Transcription
We have the beginnings of artificial intelligence, truly like some copy of our brain. And on the other hand, unfortunately, we have not yet seen this ideal computer. There isn't a second Word, right? There is a monopoly. But here, as you say, no monopoly is visible yet. In my circle, a significant number of people have started using artificial intelligence as psychologists and discussing their personal problems with it. There are many conversations about the singularity, about artificial intelligence ruling, about the dangers of artificial intelligence. The future will change significantly after the appearance of AGI, a supercomputer that will multiply the volume of tasks it can perform and their accuracy. Good day. This is Andrey Movchan and another of our shoots from the so-called Future Economy cycle, because we are, in fact, talking about future economies, about what will happen to economies next. In this cycle, we meet with very different people who, in our subjective opinion, understand something about the future of certain economies, certain industries, certain ideas. And we ask them questions. I come up with the questions, they come up with the answers. This is how long programs that talk about some economic topic are created. Today, we decided not to deviate from general trends and talk about AI, which is heard everywhere today, about artificial intelligence, to what extent artificial intelligence is an economy, whether there is a future there, and what, in fact, can be with it, what artificial intelligence is from an economic point of view. Today we will talk with Anton Saburov. He is, do not be surprised, not a scientist, not a programmer, not a specialist in artificial intelligence. He is a marketer. And this makes the conversation more interesting from an economic point of view. We will talk with him as a marketer and as a creator of artificial intelligence, including one of those entrepreneurs who have engaged in artificial intelligence. Well, it is your right to agree or disagree with his opinion, but at least you will have it after today's program. Anton, thank you for joining us. Let's start with the right question, shall we? So, why should we talk about artificial intelligence? Thank you very much for the invitation. I greet everyone. Actually, I have extensive experience in artificial intelligence. As soon as the first models appeared, I, as a practitioner, immediately rushed to test them, because as a visionary, I believe that they are changing the market, and accordingly, I was not mistaken. Currently, I have public courses on YouTube, where there are already millions of views specifically about artificial intelligence, explaining it in simple, understandable language. And this is one of the factors that you can trust me due to good audience feedback. And the second point, I am creating my own product based on artificial intelligence, working with influencer marketing, and as a practitioner, I see both problems and opportunities every day. Therefore, Andrey, I will gladly share my developments, my thoughts, and this trend in general, which is now actively going on and continues like a hurricane.
Okay, thank you. Let's start with what artificial intelligence is. Artificial intelligence is a machine. It is a machine complex that is designed to replace routine tasks, in particular calculations and other technically complex moments. For a person, artificial intelligence is currently seen as some kind of chatbot. It is a communication algorithm where you communicate just like with a friend on WhatsApp or any messenger, which makes it easy and understandable, and the person receives all the necessary benefits for them, answers to questions, reasoning, and so on. Can we say that this is intelligence? Yes and no. Currently, it is still an algorithm. It is where the machine is programmed to solve specific tasks. But now there is such a term as AGI - this is a supercomputer, a super-intelligence, which the not-so-unknown company OpenAI is ready to present next year. And this will be a new revolution, where, well, there will be computational power, and in general, the work will be built completely differently. Therefore, yes and no, we have the beginnings of artificial intelligence, truly like some copy of our brain. And on the other hand, unfortunately, we have not yet seen this ideal computer.
Is it a copy of the brain? Indeed, I have heard that it is more like recursive models that, based on some internal correspondence tables, simply select the words that fit best. That's right. That's right. Currently, these are machines that are trained on a certain volume of data. This data is very huge. The next revolution will actually happen from month to month. We are recording this video in July. The fifth version of ChatGPT will be released, where this level of trained data is billions, very large, and it is incomparable with any world libraries. Therefore, Andrey, I believe that this is something new, interesting, that will give us a new leap forward.
So you think that, so to speak, a parrot trained on more words is more intelligent than a parrot trained on fewer words? Yes. But here the parrot must also have a small function in the form of some neural networks inside that can connect these words into coherent speech and output it, in fact, in the context of the very request. Therefore, a parrot with such functions of a terminator, yes, that's a supercomputer, artificial intelligence.
Okay. Well, since you are involved in artificial intelligence, you can probably describe the state of the market, so to speak, quite accurately now, what is generally happening in this market? Yes? I will tell you here, probably from the perspective of startups, because my experience in this is about 10 years. And I actively see how the market is changing significantly with the advent of new models. How does the market for AI-based products look now? Inside, there is most often some core in the form of some model. It can be a model from OpenAI, from Anthropic, from other large manufacturers. And on top of it, there is a shell of some services. For example, let me tell you about my product. The product is called Influenance. Inside, it has a large database of influencers, about 5 million contacts, and accordingly, an AI shell on top that selects relevant influencers for the task that the client comes with. That is, artificial intelligence performs a very small task. It is to correctly interpret the client's request, select information from the database, and, accordingly, combine them and provide the result in the form of integration or some collaborations with influencers.
Okay. Well, this is like your product, right? But in general, if we talk about the market as a whole, how many players are there, what are they doing, where are they looking, where is it moving, in your opinion? There are about 10 major players in the market, such as OpenAI, Anthropic, Claude, Google. And large giants, including Meta, are creating their own models. Some models are open-source, they are free, they can be used for commercial products, some are paid. Accordingly, this is the core around which products are developed. All other products work via API models, when they simply take the core of these large corporations, large companies, and use them for their own purposes. So, in fact, they solve the needs of the end client by thinking through all the logic, and inside are the models of large companies. Why is it done this way? Because, unfortunately, small companies, small startups cannot create their models to be so effective, because they simply lack this knowledge, the base that they can upload, and the training of the system itself. And this is a problem. Therefore, there is this core in the form of manufacturers, developers, and around them are small startups, companies that create their shell around it to solve specific user tasks.
Does this core cost anything for small companies? Is there some B2B market? Yes, of course, it's an API model. That is, small companies pay for each request. It all depends on the models. But, for example, I'll tell you about my example, I currently have about twenty paying clients. Our turnover is around a thousand dollars a month for processing these requests. And these are the funds that go to, well, in our case, OpenAI.
What is the economic sense for companies like OpenAI, for example, to cooperate with you instead of making their own shells? They simply cannot do it physically, because the value of each small startup is that it solves a specific task for a specific client. And they will always do it better. Because, for example, I am a marketing specialist, I know exactly what the market need and pain is, and accordingly, I solve it. And a large company is still a corporation. Its task is to make this engine in the car. And how this car will look externally, this is all already, well, the narrow tasks of each client, each company. Therefore, it is much more profitable for them to sell access to their system in volumes through this connector. And, accordingly, for small companies to take it. Therefore, here there is, on the one hand, a partnership, and on the other hand, they simply focus on working with their model, upgrading it, and thanks to this, startups also receive great value.
Interestingly, this doesn't look much like traditional IT, where there is, for example, Windows and there is Microsoft, which made a shell for Windows all by itself, and there is almost no second Word, right? There is a monopoly. Here, as you say, no monopoly is visible yet. It is there. It is there, yes. There are vendors. As I said, there are about ten players. There are those, for example, like Link systems, yes, free to use. There are paid ones, for example, like the same, I don't know, OpenAI, you cannot use it for free for yourself, install it in your own environment. Well. I wouldn't say it's a monopoly, because access is available to other companies, but it is, uh, not very economically viable, because creating your own model, training it, is colossal expenses, and, well, you simply won't be competitive in the market with small amounts of money. That's why, yes, there is a monopoly, but these are very clear rules of the game, these are partnership rules of the game. And at the current stage, the entire market accepts this, and in principle, it looks like a sufficiently, well, clear, logical scheme.
And do you think there are enough opportunities for such a number of startups? Won't there be a situation where one startup manages to do something successfully, and a thousand startups go bankrupt because it already exists? Market competition is what it is. I believe that it is the opposite, it is good. Here, Andrey, I will give a small example. How has the startup market as a whole changed over the last 3 years? If before, according to many chats and forums, there was a situation where there was a startup founder with some brilliant idea, in his opinion, he comes to the forum and says: "I am looking for a CTO, a technical director, who can develop a product for me. I have an idea, I understand the market." Now, over the last 3 years, with the advent of large models, the market has changed. Now technical expertise is absolutely not needed. And because of this, the number of startups has increased exponentially. Here is a reasonable question, probably the next one. From the investors' point of view, how profitable is this for them, but from the market's point of view, the market is now, well, teeming with startups, there are really a lot of them appearing every day. And here's a small figure for us. A minimum viable product can now be made in less than a week and for less than $100. It was impossible even 3 years ago, because you needed a team, at least. You needed some of your own software capacities, which you could do. And now the market is simply ideal for entering it, on the one hand, and on the other hand, you have to look for these narrow passages in order to hit exactly your client and meet their need for what? In your product.
Well, if we generalize, what do you think artificial intelligence is suitable for? What is it not suitable for? Good question. I recently also recorded a video about this that got 800,000 views. It's about how deleting ChatGPT breaks your brain, right? Because you stop thinking, you get completely immersed in it, and accordingly, well, you stop being a reasonable person. I believe that artificial intelligence is ideal for solving routine tasks. Those where we didn't need to engage in creativity, think about something, and where it was necessary, I don't know, Ctrl C, Ctrl V copy, paste. Exactly the same way, for example, I solve all mathematical tasks myself, because I am not very strong with numbers, I don't like it, and accordingly, I delegate these functions to, well, the programs. Therefore, the first task is such routine, understandable things. But a second question arises: what about creative tasks? Because we see a huge number of products that help creators like us, Andrey, create YouTube videos, write scripts, work with video. And here a very interesting question arises: is this harm or is it a benefit? I don't have a definitive answer here. What am I trying to do? I try to leave some tasks for myself so that the neural networks in my head don't dry out and continue to work actively. Well. But experiments are always necessary. But here, perhaps, the important thing is to be in the moment and to fix yourself. Haven't you just become, you know, a fleshy bag for artificial intelligence, which performs very basic tasks for you? And are you carrying risks for yourself in terms of losing some of your identity and your whole, you know, reasonable human being as such? And here is an intimate question for everyone to ask themselves, so everyone answers it for themselves.
Well, you are скорее talking from a moral and ethical point of view, right? And I am from a technical one. We do use artificial intelligence for translation, for subtitles, for some simple things, even for correcting design. But there are a lot of conversations going on about the singularity, about artificial intelligence ruling, about the dangers of artificial intelligence. I don't understand any of it, and I have the impression that this is all talk and, you know, fears that people spread. In fact, artificial intelligence will remain where it is. Yes, this machine does some simple tasks and cannot move further. What do you think about this? Yes, I will share one term that is currently actively developing. These are so-called AI agents. An AI agent is essentially a copy of one human unit within our company. I'll use my own example, I have minus five people in my team thanks to AI. Previously, I had a team that was specifically engaged in integrations with influencers, this team no longer exists, because we transferred its functions to artificial intelligence in the form of AI agents. This is the kind of story that is starting to take place in business now. And here, probably, from the perspective of companies, it is necessary to consider specific business processes. For example, what is already out of work is the removal of accountants, right? Because routine tasks, taking them from the head system, sorting all the invoices, and so on, are easily transferred. Or removing salespeople? The most controversial story, which my subscribers always argue with me about, because is it really possible? And yes, we have now confirmed in practice that one bot can technically replace a junior or middle salesperson, you know, an entry-level or mid-level one, and accordingly, do these things better. Why better? It doesn't lose tasks, it does follow-ups with clients, reminders about itself, and simply systematically, time and time again, it moves your deal towards a sale. This is really true. Maybe not always ideal, but the goal is achieved. We simply see, again, the turnover of money. Accordingly, we move on. Developers, our beloved comrades, who develop artificial intelligence themselves, are now switching to it. And in most companies, knowledge of artificial intelligence tools is mandatory in order to speed up their tasks and become more efficient. Therefore, Andrey, integration is happening into all aspects with one small caveat. In my circle, a significant number of people have started using artificial intelligence as psychologists and discussing their personal problems with it. And this is a very interesting trap that people get caught in. And if you remember, there was such a movie, Her, where a person communicated with a virtual Siri, eventually fell in love with her, and it was a very interesting film. And something similar is what we are now seeing with ChatGPT in particular, that people are starting to communicate, and I already have precedents where people have fallen into this trap, and it was not very good for them. That is, after that, they had problems in life, in finances, in their psychological character. And, unfortunately, such a trend is also going on from the negative side of this issue.
Okay. Interesting, are you following what the government thinks about artificial intelligence? Absolutely. As a developer, I need to understand, because most of my clients are from Europe and America, and accordingly, I need to be within the legal framework. Well, the States, of course, are driving this topic, it is relevant, and with the arrival of Trump, this story is actively developing precisely so that, well, they want to make themselves the technological company of the world. In part, this is true. China is slightly pushing from behind, because Chinese neural networks are also developing quite coolly, and Europe is moving a little further behind. Europe is not going the technological path, but more the legal one. Initially, they are now preparing all the groundwork, which most developers, including myself, are a little afraid of, that it can be transferred to issues of ethics, all these questions. And I think that this is, on the one hand, a small misstep, because, after all, technologies need to be moved forward, and only when the technology goes into the world, then regulate it. But this is my subjective opinion. Therefore, governments are actively working, active good governments support and develop technologies. States that are a little old-fashioned are going a slightly different way. It's only understandable to them, but perhaps they see some logic in it.
Okay. And the use of state artificial intelligence, the question of armaments immediately arises, for example, right? That is, a combat apparatus with real artificial intelligence can be significantly more interesting and smarter than the current unmanned vehicles, even those that exist now? That's true. Yes, that's true. Artificial intelligence is actively used in both civilian and military purposes. Such activity exists. But the same models that are developed by large corporations, about which I spoke about a dozen, are used there. But the task is that they do not fly into the cloud, that they are not located on the external internet, so a closed internal circuit is installed there, and it is impossible to access it from the outside. Accordingly, the system itself does not communicate with the external world, well, and is tailored to its specific goals. For example, within banks, it will always be a closed circuit. And, accordingly, to avoid leaks, and so on. And the task is precisely about personal data, interference with it. So, yes, there are implementations, I see them too. But, well, knowing corporations, knowing the state, it's not very fast, but as I see it, the trend is also in this direction.
Just to understand, this closed circuit does not contradict renting capacity from large companies, it turns out. Yes, there is renting of capacity. Well, let me give an example, for instance, Microsoft's data servers, it's a huge building, in fact, where there are thousands of servers located and performing these tasks. And there are, I don't know, 5, 10, 20 servers that, for example, belong to some bank. Technically, it will look like a closed circuit, because it does not have access to the internet. It just doesn't have a plug there and cannot go out, but at the same time it has a direct connection to the entire banking infrastructure. This will also be considered a closed circuit. That is, it can be located in external data centers, in internal data centers. These are already, well, technical connections.
Okay, I understand. Well, so regulation doesn't really hinder you yet, let's say? In which country might it be better to work with artificial intelligence now? Well, besides America. Well, America is the number one country in terms of investment, in terms of state support, grants, and everything else. I haven't encountered any obstacles in Europe yet, because, in fact, compared to competitors, I am a very small company and I am not yet on the radar. But the product is global, it's doing well in all markets. Well. But we also integrate the client into our circuit, the client conducts all communication from their side. We do not take responsibility for this. And in general, yes, everything is quite comfortable. For now, well, honestly, I don't foresee any major risks for myself as a product. Well, in any case, even if they appear, artificial intelligence will help to bypass them, including by simply changing products.
I can't help but ask, since we are involved in investments, right? Our viewers usually know this, right? That we manage global, liquid investments, indeed, right? Therefore, the question about liquid investments. How do you see attempts to implement artificial intelligence instead of an analyst or instead of a manager? There are quite a lot of precedents instead of an analyst, because, as I repeated earlier, there is a large amount of routine tasks here. This is to take some object, analyze it, provide information in the required format, right? That is, in the required format. Well. And there are many such precedents. I give a green light for this to move into. But a small, but investment is always intellectual work. And I highly recommend doing a double-check here, doing a check by specific people. But if they, for example, could, I don't know, analyze 5-10 objects just with their working time, then now they can analyze hundreds of these objects, looking at specific markers that are relevant to them, well, and checking the system itself for this. Therefore, with analysts, everything looks quite positive. From the perspective of a portfolio manager, that is, a person who performs all projects from selection of objects to decision-making, this is a riskier story, because more responsibility is placed on the system here. I believe that it will come to this. There is such a thing now as AI agents. One performs all routine tasks, and the second performs control tasks. Well. And such multi-agent capabilities provide more options for implementation. and, accordingly, risk reduction. But at the current moment, well, a person is still necessary in this process, simply to control and make the final decision somewhere for themselves.
Okay. Can you name any other areas where artificial intelligence is in high demand? I mean not applied actions, like subtitles, but specifically industries, specific areas where a revolution is happening with artificial intelligence. I really like the field of medicine. Who would have thought, because medicine is an even more responsible field, but in medicine there are the same analysts, when a person, for example, gets a diagnosis and a lot of data, and you need to correctly interpret, for example, the presence of certain markers, certain diseases. Well. And here large computational power is needed, and companies come with solutions to these very narrow tasks, for example, finding cures for HIV, AIDS, and other diseases. This is very interesting, and I am really waiting for a revolution in the form of solving global human issues. Well. And then, in fact, almost the entire real sector, if you look at it, these are developers, this is the financial sector, and so on. Here, Andrey, implementation is also happening almost everywhere, probably at different speeds, because, well, there are companies, for example, the same developers, they are not very fast guys in making decisions, in implementing new business processes. Unfortunately, it happens a bit slower. And everything related to, for example, I don't know, sales, some communication, where the direction is much closer to money, here I see real activity in implementing new projects, automating business processes, because the impact is clearly visible in the moment, 2-3 months of implementation, you see how your company has moved forward financially, and accordingly, it's really, well, it's great. Well.
Okay. Well, let's talk about the future, since we are talking about the future economy, let's talk about the future. How do you see the development of artificial intelligence in 20, 30, 50 years? On the one hand, such questions always please me. As a futurist, I love looking into them, because as an entrepreneur, I try to predict where the trend is going, and to fit into it and not lose it. On the other hand, they scare me a lot, because the emergence of, say, ChatGPT, and artificial intelligence in general, appeared simply in an instant out of nowhere and became such a huge trend. The current, perhaps, 3-5, and maybe 7 years will be spent on AI agents, on replacing the functions of certain people, certain business processes of companies through artificial intelligence. A very clear applied task, which is now not at the very beginning, but this trend is just emerging. The future will change significantly after the appearance of AGI, a supercomputer that will multiply the volume of tasks it can perform and their accuracy. And this thing can fundamentally change, well, almost our entire life, I sincerely hope for the better, right? Because the uprising of machines and so on is always nearby. Here is a very interesting study. I'll deviate a bit. This is a projection for 20, I think, 35, it's called AI 2035, if I'm not mistaken with the title. This is when there are two scenarios: the uprising of machines happens or it doesn't happen. Well. And it is very detailed, broken down by year, by year. I highly recommend watching it. And here, from the perspective of futurism, I am for a positive outcome, that it will not only be our assistants, but also, in fact, perform our functions so that we can switch to some more, I don't know, creative tasks, more important tasks, well, and simply engage in our self-realization.
Okay. What else haven't I asked, in your opinion, about artificial intelligence? I would like to talk, perhaps, Andrey, about investments in artificial intelligence companies. Perhaps this will be some next block, right? Because there is enough misinformation there, and a very hot period is underway for choosing companies and so on. From the point of view of artificial intelligence itself, I would probably like to talk about how people work now and about this consumer direction in general. It is very sad for me when people miss this trend, are afraid of it, and remain, well, I will call them old-believers, who continue to do all their tasks as they did them 5-10 years ago. And, unfortunately, if you remember, after the advent of personal computers, life also changed, and it started with job postings requiring confident PC users. Now we see the same thing. And people who are in distrust of such tools may simply be left behind in life. And this is a risk that truly exists. Therefore, here I will probably maximally recommend people of middle to upper age, with young people, I have no questions for them, they live this. Well. Do not ignore and at least be interested and try to study some things for yourself in practice. Well, so that the brain doesn't freeze. And, accordingly, these technologies, well, this is our future, you need to develop in them. Well. And on the other hand, young people should not go into it headlong, because, as I gave examples with the psychologist, well, it's really dangerous, guys, and you always need to be aware, in proper consumption. This is a topic that is very close, and it is very easy to slip, to become, in fact, a slave to artificial intelligence and get into trouble.
Okay, Anton, thank you. Well, we had such a short conversation today. Perhaps because Anton gave very clear, short, concise answers. Usually, people speak much more expansively, although there may be even less information there. We actually talked about artificial intelligence for half an hour. And the second reason why it's half an hour, and not, say, an hour, an hour and a half, is possibly because there isn't much information yet that we could provide and present, except for some technical details that are unlikely to be interesting for such a conversation. Artificial intelligence is just beginning to develop. There are many stories about it, many myths. There are also many real ideas that are being implemented. Investments in AI startups are massive. The question of what they will bring, I think neither Anton nor I, none of us can answer this question. And, apparently, we will have to wait a few years to have a conversation that will last a little longer. Anton, thank you very much. Thank you to everyone who is watching us. I remind you that we have a YouTube channel where you are watching us, the Movchan's Telegram channel, and our other channels. Come, watch, subscribe. We will be glad to see you. Thank you. All the best and until the next program.