Translartion. Region: Russians Fedetion –
Source: Peter the Great St Petersburg Polytechnic University – Peter the Great St Petersburg Polytechnic University –
What is the essence of hybrid quantum-classical artificial intelligence? In which high-tech areas are quantum computing already being piloted today? And when might high-performance quantum computers appear in Russia? Anton Guglya, an employee of the National University of Science and Technology MISIS, told participants of the SPbPU seminar on artificial intelligence about this.
Anton Pavlovich in his review focused on the practical application of quantum computing for the implementation of business projects. In Russia, there are already the first experimental versions of quantum computers and a cloud interface for accessing them, although none of the domestic quantum computers can yet solve industrial-scale problems. At this stage, businesses and scientific groups have the opportunity to carry out research and pilot projects using quantum computing emulators – “digital twins” of quantum computers that use new algorithmic approaches, but operate on traditional CPU and GPU architectures.
The financial, oil and gas and other industries need business effects, they need results, not just advanced technology. If their tasks are solved using traditional methods, they will not invest in advanced equipment, new software and obtaining additional competencies. But if they see the benefit, even if it is delayed in time, from using quantum computing, then, of course, they will prefer to use it, the speaker emphasized.
Anton Guglya spoke about the technical features of hybrid quantum-classical neural networks, describing their advantages over classical ones:
quantum algorithms will be able to solve some classes of problems faster and more accurately than classical analogs; it is possible to process a larger volume of information with less energy consumption; a smaller set of hyperparameters for training a neural network when working with noisy data.
The MISiS specialist presented a typical process for developing an industrial quantum algorithm and launching it on an emulator, as well as several cases from his practice. For example, a credit scoring algorithm was implemented using quantum neural networks. The resulting model, based on data, calculates the probability of the return of funds received from the borrower to the bank.
Another popular area is image classification. Using hybrid quantum-classical neural networks, this task can be performed faster and with greater accuracy. Also, with the help of a quantum neural network, the task of generating new drugs is performed more efficiently. At the same time, the learning speed of quantum neural networks is much higher than that of traditional ones.
The second presentation at the seminar was devoted to the prospects of creating “virtual cities”. Ivan Dorokhov, a research fellow at the AI Center of Novosibirsk State University, spoke about the integrated system for modeling “virtual cities” that he and his colleagues from SPbPU and SPbSU are proposing.
According to the speaker, unlike digital twins, which directly reflect the physical environment, virtual twins can simulate a wide range of objects or scenarios for various purposes and are not conceptually limited to digital computing.
Our results show that such virtual environments can facilitate the emergence of societies consisting of AI agents, i.e. subjects with artificial intelligence, in which governance mechanisms such as decentralized autonomous organizations (DAO) and artificial collective consciousness (ACC) provide ethical and normative control, said Ivan Dorokhov.
Agent-based AI systems are already on their way to engaging in cultural, economic, and political activity, shaping the social fabric of cities, and researchers must address how to make their impact as beneficial to people as possible. This will require understanding how advanced rendering and procedural generation techniques can provide the high realism required for complex simulations, how IoT data and predictive modeling can make these environments more responsive and adaptive, and what ethical and legal frameworks will be needed when AI agents gain autonomy. Answering these questions is important to realizing the potential of virtual worlds as platforms for innovation, experimentation, and sociotechnical evolution.
Both presentations raised many questions and a lively discussion, which you can join at the next seminar on April 23, which was announced by the Head of the Department for Support of Scientific Projects and Programs Natalia Leontyeva.
Please note: This information is raw content directly from the source of the information. It is exactly what the source states and does not reflect the position of MIL-OSI or its clients.