AI will not replace humans, top scientist tells Armenpress, stresses need for responsible use
24 minute read
A top Armenian scientist based in Russia has dismissed concerns that artificial intelligence could completely replace humans.
Arutyun Avetisyan told Armenpress in an exclusive interview that AI will not replace humans because, in the true sense of the word, there is no intelligence in it.
He described artificial intelligence as a kind of exoskeleton, arguing that a skeleton cannot make decisions on behalf of people.
Avetisyan is a mathematician specializing in system programming and an academician of the Russian Academy of Sciences (RAS). He is Deputy President of the RAS, Director of the V. P. Ivannikov Institute for System Programming of the RAS, and heads the system programming departments at Lomonosov Moscow State University, the Moscow Institute of Physics and Technology, and the Higher School of Economics. He is also a foreign member of the National Academy of Sciences of Armenia.
Below is the full interview.
Armenpress – Do you think artificial intelligence is a challenge or a source of new opportunities?
Avetisyan – As always, a challenge is simultaneously a new opportunity. If we fail to meet that challenge, it will become destructive. But if we succeed, it will turn into an opportunity. Therefore, it is necessary to have a clear understanding of what kind of technology this is and how it is developing, and to build the future on this scientific foundation.
Otherwise, we will be like people building a house on quicksand. In that case, everything could collapse unexpectedly, and instead of comfort and prosperity, we will face widespread disruption. And God forbid if such a situation affects a nuclear power plant or other critical infrastructure.
– I assume that, at least for you, there are no longer any unresolved questions or areas of uncertainty regarding AI. To what extent is it artificial, and to what extent is it intelligence?
– In fact, everything is clear. There is a clear dividing line, and it is already evident that this is not “intelligence.” It has no agency, meaning the ability to act independently. This, too, is clear and has even been recognized at the international level. Beyond that, however, there is a vast “gray zone,” a fog.
Take education, for example. Why not use AI? It seems like a good idea. I am personally in favor of its application, too. But if it is done unwisely, a child may, for example, learn negative moral values along with mathematics. Society, however, is founded precisely on morality and moral values, the erosion of which can have terrible and irreversible consequences. There are examples of this in human history. Thus, moral values are important for every society.
The prominent British scientist Edward Tylor, in his studies of different human communities, concluded that there is no society on Earth—not even the smallest group living deep in the Amazon jungle—that does not have taboos, prohibitions, and similar concepts. In other words, there are always certain norms of behavior, which manifest themselves differently not only across societies but also in different spheres of human activity.
If we fail to incorporate these norms into artificial intelligence, this could simply lead to certain adverse situations. This concerns not only cybersecurity but also basic human interactions, because we may begin to sincerely believe that everything in life works in precisely that way, while the rest of society does not share this understanding.
– Cybersecurity is becoming particularly important in this regard. Has humanity become more vulnerable than ever, or is it nevertheless possible to effectively defend against virtual attacks, which can have dangerous manifestations with very real consequences in everyday life?
– Cybersecurity tools and their analysis are part of our work. In fact, the emergence of AI has not introduced anything fundamentally new. AI simply exposes vulnerabilities in the system. Accordingly, the issue here concerns the speed of response. We will need to change our approaches and use AI itself to provide protection and respond to threats.
However, this has also exposed a fundamental problem: most of our systems have been designed in such a way that they can be hacked. Effective protection of critical infrastructure is particularly urgent. And this is possible.
For example, there has not been a single technical failure in aviation control systems over the past 50 years; incidents have been attributable only to human error. There have been no software failures because strict certification and testing requirements are in place.
This experience should be applied across the board, and society's level of computer literacy should also be raised. In the age of AI, this issue is indeed particularly urgent, but the means of addressing it are also clear.
– Don't you think that the obsession with widespread digitalization could make society more isolated, gradually disconnecting it from reality and leading it to live in an imaginary, fictional world?
– Of course, digitalization can lead to that. There are entire philosophical schools of thought arguing that we are already virtual, since every person prefers to live in the reality they imagine and assess the events taking place around them from that perspective.
One can say all sorts of things about this, but the fact is that we are made of flesh and blood and use these technologies. And no matter how “virtual” we become, we still have real needs.
Today, millions of people live quite comfortably while preferring to rarely leave their homes. They have money, order food and other essential necessities online, and spend the whole day entertaining themselves. Moreover, they also make a living online. This, too, is a virtual world, a vivid manifestation of virtual life. Yet even in this case, we are dealing with a physical human being who has entirely material needs.
But that world and this way of life are built on real things: electricity and the operation of data centers. Any disruption to these immediately brings people back to non-virtual reality. After all, people also need friends in the real world. In other words, things are not that straightforward.
Decades ago, people were also saying that virtual reality would soon encompass every sphere of life: people would wear special glasses and, with the help of appropriate instructions, be able to perform the most complex tasks and highly skilled jobs. It is already the end of 2026, yet this technology does not seem to have become widespread. In other words, it is consumers who determine the extent to which technologies are used and adopted.
– Why have calls to restrict the use of AI become more vocal recently?
– It is certainly not because people fear that AI will become more intelligent and conquer planet Earth. These concerns are being voiced because people realize that its widespread deployment without proper risk management will make everything around us vulnerable. To avoid this, additional regulations are needed that will make it possible to organize processes differently.
This understanding will allow us not to fear artificial intelligence but to actually use it in our daily lives: to help educate children by addressing teacher shortages, for example. It could enable the best doctors to help a much larger number of patients, effectively multiplying their work efficiency a hundredfold. Can you imagine how extraordinary that would be? This was impossible in the past, and even today it sounds more like science fiction.
The concern that AI will replace humans because it will start engaging in creative tasks is unfounded. It will not, because, in reality, there is no intelligence there in the true sense of the word. AI is simply a kind of exoskeleton. And a skeleton cannot make decisions on behalf of people. Of course, it can be taught to harm a specific person, but nothing more. That is the crux of the matter. This also answers one of your earlier questions.
– Aren’t we idealizing AI?
– Especially now, as robotics and the energy sector begin to be enhanced and widely deployed on the basis of AI, there is a real danger that we may start trusting robots more than we sometimes even trust ourselves.
A psychological dependence is already emerging: people engage in conversations with them, for example, turning to ChatGPT with questions about their medical treatment. This is perfectly normal if the person doing so is reasonable and understands that this does not constitute comprehensive medical care. However, in the case of serious illnesses, this is clearly the wrong approach.
Of course, this does not mean that AI should not be used in healthcare. The point is that a doctor must always be involved in the process.
– So, AI is not a replacement but rather an auxiliary tool?
– It is an auxiliary tool, but it is already being introduced so actively that sometimes it may seem as though it is the primary tool. In reality, however, that is not the case; it is simply an illusion.
Of course, it can send emails on your behalf or perform other tasks. But if you entrust it with setting up and managing your business and even authorize it to make money transfers, you will most likely go bankrupt sooner or later.
– And what if it suddenly decides to launch nuclear missiles?
– I think—and, indeed, I am certain—that people will never entrust machines with the management of such matters. That is out of the question. The final decision will always be made by humans. Only completely irrational people could consider doing otherwise.
However, in wartime, when something is already flying towards you, the process of launching those missiles could be delegated to artificial intelligence in the interest of a faster response. Everything is clear here: war is war.
In the civilian sphere, however, safety requirements for the use of the same unmanned aerial vehicles are considerably stricter, especially in urban environments.
In general, I support the introduction of AI in various areas of life, but this must not be done recklessly because, as I have said, it would make our infrastructure vulnerable. This applies not only to artificial intelligence but also to the entire digitalization process. We cannot simply digitalize everything without first analyzing the possible consequences.
– Are there also serious risks in this area?
– Just one example: Sweden once wanted to transition entirely to cashless payments, effectively removing cash from circulation altogether. However, following the widespread power outages in Southern Europe last year, the Swedish government decided to abandon this idea so that people would still be able to make purchases in emergency situations.
Digitalization, nevertheless, means a transition to a different reality, so we must proceed cautiously, taking into account cybersecurity and the psychological factors that affect us.
– Could you tell us what you are currently working on and what role artificial intelligence plays in your work?
– We are system programmers. We work on the development of real-time systems, operating systems, and secure software, including the creation of reliable and secure AI tools. We are a center for trusted AI in Russia.
Our developments have been actively used in medicine for the past 10 years, with good results recorded in recent years. Relatively simple applications include the analysis of electrocardiogram data and the diagnosis of skin diseases, particularly melanoma.
There are also more complex projects. Together with Academician Evgeny Shlyakhto, we are developing a large model, a specialized tool designed to assist doctors during clinical consultations and discussions.
We are also developing solutions to improve patient adherence to treatment, aimed at helping patients follow medical instructions during treatment and improve their interaction with doctors. Our experience shows that this is a lengthy but correct, feasible, and effective approach.
In general, we are gradually expanding into applied fields, because it is impossible to create a proper application without adequate system software. We are promoting the advanced use of AI in science and, through science, its rapid introduction into industry. To this end, we are creating a joint center that will make it possible to coordinate efforts as effectively as possible and consolidate the country's entire pool of expertise.
There are also other fields of application, including physics and various branches of chemistry. Work in these areas is carried out jointly rather than through individual projects, because this can only be done collectively: gathering data, creating models, and then turning them into products that are deployed in specific industries. It is quite clear that adapting a ready-made product to the needs of a particular sector is easier than creating one from scratch.
The goal is to develop foundational models not only in Russia but also jointly on a global scale, and to host them in trusted AI repositories. This way, even the smallest countries will have access to these developments without fearing that they will lose their independence. This is also beneficial to Russia, which will ensure its own security through joint development. We have no intention of making anyone dependent on us.
– Is it possible to develop what could be called a national AI?
– Why? The question should be approached from a slightly different perspective. If we do what I mentioned in my previous answer and choose that development path, it will be possible to create secure, sovereign national AI models based on these solutions.
Why? To ensure that no one has secretly inserted something into the system at the last moment. This requires constant work. It should not be done separately from what I have described, but in a coordinated manner. This is the right approach because there is so much to be done. No one—not the Americans, not the Europeans, no one—has sufficient resources to tackle everything at once.
It is necessary to establish feedback mechanisms with all sectors and use that feedback to improve the systems. We have already seen this with large models. It turns out that open-source models are now fully competitive and are no worse than their alternatives. However, turning them into ready-to-use products requires specialized expertise and knowledge.
That is why it is necessary to have people in Armenia who understand this and can do it. There are such people in Russia. But the main work can be done jointly, using open models.
And this is not just about large models. We will soon see a huge number of specialized models. This could be one of the most interesting developments awaiting us over the next five years.
Imagine running a sophisticated model on your own laptop, tailored to your individual needs, whether medical, engineering-related, or otherwise. This is the direction we should develop together.
– Joint development implies cooperation. Do you work with Armenian partners?
– Of course. We cooperate with the Institute of Informatics. I am a graduate of Yerevan State University myself, and we are now discussing possible areas of cooperation with the rector of my alma mater. I also have a laboratory at the Russian-Armenian University, where a team is working.
We are proud to have developed a special technology for the Armenian language that is the best of its kind in the world, and it is already beginning to be used in Armenia.
Incidentally, this is also a challenge in the field of AI, because major corporations do not support the development of models for hundreds of low-resource languages, considering it economically unprofitable. With state support, this area could be turned into a national initiative, making it not a business based on selling licenses to one another, but a project driven by the scientific community.
Everything can be efficiently planned and calculated within this framework. Entrepreneurs can then turn these developments into products and deploy them. Alternatively, the state itself can introduce them, either in cooperation with businesses or independently. These are the kinds of models we should have, rather than turning everything into a market. These approaches can and should be combined.
This would also ensure the security of businesses themselves, which could build on scientific developments created within the country. To achieve this, however, we need to train specialists. People are what determine everything. But how can we train them if there is no real investment? Everything is interconnected. Development is impossible without the coordinated efforts of the state, science and business. This applies to every country.
We live in a world where there is no way to solve problems by relying on miracles. However, if you move in the right direction, you will develop along the way, alongside everyone else, and acquire new knowledge and skills.
– Do you think Armenia is moving in the right direction in this regard?
– Historically, Armenia has always been among the pioneers in science and innovation. During the Soviet period, the first computing devices were developed here. In the post-Soviet period, an entire outsourcing sector emerged, bringing specialists together in large specialized companies that perform certain functions or solve specific business problems for client companies. This helped preserve the country's human resources potential.
In other words, Armenia's technology sector has traditionally been at a fairly high level. Science has not disappeared either. All this paints a fairly optimistic picture of the future.
This does not mean, however, that we should take things lightly and think that development can be ensured simply through large-scale investments. That is not how it works. We need to develop approaches that are measured, well-calculated and, most importantly, correct.
We need to create the conditions not for the emergence of one large company, but for the development of multiple companies and an entire ecosystem. The state's role is precisely to foster the creation of such an ecosystem: an environment in which businesses have an incentive to invest in science.
I consider it an excellent and entirely correct approach that Armenia is allocating significant resources to grant programs, as well as to bringing master's and doctoral students back to their homeland.