I received an email from an Applied Linguistics student in Poland who asked
me a number of questions, uh, for the thesis that she's doing on Duolingo
and LingQ and AI and motivation and all the subjects that so often come
up when it comes to language learning.
And I'm going to show you that interview.
And I think it's very interesting that a person in Poland would contact me and
that we have this conversation and it's an indication of where we are today.
We are connected.
We can connect with people in Poland or in Japan or in Brazil or other countries.
I listened to podcasts from Persia, from Iran rather, in Persian.
And you know, I think back when, you know Montaigne would withdraw
to his tower in order to think and write his essays and so forth.
Or we have Laozi the Taos philosopher, who talks about, I can see the
whole world from my window, but the implication was that it's introspection
what we find within ourselves.
But today we are connected and languages
are what connects us.
If we can speak more than one language, we are able to connect
with more and more people.
And for that reason, I was very happy to connect with this, uh, uh,
linguistics student from Poland.
And, uh, I invite you to listen to our conversation.
So, like I said, I'm a Applied Linguistics student in my last year.
I'm doing my Master's thesis and I am, uh, analyzing LingQ
and Duolingo, talking about the differences between those two apps.
I know you've answered this question before, but for the sake of my thesis and
this interview, could you just tell me why or what inspired you to create the app?
So when we created the app, I was not aware of Stephen Krashen, was
not aware of the input hypothesis.
However, my own experience with language learning, particularly with languages
like Mandarin Chinese and Japanese, was that I spent most of my time reading,
less time listening because it was less convenient 50 some years ago.
But even then, living in Japan, I'd get in my car and I'd turn on Japanese.
So listening and reading were always a big part of my language learning.
And I was between jobs actually, and I looked at my bookshelf and you know, I'd
always been interested in languages.
So I had books in German and books in Spanish and on the pages, there
are a bunch of words underlined, and I would look them up and forget
them, or I would make a list and look them up and forget them.
And it's annoying when there's always 10 words on a page that you don't
understand because you kind of get the drift of what the page is about.
But you really would like to know what those words mean.
When I was a Chinese, like Mandarin Chinese student a tape
recorder was something this big.
And then we went to cassette players, which every show often would go.
And then you lost the cassette, right?
So once I became aware of very briefly, mini disc players, and thereafter
MP3 players, I realized, wow, this is a revolution in terms of listening.
And then the second thing was digital texts.
So now you have the ability to do things with the text, look words up,
save words, move them to a database.
So once I saw that this was happening, my experience with the massive, you
know, the importance of massive input.
When I was in Hong Kong learning Mandarin, I read far more than the other students.
I was forever buying books.
However, when you buy a reader for Mandarin, you have to have a reader
that has a glossary for each chapter, because looking something up in a
dictionary, a traditional dictionary, is a hopeless task, especially with Chinese.
It takes forever, and as soon as you close the dictionary, you've
forgotten what was there, literally.
So I said, okay, now with technology, we can do things better.
So that I think, was sort of the impetus.
There were some other issues.
We had an employee at our wood processing and, and, uh, software
company who was from China.
He really didn't understand in Canada, you know, how the culture worked.
And I realized that you need to have relevant input.
So we developed, we had a bunch of interviews with, uh, rotary Club Business
Club members talking about what they did.
A woman had a travel bureau.
Somebody else ran a, you know, whatever, it might be, a real estate
company or a construction company, and they talked about what they did.
So I said, this is relevant.
So record them, transcribe it, and give it to new immigrants so that
immigrants can actually get a sense of the society they're moving into.
So those are some of the things that contributed to the development
of LingQ in the early stages.
It was directed mainly at English for mostly immigrants in Canada.
Oh, okay.
Okay.
That's surprising.
So you thought that, uh, the textbooks or whatever are lacking
in this, like everyday use texts?
If you grab the average textbook, it's got about 10% content.
Most of it is explanations, which are often redundant in
any case, difficult to remember.
And then they have exercises, which I hate doing.
Uh, we need more content.
I have textbooks in Arabic, for example, you know, media Arabic,
and they have articles from 2015.
We're not in 2015 anymore.
So the advantage with the internet is that essentially
everything is language content.
And, and whatever you're interested in, and it's up to date.
And so that's far better than, and some textbooks are unbelievably expensive
and, and yet still dated right away.
So if, if we believe, as I do in Krashen's theory of, you know, compelling
input, that's not the way to go.
There's no flexibility.
There's no scalability as so yeah, I think LingQ is a better solution than textbooks.
Yes.
So you actually mentioned Arabic and I feel like, um,
there are languages known for,
for having many dialects.
Right.
And as far as I know, LingQ doesn't offer dialect learning.
So do you think that would ever be an option to have a separate,
you know, language learning option only for dialects?
Do you think that's relevant or not really?
When it comes to Arabic.
We call it all Arabic.
However, for some of our basic content, like our mini stories, which is sort of
where we get started, lots of repetition, high frequency verbs and so forth.
There, we offer standard Arabic, Egyptian, Arabic, and Levantine Arabic.
After that, and because I've been studying Arabic, then I have to go
to, you know, uh, MTV Lebanon, or you know, you have to go find sources.
Because at LingQ I can bring anything in.
If I have a sound file, it'll transcribe.
It makes mistakes.
Occasionally, you know, Whisper AI or whatever, will make some mistakes.
But by and large.
Dialects within Arabic.
Now we could have created three language slots for each of those.
If we ever go to do Moroccan or, or, you know, north African,
probably that would be separate.
You know, there's some discussion.
We should have Québécois French Canadian, you know, which can be different.
So the, the Québécois French varies from very standard on,
say, radio, radio, Canada.
To the average guy in the street who can be speaking a very, you
know, colloquial form of Québécois.
Norwegian and, and Swedish, Indonesian, Malaysian, uh, I mean, there's so many
languages, there's so many examples where the difference between a language and
a dialect, it's, it's not that clear.
So it's kind of arbitrary.
Uh, if we have people saying, I would like to have this language on LingQ, and
they're willing to translate the many stories and record them, we'll put it up.
We don't have any sort of fixed position on what should be on.
We have Croatian and Serbian.
Having spent a little time learning those languages.
The differences are minuscule.
So if people want it, we put it up.
It's so important to be motivated.
So if a person says, I wanna learn Croatian, or I wanna learn Serbian, so
we give 'em Serbian, and to some extent the content will be a little different.
So it doesn't matter to us, it's not any more difficult.
And another question.
Um.
Kind of connected to Arabic because again, uh, you mentioned multiple times
the importance of immersion and that's something I talk about in my thesis a
lot, however, was there ever a language where you found learning grammar to
be necessary, where you kind chosen a more traditional learning approach?
Some
of these terms are, are used pretty loosely.
So grammar is a description of standard usage, I hesitate to say correct
usage because all languages evolve.
You know, I hear a lot of English native English speakers saying I
would've went, which is not today still correct usage, but in 20, 30
years, it may become correct usage.
We no longer say to whom.
I do, but a lot of people never say to whom.
So usage evolves.
So grammar describes usage.
I think everyone who learns the language wants to use the language
in the way that's most useful.
So if again, we talk about Quebec.
If I live in Quebec, I want to use Québécois usage if I'm
doing business with Québécois.
So we all want to use the language in a way that suits our purposes.
Now, if you study rules of grammar that can help you notice things,
the question is how much time are you going to spend on that?
So is it gonna be 50%, 10%?
How are you gonna engage with the rules of grammar?
Are you gonna spend your time going over these conjugation tables?
In my experience when I was learning German, I only started to improve
when I stopped trying to ingest the declension tables because
it's, it's largely a waste of time.
So it's there, it's a part of it.
I wouldn't say it's zero.
Arabic has a complicated grammar.
Persian is not so complicated, but their forms are different, and so to develop
a habit of using that form of usage.
Reading the rules doesn't do it.
It can help you notice something that you are already aware of through
massive input, but it's not a shortcut to using the language correctly.
But yes, we all want to use the language shouldn't say correctly, but we want
to use the usage that's useful for us, which might be the usage in Lebanon.
It might be Poland, maybe, I don't know, Western Poland version.
I don't know the dialects in Polish, but you kind of wanna adapt to
the language is used around you.
Definitely.
So let's take German for example, because I also study German together with English.
So you completely stopped learning all the conjugations, all the
endings towards, or just...
I became more realistic.
I realized that I will gradually develop new habits and that.
Studying the declension tables, because in German it's more the declension, it's more
the adjectives or the nouns rather than conjugations, which is like birds, right?
I realized it wasn't doing much for me and, and there is this thing you can only
push so much in if you're trying to push the same information over and over again
the brain doesn't want it.
There is proof that when they measure the brain and how the brain deals with
language that seeing particularly verbs, but a, a number of like vocabulary items
in the sort of variety of context in which you encounter these words and especially
verbs is more important than the frequency we always think Frequency is the key.
Frequency is part of the story.
But seeing them in different contexts.
And the worst example of trying to, you know, block learn something is, is these
vocabulary lists or con detention tables.
You think you're learning, but in fact, I don't think you're learning very much.
It may help you notice things the next time years.
Oh yeah.
Right.
I saw that on the, the other problem with the tables is there's
too much information there.
It's impossible to deal with a table.
It's like when we learn the colors in one go.
It's impossible.
You have to learn one color, green grass.
In conjunction with a particular situation and the brain has
an easier time learning those.
So it's not that I never look at those tables.
I do, uh, nowadays with online dictionaries, like the ones we use at
LingQ, many of them have, you know, there are conjugating dictionaries, they all
have extensions that allow you to see that word as part of a conjugation table.
That probably does something.
I tend to look at that, but it, it's not, it's not how we get, we have to
ingest so much of the language that some of these things become
natural, you know, habits.
Yeah, I agree.
Uh, with German it's also, um, it's nice that all the endings are pretty
similar, it was something that I noticed.
Sometimes when you speak it's kind of difficult to notice which endings someone
has used, so it doesn't really matter.
And, and also we should not worry about it if I speak in German.
Or even worse, a Slavic language where there are all these different, you know,
endings, declension endings, you're gonna get some earlier than others.
The instrumental in all the Slavic languages is the easiest
because it's kind of unique.
It's, it's a bit of on its own right?
But between the dative, and the, the genitive, it's very difficult,
and masculine and feminine neuter, it's very difficult to try to
remember those while you're speaking.
But I find that I gradually, I develop better and better habits than if I don't
use the language for a while, I get worse.
It doesn't inhibit comprehension, basically.
It doesn't inhibit comprehension because the language is so redundant and, and
so if you're able to understand and listen and read, your brain is gradually
developing better and better habits.
Like I said, I am also comparing the LingQ with Duolingo in my thesis.
I talk about the like gamification methods in Duolingo, and I know you, I know that
you of course know Duolingo and you've used it and LingQ is completely different.
There's no gamification, but do you think you would like to apply or
add those methods to LingQ ever?
I mean, there are some elements of gamification because the
biggest motivation for me is
the subject matter.
So right now I'm doing a podcast series in Persian, which is very clever.
And, uh, Abbas who puts out these podcasts, he talks about subjects
that are very interesting for me.
So I, I'm motivated by what I'm learning.
It's my
boss Yeah.
By the topic.
But we have all kinds of sort of nudges, measurements, goals,
coins, maintain your streak.
So I don't know if that's communication, uh, things other than the sort of
intrinsic motivation, the subject matter.
So we use these tools.
Uh, we also have, when we go to sort of looking at it sentence by sentence.
Then we have these matching pairs, which is better than a long list of
flashcards, which I find not very useful.
So we have these matching pairs, and then we have a chance to translate the sentence
from your native language number to rebuild, which is a form of gamification.
So some of those things we do, I'm not involved in designing LingQ, but I
think that's about as far as we want go.
The biggest, I think, difference between us and duo language is that.
We allow you, whatever we do in the way of gamification is built up on
content that is of interesting to interest to you and that you chose.
Whereas the biggest difficulty I had with Duo LingQ is they impose
content on you, how you are gonna progress and what you do next and,
and none of that happens at length.
I think you can start wherever you want.
You can progress as quickly as you want.
You can choose whatever content item you want and the functionality, and to the
extent that we have gamification, it'll apply to content that's real for you.
That's true.
That's also the difficulty I had with Duolingo.
It's fun like at the beginning when you're starting with the language
I guess, but then it's kind of difficult to progress, I feel like.
Okay.
Another hot topic is, uh, artificial intelligence.
Of course, ai, so Chat GPT, and all the other AI programs
that I don't know about yet.
And I, again, again, I know that you've used Chat GPT for language learning.
So just generally, what do you think about AI in language learning as
someone who develop a language learning app that probably also uses ai?
So again, I'm not involved in the development side that I
see that we're doing there.
There's a lot of areas where AI can help.
The first thing we did was to make sure that the translations.
Of words were context appropriate because we rely on third party
dictionaries and the typical third party, any dictionary will give you 3,
4, 5 potential translations of a word.
So we wanna make sure that you get the translation that's
appropriate for that context.
I use that all the time.
That's very helpful.
I guess AI helps us to transcribe, uh, MP three files to create a lesson.
We want to use AI to help learners develop, you know, a range of
content that is compelling for them that suits their interests.
We're looking at using ai.
To help guide learners through LingQ so that they can get more out of LingQ.
I think we're, we're just at the beginning.
Whenever I start a new language, and this had was true for Polish,
for Arabic, I do our mini stories.
Now, the mini stories, the way it's structured is sort of
two different points of view.
Like he and then they, or yesterday and tomorrow.
So essentially the same words in a different time or tense?
Or tense.
Yeah.
And then you have the sort of a statement question answer, and the
answer can be negative or positive.
So you're getting interrogative, you're getting, you know, positive statement,
negative statements, a whole lot of things around essentially the same vocabulary.
So that's the model.
Well, we had to write those 15 years ago.
Now AI can.
Write those AI can localize them.
Right now we have the same story, which is translated into 50 odd languages
from Punjabi to, uh, you know, English.
But now we can localize them.
So if you're in Turkey, it'll be tea, not coffee, maybe.
I don't know.
But they, to some extent, they can be localized.
So even in creating content, I mean, I have used AI to create content in Turkish.
Gimme a thousand word history of the Ottoman.
M AI can create a thousand words on the Ottoman Empire, which I can
then import and start learning from.
So the sky's the limit.
I I, we're only beginning and the AI is evolving and we have people who are
specialists in ai and we're looking at how we can introduce it to make.
The learner experience better.
I dunno how old ChatGPT is, but I think it's only been
around for three years maybe.
So before that, the beginning of LingQ, you had no ai?
Uh, no.
I mean, LingQ is 20 years old.
AI is much more recent than that.
So, I mean, the first time I became aware of AI at LingQ was.
When we're increasingly, you know, able to transcribe audio files into a lesson.
So if you, if you're following a podcast, if you're able to get the
MP three file, which isn't always the case, then we can create the lesson.
In other words, it'll transcribe it very accurately.
So I think it's early, I think it's early days.
Uh, some people go, like to talk to, uh, you know, a, a chat box.
I mean, I can envisage a situation where person is learning polish and they're
gonna visit Poland in six months.
That chat, ChatGPT could create a lesson about walking tours in Warsaw
or crackle and you can be listening to it and then you can even ask questions
while you're driving in your car.
And this is all going over the same vocabulary in slightly different forms.
So I think we're just at the beginning, we we're scratching
the surface when it comes to ai.
Um, I agree.
I also teach and I. I haven't been using ChatGPT to create my lessons
a lot and I think it's very helpful.
Uh, sometimes it gets a bit weird, I think mainly because it's so new and it's
still learning, but it's super helpful.
Definitely
what be aware of, I'm sure you're aware of is not everything that
chat GPT spits out is accurate.
If you say, gimme 10 examples of the instrumental case in
Polish, for example, not all of those will be instrumental case.
No, definitely not.
I mean, if you ask about English, usually it's correct because I guess
it's more familiar with the language.
Right?
Uh, but yeah, the, the deeper but deeper you dig, the worse it gets.
I feel like I also wanted to talk about the future.
LingQ and ai.
So are there any new features or innovations planned, uh,
that you can tell me about?
I mean, we are adding languages in response to requests from people.
Very few of the new languages generate a lot of users, so there's not really a
very big motivation to add new languages except that our users ask for them.
So irish Gaelic, we're gonna be adding, we added Punjabi insofar as features.
Our major motivation more than adding new functionality.
The difficulty with whenever you add functionality, you increase complexity.
And we're more motivated right now to look at ways that way we can make
it easier to use LingQ more pleasant.
So the user experience.
So customizing the content to the user's needs, I don't
know, because that's not my.
Specialty.
But, uh, I think we're gonna look at increasing ways we can
use AI or other ways designed to make it more user friendly.
Again, the trouble with customizing is it has to be a, a flawless
transition to customization.
You can't constantly ask people, are you interested in A and B?
And then if then, and that, and anything that complicates things is bad.
Kind of like Netflix.
If you look up a movie, then the next movies that are
suggested to you are similar.
So, and we do that to some extent already in our library.
So I think those are the kinds of things that we wanna work on.
The other thing too is you have to appreciate that at a site like LingQ.
We're constantly working to make the site more efficient.
Uh, browsers are constantly changing.
It's the whole backend database, how it has to change, how we can
use AI to make things happen faster.
Those kinds of things are going on all the time.
Yeah.
I think that's it for me.
Again, I, I'm, I'm really, really thankful
for the interview.
Could you please sort of explain what you study, where you are?
Your interest in language is just a little bit of background, which I think
would be interesting for my viewers.
Yeah.
I study applied linguistics in Poland in pos.
I study German and English, little bit of Spanish.
I chose English because everyone can speak English in Poland, in you the most
important second language in schools.
So that was a no brainer.
But then I also moved to Germany for, for a year, a bit over a year.
I went to um, art school there.
I came back to Poland to, uh, go to college.
So I thought, I don't want to study phenology.
I'm not that interested in that.
So I thought applied linguistics is per perfect.
It's just the language, it's just translation.
I can work as an interpreter translator after that.
So that was perfect for me.
Right.
Okay.
And I should say, your English is amazing.
I always feel that people who speak another language well, who use the
language well, who totally understand everything and have a slight accent,
but certainly more impressive than someone who speaks poorly but has
no accent, which really happens.
But, so I think that what you have achieved in your English is,
is really all one can hope for.
Very, very.
Not
tonight.
But thank you very much.
Thank you very much.
But like I said, I have been learning for a long time.
Everyone has to.
Um, so yeah, like I said, thank you very, very much.
I, I didn't really believe that it would be possible to get this
interview.
If you need more information, just let us know.
