MOSCOW, Jan 5 Human and artificial intelligence hybrids will appear in 2022, says Alexander Bukhanovsky, director of the Competence Center of the National Technology Initiative Machine Learning and Cognitive Technologies at ITMO University (St. Petersburg). In an interview with RIA Novosti, he predicted five trends in 2022 cognitive technologies. Under the first number Bukhanovsky called cognitive «centaurs», or hybrid decision support systems, the principle of operation of which is based on the joint iterative solution of a complex problem by man and artificial intelligence (AI). Moreover, they will act on an equal footing, in contrast to the classical approach, when a person acts as both a «teacher» and a measure of quality for artificial intelligence, said the agency's interlocutor.
& # 34; Thus, a person and artificial intelligence can simultaneously show their strengths (a person — due to intuition, and AI — due to speed and unconditional memory), which allows you to achieve a better result than separately & # 34; — said Bukhanovsky.
These technologies are necessary in situations where the problem being solved is difficult for a person to perceive, he explained. «To recognize wearing masks in photographs, this approach is clearly redundant — classical methods of machine vision will also cope here, but, for example, to manage a large industrial company — the very thing «, — the expert noted. 
Putin called for the development of artificial intelligence systems in the military sphere The next trend, according to Bukhanovsky, will be the so-called generative design of socio-technical systems. “Generative design is usually understood as the use of artificial intelligence for automatic design tasks, when it is necessary to create a digital description of a real-world object that meets a priori requirements. For conventional technical systems, this approach is not new, «- explained Bukhanovsky. Nevertheless, a direct transfer of this approach to socio-technical systems (for example, organizational schemes for the work of enterprises) is directly impossible — people introduce (including deliberately) an additional factor of uncertainty, which is not allows you to apply existing optimization methods and even assert that the solution to such a problem is possible, the expert added. Bukhanovsky said, however, with an AI approach, he said, so in the future, it can be expected that AI, for example, will be able to design a production line along with workforce requirements, their work schedules, and an HR strategy aimed at developing and retention of qualified employees, the source said. The third trend in 2022 will be composite technologies of automatic machine learning «for human understanding, but based on» inhuman «logic,» Bukhanovsky believes. 
New Russian technology will make it possible to replace a person in another area According to him, the systemic shortage of specialists in working with big data makes one think about the possibility of replacing them with artificial intelligence systems that can perform the same creative process — form a solution to an applied problem based on existing machine learning methods Unfortunately, expert approaches based on the description of the best «human» practices turned out to be not as effective as we would like, the expert noted. there are machine learning models that select a successful design, starting from the composition of all kinds of elements (methods, algorithms, libraries) that provide the best solution to the problem, «Bukhanovsky added.» Already in 2021, a curious situation occurred when this method automatically generated a solution to the problem of preventing flooding niy, which won first place at the hackathon of the Ministry of Emergencies. At the same time, tens of times less time was spent on it than usually required by developers, «- said the interlocutor of the agency. Probably, further these trends will only intensify, and the introduction of such technologies will lead to the fact that insufficiently qualified specialists solving this or that problem simply enumeration of different methods will remain unclaimed in the market, since «automation» will do everything better, faster and cheaper, Bukhanovsky believes. 
Artificial intelligence will help doctors monitor the course of treatment of patients Bukhanovsky named metrological cognitive technologies as the fourth trend, when strong artificial intelligence «discusses» the quality of weaker AI. «Since artificial intelligence works in conditions of uncertainty and incomplete data, it, like a person, is capable of making mistakes, «- said the expert. According to him, traditionally, the quality of AI systems is discussed only on the basis of empirical validation (checks on data that were not used for training).» But such data is always small, it is fragmentary and not always representative. Therefore, the classical approaches of measurement theory are not applicable here (there is nothing to compare with), and quality guarantees are very conditional, «Bukhanovsky added. and will create the missing data for that go to cover all sorts of situations. «As a result, the quality issues of AI systems from the purely empirical area will move to a unified theoretical basis; it will be possible to speak more confidently about their standardization and certification from the point of view of the uniformity of measurements, «said the agency's interlocutor. 
Putin instructed to report on the risks of using artificial intelligence Finally, the fifth trend will be avatar-like technologies, Bukhanovsky said. “Digital avatars are now often understood as datasets characterizing a particular person in cyberspace. These data are the basis for various kinds of recommender systems, assistants and others. However, such an understanding of the avatar is “lifeless”, since there is no feedback — the “awareness” of the results of interaction with cyberspace by the avatar, ”the expert noted. reproducing a person's reaction to such interactions, Bukhanovsky explained. «Such technologies can give rise to a whole set of applied solutions related to avatars, for example, a» digital inquisitor «, which allows its user to confidently exist even in» fake universes «, or» hunt ka behind the heads, «who is able to digitally trace a highly qualified specialist (who is not going to change jobs at all) and make him an offer that he will most likely not be able to refuse,» the expert said. Thus, the introduction of avatar-like technologies should lead to a new round of personification and «individualization» of AI systems, Bukhanovsky added. 