better. So one of the biggest problems. And this happens to be 80% of the population.
People are not very good or people who are not comfortable sounding stupid are not very good at learning a language. But of course, they would still like to learn a language. And so once we realized that, once we realized that it was that pairing people up was not a good thing. And most people didn't want to be paired up because they didn't want to sound stupid. We developed a technology that actually would let you practice conversation.
And this is something that we launched about two or three months ago. And it is we launched a way to practice conversation. But when you're paired up, not with another person, but with an artificial intelligence.
大约两三个月前,我们推出了一种练习对话的方式。这种方式是与人工智能配对,而不是与另一个人配对。
So in with Duolingo, we now have a way to teach language with bots. So we're developing chat bots. And the idea is that you'll be able to practice language with this without the social anxiety. And this is one of the things we're working on.
This is an interesting map. I like to show this map. This is, we have uses everywhere in the world. This is the most commonly learned language in every country in the world. So in pink, you see English. So in most countries in the world, English is the most common in the language. In the US, it's Spanish, in Canada, it's French. There's a few interesting things here.
But probably the most interesting one is that Swedish happens to be the most learned language in Sweden. It's very strange. And at first, we were pretty confused by this. But then we started looking at it. We realized that those people are actually learning Swedish from Arabic. So those people are actually refugees. So in Sweden, it turns out in Sweden, nobody needs to learn English because they all pretty much know English. So it happens that at least among our users, they're dominated by people who are learning Swedish from Arabic and they're usually refugees.
Now, this map is interesting. It shows that learning English is really the key thing about the key language that people are learning on Duolingo. And a couple of years ago, we started getting an email from people learning English who said the following. Thank you so much for teaching me English. But now I have a problem. I need a certificate that shows that I know English. And we started getting hundreds of these emails. And at some point, we decided to look into this. Why is it that so many people are asking for a certificate? And what we found was pretty crazy.
If you live in the United States, you probably are completely unaware of this. But it turns out that about $10 billion a year are spent by people trying to prove that they know English. Not trying to learn English. They're just trying to get a certificate that shows that they know English. Now, why would you need such a certificate for many reasons? For example, if you want to apply to come to the university here in the US and you're from a non-English speaking country, you have to take a standardized test that proves that you know English. A test is called the TOEFL. If you want to get a work visa anywhere in the commonwealth, so in the UK and Canada, et cetera. If you want to get a work visa and you're from a non-English speaking country, you have to take a standardized test that proves that you know English.
So there's a lot of people taking these tests. It's about $10 billion a year. Now, when we looked at this, I know how these $10 billion has been. What we found was pretty crazy. Most of this money is spent by basically people taking standardized tests. Now, there's a few of them. One of them is called the TOEFL. One of them is called the IELTS. Another one is called the Cambridge English Test. And they're all very similar to each other. They all cost about $250. They all, in order to take them, you have to go to a specific testing center to take them there. So it's like a physical testing center. And because you have to go to a physical testing center, you usually have to make an appointment about four weeks in advance. And then after you take the test, it takes another four weeks to get your results. So basically, it's an eight-week process. You have to go somewhere to take the test and you have to pay about $250. That's how people prove that they know English.
Now, this sounds annoying because it actually is. But it's even worse because most of the people that are trying to do this are in developing countries. And there, that's a completely different story. $250, that's a month's salary. The physical testing center, they're not in every city. So if you don't happen to live in a big city, you have to travel to take the test. So this becomes very prohibitive. So what we did about a year and a half ago, or so is we launched our own version of an English certificate called the Dual English Test, where the idea is that you can take a test from an app, as opposed to having to go to a physical testing center. And it only costs $50, as opposed to $250.
Now, the reason you have to go to a testing center to take a test is to prove that you're not cheating. So you have to prove that you are you, and not your cousin who actually speaks English. And also that you didn't show up with like Google translate. Now, how do we prevent cheating when you're taking a test from an app? The way we do it is we actually turn on the front facing camera of the phone, and also the microphone to record ambient noise. And then we have a real human watch you take the test so that we can prove that you are you and you're also that you're not cheating.
So we've been doing this actually, developing the test was not that hard. But the hardest part is getting adoption. That is, we have to get organizations to recognize this as a standard for English language certification. And we've been doing this. So I'm happy to say that as of this year, a bunch of different US universities are starting to use this test, as opposed to the test they were using, in order to test their students. And some of these are pretty big name universities. So like Yale, UCLA, Northeastern, et cetera, are going to be using this test, which is much better. And that is, I think I've run out of time, but that is basically what I wanted to say. Thank you.
Thank you so much, Lewis. Thank you. So before our next speaker, I have a short announcement. There's three tables in the business area where GitHub is teaching people how to code. So there's still some spots available. If you're interested in that, it's going on right now in the business area to the left, I think.