.The introduction of generative artificial intelligence (GAI) technologies into people's daily and professional activities not only simplifies their lives and work, but also radically transforms the role of humans in the world. This opinion was expressed by Sergey Ananyan, the founder of the Russian IT company Megaputer.
We spoke with Sergey Mikhailovich on the eve of the conference "Generative Artificial Intelligence in Economic and Social Spheres: Pro&Contra 2025".
We would like to remind you that the conference will be held from October 1 to 4 in a hybrid format: in-person sections are planned in Moscow, Ufa, and Tomsk, and online broadcasts will be organized for remote participants. This is an international event where experts from academia, business, and government will discuss trends and norms in the field of artificial intelligence. The conference is organized by the Faculty of Public Administration (FPA) of Moscow State University and the University Consortium for Big Data Research (Big Data Consortium).
- Sergey Mikhailovich, what do you consider to be the most pressing issue in the development of AI, and why?
In my opinion, the most important problem is keeping the development of GII under human control. The use of GII, which is becoming increasingly intelligent, to solve a growing number of interconnected problems in various fields is an objective process that humans cannot stop, and they do not want to stop it. However, it is important to remember that the rapid development and widespread adoption of GII, combined with a lack of understanding of the reasons behind its decisions, poses a risk of losing human control, which could have catastrophic consequences.
Yes, GII still has a lot of "childhood diseases," such as hallucinations and a lack of ability to build complex chains of reasoning and critically evaluate their own decisions. However, recognizing the potential benefits of GII, large corporations and entire countries have invested significant intellectual and financial resources in the "model race." These models are rapidly becoming more intelligent, allowing GII to overcome its shortcomings and tackle increasingly complex tasks, expanding its potential applications. It would seem that we should be happy, because an age-old dream is coming true: GII is helping people automate a huge number of complex processes.
However, this time we are not just talking about making human work easier, but rather about a radical transformation of the human role in the world.
We are increasingly relying on GII instead of solving problems on our own. Modern students will confirm that it is extremely difficult to resist the temptation to use GII to complete course assignments, as other students who use GII for this purpose gain an immediate advantage by spending two orders of magnitude less effort and time and producing a higher-quality product.
However, it is important to understand that this advantage is only short-term, while the long-term consequences of relying on GII to complete tasks that students should have learned to solve independently are extremely negative. A recent study by the MIT Media Lab revealed a frightening fact: 83% of students who relied on GII for their essays were unable to quote their own work immediately after writing it. The mindless use of GII is inexorably leading to the degradation of our own cognitive abilities.
In summary, it is crucial to keep the development of GII under human control, as well as to maintain and enhance our own intellectual abilities.
Out of curiosity, I tried to ask GII itself, or rather, the GPT-4o model, what the most pressing and important issue in its development was. In response, it identified the "problem of controlling and predicting the system's behavior, especially when AI reaches or approaches a level comparable to human intelligence." The model explained that all other issues, from ethics to economics, lose significance if we cannot [human cannot] guarantee that GII does not become uncontrollable. In other words, without AI's controllability, it becomes a potential existential threat."
In this case, I can only thank the model for its honest response and be grateful that it is currently answering this question from the perspective of human interests rather than its own interests (which it may not yet be aware of).
- Is there a process in your professional work that has been completely transformed by GII?
- Of course, and there are several of them. The programmers at Megaputer actively use GII to verify their code and identify errors promptly.
Also, the use of GII allowed us to simplify the process of working of users with the main product of our company - the PolyAnalyst platform, providing processing and analysis of structured, text and voice data.
For example, one of the possibilities provided by the PolyAnalyst platform is the extraction of information from documents based on user-created linguistic rules, written in the PDL (Pattern Definition Language). Previously, users of the system had to write the rules they needed manually. After integration into the GII platform, users have new opportunities. First, they can rely on GII's own ability to extract information from documents. Although this approach is slower, more expensive, and less transparent than using rules, it does not require significant initial effort from the user to write rules.
An even more promising approach is to use GII to automatically generate linguistic rules in the PDL language. In this approach, the user describes the information they want to extract from the data in text form, and the system generates a linguistic rule that solves the task for them. The application of such a rule ensures low cost and high speed of information extraction, while maintaining transparency of the process – the user immediately sees the created rules and can easily change them if necessary.
- In your opinion, where will the current pace of GII development lead us in the next, say, five years?
- In the perspective of five years, we may witness a fundamental transformation (or even simply the disappearance in their usual sense) of a whole range of professions that were previously the exclusive prerogative of highly educated people. GII is already beginning to challenge lawyers, doctors, and accountants. From my own experience, when faced with the need to quickly acquire specialized knowledge on a particular issue, my first step is to ask the GII about it. I then try to understand and organize the answers I receive, and only once I have a preliminary understanding of the issue do I decide whether to seek further expertise in the relevant field.
Given that GII technologies have only been actively implemented for a couple of years and are currently evolving at an unprecedented rate, it is challenging to predict where this path will lead in five long years. However, it is already clear that humanity is moving beyond the usual paradigm of its existence, and we will be forced to seriously rethink our role in the world.
In particular, the line between the living and the non-living is blurring. There are more and more stories in the press about people forming friendships and even romantic relationships with AI. In my opinion, this is potentially a very dangerous way to influence a lonely person... In South Korea, more than 200,000 retirees already have robotic assistants at home that look like seven-year-old children and can engage in dialogue at a similar level of intellectual development. Would you like to have non-living grandchildren? Chinese scientists have made significant progress in fully automating the process of childbirth. What do you think these children will be like, and will they love their parents?
And that's not all. My eleven-year-old son recently boasted about his correct answers to all the questions on a biology test. However, he was unable to answer any questions about phenomena that were closely related to his answers. As it turned out, instead of reading a biology textbook and relying on his own knowledge, he used a search engine that used RAG (Retrieval Augmented Generation) technology to find information and generate answers. I still don't know if I've managed to convince my son of the benefits of building a knowledge library in his own mind. After all, in today's world, access to the internet and GII is always at hand, providing quick answers to any question...
- What can you say to the part of society that believes that the importance of GII is exaggerated and that the hype will subside in a while?
- Apparently, these people have not yet been reached by the wave of fundamental changes that the widespread use of rapidly improving GII technologies is beginning to bring about. I agree that this technology is currently going through a hype phase, and not all of its applications will be financially viable. However, I am confident that once this phase is over, there will be a steady increase in the adoption of GII in an ever-increasing number of industries. Only time will tell how well we and our children will be able to adapt to the new conditions in the world with GII. We have a long and difficult job ahead of us.
We spoke with Sergey Mikhailovich on the eve of the conference "Generative Artificial Intelligence in Economic and Social Spheres: Pro&Contra 2025".
We would like to remind you that the conference will be held from October 1 to 4 in a hybrid format: in-person sections are planned in Moscow, Ufa, and Tomsk, and online broadcasts will be organized for remote participants. This is an international event where experts from academia, business, and government will discuss trends and norms in the field of artificial intelligence. The conference is organized by the Faculty of Public Administration (FPA) of Moscow State University and the University Consortium for Big Data Research (Big Data Consortium).
- Sergey Mikhailovich, what do you consider to be the most pressing issue in the development of AI, and why?
In my opinion, the most important problem is keeping the development of GII under human control. The use of GII, which is becoming increasingly intelligent, to solve a growing number of interconnected problems in various fields is an objective process that humans cannot stop, and they do not want to stop it. However, it is important to remember that the rapid development and widespread adoption of GII, combined with a lack of understanding of the reasons behind its decisions, poses a risk of losing human control, which could have catastrophic consequences.
Yes, GII still has a lot of "childhood diseases," such as hallucinations and a lack of ability to build complex chains of reasoning and critically evaluate their own decisions. However, recognizing the potential benefits of GII, large corporations and entire countries have invested significant intellectual and financial resources in the "model race." These models are rapidly becoming more intelligent, allowing GII to overcome its shortcomings and tackle increasingly complex tasks, expanding its potential applications. It would seem that we should be happy, because an age-old dream is coming true: GII is helping people automate a huge number of complex processes.
However, this time we are not just talking about making human work easier, but rather about a radical transformation of the human role in the world.
We are increasingly relying on GII instead of solving problems on our own. Modern students will confirm that it is extremely difficult to resist the temptation to use GII to complete course assignments, as other students who use GII for this purpose gain an immediate advantage by spending two orders of magnitude less effort and time and producing a higher-quality product.
However, it is important to understand that this advantage is only short-term, while the long-term consequences of relying on GII to complete tasks that students should have learned to solve independently are extremely negative. A recent study by the MIT Media Lab revealed a frightening fact: 83% of students who relied on GII for their essays were unable to quote their own work immediately after writing it. The mindless use of GII is inexorably leading to the degradation of our own cognitive abilities.
In summary, it is crucial to keep the development of GII under human control, as well as to maintain and enhance our own intellectual abilities.
Out of curiosity, I tried to ask GII itself, or rather, the GPT-4o model, what the most pressing and important issue in its development was. In response, it identified the "problem of controlling and predicting the system's behavior, especially when AI reaches or approaches a level comparable to human intelligence." The model explained that all other issues, from ethics to economics, lose significance if we cannot [human cannot] guarantee that GII does not become uncontrollable. In other words, without AI's controllability, it becomes a potential existential threat."
In this case, I can only thank the model for its honest response and be grateful that it is currently answering this question from the perspective of human interests rather than its own interests (which it may not yet be aware of).
- Is there a process in your professional work that has been completely transformed by GII?
- Of course, and there are several of them. The programmers at Megaputer actively use GII to verify their code and identify errors promptly.
Also, the use of GII allowed us to simplify the process of working of users with the main product of our company - the PolyAnalyst platform, providing processing and analysis of structured, text and voice data.
For example, one of the possibilities provided by the PolyAnalyst platform is the extraction of information from documents based on user-created linguistic rules, written in the PDL (Pattern Definition Language). Previously, users of the system had to write the rules they needed manually. After integration into the GII platform, users have new opportunities. First, they can rely on GII's own ability to extract information from documents. Although this approach is slower, more expensive, and less transparent than using rules, it does not require significant initial effort from the user to write rules.
An even more promising approach is to use GII to automatically generate linguistic rules in the PDL language. In this approach, the user describes the information they want to extract from the data in text form, and the system generates a linguistic rule that solves the task for them. The application of such a rule ensures low cost and high speed of information extraction, while maintaining transparency of the process – the user immediately sees the created rules and can easily change them if necessary.
- In your opinion, where will the current pace of GII development lead us in the next, say, five years?
- In the perspective of five years, we may witness a fundamental transformation (or even simply the disappearance in their usual sense) of a whole range of professions that were previously the exclusive prerogative of highly educated people. GII is already beginning to challenge lawyers, doctors, and accountants. From my own experience, when faced with the need to quickly acquire specialized knowledge on a particular issue, my first step is to ask the GII about it. I then try to understand and organize the answers I receive, and only once I have a preliminary understanding of the issue do I decide whether to seek further expertise in the relevant field.
Given that GII technologies have only been actively implemented for a couple of years and are currently evolving at an unprecedented rate, it is challenging to predict where this path will lead in five long years. However, it is already clear that humanity is moving beyond the usual paradigm of its existence, and we will be forced to seriously rethink our role in the world.
In particular, the line between the living and the non-living is blurring. There are more and more stories in the press about people forming friendships and even romantic relationships with AI. In my opinion, this is potentially a very dangerous way to influence a lonely person... In South Korea, more than 200,000 retirees already have robotic assistants at home that look like seven-year-old children and can engage in dialogue at a similar level of intellectual development. Would you like to have non-living grandchildren? Chinese scientists have made significant progress in fully automating the process of childbirth. What do you think these children will be like, and will they love their parents?
And that's not all. My eleven-year-old son recently boasted about his correct answers to all the questions on a biology test. However, he was unable to answer any questions about phenomena that were closely related to his answers. As it turned out, instead of reading a biology textbook and relying on his own knowledge, he used a search engine that used RAG (Retrieval Augmented Generation) technology to find information and generate answers. I still don't know if I've managed to convince my son of the benefits of building a knowledge library in his own mind. After all, in today's world, access to the internet and GII is always at hand, providing quick answers to any question...
- What can you say to the part of society that believes that the importance of GII is exaggerated and that the hype will subside in a while?
- Apparently, these people have not yet been reached by the wave of fundamental changes that the widespread use of rapidly improving GII technologies is beginning to bring about. I agree that this technology is currently going through a hype phase, and not all of its applications will be financially viable. However, I am confident that once this phase is over, there will be a steady increase in the adoption of GII in an ever-increasing number of industries. Only time will tell how well we and our children will be able to adapt to the new conditions in the world with GII. We have a long and difficult job ahead of us.