Conference on Artificial Intelligence

Russia Vladimir Putin · President of Russia

24 November 2022 · Sber conference "Journey into the World of Artificial Intelligence"

Context
Putin addressed the annual Sber 'Journey into the World of Artificial Intelligence' conference in Russia on November 24, 2022, celebrating AI achievements and national strategy.
Main themes
Artificial intelligence development, Russia's technological sovereignty, national AI strategy, international AI competition, digitalization of public services, AI ethics and governance, public-private partnership, young researchers and scientists, Sber's role in AI coordination, economic and social development through AI
Duration
~113 min
99th percentile
Words
13,499
99th percentile
Sentences
674
99th percentile
Readability
43
43rd percentile
Grade
12.2
47th percentile
Syntax
2.95
41st percentile
Vocabulary
2.16
40th percentile

Translated from Russian

Content
Actors
Style

The broad sphere of national life a sentence belongs to. Every sentence has an Arena, so it is the denominator the other indices are read against.

Other
  • Technological 42%40%
  • Ceremonial 20%19%
  • Economic 19%18%
  • Institutional 10%9%
  • Social 9%8%
  • Environmental 1%1%
  • Geopolitical 0%0%
  • Not in focus 5%

674 sentences in 298 paragraphs

Dear friends!

I am very pleased to welcome the participants of the annual Sber conference "Journey into the World of Artificial Intelligence," scientists, engineers, entrepreneurs present in this hall today, and, of course, the entire audience – hundreds of thousands of people from different countries around the world who are following the events taking place here, as Herman Oskarovich said, in a remote format. And, you see, at certain moments, there are even more interested parties abroad than here with us. But there is nothing surprising about this. We are developing [technologies] – indeed, we have already advanced in certain areas, but in some, we still have to catch up – and the people who are constantly engaged in this subject want to be constantly informed, of course.

Such a wide representation speaks to the growing authority of your platform and the immense attention to Russia's achievements in the field of artificial intelligence. And as I said, we certainly have achievements in this area. By the way, in some areas of this field, as I have already mentioned, we are a step ahead of some other countries, and we must definitely increase our efforts. We will undoubtedly discuss specific solutions and joint steps on this matter today. But we will move in this direction, without any doubt.

I will probably state the obvious: the significance of breakthroughs in the field of artificial intelligence is colossal, and the competition between states is fierce. The results we achieve will determine Russia's place in the world, our sovereignty, the security and viability of our country, and our ability to solve tasks of economic, industrial, and social development at a qualitatively new level, creating broad conditions for the self-realization of citizens and for launching public initiatives.

I am glad that many people, including quite young ones, share these goals and make a significant contribution to their achievement.

Taking this opportunity, I would like to express my immense gratitude to domestic researchers who have recently begun their journey in science but are already shaping the technological potential of our country, including a number of young scientists I would like to highlight. Among them are Alexey Naumov and Sergey Samsonov from the Higher School of Economics, Vladislav Buzdin and Inna Minashina from MIPT, Denis Turdakov from the Institute of System Programming, Ruslan Lukin from Innopolis University, Anna Kalyuzhnaya and Nikolai Nikitin from ITMO University, Ivan Oseledets from Skoltech, Semyon Budyonny from the Institute of Artificial Intelligence, and Elena Sokolova from Sber's laboratory. I understand that some of them were just speaking when we were at a small exhibition.

I would also like to thank not only our creators and developers but also those colleagues who promote the technological agenda and set such an important vector for the future of the country. This is Sber, which has taken on the main responsibility for coordinating work on the development of artificial intelligence; this is, of course, – I cannot fail to mention my colleagues in the Government – the ministries of digital and economic development.

I would like to note that, according to international experts, Russia is one of the global leaders in the digitalization of public services, ranking among the top ten in the world. As I understand, 198 countries are being evaluated. The leaders are South Korea, Brazil, and Saudi Arabia. All of these are our good partners, and some of them we can even call our closest partners.

And of course, our business technology community shows an excellent example of self-organization. An alliance of companies in the field of artificial intelligence has been created. At its initiative, ethical principles for the development and implementation of this technology have been developed. More than 120 participants have already joined the corresponding code of ethics.

When we talked, Herman Oskarovich, about creating it?

G. Gref: Two years ago.

That's right, two years ago. And I am very pleased to note that this alliance is working – I will mention this several more times in my speech – is working well. We have made a very important step in creating the appropriate atmosphere. I consider you, esteemed colleagues, friends, to be reliable and effective partners of the state in matters of technological development. I am confident that this will continue to be the case in the future.

Thanks to our joint work, we are dynamically advancing in all areas of the National Strategy for the Development of Artificial Intelligence, supporting the development of education and science, introducing measures to support companies, and adjusting legal regulation, and this is yielding practical, significant results.

Herman Oskarovich, in the special demonstration zone – I just mentioned this – presented only some of the developments of our technological leaders in various industries. Just a few years ago, such visible achievements of human scientific and technical thought were unimaginable. The solutions created truly amaze the imagination without any exaggeration and are undoubtedly an objective basis for pride because they are made in Russia and for Russia, although, of course, they can also be used beyond our country's borders without any exaggeration, as they are absolutely competitive and, in some respects, outpace their competitors.

Domestic companies demonstrate a significant diversity of ideas and technical solutions in the field of artificial intelligence. What is also important: competition among domestic developers is developing and strengthening, and such healthy rivalry undoubtedly stimulates the accelerated growth of the industry, improves the quality of intellectual products, and their commercial success.

As for the leading companies that successfully implement artificial intelligence technologies, they demonstrate a confident increase in profits and labor productivity. And such, even if individual, examples increasingly influence the efficiency of various industries and the entire national economy.

At the same time, the potential for further growth is simply colossal, and it is essential to realize all these opportunities: to use and implement them.

Thus, according to experts, the prospects for increasing indicators in individual industries with the implementation of artificial intelligence technologies amount to an additional 20-30 percent over a decade.

Sber, as I know, was one of the first to actively implement artificial intelligence technologies and, so to speak, tested all this – not everything, but much of what could be tested – on itself. And the results are indeed very good, if not impressive. Last year, every sixth ruble of profit was earned thanks to artificial intelligence technologies, which allowed for a radical increase in the convenience and quality of Sber's services for citizens and corporate clients, for businesses.

I emphasize that we have passed an important, significant part of the planned path, creating the foundation necessary for the rapid development of artificial intelligence. Today, Russia has scientific achievements, unique researchers creating breakthrough technologies of a new era, and the first experience of their practical application has been accumulated.

The task of the new stage in the current decade is to ensure the mass implementation of artificial intelligence. It must cover all sectors of the economy, the social sphere, and the system of public administration. When Herman Oskarovich was still working in the Government, we always talked about this; it was Herman Oskarovich's forte when he spoke about the need to transition to a new stage of management in general.

What we are doing now, and what we will do, including with the support and developments of Sber, I believe should convince us that we are on the right path. We must be capable of forming a genuine technological, digital, and to a significant extent cultural, educational, and value-based sovereignty of Russia and our entire society.

I consider it necessary to fix such priorities in the updated National Strategy for the Development of Artificial Intelligence. In this regard, I ask the Administration of the President, together with the Government and the alliance of companies in the field of artificial intelligence, which I have already mentioned, to prepare the corresponding draft decree. We will review it together and adopt it.

Esteemed colleagues!

The successful implementation of artificial intelligence, as well as successful management of a region, industry, or company, in the modern world requires the use of a data-driven management model. This primarily means that the decision-making process is largely carried out not based on intuition, which is also, by the way, important. But in this regard, it is worth noting that intuition itself is formed from what? From experience and a large amount of data that accumulates in a person's mind. But in this case, even that is no longer enough, and decisions need to be made not only based on intuition but also on qualitatively processed structured data about the object of management, their deep analysis.

And here, artificial intelligence technologies play an increasingly active and significant role. At the same time, I would like to note: for the transition to such a modern model, readiness, maturity, and certain basic conditions are needed. What are they? Perhaps I will not say anything new for this audience, but nevertheless, I would like to highlight them.

The first is the use of advanced lean production approaches aimed at building optimal procedures and processes and reducing waste. It is no coincidence that they say: if you digitize chaos, you will only get digital chaos. Therefore, when engaging in digitalization and applying artificial intelligence technologies, it is essential to first bring order, as they say, to one's own household, in any own household.

The second element is a project management system. If it is absent, any attempt at change may end in failure. A clear task setting and deadlines for its completion are needed, focusing on a specific achievable result and providing it with the necessary resources.

I repeat, these are the basic conditions necessary for using a new data-driven management model, and it is extremely important to implement it both in business and, of course, in government bodies.

In this regard, I ask the Government and the State Council, within the framework of work on digital transformation, to prepare and then implement the transition of the entire system of public authority at the federal and regional levels to a data-driven management model using a platform approach.

It should be added that it is necessary to achieve the fulfillment of this task in all government bodies and across all key sectors. And I very much ask you, esteemed colleagues – I am addressing the participants of the alliance once again – to provide the Government of Russia, your colleagues with whom you are in constant contact, with the necessary advisory support.

I also propose that our Federal Competence Center in the field of labor productivity join this work. An extremely important area for joint work. We are currently only getting acquainted with the results of the work of some companies – I will return to this, colleagues, who are probably sitting here in the hall – on agricultural enterprises, the agro-industrial complex showed the results of such work.

Now, tasks are being solved here to increase the efficiency of enterprises and organizations, including through lean production mechanisms. I think it would be right if we expand its mandate so that the center [of competence in labor productivity] also focuses on the issues of implementing artificial intelligence and modern management systems in the sectors of the economy and the social sphere and, of course, in government bodies: both federal and regional.

I would also like to note what Herman Oskarovich has already mentioned regarding Moscow. I would like to point out that Moscow – and I will refer to its experience repeatedly today – has made a radical step forward in the digital transformation of many areas of urban management.

In turn, the Moscow region is on the verge of implementing a landmark project. It is about implementing a comprehensive change in the management system with the widespread cross-cutting implementation of artificial intelligence technologies. Perhaps the governor – I see he is in the hall – will also say a little about this later.

I ask both the Mayor of Moscow, Sergey Semyonovich Sobyanin, and the Governor of the Moscow Region, Andrey Yuryevich Vorobyov, to use the platform of the State Council to spread their experience of applying the latest technologies across the country – just as we are doing with you, esteemed colleagues, in some other areas.

The widespread use of advanced solutions is primarily aimed at improving the quality of life and creating new opportunities for people. Thus, the implementation of intelligent systems, industrial robots, and automated workplaces should reduce the volume of routine functions, minimizing human work in harmful and dangerous areas of production. Just below, we were talking about this, and I completely agree with this approach.

As we see, the consequence of the implementation of artificial intelligence and robotization is not an increase in unemployment, as we talked about two years ago and as experts have always said and continue to say: with the introduction of new technologies, there will be nowhere to put people who were engaged in more or less primitive labor. On the contrary, the emergence of new, more creative, substantive, interesting professions, and highly qualified jobs – this is what leads to smart and well-timed, prepared work in this direction.

And what is extremely important is that the income of the population is also growing. The growth of people's incomes is, in fact, the same result that technological revolutions have always led to in history. Companies and enterprises gain the opportunity to increase wages as they become more efficient, which, in turn, occurs thanks to effective management based on data using machine learning and other artificial intelligence technologies.

In order for domestic businesses to fully realize their potential, the state must undoubtedly create the necessary conditions. I will allow myself to briefly touch on each of the key areas separately.

First. The implementation of innovations should not impose significant costs on businesses. To this end, starting from January 1, 2023, when purchasing and implementing specifically domestic solutions, including in the field of artificial intelligence, businesses can take advantage of tax benefits and direct additional funds towards technological renewal. For example, when calculating corporate income tax, the amount that exceeds the actual expenses of the company for acquiring advanced Russian developments by one and a half times will not be taken into account.

I ask the Federal Tax Service to closely monitor how convenient and quick these and other developed and proposed tools will be.

Second, in order to establish data management and end-to-end digitalization of business processes, and, importantly, to ensure the operation of cloud infrastructure, companies must have software, hardware tools, and computing power.

I ask the Government to support the development of these areas, including defining special mechanisms within the framework of the federal project on artificial intelligence.

I would also add that we need to generally assist in the development of our own sovereign cloud technologies, that is, various platforms and services that allow companies, citizens, and government bodies to use resources via the internet: to work in applications, store information, and perform computations.

We need to establish mechanisms to work with and store data on domestic cloud platforms, and it must be made significantly more reliable and secure than keeping information and conducting document flow on one's own computer or on paper.

I emphasize that this concerns both technological and regulatory solutions. In this regard, I ask the Government to work closely with the professional community on these issues.

Third, access to anonymized data is crucial. At the exhibition, a colleague spoke about this in relation to medicine, discussing her work and what is planned to be done in the near future, emphasizing the need for access to this information. I completely agree with this. I ask Parliament to expedite the work on the relevant bill, which should be adopted as soon as possible. As far as I understand, it has already passed its first reading somewhere back in 2021 – it needs to be refined.

Of course, there are questions related to the necessity of protecting personal information, but, as I understood from the colleague who reported on the first floor at the exhibition, we are talking about anonymized data. And this will undoubtedly help improve the efficiency of medical care.

I fully understand the concerns that arise here from both the state and the business side. It is essential to take into account all the worries of citizens and entrepreneurs regarding access to large data sets, to create solid legislative guarantees for the protection of rights and freedoms, and to move forward. We must move forward, establishing all technological processes for the formation, anonymization, storage, and access to data.

I also ask the alliance to consider how best to establish a mechanism so that developers receive exactly the data they need in the volume necessary for creating machine learning solutions.

I also ask authorities at all levels to carefully study – we will return to this – the experience of Moscow. A special section has already been created on the capital's mayor's portal, where free access to data on various sectors of urban infrastructure can be obtained: for example, transport, utilities, and so on. Moreover, this information is indeed anonymized, does not relate to personal data, and does not allow for the identification of a specific person, intruding into their private life, which is undoubtedly very important.

Fourth, a large package of support for the information technology industry is already being implemented. Special grants can be awarded to startups for the development of new and existing projects in the field of artificial intelligence, for creating open, accessible libraries of best practices and technical solutions for developers.

Here, I would like to emphasize: this additional measure is working effectively; it is in demand by companies developing artificial intelligence technologies, which is why the grant amount has been increased by one-third – to 30 million rubles.

I believe it is necessary to continue strengthening support for technology companies and startups, and I propose that the alliance submit specific proposals in this regard.

I will not list all the benefits provided in this area; you are surely well aware of them. We have calculated and discussed them when implementing them, and they undoubtedly have a revolutionary character, without any exaggeration, in supporting the industry.

Fifth, significant progress has already been made in adjusting the legal framework for the implementation of the latest technologies. For safe testing of autonomous vehicles, in all environments: on land, in the air, at sea, experimental legal regimes have been created.

We must continue on the path of forming modern, globally competitive legislative foundations for industry regulation regarding the application of artificial intelligence and data handling.

We have repeatedly discussed this with colleagues from the business sector. We understand that these restrictions are significant for business development, and we will analyze the situation together with you and make decisions. It is important that they are made in a timely manner. I fully share your approach to these issues.

Of course, it is necessary to closely address the removal of so-called regulatory barriers that still remain. In this regard, I ask the Government, together with the alliance, developers, and companies in the real sector, to present proposals for changing specific regulatory requirements, technological standards, regulations, and other industry norms that hinder the implementation of artificial intelligence, and this must be done in the context of each industry. Everything needs to be specified as much as possible; then the work will be effective.

I am confident that today representatives of the business sector will discuss the necessary adjustments to the regulatory framework during your upcoming discussion. I will mention some solutions that are, so to speak, obvious, in areas where it is absolutely clear that artificial intelligence, in terms of quality, speed, and objectivity of decision-making, already surpasses and will continue to surpass humans in some cases.

For example, intelligent systems can continuously monitor the condition of drivers, including those of public transport, in real-time, literally every second. No medical worker can handle such a task.

Artificial intelligence algorithms can conduct virtual tests of devices, such as unmanned vehicles, before they appear on public roads, thereby reducing testing costs and risks to people.

Finally, we can move from scheduled repairs to repairs based on actual conditions, that is, from normative, formal repair timelines to risk-oriented, truly preventive ones, where algorithms will continuously assess the condition of parts and components and signal in advance about possible failures or breakdowns.

We also need to integrate new technological tools into standard procedures. In Moscow, for instance, the use of artificial intelligence technologies to support medical decision-making has been included in clinical recommendations and mandatory health insurance tariffs. As a result, the quality of a number of medical services has improved, as has the accuracy of diagnoses and research. This means that not only are the health and lives of people preserved, but a significant economic effect is also generated, as the need for complex, costly surgeries decreases, and individuals spend less time on sick leave. And the quality improves – that is the most important thing.

We need to implement such approaches more quickly at the federal level. And of course, I ask the Ministry of Health to pay special attention to digitalization and the use of modern technological solutions. I know that this work is being carried out; the Ministry of Health is utilizing and intends to utilize all of this, and there are corresponding plans.

This work is important for all levels of healthcare, but of course – I would especially like to emphasize – it is extremely important at the primary level, and here we need to encourage colleagues in the regions and municipalities to work together.

Next, I propose to establish requirements for increasing efficiency and mandatory use of modern technologies, including artificial intelligence, when providing budget subsidies to companies on a systematic basis. This is about not only stimulating entrepreneurs towards progress but also achieving growth in key indicators, including labor productivity, through innovations, so that we respond to this feedback during our joint work with businesses.

For example, regarding agricultural producers, we also discussed this on the floor below; the logic should be as follows: received a subsidy – increased production efficiency, which, according to what I have been told, results in increased agricultural output. Achieving such results, as we saw today, will be aided by artificial intelligence technologies and is already helping to do so.

I emphasize: what I am talking about now is certainly not from the realm of fantasy. Everyone here is prepared and knows what I am talking about. Time and technological development are accelerating. Every industry, enterprise, and organization must quickly adjust their views and approaches, work through and discard the templates that hinder progress, and undoubtedly learn something new every day, every hour.

In this regard, I believe that every national project and state program should contain specific measures aimed at implementing artificial intelligence technologies in various sectors.

For our part, our largest companies should also designate the implementation of artificial intelligence technologies as one of the key priorities in their investment plans – I want to emphasize this and hope that company leaders will hear me. They have the resources; they are substantial, if not to say significant in modern conditions, and they must be used to address the tasks we have gathered here to discuss.

I would like to once again appeal to the alliance: I ask you to involve representatives from all key sectors of the economy and social sphere in your ranks and together identify the most in-demand, most relevant technological directions and solutions in the field of artificial intelligence for use in specific industries.

Based on the results obtained, the Government will need to update the strategies for the digital transformation of industry, construction, transport, agriculture, healthcare, and other areas. This is a very large, extensive, but absolutely necessary collaborative effort.

I also ask the alliance to provide comprehensive methodological and consultative support to enterprises and organizations, including creating centers of industrial competence for the practical application of Internet of Things technologies, machine learning, data processing in each industry, as well as annually compiling a directory of the best solutions in this field.

And one more thing: I propose that the professional community introduce a user-friendly system for measuring and evaluating the products offered by artificial intelligence developers, so that entrepreneurs and civil servants understand which technical characteristics and parameters to focus on when choosing specific technologies or software.

Dear colleagues!

I would like to address a fundamental, defining issue – that of personnel. Artificial intelligence works well in the interests and for the benefit of people when it is created and applied by knowledgeable, well-prepared individuals.

Today, companies are engaged in a real hunt for specialists in the field of artificial intelligence. I do not want to provide specific examples here, but in the market, as they say, these stories are well known: a person leaves one company, moves to another – and immediately the level drops, or at least current problems arise. It is necessary to find equivalent workers, which is difficult; the shortage of professionals in this field is very significant.

We must undoubtedly respond to this challenge. This is a collective effort of the state, business, and the education system. Significant steps have already been taken to involve both students and schoolchildren in solving practical tasks. Undoubtedly, we need to go further – to implement elements of studying artificial intelligence in school mathematics and computer science programs, to increase the level of teaching these subjects, and to enhance the qualifications of teachers.

We are conducting various events that engage tens of thousands of students and, very pleasingly, millions of schoolchildren in this collaborative work, which is very good. We must certainly continue this work.

I would also note that this academic year, more than three thousand students have enrolled in advanced master's programs for specialists in the field of artificial intelligence.

As for future doctors, teachers, agronomists, lawyers, industrial workers, and those in the communications and transport sectors, starting from this academic year, they can choose a special educational module on artificial intelligence at their discretion. Of course, this involves teaching only the basic principles of machine learning and data processing. However, given the rapid pace of technological development, we need to take another step, namely: to restructure educational programs to ensure a high level of competence in the field of artificial intelligence among specialists in key sectors of the economy and social sphere. This applies both to training students and to the professional development of already employed specialists. And I ask the Government to formulate specific proposals in this regard.

Of course, it is necessary not only to implement new educational mechanisms but also to objectively assess their effectiveness, undoubtedly. In this regard, I ask the alliance companies to work together with the Government to develop a ranking of universities based on the quality of training specialists with competencies in the field of artificial intelligence. We need good examples, and it is essential that everyone sees the results of each institution's work. The most important criterion here is the starting salary of graduates upon employment. Because the willingness of businesses to pay a high salary to a novice specialist is the most accurate assessment of a person's qualities and competencies and the work of a particular higher educational institution.

Taking this opportunity, I would like to thank the companies that are creating basic departments for artificial intelligence in universities, inviting students for internships and practical training, supporting talented young people with various scholarships, and later hiring them for significant positions.

At the same time, I believe that both business and the state must continue to create comfortable and attractive working conditions in our country for specialists in the field of artificial intelligence. Once again, I appeal to the alliance: I ask you to present your proposals in this regard. There is a set of measures: preferential mortgages, deferments from military service, simplification of the employment procedure, obtaining residence permits for foreigners. There is a whole range – it needs to be developed and improved.

Today, I have already mentioned young scientists who are developing fundamental solutions, including in the research centers we have specifically created in the field of artificial intelligence. Until the middle of the decade, 7.3 billion rubles will be allocated from various sources for their funding.

But what I would like to emphasize here is that to achieve truly significant, breakthrough results, not only the state but also private business must consider the development of science as one of its top priorities, attract scientists, open their own research divisions and laboratories, and not wait for the state to sprinkle something from the sky. We need to work in a way that our competitors do, to build this system, to finance scientific work, including in the field of artificial intelligence.

The fundamental groundwork and applied solutions being created should open new areas of activity, expand the horizons of business development, and provide companies with additional tools to increase competitiveness and efficiency.

At the same time, of course, we need to study advanced foreign experience, adopt the best and most promising developments. But I would like to emphasize: we should not blindly copy – we have talked about this many times – we should not blindly replicate what is done abroad, but rather propose our original approaches and technologies that ensure the solution of tasks at a qualitatively new level. Only in this way can we build a truly new high-tech economy – a techno-economy.

For example, research is currently being conducted in the field of neural network architecture capable of performing multiple tasks in different languages – I was also told about this today – and in different modalities. That is, the machine does not just reproduce human speech but also tries to convey emotions, developing its own cognitive skills through a more advanced computational architecture. In other words, the machine constantly improves itself; I believe it even writes programs for itself.

Here, of course, we need to understand what is happening and control the entire process. This is indeed a new technological reality, a practical guideline for the development of technologies, products, and entire industries based on machine learning.

As for domestic software, we need to create and implement such developments that will allow us to be leaders in the coming years. This includes systems of automatic design that use machine learning algorithms, where it is enough to upload the initial parameters, and artificial intelligence will propose its solutions, which are more effective than those suggested by humans.

I will give another example, which concerns public and financial institutions, including at the global level – and today it is clear that such institutions must reflect the new reality of a multipolar world, be built on open, democratic principles, and exclude dictatorship, abuse, and monopolism.

We all know well that under today's illegitimate restrictions, one of the lines of attack is payments. Our financial institutions know this better than anyone else because they face these practices. The existing international payment system is expensive, and its correspondent accounts and regulation are controlled by a narrow club of states and financial groups. Essentially, they are the masters of life; they monopolistically control everything.

We need to combine advanced technologies with millennia of human experience. For example, specialists are familiar with the hawala system, which for many hundreds of years – I want to emphasize this – long before the emergence of the Western banking system, performed and, by the way, still performs the functions of international payments.

What is its essence? The sender gives money to a broker in their country. The recipient receives it in another country, also through the corresponding broker. The brokers, in turn, settle accounts among themselves, as we would now say, through a clearing system.

Based on digital currency technologies and distributed ledgers, it is possible to create a new system of international payments that is much more convenient, but at the same time completely safe for participants and, importantly, independent of banks and the interference of third countries. I am confident that something like this will undoubtedly develop because the dictatorship of monopolists is disliked by everyone and causes real harm – by the way, to the monopolists themselves as well.

Furthermore, we must pay special attention to technologies that help unlock the potential of artificial intelligence – these are quantum and photonic computing, quantum communications, where, I note, Russia has quite serious developments. I will also mention the Internet of Things and other cross-cutting and new industrial technologies. The development of these areas has been taken on by state-owned companies, which have signed the relevant agreements with the Government of the Russian Federation.

In the summer, we thoroughly examined the existing problems in specific areas at the Council for Strategic Development and National Projects. I am confident that the right conclusions were drawn, and the implementation of projects will quickly return to the required schedule, despite certain shifts.

I also believe it is necessary to pay attention to another area where Russia, primarily in terms of implementation, is still significantly lagging behind some countries – this is industrial robotics. I will return to this again – I just mentioned it in passing, but we talked about it.

In this regard, I ask the Government to develop and approve a new federal project for the development of domestic robotics, to determine the legal, tax, and regulatory regime, the necessary measures of state support, as well as the financing mechanism for developments and their subsequent implementation.

The effectiveness is obvious. Just now, a colleague spoke about this in relation to KamAZ. But this concerns not only KamAZ – it concerns the entire economy.

I also believe it is necessary to continuously monitor how work is progressing on cross-cutting technologies as a whole, to strengthen the connection between science and business in this area, and in dialogue with researchers, engineers, and technology companies, to develop the solutions necessary for these companies, removing, as I mentioned at the beginning, the barriers that hinder progress, that is, to act as we do or propose to do at our conferences, including today.

In this regard, I ask our largest companies, Rosatom and Russian Railways, together with leading research universities, with the support of Sberbank, to organize and hold a conference on new technologies for computing and data transmission next year, and the Ministry of Industry, together with industrial partners, to hold a conference on new industrial technologies. I will also try to find time to participate in this.

Of course, it is essential to invite colleagues from other countries to these events – those who wish to attend. We continue to work with many of them; they belong to our associations: both the SCO and BRICS. And those who are not members but are, as I mentioned at the beginning, leaders in certain areas will be happy to work with us.

I am deeply convinced that new solutions that change the world can only be created within the framework of close cooperation, based on trust and high ethical standards. This applies to all technological directions, but especially to artificial intelligence. There are many examples where its complex algorithms develop on the principles of open code thanks to the work of many thousands of programmers who create open libraries of solutions and developments.

Therefore, in the field of implementing artificial intelligence, we will certainly strengthen our interaction with our interested partners, as I have already mentioned – with everyone, but also including BRICS and SCO, conducting joint research and educational programs to maximize the potential of advanced technologies for the benefit of our countries. Some relevant cooperation documents have already been signed. We will continue to work in this direction.

Dear friends!

The results we achieve in the practical implementation of artificial intelligence technologies depend on our joint efforts, on the willingness of businesses to enhance and improve their technological positions, and on the desire of government bodies, on the other hand, to increase the speed and pace of change, to open the way for the implementation of advanced technologies.

Starting from next year, 2023, we will monitor the results of the application of artificial intelligence across sectors of the economy and social sphere. To this end, I propose creating a special tool – an index of the maturity of industries and regions, but to make it sound more appealing, we can call it the "intellectual maturity index," because it concerns the implementation of artificial intelligence, and based on it, we will assess the practical results of the work of each subject of the Federation, ministry, and agency in implementing artificial intelligence.

It is essential to particularly encourage those who are moving forward faster, who are creating and bringing the future closer. In this regard, I propose establishing an annual award for young scientists and engineers for scientific and design breakthroughs in the field of artificial intelligence. And another award – for achievements in the field of implementing artificial intelligence – should also be presented annually to domestic companies, enterprises, regions, and cities. The award ceremony could be held for the first time next year, in 2023, at the next conference "Journey into the World of Artificial Intelligence."

Thank you for your attention, for your patience, and I sincerely wish you success.

Herman Gref: Vladimir Vladimirovich, thank you very much. We have indeed selected three brief reports, three speakers, young scientists, who will present their proposals here. They are leaders today in the relevant applied fields of artificial intelligence development and implementation.

And the first topic – probably a topic that stimulates all researchers today. It is related not only to increasing the efficiency of the industry but, most importantly, to saving lives, and you mentioned this: it is health, it is our healthcare. Therefore, we want to start with this topic.

We also consider this topic a priority; we have a special company called SberMed AI, and with our partners, primarily, of course, with Moscow, we are implementing new modern solutions. Then these solutions are being implemented with our regional partners: 63 regions today are our partners who have implemented a large stack of artificial intelligence solutions and continue to develop this technology further.

Allow me to introduce the first scientist – Elena Vladimirovna Sokolova, who will talk about what she is doing in the field of healthcare.

(A video demonstration is taking place.)

E. Sokolova: Good afternoon, colleagues!

I represent the artificial intelligence laboratory of Sberbank, and today I want to talk about how we are already addressing pressing healthcare issues using artificial intelligence.

As of today, the two leading causes of mortality—both globally and in Russia—remain cardiovascular and oncological diseases. In 2019, cardiovascular diseases accounted for nearly half of all deaths in Russia, and due to late detection of cancer, 120,000 patients die within the first year after diagnosis. This is precisely why our solutions are aimed at the early detection of these diseases.

The main areas of our work focus on the analysis of medical images and the analysis of medical texts. I will now share an example of one such project that we implemented in the artificial intelligence laboratory—an analysis of chest CT scans, which was integrated into practical healthcare with the help of one of the companies in our ecosystem, SberMed AI.

During the pandemic, as you all know, an enormous number of chest CT scans were performed, which greatly overloaded the healthcare system. While radiologists' attention was directed towards assessing viral pneumonia, we trained artificial intelligence on this complex data to find signs of lung cancer.

In one of the experiments conducted in a real clinic using real clinical data, one and a half thousand chest CT scans were analyzed, and the artificial intelligence identified 12 patients with a high probability of lung cancer; the diagnosis was confirmed in eight of them, and treatment for these patients has already begun.

However, compared to medical images, the largest amount of patient data is accumulated in the form of texts, which are poorly structured, unstructured, and difficult to analyze. This includes complaints, medical history, laboratory test results, instrumental data, and so on. In fact, the greatest potential for the healthcare system lies precisely in the analysis of such data—textual data.

Currently, there are two main trends in medicine. The first is population analysis to identify individuals at risk of developing complex chronic diseases for timely prevention. The second is personalized analysis of vast amounts of data about a specific patient to develop an individualized approach to providing medical care.

Artificial intelligence is capable of processing absolutely any volume of information in a short time, while a physician cannot physically study the entire medical history of each patient. This is why I believe that artificial intelligence services should become regular assistants to every doctor.

One such example of a diagnostic assistant is a model that we are developing in collaboration with the Government of Moscow. This model has been trained on an unprecedentedly large volume of data—about 30 million patient visits in Moscow over two years. This model analyzes all medical information about a patient and offers the physician a final diagnosis as a second opinion.

This is an absolutely unique complex of deep diagnostics and patient condition monitoring, which has no analogs either in Russia or in the world.

What is important to note? We trained this model with Moscow doctors, who are specialists of the highest qualification. Accordingly, by implementing it in the regions, we can ensure an accessible high level of diagnostics across the country.

These are just a few examples of what we have managed to achieve. In fact, there are many examples, and time will not suffice to cover them all.

But what I would like to conclude with is what has already been mentioned before my speech—namely, the barriers that need to be overcome in the industry for the large-scale and accelerated implementation of artificial intelligence models in practice.

The first such barrier is legal restrictions. Therefore, we ask to simplify legal regulation regarding access to anonymized medical data for training and testing models, because in Data Science, in building artificial intelligence models, of course, without data, nothing will ultimately come to fruition.

The second limitation—or perhaps not a limitation, as significant steps have already been made in this direction—is the implementation and assessment of the effects of artificial intelligence models in real practice. Here again, a unique and very successful project is the Moscow experiment. Within its framework, about seven million diagnostic imaging scans have already been analyzed, which accounts for approximately 40 percent of the total volume of diagnostic imaging studies in Moscow.

The creation of such platforms should guarantee a uniform quality of such services and their safety. This, in turn, will help us overcome some skepticism from the medical community—they must understand the quality and safety of our solutions. This can only be achieved through their implementation and clinical-economic assessment of the effects post-implementation.

Here, I would like to say that such platforms need to be scaled, in our opinion: taken beyond Moscow, scaled to other regions. We believe that we can start with the Moscow region, as it is a well-developed region in terms of both technological maturity and regulatory maturity.

And thirdly—after the successful implementation and assessment of such services in practice, they need to be incorporated into clinical guidelines and standards for medical care.

In concluding my speech, I want to say that I sincerely believe that strength lies in synergy in any industry. In our industry, in medicine, the synergy of artificial intelligence and the medical community will allow us to achieve our common goal, in my opinion—extending the average life expectancy to 78 years, or better yet, to 80 within the coming decade. I believe that the implementation of artificial intelligence services will help us achieve this, and even faster than we think.

That is all. Thank you very much.

Elena Vladimirovna, I said that I would like to hear you out, note some things for myself, and then analyze them. But I cannot help but respond to what you have said.

Healthcare is the most important area of activity for any state. However, the duration and quality of human life is an aggregated, generalized indicator of the state's performance across all areas: both in the economy and in the social sphere.

Undoubtedly, entering the "80+ club" remains our task. We will certainly move towards this, and we will do so, among other things, with the help of artificial intelligence tools. There are already positive examples, and they are quite obvious. We will do all of this. We will eliminate the obstacles to using these tools—I mentioned this in my speech—eliminate the barriers.

We need to analyze all of this. I would be grateful if you could formalize all of this appropriately, and the authorities, together with you, will make the corresponding decisions.

All three components that you just mentioned, including feedback and the expansion of this practice, we will certainly implement.

I already mentioned the Moscow region, and you just mentioned it as well. I would like to ask the governor to say a few words on this matter.

A. Vorobyov: Dear Vladimir Vladimirovich!

We are also trying to collaborate with artificial intelligence, implementing it in various fields. We have all the necessary digital data for this. You visited our regional control center. The amount of data in each area is sufficient to optimize and seek effective solutions. In particular, in medicine, we do not yet read scans, but we will do that next year. However, we have a robot named Svetlana, which receives 35,000 calls every day and helps people get to a clinic or call a doctor to their home.

The key task is to further develop artificial intelligence so that it becomes a function. You once supported a decision on monitoring waste disposal—construction and household waste. How was it a couple of years ago? An inspector from environmental oversight had to stop a vehicle, write a report by hand, and impose a fine if the waste was unauthorized. After the legislative decision, a video camera and artificial intelligence now identify violations because an electronic ticket is linked to each vehicle, but the inspector still writes the report. Next year, artificial intelligence will fully become a function and replace the inspector. Thus, we will see significant progress in this area.

It is clear that transformation is needed, and for that, we need partners. We are very pleased that we have the opportunity to work with the Sber team in all these areas, with Moscow. I agree with what the speaker said about synergy—that there is a significant effect in partnership. I very much hope that we will show noticeable results next year.

In conclusion: Vladimir Vladimirovich, we must be prepared for the fact that we will seek support in changing legislative and regulatory acts, as they often pose obstacles to the further implementation of artificial intelligence and everything new.

That is generally what I wanted to add.

Thank you.

Regarding the simplified access to medical data, I also agree—with due consideration, of course. But when it comes to anonymized data, even more so. So let’s push for the adoption of these decisions.

Thank you, and good luck.

E. Sokolova: Vladimir Vladimirovich, thank you.

Please, Herman Oskarovich.

G. Gref: Thank you, Elena. Thank you, Vladimir Vladimirovich.

I would like to ask Evgeny Vladimirovich Shlyakhto, the director of the Almazov National Medical Research Center, who is one of the leading institutions not only in medicine but also in innovation, implementing both digital technologies and artificial intelligence.

E. Shlyakhto: Dear Vladimir Vladimirovich! Esteemed colleagues!

This is a wonderful project that fully corresponds to the trends in healthcare. I would even say more: it probably belongs to the area where artificial intelligence is already working better in some directions than specialists in medicine. And the question arises: will artificial intelligence replace the specialist or not? Of course, it will not replace them, but what will happen in the future is that a specialist who masters artificial intelligence will treat patients better and ultimately replace the specialist who does not master artificial intelligence.

It should be noted that the ideas presented by Elena Vladimirovna reflect the progress and depth of artificial intelligence penetration into healthcare.

If you allow me, Vladimir Vladimirovich, I would like to focus on three aspects that arise when hearing about this very interesting and innovative project.

The first direction is that we must think about how to stimulate the development of medical technologies based on artificial intelligence and how to implement them. A lot is being done in the country today: a vertically integrated medical information system is being created, and a unified healthcare framework is being established. The vast majority of regions are already participating in these projects, providing structured data in a productive environment. However, if we look at the real situation, only 20 medical technologies using artificial intelligence are currently registered by Roszdravnadzor.

Here, of course, there are questions related to regulation, as Elena Vladimirovna mentioned, questions related to promotion and business participation. We would very much ask, Vladimir Vladimirovich, if possible: not only in clinical guidelines but also in the system of compulsory medical insurance, the technologies created based on artificial intelligence should be classified as resource-intensive and evaluated differently, thereby stimulating companies to create such technologies.

It would probably also be appropriate to include in the federal project on artificial intelligence provisions or activities related to stimulating the creation of datasets that can be made available for public use. Many companies, without spending additional money on this rather labor-intensive work, could create artificial intelligence products based on this.

The second issue I would like to touch upon is related to the solutions that will truly change the organization of medical care itself.

Today, we are moving towards personalized medicine largely thanks to artificial intelligence and digital solutions, and this is certainly true because the development of genetics and personalized medicine is impossible without bioinformatics and artificial intelligence. In this regard, we certainly need integrative solutions that can unite all stages of medical care, make it seamless, and I would say, focused on the end result and patient-oriented.

Here, we need to think about what these solutions should be. Undoubtedly, they should be platform solutions, as you already mentioned in your speech. We very much support this direction—specifically platform solutions, specifically the transition from reactive to proactive management based on big data using not only the digital solutions available in each institution—most of our institutions can today be called digital clinics—but a transition to analytics, to decision support systems, and I would say, not so much predictive as prescriptive management—that is what characterizes modern trends and defines the state of the clinic that can be called value-based management.

Here, of course, it is very important to use all the data related to patients in the current activities of institutions, especially for patients at high risk of adverse events. Here, artificial intelligence can indeed help medical staff identify these patients, predict the development of adverse events, and help avoid them.

It should be noted that today we are doing a lot in this direction together with the Ministry of Health and Sberbank. This includes issues related to speech technologies, which were discussed today. These are issues related to the creation of a "smart clinic." The technological solutions we plan to obtain in the near future can indeed be scaled to the level of the entire country.

The third section, which you also mentioned, is education. Of course, we need programs for students, we need programs for doctors. But today, within the framework of the specialty project that you have allowed us to undertake, for training doctors in a scientific institution under specialist programs, we are already starting from the fourth year to form individual educational trajectories for students, based on the use of artificial intelligence to analyze each student's preferences, personal experiences, practical skills, and scientific achievements that they already have in the senior years. I think this is also very important.

Of course, there are many obstacles and difficulties in implementing artificial intelligence in healthcare: both organizational and technical. Today we see strong artificial intelligence that will help us work with "raw" data. These are ethical issues, as you mentioned, and the protection of personal data.

I believe that through joint efforts we must move in this direction, and the funds that will be invested will certainly pay off many times over in the form of reduced morbidity, mortality, and the preservation of the lives of citizens of the Russian Federation, and of course, in fulfilling all the tasks of the federal project that you have set.

Thank you for your attention.

Evgeny Vladimirovich, I know that many modern technologies are actively being implemented in your area. I heard and noted your comments regarding the creation of a system to stimulate the development of these technologies, including artificial intelligence, in clinical recommendations and in the compulsory medical insurance system. We will do all of this. I also noted your comments regarding integrated solutions and the use of platform solutions.

I don't know if Mikhail Albertovich is here or not? We will talk with him in the near future. He is a modern person, knows all about this and tries to stay on trend, and the Ministry of Health is doing many useful things. But that is why we have gathered today; German Oskarovich has brought everyone together to see what needs to be done in the near future in various areas, including in the field of artificial intelligence.

We will work on it.

Thank you very much.

G. Gref: Thank you.

Vladimir Vladimirovich, I would also like to give the opportunity to speak to a very bright young scientist, who is already a laureate of the President's award among young scientists – Ivan Oseledets. He is a researcher, a mathematician from Skoltech.

Ivan, if possible, a few words about what you are doing in terms of creating a foundation for the development of artificial intelligence and optimizing the model.

I. Oseledets: First of all, Vladimir Vladimirovich, thank you for the personal praise. It is very pleasant. I will try to briefly explain what we are doing so that it is clear what the praise is for.

In fact, the development of artificial intelligence is currently moving from a large number of small private models as weak artificial intelligence to large models that are trained on tens and hundreds of billions of examples, which have hundreds of billions of parameters, but are capable of solving a large number of different cognitive tasks.

Supercomputers are used to train such models. For example, the supercomputer "Christofari." The energy consumption of the supercomputer "Christofari" corresponds to the energy consumption of a city with about 10,000 residents. But even on such a supercomputer, training the Kandinsky model took four months.

It can also be said that the total power of all Sber's data centers is approximately equivalent to the energy consumption of the city of Vidnoye, a large city. Imagine an entire city of computers consuming electricity.

So, four months is a long time. Why does this happen? We estimate the current efficiency of open-source software for training models to be about 30 percent. We want to double this figure. What are we doing in this direction? A year ago, within the framework of the federal project to support artificial intelligence centers, supported by the Ministry of Economic Development and Sber, we started the "Green Artificial Intelligence" project, one of the goals of which is to increase the energy efficiency of model training. And already in a short time, we achieved the first result. We reduced the memory and time for training the model by about 15 percent. This translates into tens and hundreds of billions of rubles.

How did we do this? When training the model, hundreds of billions of parameters need to be updated. One number that we store takes about 32 bits – this is, for understanding, just seven decimal places. Well, it seems, why such precision? You can store two decimal places, and everything will be fine. But it turns out that if you only store two decimal places, the quality of the final model will seriously decline.

What did we do? Using mathematics, using fundamental mathematical approaches, we found those places in the complex training process where we can replace 32 bits with three, reducing memory in this area by eight times, and the total time by 15 percent. It is clear, as I said, that our goal is not just 15 percent, but to double the efficiency, and therefore we started developing our own domestic software, based on fundamentally different principles.

I will give a probably understandable example. Suppose we have a large family with a lot of shoes: big boots, small children's shoes – they need to be sorted into boxes. There will still be empty space in the boxes, and this empty space corresponds to the inefficiency of using the computing system.

How do we solve this problem? Very simply: we take computing blocks and cut them into pieces. We cannot cut shoes into pieces, but we can cut computing blocks into pieces. Thus, we already have preliminary results. We expect to obtain our own software, and the efficiency will indeed increase significantly.

But instead of four months, it will be two – still a long time for training one model. To do this, we need to increase computing power. The standard way to increase computing power is to buy more computers, but they are all based on microprocessor technology, which has already reached its limit. The size of one transistor is about ten times the size of atoms, meaning in five years, further increases in this area will be completely impossible. Therefore, companies and governments around the world are looking towards promising computing systems.

In fact, to understand what a promising computing system is – it can actually be any arbitrary physical system – one can fantasize and say: we have an aquarium with algae, we shine light on it, a chemical reaction occurs, and we read the result of the chemical reaction. Thus, the task of the modern scientist is actually to develop and select such a physical system and write a program for it.

Naturally, one of the most promising directions – here is the famous Schrödinger's cat – is quantum computing, which is based on the principle of superposition: we do not know if the cat is alive or dead until we open the box.

A quantum system with 50 quantum bits requires two to the power of 50 bytes on a classical computer – that is one petabyte. This is the memory of the most powerful supercomputer in the world. At MIT, a prototype quantum computer has already been created with 256 quantum bits. To describe it, all the atoms in the universe would be needed. This is a very promising technology.

The main problem with this promising technology, besides hardware issues, is that algorithms need to be written for it, and these algorithms need to be written now. In fact, everything we do, we write algorithms for hardware that does not yet exist, so that when this hardware is created, we are not left with a box over which we do not know what to do.

Yesterday, there was the Sber award, in particular, Academician [Alexander] Holevo received it for quantum information theory. The foundations of quantum computing were laid by Yuri Manin, Holevo in the Soviet Union, by the way. Not by physicists, but this idea actually came from mathematicians.

In fact, the key task now, in my opinion, Vladimir Vladimirovich, is large-scale investments in the development of domestic software, but, importantly, in close collaboration with hardware developers across the entire spectrum of these promising computing systems: quantum computers, photonic computers, neuromorphic systems.

Right now, we do not know which of them will be the next generation, but the state and the companies that first obtain algorithms and computers – this connection will gain a significant advantage in the computing race. This will lead to the development of science, the development of the economy, and, in my opinion, will obviously improve the quality of life for people.

Thank you. I will be happy to answer questions.

Thank you very much.

I just at the end either did not understand what you said or did not hear. You said quantum, photonic computing, and neuromorphic?

I. Oseledets: Neuromorphic. This is a system that operates on the principle, roughly speaking, of the human brain.

Yes. I heard that and noted it for myself. In principle, this is a clear matter; it is necessary to invest in domestic software, and we are trying to do this, as you know, and we will continue to increase our efforts. I would be grateful to you if you could specify what exactly needs to be done in this regard.

As for quantum and photonic computing, we have two large state companies: Russian Railways (RZD) and Rosatom – I believe I mentioned this in my speech – have supported the state in this direction. A corresponding communication line has already been established between Moscow and St. Petersburg. And we will continue to increase these efforts.

If you could specify what exactly you would expect from the state, I would be grateful.

I. Oseledets: I can say that the main problem is that we have separated software developers and hardware developers. From my personal experience: hardware developers consider software developers personal enemies because first we will make the hardware, and then you will write some software for it, which may not even work. And the key task is to have, roughly speaking, a consortium, a connection: algorithms go in parallel with hardware development, rather than having physicists separately, mathematicians separately, and machine learning specialists separately.

If we can create such a synergy point... And this is difficult because people have different mentalities. Over the years, during the 90s, when we had a hard time, everyone is very afraid that someone will come and take their money. This is a fact.

G. Gref: Vladimir Vladimirovich, fundamentally new system software will have to be created for the new systems.

Yes, I understand. But, as Ivan said, we need to find such a "point of assembly" where it would be interesting for everyone to work together.

I. Oseledets: Yes, so that all participants see the benefit, rather than seeing danger.

Yes. Dmitry Nikolaevich [Chernyshenko] is a great specialist in organizing joint work. Dmitry Nikolaevich, we need to think together about this task.

D. Chernyshenko: Yes, Vladimir Vladimirovich, we will definitely do everything, we are working with colleagues, interacting.

That immediately raises a red flag for me: "we will do everything."

D. Chernyshenko: We will exceed expectations, yes.

And we will exceed expectations.

In any case, we need to think and consult with specialists, see what is happening there.

D. Chernyshenko: In fact, an industrial center has been created.

Thank you.

Thank you very much, Ivan.

G. Gref: Vladimir Vladimirovich, with your permission, one more speaker. Ivan, thank you very much.

In fact, we are extremely grateful to Ivan's team and Skoltech because we are working on this project together, and for us, this is just a gigantic saving – a 15 percent saving in time when training the model. We are currently training our new model on supercomputer resources for a year. This means we will gain two months of savings. This is, of course, a huge saving of time.

Naturally, this is a huge saving of money in terms of electricity costs consumed by this enormous "beast" called a supercomputer.

But the task facing the team is to bring these parameters to 50 percent. If this is achieved, it will be a unique experience. Therefore, thank you very much for this work.

Another speaker. I would like to give the floor to Semyon Budyonny. Semyon is an employee of the Institute of Artificial Intelligence. There are more than 100 unique specialists in the field of artificial intelligence there. They are currently working on multimodal fusion models, which you mentioned in your speech. And here the team said that this is the latest trend in artificial intelligence – multimodal, multilingual, and multitask models, those that are already as close to humans as possible.

The team is working on 30 such projects, and Semyon is developing hybrid directions at the intersection of quantum chemistry and neural network architectures. If possible, Semyon, could you say a few words on this topic?

Semyon Budyonny. Not Semyon Mikhailovich, right?

S. Budyonny: No.

G. Gref: But a relative.

S. Budyonny: No, no.

G. Gref: No? Then I was misled; I was told he was a relative.

It is not good when the head of Sber is deceived. Pay close attention to potential losses.

Please.

S. Budyonny: Vladimir Vladimirovich! German Oskarovich!

I would like to talk about today's trends in artificial intelligence in the creation of new materials. I will try to answer three questions.

First – is artificial intelligence capable of creating fundamentally new technologies that were previously unavailable to humans?

Next, I want to discuss the existing groundwork, the existing results of the Artificial Intelligence Research Institute, some initial developments.

And then I want to propose some specific measures on how these technological seeds can be further cultivated.

When we talk about artificial intelligence in the synthesis of new materials, I would like to highlight one important property – emergence. By analogy with how living organisms, forming complex groups, transform from chaos into concise streams of groups and acquire new functions, for example, functions of resistance to predators – swarms of bees or flocks of birds, and so on, artificial intelligence, as a complex system, is also capable of acquiring new functions, generating new technologies that were previously inaccessible and non-reproducible by humans.

And the first question we tried to answer here fundamentally: can we shorten the existing cycle, the life cycle of producing new materials by orders of magnitude? A good example is the well-known material graphene, for which our compatriots received the Nobel Prize.

This material possesses a number of unique properties, and, by the way, the AIRI [Artificial Intelligence Research Institute] in collaboration with the Innopolis University, at the suggestion of Sber's research and innovation management, has taken on the creation of the largest database, a digital database of two-dimensional materials using artificial intelligence, and will make it available in the near future. This is very important for the development of data analysis in the application of synthesizing new materials.

When we talked about emergence, about creating something fundamentally new, previously non-existent in nature and non-reproducible by humans, and [about] the second factor – the question of time: can we make artificial intelligence create new technologies in a matter of months? Partially, the AIRI institute answered this question when it predicted using graph neural networks, calculated, and experimentally confirmed a new list of materials with exotic properties – so-called quasicrystals. These materials practically do not exist in nature, yet they are unique in terms of properties. For example, they can potentially be applied in the field of electrical engineering. They exhibit abnormal conductivity behavior with temperature changes. They can also be used as strengthening agents in metals and metal alloys.

As is known, previously, humans relied more on intuition and the trial-and-error method to find various recipes to achieve better results in material formulation. Then, having generalized this experience and formalized it, they obtained certain levers to control these formulations.

But now we are in a state where artificial intelligence, possessing its own "intuition," is capable of doing this in minutes and predicting, for example, a class of materials represented here by binary metals. And this is what has been calculated, predicted, and experimentally confirmed, and these results have been published in one of the prestigious global journals in the field of crystallography.

There is another example from the field of energy, for instance. Our specialists were able to calculate 650,000 different components for electrocatalysts. By deceiving calculations based on first principles, replacing quantum chemical calculations for complex compounds, we accelerated the calculations to seconds using artificial intelligence and were able to quickly predict properties for a new lattice with a new atomic composition. In this specific case, this is also not some drawn object; it is a lattice that represents a structure we calculated literally in seconds using artificial intelligence.

We understand that, aiming to create fundamentally new technologies previously unavailable to humans, we must also not forget about improving existing technologies. If we talk about materials, one of the most promising technologies is photovoltaics.

We have shown and proven the potential for an increase in efficiency of 0.5 percent. Let me remind you, efficiency is the share of energy converted from the Sun into electricity. It may seem that 0.5 percent is not much, but that technological recipe for producing solar panels using artificial intelligence, increasing the efficiency indicator by 0.5 percent, generates additional capacity for electricity generation – nine gigawatt-hours of electricity, and in absolute terms, this is already quite acceptable. This is comparable, by our estimates, to about 2,500 households. This is quite significant, and further increasing efficiency allows for an increase in this electricity capacity. Here, we, together with our partner – one of the largest solar panel manufacturers – have already proven this potential.

A pragmatic approach to synthesizing new materials requires working under conditions of uncertainty, limited resources, and contradictions. For example, the task of design research, which is relevant in engineering, involves solving a problem under the following contradictions: we want to have the lightest material possible, but at the same time the strongest, and for this, we created models that generate the porosity pattern of the material in such a way that during its 3D printing, its strength is maximized. Here are specific samples that we recently synthesized using a 3D printer, and all these calculations were conducted using generative models.

As scientists, we often draw inspiration from art, and it is worth noting that, for example, here an image was generated based on a poem by Mandelstam using a model trained on "Kandinsky" by Sber specialists. And the basis of the multimodal technology that synthesizes such objects lies in the same neural network architectures that we use in the synthesis of new materials.

I will conclude my report not with the thesis that we have problems, but rather with the thesis that we have colossal opportunities. We want to transition from the digital world to, so to speak, the physical world. The generative sum of all possible materials that artificial intelligence is capable of synthesizing exceeds the number of atoms in the universe, and this is the digital world. When I talk about the digital world, it can include alloy recipes, blueprints, chemical formulas for new chemical elements in pharmaceuticals. This volume is colossal. But the bottleneck in this story is synthesis, it is the equipment. And if we expand it, we will provide new opportunities for our industries, create new technologies, new emergent technologies previously unavailable to humans.

Here is a specific proposal – to focus on creating so-called digital technology testbeds with the ability for rapid prototyping, creating experimental samples, materials based on the results of artificial intelligence synthesis. If we create such a flow, expanding this funnel, we will provide colossal opportunities for industry. And on this platform, there should be equal access for both scientists and businesses.

That’s all from me. I am ready to answer questions.

Thank you very much.

Thank you.

What you are saying is undoubtedly captivating. As soon as you said that materials are being created that do not exist in nature – that is, do not exist on Earth, this, of course, is impressive in itself and undoubtedly creates colossal opportunities for all areas of human activity, not to mention the economy, and in healthcare, and in industry – everywhere. This is certainly very promising and very interesting.

If you say that it is necessary to create these testbeds for equipment, then, of course, my colleagues and I in the Government – I have noted this for myself – will think about what can be done, and in consultation with you, German will suggest, we will outline specific steps to move along the path you are showing us.

Thank you very much. Good luck!

S. Budyonny: Thank you.

G. Gref: Thank you very much, Semyon.

Thank you very much, Vladimir Vladimirovich.

You still have five people. It would probably be rude if we didn’t let anyone say anything at all. If possible, just briefly on the issue and the tasks you see. You have two more people there, right?

G. Gref: Yes. Guys, you’re lucky. Then, if possible, let’s skip introductions, very briefly.

The problem and what you would like to see and feel from all of us as support. Please.

G. Gref: Then Alexey Naumov, please, Higher School of Economics.

A. Naumov: Good afternoon, Vladimir Vladimirovich! Good afternoon, colleagues!

I am engaged in reinforcement learning, I will be very brief. This is actually a very important task because it arises in the design of unmanned vehicles, in production automation, in the design of cooling systems for data centers.

Reinforcement learning, generally speaking, underlies these systems. There are already many examples of artificial intelligence using reinforcement learning, for example, defeating the world champion in Go. And reinforcement learning is not just an ordinary artificial intelligence model: it learns like a human learns, that is, a person gets up, walks, falls – interacts with the environment.

Similarly, if we consider, for example, an unmanned vehicle, it also interacts with the environment, with other road users, and it receives rewards and penalties from that interaction. The goal is to find an optimal strategy that maximizes rewards and minimizes penalties.

But here the biggest problem is the volume of accumulated experience. When we, for example, teach an unmanned vehicle, it is very expensive to train it in practice, and indeed, we do not see these unmanned vehicles around us now. What many countries are doing, and what we also need to develop, is the so-called digital twins, into which we can insert an unmanned vehicle and train it there accordingly.

But the biggest problem arises, and you already mentioned this in your speech, that if we start using a trained digital twin model in practice, the cost of error here is simply enormous. This means that we need to look for algorithms, reliable algorithms, and moreover, with mathematical guarantees of reliability. Very often, there are groups of researchers who do the mathematics, but their mathematics is not applicable in practice, and there are groups of researchers who create very practical algorithms that are completely inapplicable in real scenarios.

What we have done: we have created an algorithm that has mathematical guarantees of reliability and can be transferred to real environments. This work has been highly appreciated by the scientific community.

Today, we need to develop this topic of reinforcement learning in Russia, because it arises in production automation, and we can reduce costs and create unmanned vehicles in the future. We need to develop fundamental research and, accordingly, engineering schools, because without engineers, for example, we will not be able to launch our algorithms on real "hardware."

Within the framework of the federal project, there are currently, as you have already mentioned, six artificial intelligence centers – these are support centers. This initiative is supported by the Ministry of Economic Development until 2024. We believe this is a very correct initiative, and, Vladimir Vladimirovich, we ask you to extend this initiative until 2030.

We would also be very happy if new artificial intelligence centers appeared in Russia for promising and only emerging areas of artificial intelligence.

Thank you very much.

Thank you very much, Alexey.

Regarding the extension of the program: of course, we will think about it, and I think we will do it. As for new centers – Dmitry Nikolaevich is also listening. (Addressing D. Chernyshenko.) Dmitry Nikolaevich, please consider this later, we will consult and make the appropriate decision.

Thank you very much and good luck.

As for unmanned vehicles, we have very good developments in this area. Moreover, we are even surpassing many of our competitors, and we are preparing an experiment using unmanned vehicles on the high-speed highway from Moscow to St. Petersburg in the near future.

Please.

G. Gref: Thank you.

Vladimir Vladimirovich, our last scientist is a young woman, a teacher and researcher from St. Petersburg, from ITMO University, who has significant experience in teaching as well as scientific achievements. I ask Anna to present her scientific achievements as well.

A. Kalyuzhnaya: Hello, Vladimir Vladimirovich!

Thank you very much.

I have, of course, also prepared a presentation, but I will not show it.

What am I engaged in? I am involved in the automation of machine learning. What is this? This is essentially artificial intelligence that allows the creation of all other artificial intelligence.

Why are we doing this? Because this year we see that the demand for artificial intelligence specialists is seven times higher than the level that was just three years ago. Accordingly, there are more applied tasks, more need for implementation, and more people are required to do this. This will only increase, and we will simply not be able to produce a sufficient number of specialists.

Therefore, at ITMO, we are developing algorithms that allow for the implementation of artificial intelligence to be much faster and more accessible, both in terms of time and computational resources, because objectively today, not all industrial companies have significant computational resources. With such algorithms, we can reduce the development time for artificial intelligence solutions and their implementation to just a few days.

So today, we can say that we have algorithms, and we know how to develop them. What is lacking for the mass implementation of artificial intelligence in industry?

Firstly, it is data, primarily industrial data, since unlike the banking sector and the trade sector, each industrial sector has completely different processes, completely different equipment, and this cannot be generalized so easily.

Secondly, there is a lack of technological testbeds that could exist at least for the main industrial sectors, for the main processes. Within these technological testbeds, we, as algorithm creators, could demonstrate what they are capable of, and industry representatives could assess their capabilities and evaluate the potential risks of implementation. Today, the implementation in companies is also hindered because once an algorithm has been developed and implemented, it often entails a whole chain of additional actions: entire business processes need to be changed, and sometimes even the whole system. This is a problem, and it seems to me that such technological testbeds could help address this.

That's all from me. Thank you.

Thank you very much.

I have noted this for myself as well. As for technological testbeds and data, especially in the field of industry, this is certainly a complex task; we need to work by sectors, understand with your help, I hope, how to establish this work, how to encourage our colleagues in various industrial sectors so that you and your colleagues who are working in this area have the necessary foundation for generalizations and work. We will definitely work on this.

It is extremely important, and I completely agree with your conclusions, and I also spoke in my address about the tasks in the field of personnel training. If we can put all this together, then, without a doubt, we will achieve results in all other areas of activity related to artificial intelligence.

Thank you very much.

I want to thank all of you for your work, for your participation in this work today. I want to thank Sberbank, German Oskarovich, you for paying such serious attention to this. You are not just leaving this direction, but, on the contrary, developing it, supporting it in every way, just like all the other participants in this process: both the people engaged in scientific creativity and those who organize this work in our companies, in the Government, and I wish you success.

Because, without any exaggeration, your success will determine the success of the country as a whole. This is absolutely true, and there is no doubt about it. I want you to understand that we also understand this and recognize our tasks in this area, but, of course, we can only effectively solve it by joining efforts with those people who work in the industry.

Thank you very much, and all the best. Thank you.