A Billionaire on the Future of the World | Timur Turlov | Podcast (Freedom Finance)

If you don't like something, like if your internet is slow or something else, just go and make it fast.

Yes, the feeling is mutual, I'm very glad. Hi.

Listen, it varies, actually; there are days when it's very little. I mean, when I just google a couple of facts or just ask something. I have this favorite pastime: when I'm flying somewhere and there's internet on the plane, I can actually spend several hours discussing a topic. And from that, sometimes a really cool, great, big report can be born. Sometimes they become public, sometimes they don't, but they help me at least organize my thoughts.

Listen, it really looks like this is just another industrial revolution of some kind, or even if not industrial, a technological revolution that we are witnessing right now. It is obvious that, perhaps by analogy with all previous industrial revolutions, the service economy will simply become even larger. In principle, it is almost limitless. Fewer routine operations will be performed by people. In general, human work will be focused more on communication with each other, and it will become much more creative. A huge amount of all sorts of boring stuff will be automated. And honestly, I don't believe that this will lead to unemployment. At least for now, if you look at us, we are always trying to set big goals for ourselves. For example, let's automate these particular processes. This will allow us, at the very least, not to ramp up our staff too aggressively, because we have certainly been growing quite aggressively in terms of headcount lately. This puts pressure on expenses and weighs on profits.

Listen, there are just under 10,000 in the holding. And in total, if you look at the whole group together, we recently counted about 16,000 people, including the holding plus the non-public parts of the group, which are not part of Freedom Holding Corp, where I am a shareholder. So, this is also a good indicator that, in fact, the demand for qualified people is only growing, and salaries are only getting higher. I recently heard from some of my friends and colleagues that a number of companies are quite boldly trying to experiment right now. They are, let's say, trying to fire a hundred developers and hire 10 whom they will pay roughly the same as those 100, right? Meaning, for instance, a million dollars a year, and at the same time achieve much higher productivity because they will be very effectively using all those cursors and tools that exist to speed up development. And by having fewer losses in communication, they aim to achieve radically faster development with a much smaller number of highly qualified people. To be honest, we haven't done such experiments yet, but it still seems, I repeat, that the demand for people is growing more and more. Undoubtedly, talented and bright people are probably more valuable now than ever, and we understand this better and better. I mean, we've probably never had such a deep understanding before that a company's key asset is its people.

It was always trendy to talk about this, especially at HR gatherings, but now it's becoming so obvious, in a way it never was for me before. I mean, it was clear that personnel decide everything and people are very, very important, but it seemed like client bases, technologies, and other things mattered too; now you realize it's just people, period. If you have great people, those people can actually change technologies very quickly. And, of course, I really expect this whole revolution to accelerate all processes even further, and the product itself is becoming much more complex. If you look at what a digital bank can do today, and what banks could do 20 years ago—even though they were already digital back then, at least the good ones were already quite well-automated—the functionality has certainly grown a lot. The complexity of processes has increased significantly in many ways.

On one hand, you realize we've learned to process data much, much faster, but on the other hand, we see that the complexity of features has grown so much; for example, we don't just file reports once a quarter, we do it in real-time for a huge number of things. And every time we adapt, throwing millions of dollars at automating, solving problems, and finding ways not to lose efficiency or speed while meeting a huge number of new requirements. Yet we manage; new requirements come in, the regulator comes and says, "Now I want this, this, this, this, this, this, and this, please." And it all has to work instantly. And we sit there and do it all, and we actually succeed. And this is happening all over the world. I mean, the product is becoming much more complex, and non-specific functions are appearing in huge numbers. And unfortunately, it doesn't mean we are hiring fewer people. It just turns out that those who have pulled ahead are, of course, becoming much harder to catch up with, simply because everything there is starting to head toward some kind of singularity. If you drop out of this stream and then wake up in, I don't know, 10 years, it's difficult to catch up, on one hand, but on the other, I realize now that we have actively started implementing agents for management. So we are preparing, and we really hope that we will have a full release in the fall, that we will have a fully functional assistant in our bank, and you will be able to give it tasks, including various voice commands: "Transfer money there, do this, get me this statement, tell me how much I spent on products last month? Book me a ticket there," and it will do it.

Both. We are building our own orchestrator, which is essentially a layer model that works with personal data and acts as a pre-processing step for a number of things. Then, everything that it masks—some of the data, everything it can't handle well itself—it hands over to OpenAI, it hands over to other APIs, roughly speaking, of other models. And in general, it is capable of managing various things, like for Kazakh, there is a Yandex one...and there is a model that Nazarbayev University made, and we are gathering them all to process the Kazakh language properly.

It is, of course, much better in...I mean, English is definitely the best, and there are no big problems with Russian either. But for the Kazakh language to work really well, and especially some mix between Kazakh and Russian, you have to put in the effort. And we are putting in the effort, and it seems like it's working out, so all these problems will be solved. And in the end, it turns out we are writing our own, in essence, pre-processing orchestrator. Then it triggers, among other things, some quite standard solutions. We understand that we will never build the kind of thing, well, we don't want to, at least for now, even try to do something similar to what OpenAI is doing or what Grok is doing. They are doing great, they are spending billions of dollars. We really enjoy using it. And for now, well, I mean, our task for the optimizers is, on one hand, to reduce the number of requests we send to OpenAI so that inference doesn't cost us so much, but at the same time, if the model isn't delivering something great, it still knows when to go there and force someone, someone more mature, to think about it better.

essentially, yes. Well, actually, even ChatGPT was supposedly built that way, so it could understand whether to answer you quickly or think before answering. We are, in general, trying to follow a simplified scenario compared to that. That's what I started saying all this for. Because when we started implementing this, we realized that a huge amount of functionality we used to program won't need to be programmed at all. And, well, you have, say, I don't know, creating some templates. A standard process, actually, for a bank. Setting up some standard recurring payments that you need to carry out. I don't know, paying for daycare once a month, or paying for some club or something else once a month. Now you can just tell your assistant, uh, to remember it and, I don't know, either remind you or just handle that payment yourself once a month. And you don't need to write code, you don't need to write the functionality for it to work at all. And that's really cool because it turns out that, well, OpenAI itself, in its map, already allows for all sorts of delayed tasks and scheduled tasks. Really, a huge amount of functionality; you don't need to design an interface for it or come up with the logic for it. The user can literally come up with the logic themselves, and it will work for them. And in that sense, we're reaching a point where almost all our users, well, by getting AI in their hands and interacting with it almost like a real helper or assistant, can program it for completely unique tasks without us having to guess if people will need them, or in how many months we'll test and roll out that feature. They are basically starting to write all this code by explaining to it in plain human language what they need. I don't know, like, take every time I make a payment, round the amount up to 1,000 tenge and put it in my savings jar. I mean, in a bank, this would actually be a development task that would take quite a lot of time to implement this logic. You can tell the assistant: "It has all the key methods, it perfectly understands how to move money between accounts. It can perfectly analyze every one of your transactions. And then, on the fly, it starts to implement some kind of logic like that all by itself. And it turns out that all this code didn't even need to be written in the first place."

Well, listen, traditionally one of the simplest ways is limits. So, you can tell it which transactions you are going to confirm. I mean, by default, you usually confirm any transactions that it initiates. So, it basically prepares an order for you, and you can take a look at it and press a button. Yeah, you can give it some limits and say: "Like, I don't know, up to 10,000 tenge: 'I don't want to check anything, just do it yourself.' But if you sort of understand how it all works, then for anything over 10,000 tenge, show me everything you've done." And then it can just write to you and say: "Listen, you asked me to remind you that you need to pay 10,000 tenge for this thing. The time has come, are we paying or not? Here is the payment slip for you."

Well, I really hope this autumn, we are preparing for Digital. For us, that's also a crazy challenge.

It's our conference, which the president usually attends. And all the startups and the entire tech crowd gather there, everyone shows their achievements at the intersection of Fintech, eGov, and, well, all things GovTech, Fintech, and startups. It's our main technological conference.

I read somewhere, or I think I even heard it in an interview with an astrophysicist, the one Dud interviewed, who was looking for that planet and proved it exists, back when Pluto was demoted and stopped being a planet. He proved there is indeed one more planet that nobody has seen, but it's there. Batygin, I think, yes, or Konstantin. He's somewhere at Caltech. Honestly, I don't remember his name either, but he said a wonderful phrase that humans are by nature extremely curious creatures. And satisfying curiosity for people is, in general, one of the key pleasures. I absolutely share that sentiment. I am also extremely curious, and I get great pleasure from satisfying my curiosity. I like figuring out all sorts of complex things. When artificial intelligence first appeared, I definitely got a huge kick out of how much the productivity of these learning processes increased. You can take any topic, I don't know, like, how programming languages evolved in general, where it all started, how modern screens appeared, like LCD screens, and then the whole evolution into LED screens. Take that blue LED, thanks to which all modern screens in the world actually exist, which no one could make for a very long time. I mean, there was such a problem making a blue-colored LED. And out of the entire RGB spectrum, it was specifically the blue one that had the biggest problems. And in the end, some Japanese guy invented it. And that actually changed everything. Later in screens and in how we perceive information in general. When we moved from cathode-ray tubes to, effectively, LED displays, all light bulbs became white. Before that, it was also impossible to make them white until that blue LED appeared to mix the whole thing. And again, there are, of course, tons of YouTube videos, there are lots of articles, all of this could have been done and read a long time ago, and I have been using this for a long time. But when ChatGPT appeared, when all those LLM models appeared, it became possible to dig into the essence, that is, to find answers very quickly to exactly those questions that interest you. So you have to read the whole article in order to, uh, find the answer to a specific question you're interested in. And what exactly was the problem with that? And this depth, the fact that LLMs can actually digest a fairly large amount of data and answer you in context specifically about the question you're interested in—that's a special kind of pleasure. And it doesn't matter if you're studying something like economics, monetary policy, or if you're trying to figure out some purely engineering matter, or even a physics matter. As I always, well, as we joke with my kids. My daughter likes to say that, if you, paraphrasing the story, if you, uh, don't know physics well, the world is full of magic. Well, if you don't know magic well, the world is full of physics. And so in all of this, sometimes I truly realize that, uh, I don't fully understand how this thing works at all. To understand deeply, to figure it out. I tell, well, I try to tell my kids this too, but the iPhone we're holding in our hands is an absolutely magical thing for the vast majority of people. People don't understand anything about it at all. I mean, it's already such a level of abstraction that it's become difficult even for people with higher education to grasp. And when you start digging down, uh, you're just satisfying your basic curiosity. Well, I have that. And when you're digging into a topic, well, like telecom, for example, I don't remember if we talked about this with you two years ago or not?

Well, there's this realization (not just for me, but for a lot of people, actually) that technologies are changing so fast that there's a great chance you can find a technology and master it a little earlier than others, and that will give you a colossal competitive advantage. That is incredibly motivating, because at some point you realize that, damn, well, if you understand how something works a little better than someone else, you'll win the competition, you'll earn money, you'll be able to build something faster or cheaper, and it will eventually change the whole game. And for me, that has always been the best, the very best motivation to sit down and dig into something.

It was born, firstly, from...The whole idea of entering telecom appeared when I realized that, it seems, telecom is lacking rational adventurers. It's a culture in itself. There became a lot of state involvement at some point across the entire telecommunications industry. And a very large economic concentration emerged anyway. Everything slowed down significantly. And I saw how young, talented guys would come in and start burning out quite quickly because they couldn't really fulfill their potential. They start to get bogged down in all their bureaucratic procedures. And in reality, well, everything just keeps getting pushed, carried over, and delayed. As a result, the efficiency of all processes becomes so low that at some point it started to seem like, man, what a great industry. If you can make money while procrastinating like that, then what could you achieve if you actually started doing it much more efficiently? Of course, it's not that simple. When you start diving in yourself, you begin to understand a bit better why they had all these limitations and why everything was so difficult. But the industry is definitely very interesting. And the main challenge we saw for ourselves at the moment we decided to get into this, which was late 2022, early 2023, was that Kazakhstan was ranked 100th in the world for fixed internet speed. For mobile, it was in the 60s somewhere at the time, which also, in general, wasn't very cool at all. And I remember at one point it was almost a huge hype to discuss that we have very slow internet. So we sat down with the guys and thought: "Why is our internet so slow? Are there any objective reasons for this, or should I just go and make the internet fast? I always try to instill in my children and my employees—probably in my children first and foremost—this idea that if you don't like something, say, your internet is slow or whatever, go and make it fast. It's actually cool because you can always, I don't know, move to some country where the internet is already great, or you can try to make the internet fast at home, and if it's something of value, you'll make money from it. And, well, in other countries people also, you know, kept building and eventually built themselves a fast internet. After all, it's not the gods who fire the pots. The longer you live, the more you realize just how much it is not the gods who fire the pots. And that you don't need to be the smartest, the most talented, or the richest person to change the world. In fact, very few people actually change it, and the competition in this industry isn't that great. I mean, it's obvious that Kazakhstan needs good connectivity. Obviously, we deserve to be in the top thirty for connection quality. When we started looking around, we realized that there are no fundamental problems with fixed-line internet at all. You can't say that we have a very harsh regulatory environment. That's actually not true. It exists, it's not the softest, but let's put it this way: laying communication lines is regulated in a certain way all over the world. It's not a free-for-all anywhere, but in Kazakhstan, things aren't too bad in that regard.

So, everything is actually quite livable. You can build networks, and you can build them much faster and cheaper than in many other countries in the world that have much higher connection quality. There aren't many industries where the state openly supports you. When you come and say, "I am going to build internet here." Almost everyone tells you, "Good job, go ahead, people need this." People, people really need fast internet. They complain about it. Just like they complain, I don't know, about the lack of hot water or something else. It's a basic human need. It's a basic utility that, roughly speaking, should be in every home. If it's not there, people get really, really upset—even more so than from a lack of hot water. Actually, when we started digging into the cost price, like how much it costs to lay a kilometer of fiber underground, we looked at the benchmarks. At that time, it cost something between 13,000 and 16,000 dollars per kilometer. Laying fiber underground had that kind of market price. And that's when basic curiosity kicks in. You start asking: "Why 16? Why 13? Can it be done cheaper?"