Mait Müntel: From discovering the Higgs-Boson to language learning
Kris Broholm interviews Mait Müntel, a physicist who helped discover the Higgs boson at CERN. They discuss his transition from nuclear physics to founding the Linguist language learning app. Mait explains how his scientific background informs his approach to language acquisition and designing a tool for faster learning.
August 31, 2019 · 53 min
EP 163: Mait Müntel: From discovering the Higgs-Boson to language learning
0:00 / 52:41
Guest
Mait Müntel
What You'll Hear
- 1 Podcast Welcome & Disclosure Dilemma
- 2 From Nuclear Physics to Language Learning
- 3 Mission: Accelerate Human Learning 10x
- 4 How Linguist's AI Personalizes Learning
- 5 Expanding Beyond Vocabulary: Course Wizard
- 6 Measuring True Learning, Not Gamification
- 7 Future of AI and Universal Accelerated Learning
- 8 Connect with Linguist & Future Revelations
Links Mentioned
Questions Covered in the Show
What is Linguist's core mission?- 6m
Why did a nuclear physicist create a language learning app?- 8m
How did Linguist estimate it's possible to learn a new language in 200 hours?- 13m
How does Linguist teach grammar?- 14m
What makes Linguist different from other flashcard apps?- 17m
Why doesn't Linguist use gamification to motivate users?- 25m
Related Episodes
See all →
Welcome to the Actual Fluency Podcast. Each week you'll find inspirational motivational interviews with some of the world's best language learners, industry experts, all trying to help you to learn foreign languages better, faster, and more efficiently. And here we go. If you're looking for a language teacher to enhance your language learning, then I highly recommend italki. italki is the world's biggest tutoring platform, and you can find thousands of teachers and tutors at very reasonable prices. Get a free lesson after completing your first lesson by going to languageteacher.co.
Hello everyone, and welcome back to the Actual Fluency Podcast. Really excited to be back with another episode for you. And today, the first thing I just want to chat for a few minutes about kind of the nature of today's episode, because there is a, there's a dilemma in my mind, and if you're not really interested in this kind of discussion, then I suggest you just, just move straight to the episode itself, which starts, you know, a few minutes from now, you just scroll forward and find it.
But today, I am talking to Maite, the founder, CEO of Linguist, which is a language learning app. It's an online as well, online application, also for the phone. And it's very sophisticated flashcards. And, you know, Maite was such an interesting character to talk to, because normally, in language learning, people have a pretty direct route to languages, you know, either they're in the educational space, or they're in the linguistic space, or in the, you know, even in startup space, but Maite, he was part of the team who discovered the Higgs boson at CERN. He's basically a nuclear physicist. I was like, what? How does a nuclear physicist get into languages?
So we got to talk a little bit about that, that's very interesting. And also, I, I kind of asked him a lot of questions about how a high-level physicist would see language learning. Like, how do you break that down and turn that into an application that can hyper-charge human performance, which is really what, what Linguist is trying to do. And they're trying to build a model of how we can learn things faster, because if, if humans in general can learn things faster, then we can save more time, we can be more effective, we can be more efficient. And that's kind of their model, or their mission statement, and they're starting out with languages. But you'll hear much more about it in the episode.
What I just want to disclose first of all is that I do have a professional relationship with Linguist. They are paying me to do some advertising of their product on Actual Fluency. Not this episode specifically, I just want to make that clear. They haven't paid me to, uh, you know, they haven't paid to be on the podcast. It doesn't work like that. And that's why I want to chat with you for a few minutes today, because there is a, there's a balance in, in, in the language world and in podcasting. And it's an interesting balance, but my dilemma is basically that there are a lot of companies in the language world that I want to support, and I want to kind of highlight and give them a chance to share their vision, their story. But at the same time, I want it to be interesting for you as a listener. I want you to take something away from it. I want, I don't want it to be like an hour-long sales pitch.
Sometimes the episodes that I have done with companies in the past have probably crossed over into that, just because the, the founder wasn't interesting enough, or the stories that we went into weren't really that deep. And obviously, yeah, that's, that's my failure as a host. I should really have been better. But this episode, I felt like we had a really good discussion. The only reason that I kind of bring out this disclaimer in the beginning is, um, that we do, everything we talk about is surrounding Linguist. And it wasn't my intention originally. And I, every time I invite people on, I always make it clear, you know, this is supposed to be about the person. And we can, we can definitely talk about any kind of projects. You're free to mention names or anything. It's not like, you know, we're not putting an embargo on, on sponsorships. It's not, we're not censoring, because we want to support everyone, and we want to help each other grow stronger in the, in the industry, and that way we all stand to benefit.
And actually, I think Linguists are doing some really cool stuff that, I mean, you'll make your mind up yourself when you go through it and you check out the website, of course. But I just wanted to tell you that, and just let you know that we are working together professionally. But that was after the episode was recorded, first of all. And like I said, it was, they didn't pay to be on the podcast. Um, in the future, I definitely want to feature more companies. So if you have an interesting language learning company and you want to come on the podcast and, and, you know, preferably in, in the view of an interesting story, actually also a few episodes ago, uh, Jonty, who I just met in, in Montreal, great guy, came on the show to talk about his, uh, his story of learning his heritage language, and developing an app to do that himself, and, and obviously trying to build that app into a more universal app that everyone could use for any language. He also mentions that in the episode, and obviously, gains some publicity for that. That's not something that was, um, you know, that's, that's not edited out or anything. I, I want, I genuinely want to support people with the, this podcast. But obviously, it, it can't be like a one-hour pitchfest either, you know, we want to put a good balance in.
And I, um, I have to mention italki as well, in this breath. They've, they've just been so supportive, and they're sponsoring the, the podcast, as you heard in the beginning. And, um, so there are, there are professional relationships, but just, just want to clarify, or maybe in case some people are wondering, I would never promote or talk about companies that I had any kind of doubts about. So, if they were doing shady things, if they were, if they were doing, trying to rip off other people, or, you know, they, basically, basically, if there were any kind of red flags, I wouldn't feature them on the show, and I certainly wouldn't publish an episode that was pretty much all focused around what they were doing as a company. So, I hope you enjoy it, and I hope that made sense, and sorry for waffling on a bit there, and, uh, can't wait to see you in the next episode.
Alright, Maite, well, welcome to the Actual Fluency Podcast. I'm excited to talk to you today and learn a little bit more about what you're up to. But first, do you want to give the listeners a little background to who you are, where you're from, and how you ended up in Linguist?
Uh, hello, everybody. Yes, I might. Thanks Kristof for inviting me to this podcast. Uh, I'm that kind of person who is a, a little bit embarrassed to talk about myself. So you probably have to help me with some questions.
No worries.
Uh.
So, what, how did, what's your backstory? Like, how did you end up where you are today?
And today, I'm building a startup. A language learning startup, which has actually, uh, a big mission, uh, to accomplish. Our goal is to accelerate human learning, and, uh, language learning is just the first step of it.
Mm.
Uh, But what, what got me into this journey was that, uh, uh, I didn't know languages very well. I had hard times learning them at school. I learned Russian. Everybody in the post-Soviet countries had to learn Russian for 10 years, and, uh, uh, 99.9% of them didn't succeed.
Right.
And I was, I was, I was one of them. So, like a huge social experiment, how to learn things so that you don't learn anything.
Right.
And, uh, and, uh, I had this hard journey with English as well, and, uh, then I thought that, uh, uh, I would better study physics. Something that I didn't have to learn much to do the exercises, and, uh, I became a physicist, and somehow I ended up working at the Centre of European Nuclear Research, uh, CERN, maybe many people have heard about it, because, uh, uh, four years ago, no, six years ago, that was one of the biggest discoveries in this century. Uh, we discovered the Higgs boson, the so-called god particle. And, uh, yeah, this is my background. It's a funny,
Yeah.
Funny jump from physics to a language learning application.
You know, funny story, uh, I'm totally opposite. So, when I did physics in high school, it was by far the worst subject that I ever did, and I never, because also I'm very weak at math. Like, I can do some basic calculations and stuff, but when you got into like the integrals and the calculus and all that, I was like, no way, I don't know any of this. And so, physics, as you can imagine, was just a one long struggle. But to me, languages were always easy, because, well, it's just the way people speak, you know, it's like you go out to the street, that's how it is. But in physics, it's all this like theories and you're, you're telling me there's all these invisible stuff going around, you know, it's just so theoretical that I just never really got into it. I like the cool experiments, but, uh, you know, the, the actual learning I just stayed away from. So, that's pretty funny. But also, like, that must be quite a jump then if you, but what made you return to the language? You say you want to kind of improve human learning, but you could have chosen any topic, I assume.
Hey, yes, it's, it's, it's a funny thing that, um, um, if you have a problem, then you try to solve your problem. And if I, if I didn't manage to learn languages at school, then it was such a problem to me that I thought that I have to solve it. So, maybe at some point of your life, you decide that you will start doing physics startups, you never know.
Well, maybe you make some language software down the road. That would be quite a nice turnaround.
Yes. But actually, yeah, in the long term, we may drift to that direction as well, because, um, what I really got interested in was that, um, if people learn some things for a long time and they do not succeed, and any person has some kind of, um, um, memories that he tried to learn something, it was super hard, didn't achieve, spent hours and hours, was very demotivated.
And, um, uh, if I started to think about the time that people spent on learning. I thought that, now when we have very cool technology, when we have machine learning, when we can, uh, discover unknown particles in the huge mess of data at CERN, um, we have these tools, and actually, would it be possible to utilize these tools to accelerate human learning?
And I started to calculate that if I would have a magic tool that helps me to make decision, when and what exactly I have to review and repeat, how quickly it would be possible to learn a new language. I, I did some statistics, and it was CERN was a fantastic place to do it, because CERN had to, one of the biggest computer, um, resources in the world. And, and, and even if I run some, some of my hobby projects there, then, um, I had some resources that, otherwise, people do not have, and also some skills to analyze the data.
So, I, I thought that, if I want to learn, for example, French as a new language, then it would be important for me to know, uh, what the language actually is, what actually the words that people are using. Because if you take a textbook, then there are thousands of words, but you do not know whether they are actually the most important ones to learn right now. Um, um, if you, for example, take a word, and this is a real example that I, I bought a, uh, a book, Learn French in 24 Hours. And in the first chapter, there was a conversation about working in a garbage recycling factory. And, uh, it, it sounded very odd to me. How often do I use this garbage recycling factory in my native language, or in English? Almost never.
Wow. Yeah, exactly.
And if I use the same time of learning this for something that is, like, beer, or even some curse words, then I would spend my time million times more efficiently. And if I always make this decision, then the total learning time should be much, much smaller.
And I, and I calculated that given all of the words which are out there, uh, how frequently they, uh, um, appear in the language, um, what is the capacity, my learning capacity, ideally? I figured out that it, it would be possible to learn a new language in 200 hours. And it was so small compared to the my previous experience of 10 years learning Russian without learning anything, and 20 years learning English, nothing still fluent. It's, it seems so amazing that it actually can be so short that I thought that I have to build a product for it, or test it out for myself. And then, then, then I started to build the software for myself. So, this is how hobbies take off.
Right. Yeah, exactly. But how did you, how did you account for things like grammar and like the other parts of the language, like the, that wasn't just the words?
I, I did, uh, um, pretty complicated stuff. Uh, uh, not all of it didn't work out. Um, uh, but basically, um, I can count the grammar in the same way that I, for example, take all of the sentences, and I translate them into grammatical concepts. So that I replace every word with some grammatical concept, and I, I get some, some, like a meta-sentence of grammar.
And I can do these frequency analysis about these grammar concepts as well. To see what kind of different grammar constructions are used. And I found out that actually the most frequent ones are not those that, um, uh, are used for the learning in the first year, normally. Like, if sentences and things that really help me to speak.
Right. Well, that's very interesting. And, and, and I'm sure a lot of people who are, who are in the technical space and things are perking up now with this discussion, because I, I, I think the flaw often when you see, let's say, language learning trying to be decoded or, or like simplified into a system or some kind of rule, often it seems to be very much heavily focused on individual words. And, and you're saying that, well, okay, if you, if you learn 2000 words, you are at a sort of a fluent level. But obviously, just learning 2000 words doesn't mean anything. It, you know, that's just facts you can pull from your brain, you know, anyone can learn 2000 words in probably a couple of weeks. But the, uh, the challenge is really producing language that you can open your mouth and use, and then understand what's coming back. So, it's quite exciting that you, you, you, you thought about that early on. Um, and, and that's obviously a while ago. So, so, what's, what's sort of the, how's the progress been? When was this, you said you, you started this?
Um, I started it six years ago, right after the discovery of Higgs boson. So, um, it happens when there is a big, big, uh, achievement that, uh, you have worked hard on and waited for for a long time, then after achieving this, you kind of need a different challenge or,
Right.
Try to explore new paths. Uh, but obviously, what I was doing back then, um, was a, like a hobbyist, uh, work. Um, because I was not professionally trained in language learning or natural language processing. Um, and now we have 100 times better people working on these problems.
I see.
It's, uh, uh, now looking back, it looks very childish what I did, but this is how, uh, uh, companies get started, that initially, some, you do something quite stupid, and then much better people come on board and they make it professionally.
Right. It's like the scientific method, you, you start out with the science and then you add all the, the professional models. But you wanna, can you break down sort of how Linguist works today in sort of simple terms? Like, how, how much does it differ from, let's say, I have a, a flash card app that shows me the result of a query on the other side, and then it shows me that card, you know, every, I don't know, four, and then 10, and then 20, and then 60, 90 minutes, you know, it keeps kind of a fixed program. How, how does Linguist differ, and, and, have you, how much has it evolved since that prototype?
Uh-huh. I think the main differentiator is, um, uh, the, uh, idea behind it, why we are doing it, and what we are trying to achieve, that, um, uh, what we're really building is, we're building technology to make human learning faster. Uh, and this is something that nobody else is doing, uh, uh, and if you look at the functionalities, it looks like a flashcard game or, um,
Yeah, exactly. On the outside.
Yeah, it, outside looks very, very simple. Uh, but, uh, what happens inside is that actually every single thing that people learn, uh, is analyzed so that, uh, uh, for every single item that you have learned, we know what is your memory profile. And these memory profiles are very different from person to person. Um, they, uh, depend on your previous knowledge, uh, uh, your memory parameters, every single review that you do changes the memory retention curves.
And all of the words, they are different difficulties, um, as well. So every, every single word has different difficulties of memory retention curves. But if people are learning with this, uh, in our product, then, uh, uh, we have all this data, and we can calculate exactly when and what they have to review to learn most efficiently.
And currently, we're very much focusing on vocabulary acquisition. There are more things to come into the product, but I can't tell when, because, uh, currently, it's supposed to be more like a additional tool when people, uh, are taking language classes, uh, or, uh, or they're living in the foreign environment, um, uh, it's, it's not really meant to be a stand-alone tool yet.
Right.
Um, But it, the main main differentiator is that, uh, uh, we use machine learning algorithms to make to accelerate the learning as much as it's humanly possible.
Personalized for every user.
Yes, yes.
Right. So I'm, I'm not getting recommendations based on the pool of, of data. I'm getting because I failed this 10 times in a row, it's going to slow it down and show it to me 100 more times.
Um, yeah, it's, it's very, very individualized. Um, but if you do not, uh, if you start learning, and we do not have much information about you, then, of course, we use other people's data as well to help to get you going faster.
Of course.
On this podcast, I've interviewed hundreds of language learners, some of the world's greatest polyglots and industry experts. And one thing they all agree on is the value of one-to-one tutoring lessons. And for this, I highly recommend italki. They have thousands of teachers in all price ranges, and they even have certified teachers who have taken diplomas or have degrees in the language you're learning. So, whether you're just brushing up on your Italian ahead of a trip to Rome, or you want to master Russian to take the, uh, exam or whatever your goals are in languages, italki has a tutor suitable for you. And compared to private tutoring offline, it's really affordable. You can find informal tutors down to $5 an hour, or and you can have trial lessons for even less. So, if you want to master a language, uh, from the comfort of your own home, and you even get a $10 credit when you complete your first lesson, go to languageteacher.co and check out italki. It might be the best thing you do for your language learning this year.
What are some of the challenges of, like, growing it into like a, let's say, a full language learning tool that you can use just this one, and you don't have to use anything else, and you can go from complete beginner to advanced? What are the biggest challenges there?
Um, I think the biggest challenge is that, uh, uh, there is a, it's such a huge space, uh, people have very different skill levels, they have very different needs, uh, and, uh, uh, uh, some people need language for writing, some people for speaking. They have different learning styles. And you cannot address all those, all of those things, uh, right away with one product.
Of course.
Um, and, uh, companies who try to do it, they, uh, do everything in a mediocre way, and finally fail. Uh, so that's why we have taken our goal to build a really good vocabulary acquisition tool that can teach you vocabulary 10 times faster than anybody else, and then we can expand from there.
And currently, we're already introducing, uh, new features that we call a course wizard that actually opens up the vocabulary learning for the long tail of learners, the advanced learners, intermediate learners. So, if you, if you, uh, uh, think about, uh, like an average language learning application, then they address only first 1000 to 2000 words.
Yeah.
It's, it's quite useless for intermediates or advanced learners who want to learn it for medicine or physics or some other professional needs. Uh, this course enables to do it in, in a couple of seconds, adapted to your personal needs.
You can change the, yeah. And, of course, if you go through learning vocabulary, then you also kind of get a, it's in context, you know, it's fill in the blanks. So you get exposure to grammatical structures and, and, and that kind of thing. Does it, does it take that into account when it decides what to show you?
Yes. Yes. Of course. Yes, of course.
Okay.
It, it actually teaches a lot of underlying grammar as well, but it's not taught explicitly. So, you, you, uh, as you notice these grammar concepts in, um, repetitively in different sequences, then your own machine learning brain puts thing, puts things together and creates your own rules.
Yeah.
Without us explicitly saying why it happens. And it's, it's very fascinating how people, people's brain actually start to construct irregular verb conjugations without explicitly teaching them.
Right.
Because,
That's a beautiful part of language. I think the brain sort of looks, it just recognizes the patterns and then uses them. And sometimes it's wrong. Like, kids get it wrong all the time, you know, when they're learning their first language. They, they say things like go-ed and run-ed and, you know, that's, that's fine, that's part of the process. But adults also obviously do it too. So, one thing I want to know as well about this is, if you have all this data and you, you're personalizing it in a way that makes me learn the fastest, how do you kind of encourage or set up a parameter where I actually log in enough, or at the correct time? Because if I only log in once a week, but your algorithm has concluded that I should be learning, let's say, a new word every seven minutes, or some random example, how do you, because you can't control how much I use the app. Do you make any recommendations on that? Or is there any way of tracking that? Or how does that work?
It's a very good question, because if you, if you would have an application that teaches you things most efficiently, uh, but you do not spend any time using it, then, obviously, you don't learn anything.
Right.
So,
That's what I mean. Yeah.
And, um, uh, it's a challenge for all of the startups that, um, uh, they need users to come back, and they build, start building, uh, features that help with retention. They do push notifications, email notifications. They start gamified. Because if they do it, then users come back more frequently. Uh, they have bigger user base. Um, but there is a catch that, uh, uh, some of these gamification elements, they bring back people based on external motivation, but not because of the internal motivation.
Right.
They, they, they want to fill, they, they want to get the points, or play the game. Um, but, uh, the learning value gets deprioritized, and this is what has happened with many, uh, educational applications, that they have started to run for less motivated users, try to attract them with gamification, but they have ended in being not in the education company, but game company.
And we took a opposite approach, because we thought that if we really, really are serious about making human learning faster, then we can only do it, make human learning faster if we initially built the product for those people who really want to learn, who take care of their internal motivation by themselves. Um, and, uh, that's why intentionally we didn't build much motive, motivation design elements, because we wanted to capture these user base, uh, and train our algorithms based on them, and try to unlock the human brain, how the brain really learns new stuff.
And now, when we have trained our algorithm, now, now we are starting to build some motivation design elements into the software as well. But we're not going to do, we're not going to gamify it, because, uh, our ultimate goal is to build technology that accelerates human learning. And game is a nice, but it doesn't help to reach the final goal.
Yeah, I agree with that. I think it's way more important to have the technology in place and an efficient method. But I'm still not sure about, let's say, I, I just signed up for the service, and I wanna, I wanna maximize my learning, but I'm also kind of busy, you know, I've got a family, I've got, uh, working a lot, and, and, and, so, what, how much should I do? Or does it not matter? So, if I do less, it'll just be, I'll just be less efficient. Or does the system kind of compensate extra for that? Like, would I, would I get slowed down more if I spent less regular time on it?
Um, actually, uh, indeed, this is the case that if you, um, uh, learn more regularly, then, uh, your learning efficiency per unit time also increases. Um, so that, because you reinforce more things, you, you repeat in vain things less.
And, uh, uh, we do not have yet in the product a full supporting motivational design framework around it. But, uh, we have figured out something that is really, really important for building educational software. Uh, that is very precise measurement of learning itself. Um, and we want to bring it to the user so that you will actually see that if you spend 10 minutes learning today, what was the value in the terms of learning for you?
Yeah.
Or, or if you spend one hour, and we think that when, when we have, uh, brought it to the user, then some users will start making experiments on their own with their different learning styles that we can use, uh, to make the product better.
I think that's a great idea. Yeah, because that's obviously also a problem if you're creating an app or some, some software is how do you actually get people to continue to use it? Because it, people probably can do it for a while, but then maybe the, the, the motivation drops a little bit, or they just, it, it maybe becomes too repetitive as well, and people just kind of let it go. So, so that, I, I suspect that would be a challenge as well, uh, but, you know, you definitely want the engaged users, because, you know, an app like Duolingo, right, is,
Mm-hmm.
I mean, I guess if you're waiting for the train or something and you don't want to play a real game, then you can work with Duolingo and get barely any language learning done. But, um, you know, and they've focused just too much on the, uh, the gamification, right? So, they get all these people who are not really motivated, they don't really care about it. And if they even finish it, they didn't really have anything to show for it either. It's a bit like going back to the school example, where, you know, people are in school for 10 years, and they don't,
Yeah. Exactly. Exactly.
Exactly. They don't have it.
Exactly. And there are no good technologies actually, uh, uh, out there, because, um, all of the companies have had drive to do, uh, um, the main difficulty has been that, uh, if you want to improve something, then you have to be able to measure it.
Right.
But, uh, but it's incredibly difficult to learn, to measure learning. Um, if you think about final exams, or exams at school or university, then these exams are made once in a year, probably, or, or once in a couple of years. And, uh, only very small percentage of your total knowledge is proved. Um, and if you get your exam score, this score doesn't actually tell you much, what could you, could you have done differently, uh, half a year ago, learning this or another thing, uh, slightly differently.
Um, this is a, uh, if the measurement is done, uh, after a very long time, then you do not anymore know what were these things that influenced your learning efficiency. Um, but, uh, if you can measure learning in every, every, every minute, every second, every interaction that you do, then actually it creates a very precise map, um, and we can say that if you learn today five minutes, we can, we can tell you exactly that these words you will forget in two weeks, these words you will, you will forget in two hours and 50 minutes. These words you will, uh, forget, uh, in three months. So, we can actually calculate what was the total value of your learning time.
What unit do you use for that then?
Sorry?
What unit would you use for that then to show me?
We, uh, um, we use, uh, um, unit that we have developed by ourselves. We call it, uh, uh, memory curve improvement. Uh, okay. And, uh, and, uh, but it's, it's a, it's a very nice, uh, um, way to understand how much, um, have we delivered value to the user. Um, and, uh, if we can have this, uh, um, if we can measure it, then, then we can improve the product and improve the company so that, uh, we can maximize this value. We can make choices that, uh, okay, people retain more, but they don't learn. Actually, so, this is not a valuable path to go.
No, that's interesting. And what kind of, what kind of future, uh, kind of tech, are there any technologies upcoming? Or any improvements? Because we've, you know, machine learning, the, the ways to use it, I think, has kind of developed, and it's, it's new stuff being developed all the time. But has the fundamental technology and innovation actually changed in, in, in terms of language learning? Are there any, are we doing new things with machine learning, or maybe AIs, or what, what do you see kind of coming that could maybe make all this even better, or change it for the better?
Um, as much as I have asked, um, I have talked to many, uh, venture capitalists who have a very good picture of, uh, educational companies, uh, including language learning companies, and I have asked this question, do you know any company who can, who can measure learning, who, who measures learning for, um, uh, to have a north star for the company, what, what do they want to achieve? And I have not yet got any answer.
Wow.
Like, uh, anybody is doing it. And investors complain a lot that there are educational companies, but actually, they do not turn out to be delivering learning value. And it's very obvious why it's happening, because, um, learning needs effort, and if they start to run towards bigger user base, they, they start to run towards less and less motivated users, who actually do not want to learn. And that's why they cannot build a product that helps learning, because the user base doesn't want to learn.
Yeah, that's pretty terrible. Uh, you know, but that's, that's the reality, though.
It's a reality, but this is also an opportunity for us, because, um, we have developed this technology with this mindset that we should be able to measure learning, we should improve the learning. Um, there are university studies, for example, University of Tartu, they recruited 500 students, they made them to, use different softwares, including Linguist. And, um, uh, the conclusion was that people can learn with it four times faster. And, uh, it was a couple of years ago. Uh, we have improved our products, um, and I believe that it's possible, with this technology, it's possible to make human learning 10 times faster.
And if you imagine, what does it mean? That if you can learn a language 10 times faster than conventionally, it means that you can learn a new language in, in a couple of months. You can learn a lot of new skills, um, because this is a generic technology. We, we can apply it on history, or medicine, or, uh, pilot studies, or whatever. And, and this is the world that nobody has more time. So, we have to use our time more efficiently, and there are always more things to learn.
Right.
Of course, yeah, that's, that's universal for sure, we can learn anything.
I'm just gonna learn anything.
But, uh, uh, yeah, but, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, Uh.
that's why I think this technology is super important that we can, uh, we're building technology that accelerates actually human learning. Language is just the first step.
Yeah. And also, how cool would it be if we could cut the mandatory school time in half? Or reduce it? Or?
Definitely.
Yeah, just make learning new skills much easier. That would be super cool. But do you see any kind of, you know, we talk about language now, there's been a lot of kind of, uh, uh, reports about translation software, translation AI, and all this kind of, you know, obviously machine translations are getting better and better. But, do you think that it will ever replace kind of training of translators or interpreters or people actually learning foreign languages for the, for the sake of it?
I think it will definitely change the market need for translators, and, uh, and interpreters, in written and in, in spoken. Um, and it definitely shapes the language learning market as well. Uh, nevertheless, there are some things that we probably doesn't influence a lot. Because if you need another language to, to live in another country, then you can't do it with a translator. If you need a language for better education, you can't do it without a translator.
Right.
Or,
But you can have like a translation earpiece, you know, that live translates.
Yes, but you, uh, you can't build human relationships based on that.
Right.
So,
That's true.
You never know. And, uh, um, these need to learn a new language will definitely stay for a very long time until we have the one language on, on, on Planet Earth, that probably hopefully will not happen. But, uh, um, the language learning definitely remains. But what gets even more important is learning other things. Uh, and that's why I think this technology is super important that we can, uh, we're building technology that accelerates actually human learning. Language is just the first step.
Yeah. But how, how long do you think that will take before you have the software to a point where you're, okay, now we can kind of roll it out to other, other areas, or other parts of learning?
We do not know. It's, it's, it's very difficult to predict how long things take. Um, we are already experimenting with some things. But the main focus is language learning at this point of time.
Right. Well, glad to hear it, you know, that's, that's why we're here and, and that's why people listen to this podcast. So, so thank you so much for, for coming on the show and and talking more about machine learning and the future of language learning as well. And and and Linguist and it's very exciting to see such a, I mean, it looks so simple on the surface, but obviously, there's a lot going on underneath and and I can't wait to see how you can build that next level that we mentioned, that how to go from vocabulary acquisition to actually language acquisition, which is for me, the big puzzle. And I think there's so many ways to do it today, and I haven't seen any integral way to do it in just one place, or with one method. So, that's going to be very exciting to, to see going forward. So, so thank you so much for, for coming on. And, and, um, where can people find out more about you or Linguist?
Um, yes, thanks for inviting me into this podcast, and I'm happy people to coming to our site, linguist.com, and, uh, if they, if they learn, they also know that they contribute to the development of application that accelerates learning.
That's great. And, uh, do you ever publish any of this data? Like any aha revelations?
We will do, but, uh, we will definitely do. We have some collaborations with universities who want to publish things based on our data. But I cannot tell you exactly when and what. But, uh, stay tuned.
That's all right. We'll, linguist.com.
Yes.
Perfect. Well, thank you so much.
Thank you so much, Kristof.
Thank you so much for listening to this episode of the Actual Fluency Podcast. I really appreciate having you here today. Just before you leave, I just want to give a quick shout-out to today's sponsor, which is italki. italki is a tutoring platform where you can find affordable tutors for every language in the world, pretty much. So, get started today and get a free $10 credit when you book your first lesson. If you go to actualfluency.com/italki, that's spelled I T A L K I. So, give it a go and feel how tutoring can really boost and enhance your language learning.