Botond Seres, Dave Erickson, Ido Cohen
Dave Erickson 00:03
Speaking English fluently is one of the most important business skills you can master. If you want to grow your business globally, there is a new tool that can make you sound like a million dollars. This app uses artificial intelligence to help your pronunciation and speaking of English. We're going to put every word about fluency AI under the microscope and this month's ScreamingBox podcast. Please like our podcast and subscribe to our channel to get notified when next month's podcast is released.
Dave Erickson 00:55
In this podcast I, Dave Erickson and my co-host Botond Seres, are going to try to understand language fluency with our guest Ido Cohen, CEO of fluento.ai. Leading up to the founding of fluento.ai Ido was vice president of products for Dream data and after, working for Permutive and Microsoft on product development. He is also a mentor on Plateau, helping professionals at different levels, build their careers, and helping them focus on collaboration, prioritization, roadmap building, vision, and leadership skills. Ido's latest product is fluento.ai, which is here to assist you throughout your language and your English fluency journey. It creates a similar learning situation as if you are in an English class. So I know, is there anything you want to add to this? Sounds good? All right, well, can you first kind of give us a rundown of fluento.ai? How does it help people to become more fluent in English?
Ido Cohen 02:01
Yeah, so, so Fluento is here to focus, first of all on your fluency and confidence that everything we do is here to build, you build better conversational skills. So we don't do any grammar or exercises that you can get for free somewhere else. And we do that by combining the latest tech and the best methodology that's available out there. One way I like to explain it to, I guess talk about it is that it really, it mimics immersion. Immersion is like throwing you in a country where English is spoken. And that thing you try to adapt to the environment in a simple way, probably a better way to explain it. But anyway, so, so the way Fluento works, in simple words is, is using AI to mimic real life situations. Like for example, getting you to get a cup of coffee at a restaurant or something that's more complex, like going through a job interview. And the way we do that is by pairing up two learners from different parts of the world together and into this immersion-like environment. It's essentially a one to one video session where you get to roleplay, through, through practice. And then we use AI to mimic a teacher that is there to help you maintain the flow of the conversation. We do that with suggestions of what to say, tips on things that you might have, you might struggle with. But also after every conversation, we analyze the entire speech and then give you an idea of where you're going right and of course, what you need to do to improve. So I feel like this is a, this is a topic especially near and dear to my heart. I mean,
Botond Seres 03:45
So I feel like this is a, this is a topic especially near and dear to my heart. I mean,
I do, if you don't mind, I'm gonna start off by sharing my super secret strategy for learning fluent English, and that is to watch everything in English with the subtitles on, that way. You don't know the pronunciation and the grammar, and that has been the one thing, that ,the singular thing that has helped me achieve fluency. I mean, everything else, sure teachers were great and all that, but, to truly expand your vocabulary, you're gonna need exercises. And the one thing that I feel is, as you said, immersion is truly a huge jump from just watching movies with subtitles, because that's when you have to think on your feet, complete sentences on the go, which is honestly the most difficult parts and just leaving that out with all those ums and ahs that especially new learners tend to lean on quite heavily. And I do wonder how, how do you come up with this idea to assist new learners with both pairing them up but also how does AI come into the picture because I kind of feel like that, putting two people in a virtual room and giving them an exercise is already a great idea. So,
Ido Cohen 05:12
Thanks. Well, thanks for sharing your experience, it's connected with what we're doing in many ways. It's firstly, the idea of mimicking, right. And secondly, it's the, it's immersing yourself, in your case, fully immersing yourself in a, in a movie, or show. I spent five years in a school called Berlitz, which is a language learning school with a bunch of locations around the world. And two things I've learned from it is, well, first a language, now I speak Turkish because I spent those five years in Turkey. But the second one is how, how to teach English from day one; how to, how to get you to speak in less than an hour, are personal, about politics classes, just introducing yourself or getting something in the restaurant with how to make it super, super practical. And I haven't really invented this or anything like that. It's just that I thought, well, you know, when you're building something that you want, when you're building something new, a lot of the apps that you see out, out, there are apps that build something that already exists in real life, but just better. So if you think about Revolute maybe, it's a banking application. They haven't invented banking, they haven't invented wiring money across the country, but they have made it a lot cheaper and better and faster and more accessible. And that's kind of like what guided me when I started Fluento. It's about thinking about elements or methodologies in class that are super, super effective, but also super, super fun. And in five years of doing this, taught me that I'm not there to replace anyone else in class, except the teacher, I'm there to enable conversations, I'm there to help people boost their confidence, and then kind of leverage on what they do to help them get better. Obviously, there's a target behind, but the entire point is not necessarily speaking to me as a native speaker, but it's also getting you to actually make most of the conversation. And that's where Fluento will guide me, but ,and so that's kind of how I thought about Fluento when I started it. And this is what works with, think about role playing and classes, it's a great example of this methodology. It's about getting you to go and execute something that you might do in real life, your job interview, and letting you try and achieve it and boost your confidence. And of course, after or along the way, like a teacher help you when you're stuck, or help you, help you pay attention to the errors that you're making in a way that, that's not only good, but also digestible. Translate this into what the AI does and this is it right? We're not here to replace the speaker that you're talking to with AI, which I know a lot of companies are trying to do. What we're doing is actually more assistive, the AI is there to listen to everything you do to begin with, which is great, because it can pick up on the tiniest mistakes that you're making better than any, any feature. It's also there to remember these things better than any other feature on one side and on the other side is when you think about what's going to be valuable at each given time. So for example, think about moments where you're, you're trying to start a conversation with someone you probably don't know. You have the idea; you're there in the back of your mind, most people, or at least most learners, I know we try to make that sentence in the back of their mind and then try to translate it over to English and along the way they get stuck because it's too complex. And you can click a button on Fluento and it will analyze the context and then complete your sentence so that you can maintain the flow. All the ways are well listening into previous conversations you've made and pick up on errors that you're making, you know, to how you try and apply them. So you went into vocabulary, right? Memorizing vocabulary is, is not as easy as people think. I see a lot of people for example.
It's completely worthless
What is that?
Botond Seres 09:28
It's also completely pointless, like without context, vocabulary. (exactly) that's exactly the rulebook for first timers.
Ido Cohen 09:36
Exactly NPP, I see, I remember seeing a lot of students preparing for tests like the TOEFL, which is an English proficiency test. And the requirements there is that they tell you to memorize 500 words, it’s something ridiculous. And the reality of it is that you can't memorize that many words in a short amount of time. And like you said, what's really important is to apply those words. So the AI actually picks up, picks up on these words and lets you apply them in the right context or the right conversation. So that's one way, right, real time. And the other one is, well, after you've had the conversation, try your best, and what do we do? How do we help you? First of all understand whether you're going well, and then the second one is, what do you do to improve? And this is where we pick up on anything that learning a language comes from grammar, to pronunciation, to your pace, to your sense of confidence to what else is there, obviously your fluency? And then tell you first of all, where you're going right, which is really important. It's the, for us the idea of telling you, are you able to actually go and order a cup of coffee, regardless of your grammar? And of course, then what do you need to do to sound a little bit better? What do you need to do is to sound like Brad Pitt, And on set?
Botond Seres 10:58
Oh, I mean, I think that's a noble goal. But I feel like we lost over part of what is really a genius idea on your part. And, as you said, one of the most effective exercises in the classroom is to get two people to roleplay different situations. But at the same time, if we're looking at the entire class, it is one of the least effective tools because like two people are talking and 30 are listening. And yes, that's I think that is absolutely genius to just take those two people and take all the other 30 listeners out of the equation.
Ido Cohen 11:34
Your Well, thanks for the genius, I stay humble, I guess. I think the reality behind it is that I see, obviously, groups like products where you have a group of people, and it's exciting and sounds like it's great. But then you have different kinds of people, some of them are shy, and some of them are more, you know, outspoken. And that's where you get this imbalance in terms of like,people that who are being able to listen and not and if you're if your goal is to be able to speak, then you want to get to a point where the conversation is about 50% listening and 50% speaking on average. And that's where you need a one to one environment so that it's safe a nd you're allowed to go out and actually feel under pressure to speak, which is another point I can add on later.
Botond Seres 12:22
One of the things that is absolutely baffling to me is how good AI is getting at speech recognition these past years. Because five years ago, I had trouble asking Siri for anything. I asked like, hey, I want to watch this movie and see would be like, okay, fly to New York. Okay. That's an example of a language recognition gone wrong, but something else. Anyways, I distinctly remember, I used to have so many problems with recognizing complex sentences, and even just understanding what I'm trying to say. And not necessarily because my pronunciation was terrible at the time, but I see, Irish people have this problem a lot. They even joke about it. I don’t know if you saw that skit with two Irish guys are trying to control the new voice activated elevator?
Ido Cohen 13:25
No, not Yeah. I mean, the funny thing is, my partner is Irish and she's struggling with Siri as well. So I know I can see this every day where I have mostly stuff for her. No offense to Irish people, obviously, this is also, in real life. But I mean, when I moved to, to the UK, my partner introduced me to her parents, Northern Irish, really, really nice people. I used to sit in dinners with them. And they used to speak to me and my partner had to translate everything. I did not understand a single word. So not only theory, but on that like on a serious side. Let's break it down into three, I guess three areas. One is Watson and Siri. And the other is understanding sentences. And the third one is speech to text is to separate three separate areas. Why? Because Siri and Alexa today still struggle with that sort of stuff. And that's because I assume, I have, I don't work for Amazon, but I assume it's really about investing in the right technology that is super, super expensive. And that means diving into different use cases and developing the AI that sits behind it. At the same time there is oh I mean, AI speech to text has been developing really, really fast especially recently for obvious reasons as AI, generative AI is picking up and the level of understanding is very, very high at this point. It's not 100%, specifically, when it comes to non-native speakers, it's still hard. And this is part of the tech that we're building as well. It's about training our machine models to pick up on these accents. But even at a certain level, with API's that exist out there, it is possible. And so that's one text, understanding text is a different story. And that today, I mean, a good example for this is GPT. You can, you can say anything to GPT, even if your English is broken, or you type too fast, and your words are not really words, and it will still understand you to a very high degree. So that's developing really fast. The last one is, so we talked about Siri, we talked about GPT and then the third element is what we're doing. Which is, I think a combination of all of those things, we, the way we work is a we do a flow into works, it breaks down those, those, those technologies into separate areas while improving on, on speak to text, speech to text is one thing. Understanding the text is second thing, and of course, what to do with it is a third thing, like thing, and they could develop at the same time and improve and still provide high value.
Dave Erickson 16:29
Yeah, I think on the, you know, on the Alexa thing, I think part of it is also they're trying to stick, you know, that they're working with a small hardware platform, is trying to do a lot of things. So that's part of the processing power. I think that the text or the word, you know, vocal to AI to understand speakers, and what they're actually saying. I think that integrating the AI to detect when somebody says a word incorrectly, or they say it was such an accent that is not that understandable, is a really big thing in your app, because that's how you provide correction for people. I know that I was learning a second language, actually Hungarian, and I'm kind of tone deaf, and they have vowels that are high vowels and low vowels. And for me, I can't tell the difference when I'm hearing them or when I'm saying them. So I was missing words all the time. And I wanted to say something like, you know, dog, and it would come out very different sometimes, because I didn't put the accents on the right vowels. So that's probably an important aspect of what the AI is doing in Fluento. Is that correct?
Ido Cohen 17:48
Yes, that is correct. It's your, I think you're spot on, it's a difficult thing to understand it. And it's obviously a big thing for, for learners as well. Because no one no one, well, it's very hard for them to get access to a platform or a person that can actually help them understand that, that as a trend. So for example, many of our users don't don't know how to, how to pronounce the combination of t and h. They pronounce it in different ways d t. They don't Z. There's lots of ways to pronounce it. And it's hard for them to understand that because the feedback that they get from different people is, is, is not in a way that's that digestible. It's maybe you have a problem pronouncing this word, but, but but they don't understand that this is a trend, for example. So like in Fluento, it's not only about picking up those accents, it's also about showing you the pronunciation feedback in a way that goes, you know what, I have a problem with th across the class, a clot across my cross across different words. And that's something I need to work on. So there's like two, I guess, two sides to this coin.
Dave Erickson 19:10
Yeah, the other one is also emotional. The AI kind of takes the emotion out. I was, I had, I struggled when I was learning Hungarian, because the teachers that I had, were humans and empathetic and so I'd say something and it wasn't said correctly, but they understood it so they wouldn't correct me. Right? Whereas the AI doesn't have that kind of sympathy or emotion. If you say it wrong, you say it wrong and therefore the AI is gonna say you said it wrong, correct?
Ido Cohen 19:41
Well, this is yes and no, I mean, yes. That's not a that's, that's a, that's a that's how I see platforms acting today. So for example, if you are the most popular language app to hear a lot of what they do is is, is, I guess, judging you based on whether you ,whether you're doing 100% Correct or not. So you have an exercise that you need to do right type of word or type of scent, scent, make a sentence, and it has to be 100%. Correct for you for the app to tell you, let's go the next exercise. The problem with this is exactly what you said, is that in real life, if you go and, you know, if I order a cup of tea at a restaurant, you can order a cup of tea with wood, wood, wood grammar that's broken, right, you can say, Hey, I'd like a cup of tea, please. Or you can say tea, like a caveman, all of them are fine. And so that's not really, it's not realistic, but also breaks your confidence because you, you're thinking in black and white terms, if it makes sense, Fluento, tries to do the opposite. It tries to assess the context of things, first of all, and tell you, you know, you may have had a few mistakes there. But first of all, you're able to get a cup of coffee, which is the most important thing. Why? Because this is really the bottom. This is the reason why people are learning languages. But also it's confidence boosting, because you come out of it thinking, oh my god, I can't do this, it's fine. And, in order to do that, we're building machine learning models that pick up on speech patterns, rather than thinking about the way maybe after Grammarly or doing it, which is more you know, about making sure that your sentences are fully perfect, there's a different way to do it. And part of it is picking up on these things that you're talking about. You're making out mistakes, but I can still get what you're saying. Maybe I can suggest different words or different ways to stress it or whatever works, depending on the context.
Botond Seres 21:46
So, I think this was a creating truth kinda topic, that we'd like to touch on is how does Fluento actually measure fluency or proficiency or whatever other metric that it has.
Ido Cohen 22:01
Well, it's still in I guess, it's still in, in development. But there's different elements that we're thinking about, starting from your ability to achieve a goal, which is, I think, the most important thing, right, your level of fluency will change, and your score will change depending obviously, on what you're trying to achieve and your level of English, right. So, you know, ordering a cup of tea is very different from going to have a job interview, and that will affect the score. And so it's your ability to function or, or achieve a goal depending on the complexity. But there's all the things that feed into this, from your grammar to your vocabulary. So things that are more semantics, and on top of it things that would I guess, show Fluento that you're building confidence. So that's from you know, how fast you're speaking to how many pauses you're making along the conversation, your intonation, so you do sound robotic, or you go up and down and sing it to sound like you're confident. And the, the your ability to make sentences that are shorter, more coherent, depending obviously, on, on what you're saying, you can't use the vocabulary you're using in context. So for example, you're going through a job, job engineering, interview, engineering, engineering job interview, sorry. Are you using words that are applicable in context? Are you going around to try and explain it? Things that would make us as a, as a teacher believe that you're fluent and things that would obviously lead to things that we translated into how the AI analyzes your speech?
Botond Seres 23:50
Right, so speaking of specifically reaching a given goal, should we imagine it like, kind of a double blind test, like, one person would get the task, order a coffee, and the other person would get the task? You already know whether you call the cashier at Starbucks, okay? And then you have to write down, like what you think the other person ordered.
Ido Cohen 24:25
It's a bit more advanced, actually. So I'm not saying so let's start with the goal. So the way we determine what your goal is depends on what you want to achieve in real life. So if you're a person that wants to emigrate to a new country, well, we have a set of missions that are relevant to moving to a new country. If it's about getting a different job, and just different sets of missions, we can actually get very, very specific so for example, if your work for a hotel and your guest services in the hotel, the manager of the Hotel wants you to be very good at helping guests and you have missions that are specifically written for this. And the way we let learners practice them is that we create, like you said, different roles for each learner, usually contradicting. So let's say, let's take a simple one, ordering a cup of tea at the restaurant again, one of them is going to be a customer, the other one is going to be the waiter but will enter the mission with wide goals that would generate some sort of a conflict. So that process of negotiation is not linear. Hey, would you like, I would like a cup of coffee, 10 pounds, please Good luck. It's, it's there to continue the conversation trying to try to put you in positions where you probably experience it.
Botond Seres 25:50
So for example, one person could get, You want to pay by cash and the waiter could get, You can only accept cards.
Ido Cohen 25:58
What do you do, then? That's cool. And we also support right, so we're not throwing you into the water without giving you the tools to float. So we'll give you phrases and vocabulary along the way and break down the higher level goals into milestones that you can achieve. So that you understand really, what you need to do and go along with it. And the AI detects what you're saying. So when you achieve a goal, it will take and go, ‘DING’ you will achieve that milestone, go on to the next one. In fact, again, if you're stuck, you can hit a button, it will tell you what to say and then you can say it in the way I suggest or you can do it your own way it doesn't matter, the AI will still pick it up and be able to understand whether you're achieving that goal.
Botond Seres 26:48
So like how do partners get assigned? Is it like Chatroulette? Or is it more like Hey, I wanna work with this guy because he's sympathetic or whatever.
Ido Cohen 27:00
Chatroulette is probably a good way to it's a bit smarter than chat or that but it's a good way. What we do is we use smart criteria to match people. So it would be understanding their level, understanding their goals, understanding their native language. Other, any, interested AI picks up on so that we, we, make sure that the two learners are, you know, they're gonna get along, they're going to be friends, this is what we want to achieve. Right now, because we have a small number of users relatively, we let users tell us when they're going to be available. So they submit their availability a week in advance. And then on Sunday, we run this matching process and then picks up on all the availability and then lets you know when the next meeting is. We've also started now with kind of Chatroulette, so there's real time in some, some in some days, where you just click a button, and it will pick, it will search for someone who is available at the same time and then immediately match you with them. At some point of time, we estimated at around 5000 users that that is going to be available all the time. So you'll be able to click a button and get matched with someone immediately.
Botond Seres 28:19
Right? So Can, Can we customize the app to focus on some specific aspects? Like, let's say I want to expand my vocabulary or want to sell more Posh or I just want some more confidence. So can I just tell the app like, hey, I want to be this?
Ido Cohen 28:37
I mean, you can customize it into the level of saying, I want to be confident in a specific area. A lot of our users are there to build confidence among engineering. And so you can do that. And the mission is you get our engineering specific. However, that's what we're building in the future. So you'll be able to tell us specifically what you want, and see what's relevant to you.
Botond Seres 29:04
So would you like to expand on that a bit? Like, we see that users want to gain more confidence in engineering? Does it mean specifically like IT engineering or just in general, like learning technical terms or just business communication? Are they trying to learn how to say, instead of I think this is bad, they could say, I see this as an absolute win. It's an improvement for opportunity,
Ido Cohen 29:35
Botond, I love that, right? It's a positive argument, right? It's not telling you what's wrong. It's hard telling you what to do better next time. And it's it's interesting right now, but I think it's because this is we found most of our users on on Reddit, that most of our users are software engineers at different levels and expertise,
Botond Seres 29:58
Reddit is the dwelling place of software engineers, for sure,
Ido Cohen 30:02
it is. This is why I mean, I remember thinking about this with my co-founder for a little bit going, how, why is it everyone is software engineers? But now we know that. So it's, it's just because of the channel. And so, yes, most of them are, well, the, I can break them down into different use cases, some are there to get the next interview and they want to be able to speak about what they know and what they don't know and their expertise, some are there to perform better. So they already have the jobs when it comes to speaking to people across different countries or different teams, they obviously have to speak English and they feel super shy when it comes to expressing themselves and sharing ideas. And they need to feel more confident. And some are there to try to get the next job. And they're building it up. It's not like happening tomorrow. So you don't have a deadline. But they know where they're aiming for or the company that they want to work for. And when that happens, again, all of these use cases are interesting, because we are able to generate specific missions, depending on what they think is going to happen, or better, they come back in the community and go, Oh, my God, I had this conversation with this team. And in America, I had so many ideas, I couldn't say anything. And then we actually work out work with them to understand what that is and build specific missions that will help them make the next conversation much, much, more better. And the first thing that comes out, don't if you want to ask that, but it's the confidence. Fluento will boost your confidence really, really fast. Why? Because unlike apps where you can decide when to practice, you can obviously decide on Fluento when you want to practice. But when you are in practice, you're being put under this pressure to speak and deliver right now. There's no way to come out of it in a way. And you might feel extremely shy at the beginning, like a lot of our users are saying, but when you come out of it, you'll go oh my god, I did this, and then your confidence will go up. And because of the social interaction, you get a lot of it.
Botond Seres 32:18
So if I understand this correctly, you can kind of set up Fluento to be like your daily stand-up after work.
Ido Cohen 32:25
It's a daily standup, where no one is judging you on whether you've delivered or not. You're not talking to your boss, boss so there's no you know, you know, there's less, I mean, less anxiety involved, I think. And it's fun. It's fun, you're talking to people who you know, you might share a lot with, in fact, some of our people have developed relationships, so.
Dave Erickson 32:47
One of the issues I had with learning a second language was that the only people I could actually practice the language with happened to be high schoolers. And as a result, the language that I learned was very slangy, right? Because they spoke in, in a very slangy form of the language. So I was just learning and as I was developing conversation, I was doing it with high schoolers. So my language was all slang. I didn't know that. Because as a foreigner I couldn't really tell it was slang. They would just say, say it this way. And I would say it that way. And that was very slangy. And so when I started going into business meetings, speaking this language, I will you know, in, in Hungarian, there's a very formal Hungarian and then there's informal, I was speaking even one step below informal, which was slangy informal, to these people who want we're expecting to hear a very formal language so it was quite hysterical on many occasions, but with Fluento that that's one of the questions is, can you adjust or have it helped you with slang because sometimes people do want to add slang to add color to something or they want to understand how to use slang correctly. How does Fluento approach that?
Ido Cohen 34:16
That's a really good one. It's funny that you're mentioning this because that's the way I learn Turkish. I learned Turkish, Fluento did not exist at the time so I learned it on the street. It, I was a teacher with books and I did not open a single page. I just learned on the street and now when I speak to Turks they go are you Turkish? Are your parents Turkish or anything like that? And I'm not just because of the way I speak. I think that in Turkey there's less formality so people do enjoy the fact that the non native speaker knows all of these words and phrases. We don't have that yet in the end, honestly. We use, obviously, slang or, you know, terminology when it comes to conversations, like for example, whatever the terminology is for software engineering, we don't have classes specifically that are designed yet to, to build that sort of language. It's an interesting concept, though, because I do think that at some point, this is what people might want to do, and then we'll help, we'll help them achieve it. The foundations already exist. So a good example for that would be the fact that we, some, some of the feedback that we give users is how to, how to sound a bit more natural. So an example of that would be users or learners, I guess, trying to say things, maybe forgetting a word or maybe using a word that they've learned from the dictionary. And in many cases, the word is technically correct, but it just doesn't sound right in the context. So for example, the language is super, super informal and suddenly they find a word super formal and it sounds, like a little bit like, Tourettes because it just doesn't fit. And, and what Fluento does is pick, pick up on there, say, hey, this was great. But the next time you're trying to talk about this, use this word. And that's depending on what you're trying to achieve as a software engineer, or as a cup of coffee. And of course, that means that your level of formality changes. And that's worth using slang might might be a really good use case, because
Botond Seres 36:35
Or, to drink coffee might be more formal than software engineering.
Ido Cohen 36:39
What users do is that they do tell us whether they like to sound a little bit more formal or informal, or whether they'd like to sound a bit, it's funny, some of them want to sound more British, and then we adapt the feedback based on that.
Botond Seres 36:53
Oh, yeah, that's a thing.
It is a thing.
I mean, a lot of people I see well, some more Aussie, for example. That's, that's big.
What is it?
It's a trend to sound more Australian. It's a big thing.
Ido Cohen 37:12
I mean I'm guessing it is. Most of our users want, for some reason, to sound like the Queen of England. This is the, this is the, like, the Holy Grail. And I keep saying no, it's fine you don't need to sound like it's actually, but, but we can do that, we can adapt the feedback depending on where, what you'd like to say.
Botond Seres 37:34
I mean, in the end, people are going to sound like they would, they are going to always have an accent. I mean, I've learned English from all sources. Sometimes I pronounce words in the Australian accent, sometimes in British, mostly in American. Thanks Hollywood, by the way. But yeah, in the end, I'm always gonna sound like I'm Hungarian. It is what it is. I never understood the quest for perfect, immaculate British English.
Ido Cohen 38:10
It depends, you know, how you see it. As a native speaker, I don't really care. I care about other things about your ability to deliver a message. What I mean, I'm, I can see how for native speakers, for non native speakers it, that's the Holy Grail, whatever they decided that they wanted to sound like, and the Queen's English is, I guess, is the best. And this is where you want to be. The reality of it is, I think it's part of it is a question of confidence. Because building fluency is, is a combination of building your actual language, but also feeling confident. You can't have one without the other. And being able to speak like that makes you feel like you’re confident. What you should probably realize if you're learning English, is that it doesn't. Yeah, like you said, the accent doesn't matter. Even your grammar isn't the most important thing. And evidence of that is to look at the people who run Microsoft and Google today. They're not native speakers. They have some of, I mean, they both have pretty heavy accents, but they're doing amazing things. And so the, you know, the reality is that you can get really really far even with the, with an accent that isn't, isn't native. In fact, I think it gives you character.
Botond Seres 39:35
Speaking of accents, I distinctly remember, like, you're both native English speakers so I'm going to ask you, how do you pronounce the number 3 Right? And I distinctly remember that I pronounce it like the plants like tree because there's only way AIs five or 10 years ago would understand. If I said, three, it just wouldn't, it would never get it for some reason.
Ido Cohen 40:08
We should probably test it on Fluento.
Botond Seres 40:10
Dave Erickson 40:11
It's that, it's that th sound again, right?
It is .
Ido Cohen 40:17
Oh, yeah. I mean, if you're Irish, it's totally fine. But the only ways the anim, the, the AI, picks it up is not just based on how you're saying it. This is for obviously, pronunciation. But also the context. If you're saying I have three fingers, it's probably not a tree, it's fine on the plant, right? Make sense?
Botond Seres 40:42
Those do try to mix people of different levels, like, because I think that could be a great feature. That happens all the time.
Ido Cohen 40:54
Fluento, for nowadays, doesn't really work if you're a zero beginner, simply because we are an early stage startup. And we decided to focus on the and what we know how to do best, but also the largest chunk of the market. So, funnily enough, most learners are actually not zero beginners, they have some background behind them, perhaps not speaking, mostly grammar, but they can say their names and ask where you're from. And that's why we also like, we can touch that in a second. But our B2B, the organizations that work with us are not software engineering, or at least not all of them actually travel right now. So a couple of hotels, and an airport to begin with. So there's different use cases, however, you asked, Hey, can I be matched with somebody who was in a slightly different level. And we've seen that, with users asking for it. Of course, it's a complex thing to, to think about. Because we've there's too many things that we want to build. But we had had had, we have had cases where the AI had a bug, and it matched actually somebody who is slightly better than the partner. And the interaction was really interesting, because one of them felt like they wanted to teach the other. And teaching is a really great way to learn. And so that turns out to be a successful, like the bug standard into a successful potential feature. And now we're thinking about how to do it in a way that's, that's useful.
Botond Seres 42:36
Yeah, I mean, that definitely should be a feature, especially because as a non native speaker, I do realize that based on who I'm talking to, I have to speak different levels of English. So if I go to the Netherlands to a little shop, I'm just gonna say, hi, one coffee, please. Or if I go to the UK, I can just go like, Hey, how are you? Beautiful day, that's, that's a whole different thing. And even professionally, maybe you have some colleagues who speak quite good English, and maybe you have some colleagues who don't. And this requires a whole different vocabulary to communicate with them.
Ido Cohen 43:23
Fluento is not just there to help you with language learning, it’s obviously, the main, the main drive, but it's, it fosters social skills, and this is part of it. Something that you can't do yet with an AI is developing social skills, adapting your language, depending on who you're talking to. Just putting yourself out there. Which I think it's super, super cool.
Botond Seres 43:48
So this AI pulley rule in a generative manner, does it generate the task? The assignment?
Ido Cohen 43:55
Yes, yes 100%. It's super, super personal to whatever you want to do. So yes, that's, that's, uh, that's how missions get generated. We have a set of missions that obviously will also generate by AI. But now we're at the level where you can tell me, Hey, I have this job interview tomorrow, Give me some more specific and we'll generate a scenario that's specifically designed to tailor to what you're trying to do
Botond Seres 44:25
And what type of learning materials are included. So we already spent a great deal about talking about these exercises, but are there any additional learning resources?
Ido Cohen 44:37
We kind of talked about the fact that everything we do is for here to try and set you up for success. So the first material is the prep material. It's a set of goals, set of milestones that will help you understand what you need to do in a conversation, and it's a set of phrases, vocabulary, an example of a conversation that will help you understand what it could sound like. And then in the mission itself, there is a set of different cards that you get that will help you stay focused on what, where you need to put your attention, it could be your pace, and it could be a word or a type of grammar that we want you to help to focus on. And of course, after the conversation, then there's a set of materials or a set of feedback, I guess that will help you understand where you're going, why it's and where you need to improve. We're building drills. So that doesn't exist yet in the platform, but let's touch on that. It's the idea of, okay, I have a conversation and my next conversation is in a week, what do I do between, in between, and that, those are drills that we're building, and that will help you practice in between conversations. And the second one that exists right now is the community, it's about your ability to connect with those people that you either have spoken to or maybe other people and continue the conversations with them through the platform. And that helps you obviously, it's the idea of sharing pain and being supported, but also maintaining the momentum that you already built to continue the conversation.
Botond Seres 46:22
And this creates fiery crucible in which true relationships are forged.
Ido Cohen 46:28
Exactly, I mean, the community is amazing. It's a, it's extremely active to begin with. And it's like I, as a founder, I'm finding it extremely valuable, because I can ask a question, the most random question that me and my co founder has just discussed, and type it in and throw it out there. And you will get immediately people saying things in real time, and then continue and building that discussion on top of it, which is much further better than the survey that you send out there. It's super, super engaging. And I'm Yeah, I love that. But is this me and a founder? I think that's obviously for for learners. It's really about a lot of it is a lot of building connections and connecting with people on, you know, to try and solve similar pains. Yes. Especially when it comes to software engineering, in this case, they are sharing the same things.
Botond Seres 47:25
What are the greatest roadblocks to learning a foreign language? Yes, finding a teacher. And beyond that, actually paying that teacher? So, Ido How does Fluento Compare?
Ido Cohen 47:40
What, what makes them a roadblock? Or? In your opinion, let's address them directly?
Botond Seres 47:44
What makes lack of money a roadblocker?
Ido Cohen 47:49
It's money you're saying? Is it?
Botond Seres 47:51
Oh, yeah. Yeah. So I think that's, that's one of the biggest roadblocks. So I wonder if, in comparison to the teacher, you could get a lot more time out of Fluento? Possibly?
Ido Cohen 48:06
Well, okay, let's, let's address some of the pains. I think Dave also mentioned, you know, someone who had a teacher that understands, you know, that the mistake you're making is something that AI can't do, and he said something about fun as well. So the story behind it is maybe the fact that when I was teaching,actually teaching, the company I worked for built a set of materials, teaching materials that would ideally make all the lessons with any teacher with the same level of quality. And the idea behind it was just a slip, you will build material that would let, will allow the teacher just look at the page and read out whatever the question or the answer is, and without having to think about it. It's like maintaining high quality at a scale. And by on I guess, in theory, it sounds like a great idea. But the reality is that we discovered that what trans…what, you know, what, get what we want to say, what keeps users engaged is not just the material, but the experience and the teacher. So a teacher who's boring, and doesn't have the ability to actually get people engaged. It doesn't matter how good the book is. It's not gonna work, people confirm.
I can confirm it.
Fair enough. And if it's a teacher who’s fun and like, you know, fast paced and funny can get better results than the best book. Why? Because people you know, they're looking forward to being in the lesson. And, and so, like part of it is, is something that guides us in building in Fluentol. What do I mean by that? Well, first is it's, it's, it's making sure like working on a fun, engaging experience. It's for us. It's mimicking immersion. I'm simplifying it. But of course, it's about making it,making sure it's fast paced, mimicking immersion by getting you to a point where, you unless you speak about it, you're not going to achieve goals gamifying it? Anything that could make it fun, right? Roller Coaster? Roller Coaster? But the second part is, like, how do you actually get a good teacher? And this is where I think Fluento excels. It's, uh, you know, if you wanted to find an English teacher that will teach you how to do software engineering very well, well, we'll speak about software engineering, I don't know where you can get that. I mean, it, you might, I am.
Botond Seres 50:51
It’s gonna cost like 200 bucks per hour, at least.
Ido Cohen 50:55
So that's price. But before that, right? If your ability to find a teacher that will be engaging enough and understand whatever you, whatever you want to do that. And the third element is cost. And like you said, doesn't matter what you're learning, it's going to be expensive. Especially if you're in a country where English isn't spoken, if you're from Argentina, paying a teacher about $20 to $40 an hour. It's not, it's not achievable. It's not possible, especially you want to sustain it for 60 hours or 120. whatsoever.
Botond Seres 51:31
Same thing in Hungary, what are the teachers for 80% of the population is just out of the question. They cannot afford it.
Ido Cohen 51:39
Botond Seres 52:22
Now that’s, that's a reasonable price, it's the most reasonable price I've ever heard for learning English anyway. I mean, even Netflix is more expensive.
Ido Cohen 52:34
That's the way I was, you get the price. Otherwise, I'm not so sure.
Dave Erickson 52:41
We've talked a little bit, we've mostly talked about how individuals are using Fluento, you did mention that you're working with some organizations, maybe you can go into some use cases of how an organization might use lento. Because I'm curious about that as well.
Ido Cohen 52:59
One year, one of the use cases is travel related. So a certain airport, for example, has a lot of employees who are customer facing, they're interacting with passengers. And they need to be able to guide them to the right gate or solve things deeper and more complex, like Visa issues. Or obviously helping them with instructions or how to, how to get wherever they need to get in where they flew into. And the issue is that well, most of the employees in the airport can't really speak English at a level where they can have conversations. Teaching them is super difficult. Why? Because there's 1000s of those employees. And there's not enough teachers that they can find and like in a month or a couple of months to teach them all, let alone obviously spend the time to get them to a certain level. So the way Fluento addresses it is, first of all focusing on the specific use cases, right? We're not talking about cats that sit down there under the table, like in other platforms. None of it is, a theoretical level. It’s really getting into those conversations, those specific use cases and swiney to getting you to speak about them. Right. So I don't know where my gate is. Okay, and then we continue to the conversation or how do I get to the hotel, things that directly, specifically specific to what you're going to do. That’s one, and of course, the second one is scale. You can start right now. We have enough people to get them to start talking. So that's one and the other one is also travel related, but a different use case it's Hotel Management, or hotel guest services actually a better, better said. These are people who, again, interact with, with guests. And what we see is that in many hotels are not based in the US are based, for example, South America, there's a limited amount of people that can actually interact with customers. And whenever it comes to into conversations or at a deeper level, complex problems, then the problem creates a bottleneck; like, these passengers, these guests have to wait for the supervisor who speaks really good English, and to solve the problem. And what hotels want to do is get everyone to be able to speak and interact fast, and obviously cheap. And Fluento comes into like comms to save the day by making it affordable. And again, helping you focus on those specific conversations. And…
Dave Erickson 55:45
I assume another use case might be like a call center in the Philippines, where they want to improve the quality of calls, they train their whole department to get a little bit more fluent in English. Is that a correct use case?
Ido Cohen 56:00
Yes. So funny that you're mentioning it, like one of our users, who introduced us now to an organization is someone who works at a call center that addresses technical, technical support. And he, he joined Fluento because he wanted to improve on his performance. And now it's the management that wants to improve on the performance of the team. And increase the level of satisfaction, customer satisfaction.
Botond Seres 56:27
What does the future of learning look like?
Ido Cohen 56:30
Well, the future of language learning is, first of all, much more practical. I think what I've realized is, most learners that I taught in America, they're less about the theoretical level, they want to get practical right away. They want to be able to speak about whatever that is introduced themselves, and they don't really care about the grammar. And I think the future of language learning or any learning specific is about getting practical immediately. That's one, the second one is peer to peer. If this seems to be a trend, everyone thinks this is going to be part of language learning. It's about collaborating with a peer that is trying to do the same thing as you. That's not necessarily just language learning, but learning in general. And of course, the most exciting part is AI and how AI comes into place. For us, I think it's first of all, it's the level of accuracy and ability, your ability to pick up on any, any details that a human cannot do. But the second question is, I think whether it's VR or AR, not necessarily in a way in terms of learning, you're using Google, but it's whether it replaces someone or whether it's assistive. And I think that at least for the next few years, it's, it's really about, it's about assisting you with your goals, rather than replacing the environment. So this is why for example, VR, I think it's too early. Why? Because it's not good enough. Most people can't actually afford it. But it's just trying to replace something that doesn't have like Dave said, any human feeling into it. Whereas in assistive technology, you are actually doing the real thing, just an AI that supports you along the way to achieve that better. And you can see this, and not just in language learning but also other apps. Do like one of them is called car, which is a skiing app with a guy, AI guy that lets us go on a slope and like ski down the slope, and helps you along the way. With, with, with the feedback and analysis. I don't know if that was very coherent. To be honest. It's you put me on the spot.
Botond Seres 58:43
Oh, totally. I, I 100% agree. As soon as I asked the question, my mind jumped to VR it's gonna be VR. One guy is going to have one of these HUD’s get coffee and the other dude is gonna get denied coffee at all costs. It's gonna be beautiful.
Ido Cohen 59:02
I think VR is kind of far right? Because if you're talking about a language learner that is having trouble affording a 10 pound $10 app, getting Google and then the internet that's fast enough to do this and obviously, paying for whatever that cost is Mission Impossible right now.
Botond Seres 59:24
Right now within 10, 20 30 years,
Ido Cohen 59:28
but maybe a few years from now. And you know, when it gets a look there on the Metaverse, actually, somebody actually builds it. Perhaps why not?
Dave Erickson 59:37
Well, III do thank you so much for taking the time to fill us in on Flueinto.ai and how it can help everyone speak better English. For our listeners. Please join us the first week of next month for another ScreamingBox technology and business rundown podcast and until then may always find the right words. Thank you very much for taking this journey with us. Join us for our next exciting exploration of technology and business in the first week of every month. Please help us by subscribing, liking and following us on whichever platform you're listening to or watching us on. We hope you enjoyed this podcast and please let us know any subjects or topics you'd like us to discuss in our next podcast by leaving a message for us in the comment sections or sending us a Twitter DM. Till next month. Please stay happy and healthy.
ai, speak, learners, teacher, conversation, language, english, accents, learning, words, grammar, non native speakers, users, vocabulary, sound, building, level, understand, achieve, making
Words = 10,108