AI allows us to solve problems in the style of Alexander the Great, says Artem

An Armenpress correspondent spoke to Artem Oganov, one of today’s prominent Armenian scientists based in Russia, during his latest visit to Armenia. A professor of the Russian Academy of Sciences, a foreign member of Armenia’s Academy of Sciences, a crystallographer, chemist and materials scientist, Oganov took part in discussions in Yerevan entitled “Artificial Intelligence in Science and Industry: Opportunities and Challenges” and “Artificial Intelligence in Biomedicine, Pharmaceuticals and Materials Science: From Genomic Data to New Medicines and Materials”.

Our conversation focused primarily on one of today’s most pressing subjects: the role and significance of artificial intelligence (AI), its influence on scientific development and its impact on human life.

We also discussed a welcome and significant event this summer: his receipt of China’s prestigious international state award in science and technology. Our compatriot was presented with the award by Chinese President Xi Jinping.

Before beginning the interview, I would like to congratulate you on receiving China’s state award. Which of your scientific achievements was it awarded for?

I will probably never know precisely myself, although, of course, I could find out. The point is that my Chinese colleagues nominated me, but I have not seen the nomination documents or the justification, so I do not know which of my works they considered.

I can only speculate. Most likely, they nominated me for the work I also regard as my most significant: predicting crystal structures, developing methods for designing new materials, and research into high-pressure chemistry. These are my most significant contributions, although there are other important studies too. In principle, I could ask my Chinese colleagues what they wrote. Judging by everything, they wrote something good.

Let us consider it an award for your overall scientific work. As this interview is taking place within a discussion about the role and significance of artificial intelligence in science and industry, how much does AI help scientific work, and how much does it hinder it?

Of course it helps. Artificial intelligence is an immensely powerful tool that allows us to solve problems previously considered entirely unsolvable. A scientist’s work today is different from what it was even a few years ago. AI has changed everything fundamentally. Before that, computers had already changed everything substantially.

Just think about how scientists put their work on paper 40 or 50 years ago: using a typewriter or writing by hand. There was far less automation in scientific research. Now we are reaching the other extreme.

The automation of experiments has advanced to the point where robotic systems are being created that conduct experiments around the clock, practically without human involvement. A scientist gives AI a task. It devises an experimental procedure and passes it to a robot, which carries it out, measures the resulting data and sends them to a computer. The AI model is then refined, and the work continues.

This is a closed loop that can increase an experimental laboratory’s productivity around a hundredfold. Imagine that work which might previously have taken 100 years can now be done in one year. There is your comparison.

What has AI changed in your field, materials science?

Previously, scientists believed that a material’s properties could be predicted by understanding the physics of the processes that determine them. Knowing the forces of interaction between atoms, they calculate derivatives and obtain information about the material’s stiffness, its elasticity and so on.

But if you do not understand those properties, everything becomes very complicated and tangled. For example, what is a material’s radiation stability or resistance? How can you know what it is? There are so many different processes involved—the breakdown of chemical bonds, ionisation, the formation of defects and so on—that they create an incredibly complex mixture. How can you predict such a property? You cannot. If you do not understand the properties, you cannot model the necessary parameters.

You have probably heard the story of Alexander the Great cutting the Gordian knot with a single stroke of his sword after many others had unsuccessfully tried to untie it. He solved the problem without delving into the intricacies of how the knot had been tied.

In this sense, AI allows us to solve problems in Alexander’s style. We examine a database of a material’s properties, and you could not care less about what causes them.

Take the radiation properties of steel, which one of my students modelled for Rosatom. He was able to identify new types of steel with high radiation resistance. These are for nuclear reactors. This is a genuinely important practical problem that you would never solve by following the traditional route of a theorist. But through machine learning, the problem has been solved.

Those results are now undergoing experimental testing. Other, equally important properties, such as corrosion resistance, have already been confirmed experimentally.

How much can we trust artificial intelligence when solving such serious problems?

Let me ask a question in return: should we not also double-check data obtained by a human? I think checking is very important in every case. Artificial intelligence provides a quick, inexpensive prediction. That does not mean you should trust it unconditionally. You must verify it.

AI can, of course, make mistakes. To avoid such situations, you must rely on your own knowledge and use AI’s capabilities properly.

Every self-respecting scientist—and I count myself among them—publishes research results only after checking and rechecking them repeatedly. That is necessary; indeed, it is compulsory.

Similarly, every self-respecting writer reads a text several times before publication, rewriting some passages and editing others. How do predictions differ from the solutions provided by artificial intelligence? In principle, they do not. So they also need to be checked.

Can we therefore say that a distinctive exchange of knowledge takes place between humans and AI?

Certainly. In the same way, when solving a problem, we draw on our own experience, our teachers’ experience and the knowledge we have acquired. Artificial intelligence similarly learns from other experience. In that respect, there is not even a fundamental difference.

Could AI ever completely replace experiments?

It will not. Experiments will always play a primary role, particularly in the physical sciences. AI models will continue to improve. Their likelihood of success and the accuracy of their predictions will increase, but will never reach 100%.

A device or programme can never completely replace a doctor, but it can allow you to monitor your health every day and consult a doctor much earlier than you might otherwise think of doing.

Do you think AI could develop to a level at which it becomes possible to devise a means of living forever?

Scientists have long been working to extend human lifespans. This is no longer an alchemical problem, as it was in the Middle Ages, but a mainstream scientific challenge.

There is a widespread scientific field, gerontology, which studies ageing in living organisms and humans and the possibilities for slowing it down. Of course, we will not live forever, but we can substantially extend our lifespans.

We are already seeing this happen: people now live longer than ever. Most scientists estimate that the maximum human lifespan is currently around 120 years. Some people live approximately that long. Can it be extended further? I do not know. But if the maximum is 120 years, that means average life expectancy could be brought close to that benchmark.

What is human health? Where can it fail? What is disease? These are important and very complicated questions. When discussing a person’s DNA profile, we must understand that we are talking about potential diseases, not guaranteed ones. A high probability of developing skin cancer, for example, does not mean it will necessarily happen.

Incidentally, DNA testing does not replace tests of blood, other biological tissues and so on. In this context, I think it is important to mention that I recently tried a new ATM that will be used in Russia. Customers simply place their fingers in the designated position and look at the screen to obtain information about their heart rate, blood pressure, cholesterol and glucose levels, and so on. The data have a reliability rate of 80%. Now anyone can check all of this completely free of charge without visiting a medical facility.

Why is this important? My working hypothesis is that lifespan does not depend very heavily on living standards. It depends on two factors: dietary habits and early disease prevention. The information obtained from that ATM can help people change their lifestyles and extend their lives to some degree.

Have you tried developing an elixir of longevity?

No. Everyone should do their own job. There are professional gerontologists whose task is precisely that. I have a different specialism.

Do you have a scientific dream that you are absolutely determined to achieve?

I want to understand this world better.

Are you succeeding?

Perhaps I have made about 1% progress in that direction.

Do you have scientific contacts with Armenian colleagues?

We have developed excellent cooperation. There are many colleagues with whom discussing scientific subjects is always interesting and enjoyable.

But there is also direct scientific cooperation with the laboratory led by my good friend Hayk Zakaryan, who works at Yerevan State University. That cooperation is developing very well: we have joint projects and publications.

You probably know that two data centres have already opened in Armenia. Could they attract leading scientists and help important research take place here?

They could, but this is quite a complex issue. Attracting talent to a country requires political will above all. It requires financial resources, scientific infrastructure and a suitable environment.

There must be a scientific system that does not burden scientists with bureaucratic hurdles and gives them freedom. That means freedom to spend most of their time pursuing their dreams and creating, rather than doing nothing.

There must be a stimulating environment based on meritocracy (rule by the most deserving, with leadership positions occupied by the most capable—Ed.), where people are judged by their talent and intellectual ability rather than their political views or closeness to influential people.

Then it will be possible to retain talent in the country. But one further condition is necessary: political stability. It is very difficult for people to make plans in a country with an uncertain outlook.

We can see this in Israel. Salaries are high, the economy is developed and scientific infrastructure exists, but I find it difficult to imagine a foreign scientist moving there given the current situation. Personally, I would not move to Israel because I do not see a future.

What about Armenia?

Honestly, the same applies.

Is there more certainty in Russia?

Yes, but there are no guarantees anywhere in the world now. The world is a hair’s breadth away from a major war. So what can guarantee the future? As you can see from the news, first and foremost, probably the reliability of air defence. That is what a guarantee of the future looks like today. In that respect, Russia’s air defences are better than anywhere else.

As for the rest, time will tell. Russian science also faces enormous challenges now. We are managing for the time being. What will happen later, I do not know.

Can we say that a distinctive community of Armenian scientists has emerged in Russia?

It emerged a long time ago. There is an enormous number of scientists of Armenian origin in Russia.

The greatest Armenian scientist living in Russia today is Yuri Tsolakovich Oganessian, a man who discovered 10 elements of the periodic table. In science, you cannot place anyone alongside him on that pedestal.

I could give a long list of outstanding Armenian scientists living in Russia. Of course, Armenian scientists are in contact with one another, and those contacts extend beyond any one country.

When were you last in Armenia?

About a year ago. I usually come once a year, sometimes several times. My visits sometimes include meetings with young scientists or students and lectures. This time, I gave two talks at different round-table discussions.

What change would you like to see in Armenian science by your next visit?

Recently, I watched an interesting video on YouTube that ranked the Soviet republics according to how useful or unhelpful they had been during the Soviet period. It discussed each republic’s strengths and weaknesses at the time. I will not say which were the most useful or the opposite, but will focus on what was said about Armenia.

It noted that, despite its small territory, population and limited natural resources, Armenia had been a republic with substantial industry and enormous scientific and engineering potential during the Soviet years.

In effect, Armenia occupied roughly the same position within the USSR as Greece did within the Roman Empire. Rome’s most educated and progressive emperors were sincere admirers of Greece and Greek culture. Greece still has well-developed science today, but many of its best scientists live in other countries, including Russia.

In this respect, too, there is a similarity between Armenia and Greece. Armenia probably has not lost its scientific potential either, but many Armenian scientists, like their Greek counterparts, are described with the prefix “diaspora-”.

I very much want people to say of present-day or future Armenia that this small country has such enormous intellectual potential that all the surrounding states aspire to match it.