ACM ByteCast - Luis von Ahn - Episode 14

Episode Date: April 9, 2021

In this episode of ACM ByteCast, Rashmi Mohan hosts past ACM Grace Murray Hopper Award recipient Luis von Ahn, co-founder and CEO of Duolingo, the world's most popular language-learning platform. He is also a Consulting Professor in the Computer Science Department at Carnegie Mellon University. Known as one of the pioneers of crowdsourcing, his many recognitions include the MacArthur Fellowship, MIT Technology Review's TR35, and the Lemelson-MIT Prize. They discuss how he, Manuel Blum, and others at Carnegie Mellon conceived the now famous technology behind reCAPTCHA, the company he founded before Duolingo, and sold to Google in 2009. Von Ahn gives insight into his journey toward harnessing the power of crowdsourcing to provide free, globally distributed language learning. They discuss the dominance of the English language in computing, the benefits and challenges of starting a company in Pittsburgh, some Duolingo user stories Luis has found particularly gratifying, and more.

This is ACM ByteCast, a podcast series from the Association for Computing Machinery, the world's largest educational and scientific computing society. We talk to researchers, practitioners, and innovators who are at the intersection of computing research and practice. They share their experiences, the lessons they've learned, and their own visions for the future of computing. I am your host, Rashmi Mohan. If you want to boost your brainpower, improve your memory, or enhance your multitasking skills,

then you're often recommended to learn a foreign language. For many of us, that option has become a reality thanks to our next guest and his creation. Luis Juan An is a serial entrepreneur and founder and CEO of Duolingo. An accomplished researcher and consulting professor of computer science at Carnegie Mellon University, he straddles both worlds seamlessly. He's a winner of numerous awards, including the prestigious Lemelson-MIT Prize and the MacArthur Fellowship, often known as the Genius Grant. Luis, welcome to ACM ByteCast.

Thank you. Thank you for having me. Wonderful. I'd love to lead with a simple question that I ask all of my guests. If you could please introduce yourself and talk about what you currently do, and also give us some insight into what drew you into the field of computer science. Sure. So, well, okay. So my name is Luis. I am currently the CEO and co-founder of a company called Duolingo. Duolingo is a language learning platform. It is now the largest language learning platform in the world. It's the most popular way to learn languages in the world. We have more than, there are more people learning languages on Duolingo in the world. It's the most popular way to learn languages in the world. We have more than, there are more people learning languages on Duolingo in the United States than there are

people learning languages in the whole US public school system. I am, of course, a computer scientist by training. I have a PhD in computer science. Before Duolingo, I used to be a professor of computer science at Carnegie Mellon University. And there I worked on all kinds of different research. I have also started two companies before Duolingo. And what originally drew me to computer science or computing in general was it was early on in my life. I was eight years old. I wanted a Nintendo and my mother instead got me a computer. And I was pretty upset. I just wanted to play games like all my other friends. But that's, you know, all I had was a computer. So I figured out how to use it. And that's kind of what did it.

That's wonderful. Very creative of your mom to start you off on that journey. But also, you know, it's incredible what you say about Duolingo being the largest platform for learning languages. I know I have two children in high school and they do a number of years of learning language. It's mandatory. So it's pretty incredible the kind of impact that you've had via Duolingo. But I'd love to get into that in a little bit.

I wanted to sort of tap into your history when you said you started two companies prior to this one. I know a lot of our listeners probably know that you invented Captcha. Could you give us a little bit of history on what sparked that idea? Were you interested in a related field in computing when you hit upon this idea? How did that come about? Sure. So CAPTCHA is these distorted characters that are all over the internet when you're, for example, trying to get an email account or buy tickets on Ticketmaster. So that's a CAPTCHA. The reason that's there is to make sure that you, the entity filling out the form, are actually a human and not a computer program that was written to submit the form millions of times.

And the reason it works, or at least, you know, used to work much better, it still works, but just not as well, is that humans are better than computers at reading these distorted characters or have been historically. The idea originally came up, this is a long time ago, it was 20 years ago. I was just starting my PhD in computer science. I was a first year PhD student at Carnegie Mellon. And I went to a talk that at the time, it was given by the person who was the chief scientist of Yahoo. Now, again, this is the year 2000. Yahoo was kind of the biggest internet company of the time. The guy who was the chief scientist came to give a talk. And in the talk, you know, he described 10 problems that they didn't know how to solve at Yahoo. And one of them was that they had a bunch of people who were writing programs to obtain millions of email accounts.

Just, you know, and the reason they wanted to do that was to send spam. So, you know, the idea was they wanted to send spam from Yahoo accounts, but each, each Yahoo account only allowed you to send like a hundred messages a day or 500 messages a day or something. And people who wanted to send spam wanted to send millions of messages per day. So what they did is they would write programs to obtain millions of email accounts. And from each one, they could send like 500 messages per day or something. So they didn't know how to stop them. And I, you know, I listened to this talk. I, you know, went, you know, home and started thinking about it a lot. Then my PhD advisor, Manuel Blum, who's a very celebrated computer scientist. And I started thinking about this,

this idea of, you know, how do we prevent, you know, this, this problem. And we came up with this idea that, and when we came up with the general idea first, that one way to prevent it would be to come up with a test that can distinguish whether something is a human or a computer. This may sound familiar. You know, this sounds like a Turing test, which 1940s, 1950s, Alan Turing came up with this thing called the Turing test, which was trying to, you know, a test that could figure out whether something was a human or a computer. So we realized we needed something like that, but there was a twist. We needed something that was also graded by a computer.

So the original Turing test, there was a human kind of judge that would decide whether they were talking to a human or a computer. In this case, we needed a computer judge that was deciding whether it was talking to a human or a computer. And so, you know, so that was kind of the general idea of what we needed. At some point, we came up with this idea that, well, it turns out that you can write a program that can put some letters on a, you know, on a canvas, distort them a lot and display them and then ask, you know, the entity on the other side to read those letters. And it turned out that humans could read those letters pretty well, but computers could not. And so that's what turned into Captcha. Pretty quickly after that, Yahoo started using it. And then essentially every other website in the world started using it.

Yeah, that's what is incredible about this incident is the fact that you were able to find a real world problem and work on that as a part of your PhD. I mean, that interaction between what is considered an industry problem that somebody's trying to solve and bringing that into the academic world. Do you see that happening often or often enough, Luis? It happens. I would say it doesn't happen that often. It's, you know, I find, you know, now that I am running a company, I find there's, there's a little bit of a disconnect between, you know, what companies care about, and what, you know, academia cares about. It's also difficult to share data, a lot of times, you know, many problems nowadays require quite a bit of data to solve. And it's just,

it's just difficult to share the data, not only there privacy problems, but also a lot of times there's just quite a bit of engineering is required to even make data sets that are clean enough for somebody, say, in academia to use them. So I do think it happens, but I think it probably could happen more. Now, I don't know if it should happen more, but it probably could happen more. That was going to be my next question. Do you find that there is inherent value or like a missed opportunity there if we don't sort of collaborate between the two, you know, streams, as it were, in this specific

manner? I'm sure there are many other collaborations, but you know, solving problems in academia that are surfaced in industry, but are being solved with, you know, the brightest minds in academia. Do you think that that's a missed opportunity if we didn't do more of that? I mean, I definitely think there's value. It's hard for me to know how much of this should happen. I definitely think there's value. I think generally my view is that, you know, academic research has a longer time horizon. You know, when you're running a company, particularly unless you're an extremely large company, if you're a company like Duolingo, you know, we do some research. We have people, you know, we have maybe 30 people with PhDs, we do some research, but our time horizon is a lot shorter, we just, we just cannot afford to do things that are gonna,

you know, be valuable in 15 years. That's just, that's just not a timeframe. Whereas I think in academia, depending, of course, on the field, you know, some things are, you know, things that are going to be valuable in 100 years, you know, you think about or maybe 50 years or something, you think about a lot of the research in quantum computing, it may to be valuable in 100 years, or maybe 50 years or something. You think about a lot of the research in quantum computing. It may not be valuable in the next five years, but it is still very valuable. So I think there's a little bit of a difference in the time horizons. However, I do think that when things do match up, I think some of the best technology in the world has come from, you know, collaboration between industry and academia. Got it. Yeah. No, I think that's a very valuable point that you bring up. The timelines always, you know, may not always align. But how easy was it for you? I mean, you

know, coming from an academic career to running a business or working for a company, you know, how do you build that muscle in other areas of running a business other than just sort of, you know, solving the technology problem? Yeah, that was, that's a good question. I mean, so like I said, this dueling was not my first company, I would say my, where I really learned a lot more about how to run a business was my previous company, which didn't grow that much in terms of number of people. I mean, it just got to maybe about only 15 people. But I did learn quite a bit there. And this was, I never really wanted to be an entrepreneur. I didn't really want to start companies.

This was not a desire. I was not against it, but this was not a desire. It just kind of happened for me. I was, so the previous company that I started was related to CAPTCHAs. It was not exactly CAPTCHAs themselves. It was kind of a second go around of CAPTCHAs. So, you know, I can tell you a little bit about how that happened. So this was, you know, the original CAPTCHAs came out in the year around 2000. Then, you know, by the year, maybe 2006, I was no longer

a PhD student. At that time, I was already an assistant professor at Carnegie Mellon. And I was just thinking about how many CAPTCHAs were typed by people around the world. I thought a lot of people that knew me would send me angry emails every time they saw a CAPTCHA and they saw them pretty often. And so I just did a little back of the envelope calculation. And the number I came up with was maybe something like 200 million times a day, somebody types a CAPTCHA around in the the world. I started thinking, okay, that's, you know, that's, that's a lot of times. And each time you type a captcha, you waste about 10 seconds of your time. So if you, because, you know, it takes 10 seconds to type a captcha. And if you multiply that by 200 million, you get the humanity as a whole is

wasting about 500,000 hours every day typing these, these annoying captchas. So, so I started thinking, is there any way in which we can make good use of that time? See, during those 10 seconds while you're typing a capture, your brain is doing something that computers cannot do. So could we make you do something useful? And this is, you know, I had this realization which led me to create a company, which was that while you were typing a capture, you could be helping us digitize books. So let me explain how that worked. At the time, there was a lot of projects trying to digitize all of the world's books,

basically take all the physical books that had been printed and then putting them on the internet. And the way the process works is, the digitization process works, is you start with a book, then you take a digital photograph of every page of the book. So then you're left with a bunch of, you know, pictures of words, and then the computer needs to decipher all of the words in these pictures. Now, the problem was, especially at the time, for older books, the computer could not recognize many of the words, like 30% of the words it could not recognize, but humans could. So the idea is we started taking all of the words that the computer could not recognize, and we started sending them to people on the internet

while they were typing captchas. So when you type the captcha, the words that you would type as opposed to being these randomly made things were actually words that were, that had been, that come from a book that had been digitized that the computer could not recognize.

And we would use what people would enter to help us digitize the book. Now, this was, at first, it was a research project at Carnegie Mellon, and I was pretty proud of the research project, and it was good. And at some point, we started, you know, I went and gave a talk somewhere,

and I had this whole idea, but I didn't have anything to digitize. It's not like I had a bunch of books to digitize. I just had the idea of how to do it. And when I was giving the talk about this, it happened that at the time, the CTO of the New York Times was in the audience. And then he came to me and he said, look, we have this huge archive of all editions of the New York Times that, you know, for the last 140 years. And, you know, we've tried digitizing it.

And, you know, just computers cannot recognize most of the words. But it seems like you can help. And so I said, sure, we can do it. And then we signed the contract to do it. We started digitizing the New York Times with people typing captions on the internet. And it turned out that we really, that it was working really well. And at some point they were paying us, they were paying us $42,000 per year of content. And it was taking us like a week of, like a week of time to digitize a whole year of the New York

Times. And so we started getting these checks of like, you know, $42,000 every week or every other week. And at that time, Carnegie Mellon kind of found out that we were doing this. And I didn't really know the legality of any of this, but what, you know, they came and they said, look, hello, professor, you, it's nice that you're doing this, but here's the thing, you got to get out of Carnegie Mellon. You got to start your own company. We cannot do, this is work for hire. You know, we're a nonprofit. The university is a nonprofit. You got to go start a company.

And so I did that. I went and started a company, but it was, this was not something I wanted to do. And it just, it just kind of happened. And from then on, you know, I started kind of, you know, figuring out what it is to, you know, that is required to start a company. I now realize the first thing you need to start a company is you need to find a lawyer that will start the company for you, like actually do all the papers. That's this I know. And then you start learning all kinds of things. But yeah, that was kind of the transition for me. That's, you know, that's

incredible. I mean, I'm also amazed at Louise at how you've found yourself in these situations where the work that you're passionately doing in terms of solving a problem has an immediate sort of application that has such far-reaching impact. I mean, harnessing the power of crowdsourcing with this project, I mean, it's incredible. It sort of probably transformed what New York Times wanted to do, as well as many other companies that have benefited from this. Yeah, in some sense, I've always been on the more practical side of things when it comes to my research. So that's what happened. I was, I was also fortunate that it immediately had applications, I think. That's great. And so how did Duolingo come about? Where did that idea spark from? Yeah, so Duolingo. So what happened? Okay, so I was working on this company to digitize books. And you know, it was was doing pretty well. At some point when we had, you know, about a dozen people, a little over that, actually

Google came and just acquired the company because since they were digitizing so many books, this was useful for the book digitization process. So they acquired it. It was a company. The company was called ReCaptcha. To this day, you know, when you get many Captchas online, they're served by Google and it basically comes from the same team that they acquired many, you know, many years ago. So, you know, after that, I went back and I was a professor at Carnegie Mellon and I was just, uh, you know, thinking about what big project to work on next. And I knew I wanted

to do something that was related to education. You know, that's always been my passion. That's kind of why I became a professor. I wanted to teach. So I wanted, I knew I wanted to do something related to education, uh, but I didn't know what I just wanted to know professor. I wanted to teach. So I knew I wanted to do something related to education, but I didn't know what. I just wanted to know. I just wanted to do something where computers would teach people something. One thing that was really influential in my thinking was the fact that I'm from Guatemala. So I was born and raised in Guatemala, and it's a very poor country. And a lot of people talk about education as something that brings equality

to different social classes. But I always saw it as the opposite, as something that brings inequality. Because what happens, particularly in countries like Guatemala or generally poor countries, is that people who have money can buy themselves really good education.

And therefore, they continue having a lot of money. Whereas people who don't have very much money barely learn how to read and write. And because of that, they never end up making money. So I wanted to do something that would give equal access to education to everybody. And, you know, I started thinking about that. Then I had a PhD student, his name was Severin. And together, you know, we were like, okay, let's try to do something related to education. You know, what should we teach? At some point, we came up with the idea that we should teach

languages. And, you know, at first, we came up with the idea that we should teach languages. And at first, we probably wanted to teach math. We're both kind of really math nerds. But we realized that actually teaching languages, in particular, teaching English, can be really transformative in people's lives. There's something like one and a half billion people in the world wanting to learn English. In my case, having learned English allowed me to come to the United States. It completely transformed my life. And in most countries in the world, knowledge of English can immediately double your income potential. So we thought, OK, let's do something that teaches languages and teaches them for free. And so that was the idea. And so we launched Duolingo in the year 2012.

And this time I knew better and I knew that I should just start a company at the time. And so I went on leave from Carnegie Mellon and started this company. And we were very fortunate that soon Duolingo started really growing a lot and pretty quickly became the most downloaded app in the education category. I mean, we were fortunate.

We also did some things right, which in retrospect, we didn't know that we were doing them so right. But in retrospect, you know, it makes a lot of sense that we made some good decisions kind of by accident. And yeah, and it was it's been going up since then. Oh, fantastic. You know, I recently heard a TED talk by Adam Grant on original ideas. And he said, like, the first idea we usually encounter tends to be raw or unfinished. So a little bit of procrastination in this whole ideation process helps. It sounds like you started off with math, but then, you know, you kind of hit upon this languages idea, which has obviously

become, you know, incredibly, incredibly popular. But what I'm, you know, also hearing so much from you, Luis, in terms of passion for the product and the idea, how did that, you know, from being, you know, deeply involved in solving computing problems, how did that translate into sort of running this company? Like, what are the largest pressing computing problems that you're trying to solve? And was that, you know, when you become a CEO, or when you become a co-founder, do you sort of have to focus on so many other things that you kind of leave those solving those really deep problems to others? What is your role in that whole space? Yeah, that's a great question. I mean, early on, I was very involved in the things we're doing now. Early on, the main thing we needed to

do was make a thing that works, a thing that actually teaches you languages. There, one of the things that we realized quickly was the hardest thing about learning a language is staying motivated. So we spent a lot of effort, and I was very involved in this, spent a lot of effort in trying to make the process of learning a language through and out as fun as possible. So we basically used a lot of tricks that games use. So we made it really feel like a game. And that was kind of the first order of business is just make a thing that a lot of people use. Over time, now, eventually, a lot of people start using it, et cetera.

By now, the problems that we face, there's a number of problems that we face that are really interesting. I mean, of course, I'm not the one having to solve them or even able to solve them just because there's just many other things that I have to do. And so we have people whose job it is to solve them just because there's just many other things that I have to do. And so we have people whose job it is to solve them. But the types of problems that we are solving at Duolingo are things like, for example, we have access to probably the largest data set of people learning anything in the world. There are somewhere between half a billion and a billion exercises are answered by our users every day. So every single day, we get somewhere between half a billion and a billion answers to exercises every day. The question is, how can we use this data to teach better? And, you know, we've been

able to really improve how well we teach by doing all kinds of things. For example, you know, this is kind of one of the first few things we did. We do A-B tests on our content. So, for example, we imagine that we are, you know, when we're teaching a language, we're teaching, imagine we're teaching Italian to Portuguese speakers. And we don't know whether we should teach plurals before adjectives or adjectives before plurals. For many of these concepts, nobody really knows what the best ordering is to teach. The beautiful thing about something like Duolingo is that we now have the data to figure this out ourselves. So if we want to figure out

whether we should teach plurals before adjectives or adjectives before plurals to these users, what we do is we just run an A-B test. So for the next 50,000 people that sign up, to half of them, we teach them plurals before adjectives. To the other half, we teach them adjectives before plurals. And then we figure out, you know, once and for all, at least for our user base, which ones learn better, which ones are more motivated, which ones stick around for longer. And, you know, we can actually figure it out.

And by now, with Duolingo, we have a pretty sophisticated system that just continually is getting better and better at teaching by using the data from our own users. So I think those are some of the more interesting problems that we're solving. And I think we've only scratched the surface of what we can do. I really think that over time, computers are going to be able to teach a lot better than humans. That's not true right now. We have a lot of data. Right now, I think for certain aspects of the language, Duolingo is about as good as a classroom, but not as good as a one-on-one human tutor. So a good one-on-one human tutor is known to be better than a classroom. And right now, Duolingo is about as good as a classroom. That's great.

You know, I mean, when I was sort of preparing for our conversation, I actually spoke to my daughter, who does use your product. And I was asking her, I said, you know, can you tell me a little bit more about what you find most exciting about it? And, you know, a couple of pieces of feedback that she gave me was, you know, it's great that it adjusts to the learner's level. So you get content that's appropriate to the learning level that you're at. She said, you know, there's a great amount of information that's packed into a very small unit of time. So especially for a high schooler, it doesn't feel like an overload of, you know, information at one time. It feels like, okay, you have just a small unit of time. So especially for a high schooler, it doesn't feel like an overload of, you know, information at one time, it feels like, okay, you have just a small unit of information that you can absorb, and then retain and then go back for more, you know, so those are a couple of

things that she brought up. But the other thing that she brought up was like she said, the user experience of Duolingo is, is so engaging, right? And this is what you were referring to earlier, also, in terms of making it like a game so that you bring your users back. Was that a conscious choice based on like, you know, previous research that you did on terms of, you know, is this is how people learn languages? Or how did you hit upon that? It was a very conscious choice. It's not just about languages. It's just generally keeping people motivated. So one of the things if you're going to teach something with an app, one of the things that you realize pretty early on is, you in a classroom people are almost held hostage right i mean you sort of you can't really leave the classroom and also your parents probably paid a lot of money to take your

class or whatever it's socially unacceptable to leave high school it's just not acceptable so generally you know when you when you're in a class you kind of have to be there when you're learning something with an app it is so easy for for anybody to just leave i mean you know, when you're in a class, you kind of have to be there. When you're learning something with an app, it is so easy for anybody to just leave. I mean, you know, a lot of times people ask us who's our major competitor. You know, our major competitor is not another language learning app or anything. Our major competitor is like Instagram or, you know, TikTok or something. It's just so easy for people to say, well, yeah, okay, I'm just, you know, I'm just going to go somewhere else. So we've, we spent a ton of effort trying to keep people as engaged as possible. And the way we do it is through gamification. And we do, and gamification and a lot of things. I mean, for example, a lesson on Duolingo doesn't take one hour, a lesson on Duolingo takes three minutes.

And again, it's because it's just because we want, we want people to be as engaged as possible. And this hasn't come, you know, this didn't take five minutes. I mean, we've, we've been iterating And again, it's just because we want people to be as engaged as possible. And this hasn't come, you know, this didn't take five minutes. I mean, we've been iterating on Duolingo for the last eight years to make it more and more engaging. And again, we use the data from all of our users to try to figure out, you know, we do all kinds of measurements about should each lesson be three and a half minutes or three minutes or two and a half minutes? What's the optimal? Also, we have a pretty sophisticated system. So for every exercise we give you on Duolingo, we actually know what the probability is that you, as a particular user, are going to get it right or wrong. So we have a user model for everybody that is using Duolingo.

And we know for this particular user, this particular exercise, that user has a 72% chance of getting it correct. And we use that model to give you exercises that we know are going to keep you as engaged as possible. They can't be too easy, because that's just not fun. If you have 100% chance of getting everything right, it's just not fun. But also they can't be too hard hard because then you get frustrated. So we actually target, you know, to give you things that are like 80% that you have an 80% chance of getting correct. And so we do things like that and it really makes a huge difference. It really keeps you engaged. And in addition to that, with our user model, we also give you, not only are we trying to make sure that you have an 80% chance of getting everything

right. We also give you things that try to exercise things that we think you're about to forget. So, you know, we know kind of what you know and how well you know it for everything you've done on Duolingo. So, you know, we know that, well, maybe the past tense, you're kind of struggling. So, you know, whenever we give you a lesson, we try to find an exercise that is in the past tense, because we know you're kind of struggling with it, that you also have an 80% chance of getting right, then we give that to you. And it turns out that that makes a huge difference in engagement, because it just feels good to get things right most of the time, but for them to be challenging enough. Right. Yeah. I mean, I think that you hit upon the nail on the head, just challenging enough,

enough to keep you wanting to come back, but enough to be motivated to actually say, okay, I think I'm moving ahead. I'm actually progressing. Yep. But the point you brought up about, obviously, you're collecting data, you're modeling users, et cetera. So what are your thoughts around or data privacy concerns in the space that you are in? I mean, we obviously take data privacy very seriously.

The fortunate thing is, you know, we don't collect very sensitive information. We don't know like you're almost anything that a normal social network would know. So we don't collect very sensitive information. So, you know, the types of things that we have, if somebody were to like get all our data, the types of things they may find out

is that you're not particularly good at the past tense of Spanish, which, which again, you know, it's obviously we take it very seriously, but it is just not quite the same as figuring out that you've been, you know, Googling how to, I don't know how to build a bomb or something like that. That's a very different thing. So, yeah, that's kind of how we see it. Got it. Yeah. And that makes a lot of sense. I think the value that you see in data that you collect is valuable for your product alone, but does not compromise another person's privacy in too deep a manner. Yeah.

At least, I mean, I'm sure there are people who wouldn't want this to get out. And that's why we take it seriously. But it's just not quite the same as your social security number is out now. Fair enough. Yeah, fair enough. And do you find that with this whole pandemic situation, and everybody's now, of course, learning things online, and that was the model that you started with, but have you seen ways in which Duolingo is now getting used that you didn't anticipate, or any aha moments through this madness that we're all living through? Yeah, I mean, we've definitely seen increased demand

due to the pandemic. Actually, this was an interesting thing, you know, when this whole thing started, because we have users in every country in the world, you could you could see which countries were going into lockdown by looking at our traffic. Because whenever a country goes into lockdown, our traffic does go up and you know and it's for multiple reasons so one of them is just you know kids who were using edits or who were going to learn in school now are more heavily relying on on duolingo because they have to do it at home or just just generally adults maybe you know they're they're more bored at home because there's a lot less to do now that everybody's in lockdown. So we do see that, you know, in terms of in terms of aha moments or anything, I don't know.

I mean, this has been our vision all along that we can learn by themselves. And it's we've been very touched to to see how much impact we've been able to have in the world. I mean, we're for a lot of people, we made lockdown a lot more meaningful because, you know, at the very least, they're not completely wasting their time. And, you know, lockdown passed. And at the very least, I got a lot better at French or something like that. So we feel pretty good about that. Yeah, no, that's terrific. I think that's definitely something that, you know, a lot of us have felt that, you know, the time that we have gained from maybe not spending on the road commuting to work or, you know, other means that has been put to,

you know, ways in which we want to improve ourselves, whether that's learning a language or any other sort of hobby that we want to pursue. You know, I'd like to go back to something that you said earlier about, you know, languages, especially English, learning English, you know, being the gateway to jobs in many developing countries. You know, one thing, and, you know, I was recently chatting with a friend who was not a computer science professional, not in the tech industry. And one of the questions she asked me was, you know, do you believe that native English speakers have an advantage in computer science? And we see mainstream programming languages are all in using the English alphabet. Do you think that's a barrier for people entering, you know, the technology space? So yes, I believe that English speakers have an advantage.

Is it impossible to be able to program without knowing English? No, you can become a very good programmer without knowing English. But I do think that English speakers have an advantage. There's not only other programming languages in English, essentially, but there's just a lot more documentation. I mean, if you know English, you can read how to do things with Stack Overflow. There's way more content in English like that, way many more books in English than everything else. So I do believe there's an advantage. I think there's a similar advantage when starting a company. I know that's not quite the same as computer science, but I think knowing English just gives you access to just kind of the world.

Whereas if you don't, you only have access usually to your country. And so I think that's a pretty big difference. And so, yes, I do think there's an advantage. Again, that's not to say that it is impossible to go through life without knowing English. But just like with income potential, I mean, knowledge of English in most countries doubles your income potential. It doesn't mean you can't make money without knowing English, you just can make more if you know English. And I believe the same is true for, you know, in computing or in starting a company. Right. No, that's very true. And I think that, you know, it's incredible that, you know, Duolingo is actually helping people sort of bridge that gap

and in some ways address that inequity that we may see. And also, it's a great responsibility that you are undertaking in order to be able to, you know, give these people the ability to enhance their skills and, you know, hopefully make a better life for themselves. One of the questions I had for you was also, you know, when you look at both for yourself as a technologist, as a co-founder, as a CEO of a company or for the company itself, how do you measure success? For me, it's always been impact. How much can you positively impact the lives of people and how many lives can you positively impact? That's always been, I've always just wanted to impact as many lives as possible in a positive manner.

That's impact been, I've always just wanted to impact as many lives as possible in a positive manner. That's impact for me. I mean, you know, I can tell you probably the proudest I've been of our work at Duolingo was a couple of years ago. I learned in the same week, I learned two facts that just, you know, juxtaposed made me very proud. On one side, on one end of the spectrum, I learned that Duolingo was being used by a ton of Syrian refugees all across Europe to learn the native language of each country. And in fact, in many refugee camps, they had, you know, like actually dedicated Duolingo programs because they didn't have anybody to teach them the language. So they would just set them on a computer to learn with Duolingo. So on one side of the spectrum, refugees were using Duolingo to learn a language. Now, if you're a refugee,

you usually don't have much money to your name. So it's usually kind of sort of some of the poorest people in the world using Duolingo to learn a language. That same week, I learned that Bill Gates was using Duolingo. And so that to me was really amazing. It's like, look,

this is one of the richest people in the world. They can afford anything they want. And they happen to be using the same educational system as Syrian refugees. And, you know, it's not just Bill Gates. I mean, you know, a ton of famous people that have used or that used Duolingo, like, you know, Tom Hanks and the Jonas Brothers and stuff like that. So the fact that these people for whom money is just no issue, they could do whatever they want. And they happen to choose the same tool to learn a language as somebody who doesn't have very much money. That to me is exactly what motivates me, that basically more money cannot buy you a better system. To me, that's what makes me the proudest. That's such an incredibly heartening thought, Luis, because I think that's, that impact, I think, not just as a technologist,

but just as a human being to think that, you know, the product that you're putting out there can be of value to anybody in the world and doesn't have any barriers for entry is pretty amazing. So thank you for sharing that. The other, you know, thing that I, I mean, it feels like, you know, you do so much pathbreaking work. The other thing that I mean, it feels like you do so much pathbreaking work. The other thing that I found very interesting is also I happened to be in Pittsburgh sometime last year, and I saw hoardings for Duolingo around the area. You started a company, a technology company in an area that was not traditionally a stronghold as a tech hub. What were the advantages or what were some of the challenges that you faced with that decision? That's a tech hub. What were the advantages or what were some of the challenges

that you faced with that decision? That's a great question. I mean, you know, the reason we started here was because of Carnegie Mellon. I mean, we were, you know, I was a professor there and we were here anyway, so we started. And if I were to go back and do it again, I would do it again in Pittsburgh. I think it's been pretty advantageous to us. And I'll tell you, it's not all good. There's some good and bad things. The good thing has been, we have been able to hire really excellent computing talent, particularly coming out of Carnegie Mellon, but it's not just Carnegie Mellon. I think, you know, people have moved here from a lot of really amazing universities. So that has helped a lot. Another thing that has helped with a lot is we have little competition in terms of hiring. You know, Pittsburgh is a relatively large enough city that there are, there's a good

number of people that for one reason or another just need to be in Pittsburgh. A lot of times it's because their family's here or, you know, their aging parents are here or something. They just need to be in Pittsburgh. When there isn't that much competition, I mean, there are other tech companies in Pittsburgh, for sure. There's just not very many. But when there isn't that much competition,

it just happens that we have been able to hire people kind of out of our league. But it's because, you know, Duolingo is kind of the best job in town. So many of the people in our executive team, for example, a good fraction of our executive team is here because of aging parents, because they have aging parents that live in Pittsburgh. They usually have been in either New York or Silicon Valley. And if they were there, they would be part of an executive team, a significantly larger company. But they just happen that they need to be in Pittsburgh. And, you know, we're kind of the better job in town for them. So they choose to work with us. So that has been pretty advantageous

for us and it has worked. Where it hasn't been so good is there's certain roles for a company that there's just not very much of that role in Pittsburgh. The one I'm mainly thinking about is marketing. There is just not a lot of marketing talent in Pittsburgh. There's some, there's not a lot. And we've had trouble getting people to move to Pittsburgh, in marketing in particular. In other areas, we don't have trouble with people moving to Pittsburgh, but in marketing, they say, look, part of my value is my network. I have a really good network here in the city that I live in,

be it New York or San Francisco or something. I could move, but honestly, I'll have much less value, not just to you, but to the world, because my network is just not going to be where I am. So in certain roles, we've had trouble hiring. What we've done, of course, is we just have, we've opened other offices. So by now we have our largest office by far is our Pittsburgh office, but we do have a, we do have a relatively sizable office in New York city. We have another one in Seattle and we have one in Beijing. The Beijing one is not, you know, a lot of times when I talk to people, they say, oh,

you have cheap programming talent in Beijing. You would be surprised how expensive it is to hire programmers in Beijing. It is not because of trying to save money. It is because we're trying to grow in China. And we figured that the best way to do it is to have a local office there. So we also have an office in China. It's wonderful. And I hope that with, you know, with the one thing that COVID has taught us, as horrible as it's been otherwise, is that, you know, talent can really reside anywhere and add

tremendous value to companies across the world. So I hope that that solves some of your marketing talent issues as well. I hope so too. Although, I mean, we really did solve it with the office in New York that pretty much solved it. Right. I have one more question for you, Louise, which is a lot of the times, I mean, you know, we're talking to an audience of mostly folks who are practitioners and folks who are in industry, but there is always, you know, as we're sort of going through our lives and careers and trying to sort of, you know, add impact to the problems that we're working on, there is always sometimes a spark of an idea that comes in and saying,

oh, I'm going to go out and start my own company. What would be your advice, both to people who are in academia or in industry, as to, you know, how does one sort of scratch that itch, if you will? Well, my biggest advice, I can tell you a few things, but I would say my single biggest

advice I would give is just do it. The number of people that I know that have come to me and have said, look, I, you know, I've been thinking about it. I think about it for years and just never do it. You know, the hardest thing is just to get started. Just, just get started. Your idea doesn't have to be perfect. Your idea, like you said, your idea will probably change 50 times before you actually, you know, make it big. So the biggest thing is just get started. Others, you know, other pieces of advice. I personally have found it a lot easier to start companies with other people because they kind of keep me in check and also keep me motivated because then there's somebody else that, you know, I will disappoint if I don't work on it.

So if you find yourself a co-founder, I think that would be good. Another thing that I think is important is if you're going to start a company on something, you probably either you or your co-founder should have a good amount of expertise on that. Whatever the hardest problem is in your company. For example, if the hardest thing is in your company is something related to, you know, computing, you probably should be, you or your co-founder should be in computing. The companies don't work super well are companies where it's like two MBAs wanting to do something really deep in computing and they think that you're going to be able to hire some lackeys to just do it for them. That doesn't work super well. But on the flip side, if the hardest thing that your company is going to do is like a sales problem and you're in computing, you probably want to start the company with somebody who knows a lot about sales. So I would say figure out what's the hardest thing related to the company, to your idea, and make sure that you're in the founding

team, you have somebody that knows about it because otherwise it just doesn't work super well. I would say another piece of advice is be very, very picky on who you hire. This really matters, particularly early on in terms of, you know, there's a huge push at first when you say, oh, we just need to hire somebody who will code. You'll pay the price dearly if you just don't have a high bar of quality for the first few people you hire because the thing to realize is the first few people you hire are going to set, it's like the seed that is going to set your company for years to come.

And if you hire a bunch of really good people, your company is probably going to be full of really good people. Similarly, this even accounts for diversity as well. I mean, if you hire a bunch of men in your company, the first 10 employees in your company are men, it's very hard to hire the first woman if that's the case. So I think you want to make sure that the early team is as diverse as possible, and is as high quality as possible, because that's the seed for the rest of the company. I really strongly believe in this. I think that's excellent advice. You know, whether we're starting our company or not,

I think making sure that, you know, we're hiring talent that also questions and, you know, brings in that diversity of thought to make sure that we're actually solving for all parties involved as opposed to just a small section of society. Wonderful. Luis, and for our final bite, I mean, this has been a great conversation, but what are you most excited about either in the field of computer science or in technology over the next few years? Related to what I do, I am pretty excited about our computers being able to teach most everything, you know, just with a computer that you'd be able to learn a lot of really meaningful things just with a computer. I'm pretty excited by that. We're not quite there yet for most subjects, but I think we will be able to figure this out. That's something that I'm particularly excited about. I would say I'm pretty excited about virtual reality. Again, something that's not quite there yet, but I think it's going to be pretty transformative. I guess those are the two things that I'm pretty excited about. Wonderful. Well, thank you so much for joining

us today. This has been an incredibly fascinating conversation. We really appreciate that you spent time with us here at ACM ByteCast. Thank you. And thank you for the great questions. ACM ByteCast is a production of the Association for Computing Machinery's Practitioners Board. To learn more about ACM and its activities, visit acm.org.

For more information about this and other episodes, please visit our website at learning.acm.org slash bytecast. That's learning.acm.org slash b-y-t-e-c-a-s-t.