WEBVTT - Chatting with the Machine 0:00:15.316 --> 0:00:24.436 Pushkin. Artificial intelligence is this weird, big phrase that suddenly 0:00:24.476 --> 0:00:26.876 seems to be everywhere, and it can be hard to 0:00:26.916 --> 0:00:30.116 know exactly what it means. But when businesses say they're 0:00:30.196 --> 0:00:34.516 using artificial intelligence, they usually mean one particular thing. They 0:00:34.556 --> 0:00:37.756 mean automated systems that can take in lots and lots 0:00:37.756 --> 0:00:41.276 of data and use the data to make predictions. This 0:00:41.356 --> 0:00:45.596 is called machine learning, and it's spreading everywhere. Drug companies 0:00:45.716 --> 0:00:47.836 use it to predict which molecules are likely to work 0:00:47.876 --> 0:00:51.076 as medicines. Hedge funds use it to predict which stocks 0:00:51.076 --> 0:00:53.396 are going to go up or down. Instagram uses it 0:00:53.436 --> 0:00:55.596 to predict which adds I'm most likely to click on 0:00:56.316 --> 0:00:58.356 For the record, the machine has learned that I will 0:00:58.396 --> 0:01:02.676 often click on ads for overpriced workout clothes. Anyway, if 0:01:02.676 --> 0:01:05.996 you want to understand what's happening with business and technology today, 0:01:06.396 --> 0:01:14.076 you really have to understand machine learning. I'm Jacob Goldstein. 0:01:14.236 --> 0:01:16.556 This is What's Your Problem, the show where entrepreneurs and 0:01:16.596 --> 0:01:18.756 engineers talk about how they're going to change the world 0:01:18.956 --> 0:01:24.116 once they've solved a few problems. My guest today is 0:01:24.236 --> 0:01:29.236 Luis Vonon, the founder and CEO of Duolingo. Duolingo is 0:01:29.276 --> 0:01:32.596 both a wildly popular language app and also a hardcore 0:01:32.636 --> 0:01:36.196 tech company built on machine learning. Luis used to be 0:01:36.276 --> 0:01:39.356 a professor of computer science at Carnegie Mellon, and in 0:01:39.356 --> 0:01:42.796 our conversation he was really candid about the technical limits 0:01:42.836 --> 0:01:45.996 of what Duolingo can do today. The app is good 0:01:46.036 --> 0:01:48.676 at teaching people to read and to understand, he said, 0:01:49.196 --> 0:01:51.836 but Duolingo is not as good at teaching people to 0:01:51.996 --> 0:01:55.596 speak a new language. And solving that problem turns out 0:01:55.636 --> 0:01:59.556 to be part of this great, big, interesting frontier problem 0:01:59.596 --> 0:02:02.316 that is relevant not just for Duolingo, but for the 0:02:02.316 --> 0:02:07.436 whole field of artificial intelligence. We started out talking about 0:02:07.476 --> 0:02:10.476 the origins of Duolingo, which go back to a different problem, 0:02:10.796 --> 0:02:13.556 one that Louis discovered before he'd ever heard of machine learning. 0:02:13.796 --> 0:02:15.636 It was a problem he saw all around him when 0:02:15.636 --> 0:02:19.836 he was growing up in Guatemala. I was fortunate that 0:02:19.916 --> 0:02:23.236 my mother basically spent essentially her entire net worth on 0:02:23.396 --> 0:02:25.996 my education, and so I was fortunately that I got 0:02:25.996 --> 0:02:27.836 a good education. But then I could see the people 0:02:27.836 --> 0:02:31.476 who would get public education barely learn how to read 0:02:31.476 --> 0:02:33.196 and write. This is just what would happen, And you 0:02:33.196 --> 0:02:35.436 cannot expect that these people are going to become, you know, 0:02:36.796 --> 0:02:38.996 the CEO of a public company or anything like that, 0:02:39.156 --> 0:02:42.996 because they kind of wont. Often people talk about education 0:02:43.276 --> 0:02:47.236 as an engine for reducing inequality, but what you're describing 0:02:47.316 --> 0:02:50.916 is the exact opposite, and it's education, when you have 0:02:50.956 --> 0:02:54.156 to pay for it, is a mechanism for perpetuating inequality. 0:02:54.196 --> 0:02:55.956 And I really believe that, and I believe that's true 0:02:55.996 --> 0:02:57.996 in most countries in the world. There may be some countries, 0:02:58.036 --> 0:03:00.476 you know, like the Scandinavian countries, where pretty much everybody 0:03:00.476 --> 0:03:03.436 gets the same education everything, right, Yeah, there may be 0:03:03.516 --> 0:03:05.796 some countries like that, But in the vast majority of countries, 0:03:06.276 --> 0:03:09.116 if you have money, you can get a much better education. 0:03:09.316 --> 0:03:11.156 So I wanted to do something that would give equal 0:03:11.156 --> 0:03:14.156 access to education to everybody, and so we started with that. 0:03:14.356 --> 0:03:18.116 But then we started thinking, Okay, if education is pretty general, 0:03:19.356 --> 0:03:22.476 let's start by teaching one thing. Eventually we settled on 0:03:22.516 --> 0:03:26.076 teaching languages for a number of reasons, the biggest one 0:03:26.076 --> 0:03:31.196 of which is that learning English in particular can completely 0:03:31.276 --> 0:03:34.116 change people's lives. If you know English, you can double 0:03:34.156 --> 0:03:37.276 your incompotential in most countries. Wow, it's just as simple 0:03:37.316 --> 0:03:39.596 as that. And so it's it's why why is that. 0:03:39.836 --> 0:03:43.596 Basically it opens up for almost any job. You can 0:03:43.636 --> 0:03:46.036 get a better version of that job. For example, you 0:03:46.036 --> 0:03:48.396 could be a waiter, or you could be a waiter 0:03:48.436 --> 0:03:52.356 at the five star hotel. You could be an executive assistant, 0:03:52.556 --> 0:03:55.356 or you could be an executive assistant for a multinational 0:03:55.676 --> 0:03:59.676 ceoyeah yeah, okay, So really teaching English the core, the 0:03:59.716 --> 0:04:05.116 core sort of reducing inequality dream of a language is 0:04:05.196 --> 0:04:10.316 really teaching English to people in largely in poorer countries. Yes, 0:04:10.516 --> 0:04:12.676 so what we wanted to do was teach teach English. 0:04:12.676 --> 0:04:13.916 But you know, if you're going to teach English, we 0:04:13.916 --> 0:04:16.476 may as well teach other languages, So teach those and 0:04:17.196 --> 0:04:21.316 do so for free. So that's what Luis did and 0:04:21.396 --> 0:04:24.796 it worked. Today, tens of millions of people use Dual 0:04:24.876 --> 0:04:27.716 Lingo every month to learn English and dozens of other 0:04:27.796 --> 0:04:31.116 languages for free. The company makes money by selling ads 0:04:31.116 --> 0:04:34.636 and premium subscriptions. It went public in twenty twenty one 0:04:34.676 --> 0:04:37.476 and is currently worth billions of dollars. And the company 0:04:37.556 --> 0:04:40.276 really is built on machine learning. Luis gave me a 0:04:40.276 --> 0:04:43.236 few key examples of the way the company uses the technology. 0:04:43.716 --> 0:04:45.196 So let me tell you a few of the things 0:04:45.196 --> 0:04:48.676 that we do. One of the things that we do 0:04:48.796 --> 0:04:53.556 we like very much is we have data on whenever 0:04:53.596 --> 0:04:57.796 people use dual Lingo. We record every exercise that they 0:04:57.876 --> 0:05:00.756 do and whether they got it right or wrong, and 0:05:00.756 --> 0:05:02.476 if they got it wrong, why they got it wrong. 0:05:02.516 --> 0:05:04.956 With all of this data, we're able to do certain 0:05:04.996 --> 0:05:08.196 things with artificial intelligence. For examples, for every exercise that 0:05:08.236 --> 0:05:10.196 we're about to give you able to predict what is 0:05:10.196 --> 0:05:12.316 the probability that you're going to get this exercise right 0:05:12.396 --> 0:05:14.876 or wrong. So, in a sense, that is a thing 0:05:14.956 --> 0:05:19.436 that a teacher in a classroom could do fairly easily, 0:05:19.516 --> 0:05:23.916 right a teacher with twenty students, But you're able to 0:05:23.916 --> 0:05:26.996 do it with whatever how many people use your app 0:05:27.036 --> 0:05:30.396 actively forty two million per month, so the machine can 0:05:30.436 --> 0:05:32.956 do that for all forty two million people at the 0:05:32.996 --> 0:05:36.356 same time. More or less, yes, and very accurately. Part 0:05:36.396 --> 0:05:38.996 of the secret source of Duelingo is that we realized 0:05:39.076 --> 0:05:41.716 if we were to only give you things that you're 0:05:41.756 --> 0:05:44.476 not very good at, we'd basically be giving you lessons 0:05:44.516 --> 0:05:47.236 from hell every time. So we can't do that because 0:05:47.236 --> 0:05:50.316 that frustrates users. So what we do is, whenever you 0:05:50.356 --> 0:05:52.436 start a listening and doing we're actually trying to optimize 0:05:52.436 --> 0:05:54.236 for two things at the same time. We're trying to 0:05:54.276 --> 0:05:56.916 teach you things you know that you don't not very 0:05:56.916 --> 0:06:00.276 good at, but also we're trying to keep you motivated 0:06:00.316 --> 0:06:03.316 and engaged. Yea. And the way we do that is 0:06:03.396 --> 0:06:06.276 we try to give you exercises for which we know 0:06:06.716 --> 0:06:09.676 you have about an eighty percent chance of getting them right. Huh. 0:06:10.396 --> 0:06:12.156 And have you found that to be the sweet spot? 0:06:12.156 --> 0:06:14.396 I mean, have you done like experiments and sort of 0:06:14.436 --> 0:06:16.356 turn the dial. We've done that, and we're you know, 0:06:16.356 --> 0:06:17.716 we're not the first to figure this out. I mean, 0:06:17.756 --> 0:06:20.236 there's a lot of literature and psychology, etc. Just and 0:06:20.276 --> 0:06:22.076 the number is not exactly eighty percent. It's a little 0:06:22.156 --> 0:06:23.756 higher than that. It's like eighty three percent or something. 0:06:23.756 --> 0:06:26.116 But there's there's a number, and it really is the 0:06:26.156 --> 0:06:28.116 case that if that number is higher, that means these 0:06:28.116 --> 0:06:30.276 things are a little easier for you. Then you get 0:06:30.316 --> 0:06:32.236 a little bored. You feel like you're not learning right. 0:06:32.236 --> 0:06:34.396 If I'm getting ninety five percent right, I'm like, what, 0:06:34.476 --> 0:06:36.676 I'm just wasting my time? And you feel bored because 0:06:36.756 --> 0:06:38.196 it's like it's like a game that you always win. 0:06:38.316 --> 0:06:40.396 I mean, that's that's nice. At the very beginning, but 0:06:40.436 --> 0:06:42.996 then you're just not going to play it, and then 0:06:43.036 --> 0:06:45.276 if it's lower than that, that means that things are 0:06:45.276 --> 0:06:47.236 too hard for you. You get very frustrated and you 0:06:47.276 --> 0:06:49.996 go away. And there's a lot of tricks that you know, 0:06:50.596 --> 0:06:53.476 certainly app developers play, and you know we play as well. 0:06:53.516 --> 0:06:55.676 So I'll tell you another kind of similar trick. You know, 0:06:55.876 --> 0:06:57.996 we end up applying it to language. But the easiest 0:06:57.996 --> 0:07:00.796 way to understand this trick is with a slot machine. 0:07:02.556 --> 0:07:05.116 When you get two out of three, it's you almost 0:07:05.156 --> 0:07:06.956 got it, you gotta do one more, you gotta do 0:07:06.996 --> 0:07:08.716 one more. You just gotta do one more because you 0:07:09.236 --> 0:07:11.316 got it. So there's this, there's this you're so close 0:07:11.356 --> 0:07:13.516 to psychological trick that we played. It's like, oh, there's 0:07:13.516 --> 0:07:15.876 two out of three, almost got it, But you knew 0:07:15.916 --> 0:07:17.796 I was going to get two out of three. Yes, sure, 0:07:17.876 --> 0:07:19.756 you gave me two easy ones and when there was 0:07:19.796 --> 0:07:22.796 super hard that's exactly right. So so we we played 0:07:22.796 --> 0:07:25.556 this type of trick where just people are like, almost 0:07:25.636 --> 0:07:27.436 got it, and that gets them to do another one. 0:07:27.476 --> 0:07:29.476 So you know, in our case, we just we basically 0:07:29.516 --> 0:07:33.356 spend a lot of time training computers to figure out 0:07:33.396 --> 0:07:35.916 what it is that makes people use Duolingo for longer, 0:07:35.996 --> 0:07:39.596 and also that we teach them more so that that's 0:07:39.596 --> 0:07:42.596 a major use for artificial intelligence. Main use is just 0:07:42.676 --> 0:07:47.676 in teaching better. After the break a big problem, Louise 0:07:47.756 --> 0:07:50.596 and dual Lingo are still trying to solve a problem 0:07:50.596 --> 0:07:53.196 that turns out to be a big frontier problem for 0:07:53.316 --> 0:08:02.516 all of artificial intelligence. That's the end of the ads. 0:08:02.956 --> 0:08:05.316 Now we're going back to the show. So let's talk 0:08:05.436 --> 0:08:08.716 now about problems you haven't solved yet. You know, like, 0:08:08.756 --> 0:08:10.796 what are you what are you trying to figure out? 0:08:10.836 --> 0:08:13.236 What are you working on that that isn't quite working yet. 0:08:13.396 --> 0:08:15.476 So dual linguals is very effective at teaching you all 0:08:15.556 --> 0:08:18.236 kinds of things. But if you go look under the 0:08:18.236 --> 0:08:19.756 hood or you know, what is it that you're learning. 0:08:19.796 --> 0:08:23.636 You're learning reading really well. You're learning writing pretty well, 0:08:23.636 --> 0:08:26.796 but not as well as reading. You're learning listening pretty well, 0:08:27.916 --> 0:08:31.876 but you're not learning spontaneous speaking very well. This, by 0:08:31.916 --> 0:08:33.916 the way, is also you're not something you're not learning 0:08:33.996 --> 0:08:36.076 very well in university semesters, Like you're basically not learning 0:08:36.116 --> 0:08:39.316 that well either in due lingo or in university semesters. Okay, 0:08:39.716 --> 0:08:42.196 it's just harder to teach in a sort of classroom. 0:08:43.916 --> 0:08:47.276 What you need to do to teach that is basically, 0:08:47.316 --> 0:08:51.276 have you really interact with wealth? For now another human 0:08:51.756 --> 0:08:55.076 and just you just practice that a lot. Now, here's 0:08:55.156 --> 0:08:57.596 the here's the thing about that. I know how to 0:08:57.596 --> 0:08:59.756 get you to interact with another human. Just put another 0:08:59.836 --> 0:09:04.676 human there. The problem is about eighty percent of our 0:09:04.756 --> 0:09:07.596 users just does not want to talk to a stranger 0:09:07.796 --> 0:09:10.116 in a language that they're not very good. So the 0:09:10.716 --> 0:09:13.676 problem that we're trying to solve here is how do 0:09:13.756 --> 0:09:18.756 we practice kind of spontaneous conversation but without having a 0:09:18.836 --> 0:09:22.076 human on the other side. And we've been working on that, 0:09:22.396 --> 0:09:24.796 and you know we're not there yet. Can I just interrupt? 0:09:24.836 --> 0:09:28.076 Because I mean we were talking about artificial intelligence? Right? 0:09:28.156 --> 0:09:30.756 The most famous test that I know of, the most 0:09:30.756 --> 0:09:33.756 famous idea I know of of artificial intelligence in a 0:09:33.796 --> 0:09:38.356 computer is can a computer hold one end of a conversation? Right? Like? 0:09:38.396 --> 0:09:41.876 That's the classic touring test is like you're going to 0:09:41.956 --> 0:09:45.796 have this like chat conversation and can you tell if 0:09:45.836 --> 0:09:48.156 the person on the other end is a person or 0:09:48.156 --> 0:09:51.556 a machine? Like? That's the og artificial intelligence idea, right, 0:09:51.556 --> 0:09:53.196 I mean, are you telling me, that's what you're trying 0:09:53.236 --> 0:09:56.716 to solve. Not quite. I mean, it would be awesome 0:09:56.756 --> 0:09:58.836 if we would solve that. I mean, but that's the dream, 0:09:58.956 --> 0:10:01.476 right solution to that, that is the dream. But notice 0:10:01.516 --> 0:10:03.516 in our case, we don't actually care if the human 0:10:03.556 --> 0:10:06.036 can tell that there's a computer on the other side. Okay, 0:10:06.516 --> 0:10:09.516 it's okay. As long as it practices thing and as 0:10:09.516 --> 0:10:11.756 long as you're able to carry on a conversation in 0:10:11.756 --> 0:10:14.196 a way that seems a little natural or something, it's okay. 0:10:14.236 --> 0:10:18.076 If if it, you know, goes off the rails every 0:10:18.116 --> 0:10:20.516 now and then, So tell me, what is it that 0:10:20.556 --> 0:10:22.836 you're trying to build. This is exciting, Like what are 0:10:22.836 --> 0:10:25.276 you trying to do? We're starting with text, by the way, 0:10:25.476 --> 0:10:28.796 so either just basically a texting conversation. So think of 0:10:28.836 --> 0:10:31.436 it as like a chat bot in uh, you know, 0:10:31.436 --> 0:10:34.196 in Spanish, where it just you're just having a real 0:10:34.636 --> 0:10:36.676 a little conversation. You've had lots of people have had 0:10:36.676 --> 0:10:39.596 experience with chat bots. Yes, they're right, Like you go 0:10:39.636 --> 0:10:44.756 to whatever, cancel your cable and they want you to text, 0:10:44.796 --> 0:10:46.876 and then you realize you're texting with the machine. So 0:10:46.956 --> 0:10:50.756 like that's the that's step one. So that's the idea. 0:10:50.956 --> 0:10:53.436 That's step one. We of course, I mean a lot 0:10:53.476 --> 0:10:58.116 of those experiences with chatbots are are very um, they're 0:10:58.156 --> 0:11:00.796 just very geared at whatever it is you're trying to do. So, 0:11:00.836 --> 0:11:03.236 for example, that chapel maybe very good at at canceling 0:11:03.316 --> 0:11:06.036 your cable, but only that in my experience, they're not 0:11:06.076 --> 0:11:09.316 even good at that. No, they're not that great. So 0:11:09.316 --> 0:11:10.916 we're trying to do that, and you know, we're not 0:11:10.956 --> 0:11:13.756 there yet. I don't think so you're like out on 0:11:13.796 --> 0:11:16.156 the frontier. We are. We are like you're trying to, yes, 0:11:16.276 --> 0:11:18.396 and we're not there yet. I mean, this is something 0:11:18.396 --> 0:11:20.276 that's going to take us, not just us, I mean 0:11:21.396 --> 0:11:25.076 the whole academic community and technology just a few more years. 0:11:25.516 --> 0:11:28.156 But so let me ask you this. Can we talk 0:11:28.196 --> 0:11:30.556 about that in a way that would be like, can 0:11:30.596 --> 0:11:33.596 we try and just go one level into sort of 0:11:33.636 --> 0:11:35.756 what you're trying to do and like what works and 0:11:35.796 --> 0:11:38.796 what doesn't work, and like why it's hard? Yeah, I mean, 0:11:39.116 --> 0:11:41.076 you know, the first, by the ways, the first way 0:11:41.116 --> 0:11:42.636 you think of if you're trying to make a chap 0:11:43.116 --> 0:11:46.156 the first thing you think of is, okay, I'm just 0:11:46.196 --> 0:11:49.396 going to program the computer. Forget about artificial intelligence. I'm 0:11:49.396 --> 0:11:54.716 just going to program the computer to respond to specific questions, 0:11:55.036 --> 0:11:58.796 and how many possible questions could there be? You start thinking, okay, well, 0:11:59.236 --> 0:12:02.516 when the person says high, we're going to program a 0:12:02.516 --> 0:12:04.996 think to say hi back. When the person says, how 0:12:05.036 --> 0:12:06.756 are you doing, We're going to program I think to 0:12:06.756 --> 0:12:09.476 say I'm doing pretty well? How about you? Yeah? V 0:12:09.716 --> 0:12:11.716 zero of a chatbot. And this, you know, this comes 0:12:11.756 --> 0:12:13.396 from you know, fifty years ago. This is what you 0:12:13.436 --> 0:12:15.836 start doing. The problem is there's billions of things that 0:12:15.836 --> 0:12:18.716 people can say, and so we may have programmed the 0:12:18.756 --> 0:12:20.516 thing of what to say, but you know how you're doing, 0:12:20.876 --> 0:12:23.396 and we can respond. But if instead of asking that, 0:12:23.436 --> 0:12:26.596 they may ask like, hey, did you watch the game 0:12:26.676 --> 0:12:28.956 last night? And we just have no idea how to 0:12:28.996 --> 0:12:32.236 respond to that. About a decade ago, Louis says Ai, 0:12:32.276 --> 0:12:36.276 researchers started trying a really different approach. Rather than trying 0:12:36.276 --> 0:12:40.516 to teach computers every rule, they started throwing massive amounts 0:12:40.516 --> 0:12:44.036 of documents and texts at computers and essentially telling the 0:12:44.036 --> 0:12:48.116 computers figure out the patterns in all these documents. So 0:12:48.196 --> 0:12:50.916 when somebody writes something like did you watch the game 0:12:50.996 --> 0:12:53.796 last night? The computers should be able to predict what 0:12:53.916 --> 0:12:57.516 kinds of answers might follow. This strategy clearly has not 0:12:57.836 --> 0:13:00.836 entirely worked yet. That's why it's still a problem solving. 0:13:00.836 --> 0:13:04.156 It will take both more text and more clever algorithms 0:13:04.196 --> 0:13:08.036 to help computers make sense of that text. But Louis says, 0:13:08.316 --> 0:13:11.636 you can see progress every time you open your Gmail 0:13:11.796 --> 0:13:13.636 or a Google Doc. And I don't know if you've used, 0:13:13.676 --> 0:13:16.756 for example, you use Google Docs lately or Gmail like 0:13:17.276 --> 0:13:20.236 it finishes off your sentences now. And basically the way 0:13:20.276 --> 0:13:23.516 this works is, you know, this system has looked at 0:13:24.356 --> 0:13:26.876 a ton ton of text that has been written by 0:13:26.876 --> 0:13:28.196 a lot of people. In the case of Google Docs, 0:13:28.196 --> 0:13:29.676 I actually don't know what they look at, but I 0:13:29.676 --> 0:13:31.676 wouldn't be surprised if they look at everything that has 0:13:31.676 --> 0:13:34.116 ever been written in Google Docs. I'm going to tell 0:13:34.156 --> 0:13:36.716 you one that happened to me in Google Docs today 0:13:36.756 --> 0:13:39.476 when I was typing notes for this interview, I typed 0:13:39.956 --> 0:13:43.556 zone of pr and then you know what, you know 0:13:43.556 --> 0:13:48.676 how it completed it proximal development. Yes, it knew I 0:13:48.756 --> 0:13:52.196 was going to write zone of proximal development. Yep. No, 0:13:52.316 --> 0:13:54.316 this is amazing, And they just see that if you'd 0:13:54.316 --> 0:13:57.516 write the zone of per there's like a ninety five 0:13:57.516 --> 0:14:00.996 percent chance that it ends in proximal development. What is 0:14:00.996 --> 0:14:05.516 the zone of proximal development? You know? In teaching, you know, 0:14:05.516 --> 0:14:07.956 there's this concept of just keeping you at this zone 0:14:07.956 --> 0:14:10.676 of proximal development, which is always kind of challenging you, 0:14:11.076 --> 0:14:12.716 giving you things that you don't know. But but there 0:14:12.756 --> 0:14:15.196 are all things that are fair to give you. Proximal 0:14:15.276 --> 0:14:17.676 means like close to or next two. Right. So it's 0:14:17.716 --> 0:14:19.716 the idea is like you know a thing, yes, then 0:14:19.756 --> 0:14:21.476 like what you want to teach the person is the 0:14:21.596 --> 0:14:24.396 very next thing, right, It's like that's right, It's like 0:14:24.436 --> 0:14:26.316 the frontier of your knowledge. I like it because it 0:14:26.356 --> 0:14:28.476 applies to like the way you teach, but also to 0:14:28.516 --> 0:14:32.036 your work, right, And like it's just a nice life idea, right, 0:14:32.076 --> 0:14:34.476 It's like the next thing you want, the next thing. 0:14:34.516 --> 0:14:36.276 And I feel like the chat bot is maybe a 0:14:36.356 --> 0:14:39.476 version of that at the level of your company. Yeah, yeah, 0:14:39.476 --> 0:14:41.156 it really is. It really is a nice idea, and 0:14:41.236 --> 0:14:43.196 it is I mean, and if you think about it, 0:14:43.556 --> 0:14:45.516 this is what a great teacher does. You know. I've 0:14:45.556 --> 0:14:48.916 said this inside the company at due Lingo. UM. All 0:14:48.996 --> 0:14:51.196 we need to do is first figure out what you know, 0:14:52.396 --> 0:14:53.876 by the way, not that easy to figure out what 0:14:53.916 --> 0:14:55.436 you know. But let's first figure out what you know 0:14:55.596 --> 0:14:58.116 and then just take you to that zone of proximal development. 0:14:58.156 --> 0:15:00.396 Because now we know what you know, just take you 0:15:00.436 --> 0:15:03.516 to the frontier and then just keep expanding it as 0:15:03.516 --> 0:15:05.476 fast as possible. That's all we need to do. Of course, 0:15:05.516 --> 0:15:09.196 this is easily said, hard to do. Yeah. Um, And 0:15:09.276 --> 0:15:11.716 is there a limit to what you can do with 0:15:11.756 --> 0:15:13.836 a computer? Is there anything a teacher can do? Is 0:15:13.876 --> 0:15:17.196 that a computer will never be able to do? You know? 0:15:17.236 --> 0:15:19.236 Of course I do lingual We love teachers. If they 0:15:19.236 --> 0:15:22.156 are a good teacher and also have the time, they 0:15:22.196 --> 0:15:24.116 are much more able to adapt to their students than 0:15:24.156 --> 0:15:28.556 a computer is. Um. But I don't believe that will 0:15:28.556 --> 0:15:30.596 always be the case. I mean, I think at some 0:15:30.636 --> 0:15:32.756 point it's not just teachers. I mean teachers, this is 0:15:32.796 --> 0:15:35.756 one thing. I mean, at some point my belief and 0:15:35.756 --> 0:15:39.116 this is of course just my belief. People, not everybody agrees. 0:15:39.556 --> 0:15:41.396 They believe that computers will be able to do every 0:15:41.436 --> 0:15:44.436 single thing that humans can. Now you may start asking 0:15:44.476 --> 0:15:47.516 really tough questions like can they love? Yeah, I don't know, 0:15:47.756 --> 0:15:49.076 I don't know what they can love or not but 0:15:49.196 --> 0:15:51.636 from the outside it will look just as if they 0:15:51.716 --> 0:15:54.836 love so who knows who knows what's going on inside? 0:15:54.836 --> 0:15:56.636 Who knows that they that's like a big yeah, we're 0:15:56.756 --> 0:16:00.636 big philosophical questions that I'm not here today, and that's right, 0:16:00.676 --> 0:16:03.036 nor am I. But I do think from input output behavior, 0:16:03.116 --> 0:16:05.356 I don't see why. I don't see any reason why 0:16:05.396 --> 0:16:07.316 computers won't be able to do everything that humans can. 0:16:07.876 --> 0:16:11.036 So they can teach, but they can also write a 0:16:11.156 --> 0:16:13.516 computer code. They can also run companies, they can also 0:16:13.556 --> 0:16:16.356 make podcasts, they can do everything. Should be able to 0:16:16.396 --> 0:16:17.996 do that. I think they should be able to do that. 0:16:18.236 --> 0:16:21.396 I don't know when that'll happen, but they should be 0:16:21.436 --> 0:16:25.796 able to do that. In a minute, the lightning round, well, 0:16:25.836 --> 0:16:28.196 hear what job Luise would love to do but thinks 0:16:28.196 --> 0:16:30.836 he wouldn't be very good at. And the real reason 0:16:30.956 --> 0:16:39.916 treasure Chess keep showing up in duo lingo. And now 0:16:40.036 --> 0:16:41.876 back to the show. We're going to finish with a 0:16:41.956 --> 0:16:45.036 lightning round, not counting duo Lingo. What's your favorite app 0:16:45.076 --> 0:16:48.156 on your phone? Spotify? What have you been listening to 0:16:48.276 --> 0:16:51.396 on Spotify? I'm always a huge fan of the band 0:16:51.436 --> 0:16:56.556 called Churches with a v to Virchase. So that's what 0:16:56.636 --> 0:16:58.596 I was listening to this morning on my work at work. 0:16:58.716 --> 0:17:00.236 If you have a ten minute break in the middle 0:17:00.276 --> 0:17:03.356 of the day, what do you do to relax? Played 0:17:03.356 --> 0:17:06.956 this game called Class Royale. We are a lot of 0:17:06.996 --> 0:17:09.596 the gaming mechanics that we use for duel and will 0:17:09.636 --> 0:17:14.596 come from gaming companies, like the treasure chests, exactly right, 0:17:14.796 --> 0:17:16.956 the treasure chests. If you ever played Class Royale, they 0:17:16.996 --> 0:17:19.996 have the treasure chests. If somebody's going to go to 0:17:20.516 --> 0:17:22.876 visit Guatemala for the first time, what's one thing they 0:17:22.916 --> 0:17:28.476 should definitely do? Oh? Um, Decal is the Mayan ruins. Um. 0:17:28.756 --> 0:17:30.716 You know, if I feel very strong, I've been to 0:17:30.956 --> 0:17:34.276 southern Mexico where they have chi Chenitsa. It's a joke 0:17:34.396 --> 0:17:37.436 compared to the Mayan ruins in Guada. There's there's one 0:17:38.316 --> 0:17:41.916 pyramid in Chichenitsa. There are four hundred in Guatemala in Decal, 0:17:42.156 --> 0:17:45.836 So yeah, they should like I like that. Not only 0:17:45.836 --> 0:17:51.356 are you recommending Tikal, you're also taking as I have 0:17:51.436 --> 0:17:53.876 no trouble with Chichenitsa. It's just they are very good 0:17:53.916 --> 0:17:58.356 at marketing. Amazing. What would you do if you couldn't 0:17:58.876 --> 0:18:01.836 do the job you do. Now, well there's what what 0:18:01.956 --> 0:18:03.956 would I actually do? On? What would I'd like to do? 0:18:03.956 --> 0:18:05.396 I would love to be a writer. I don't think 0:18:05.396 --> 0:18:07.956 i'd be a very good one. Um So if I 0:18:07.956 --> 0:18:10.836 if I wasn't doing the job that I'm doing right now, 0:18:11.396 --> 0:18:13.436 you know, I'd probably be back to being a professor. 0:18:14.236 --> 0:18:16.956 How will you know when it's time to retire? I'm 0:18:16.996 --> 0:18:22.556 never retiring, That's what everybody says. Well, maybe I will, 0:18:22.556 --> 0:18:24.316 but I mean right now, I don't. I don't want 0:18:24.356 --> 0:18:29.276 to do that. Luis Vaughan is the founder and CEO 0:18:29.436 --> 0:18:33.596 of Duelingo. Today's show was produced by Edith Russelo. It 0:18:33.636 --> 0:18:36.436 was edited by Kate Parkinson Morgan and Robert Smith, and 0:18:36.476 --> 0:18:39.276 it was engineered by Amanda kay Wong. Theme music by 0:18:39.356 --> 0:18:42.116 Louis Kara. Our development team is Lee, Tom Mulad and 0:18:42.196 --> 0:18:45.076 Justine Lang. A huge team of people makes What's Your 0:18:45.116 --> 0:18:48.276 Problem possible. That team includes, but is not limited to, 0:18:48.916 --> 0:18:52.236 Jacob Weisberg, Mia Lobel, Heather Fain, John Schnars, Kerry Brodie, 0:18:52.276 --> 0:18:56.636 Carli mcgleory, Christina Sullivan, Jason Gambrell, Brand Hayes, Eric Sandler, 0:18:56.676 --> 0:19:00.276 Maggie Taylor, Morgan Ratner, Nicolemrano, Mary Beth Smith, Royston Deserve, 0:19:00.356 --> 0:19:03.996 Maya Kanig, Daniello, Lakhan, Kazia Tan and David Clever. What's 0:19:03.996 --> 0:19:07.676 Your Problem is a co production of Pushkin Industries and iHeartMedia. 0:19:07.796 --> 0:19:11.036 To find more Pushkin podcast Us, listen on the iHeartRadio app, 0:19:11.196 --> 0:19:15.196 Apple Podcasts, or wherever. I'm Jacob Goldstein and I'll be 0:19:15.236 --> 0:19:24.156 back next week with another episode of What's Your Problem.