LETI: On the Leap to Industry 5.0 and Trust in AI: Scientific Discussion at LETI

LETI hosted the 25th International Conference on Soft Computing and Measurement (SCM’2022) from May 25 to 27. More than 90 researchers from Russia, Canada, Spain, the Philippines, China, and the USA participated in the event.

Mikhail Kupriyanov, Head of Scientific and Educational Areas at LETI, opened the plenary session. In his welcoming speech, he noted that for 25 years of its existence, the conference brought together Russian and international scientists in the field of artificial intelligence, and brought up a whole generation of researchers in the field of soft computing and measurement.

“25 years is a long time. During this time, new schools and areas of research have emerged within the conference. Some have graduated from schools and universities and are now participants in our conference. This conference lives its own life and keeps up with the times, constantly improving its agenda according to the most advanced research. It is adaptive and reflects the true picture of the latest cutting-edge research.”

Mikhail Kupriyanov, Head of Scientific and Educational Areas at LETI
Svetlana Prokopchina, Professor of the Financial University under the Government of the Russian Federation, Co-Chair of the SCM’2022 International Program Committee, made a presentation on Intelligent Sensor Networks. The speaker demonstrated a generalized concept of creating digital control platforms for complex systems based on a Bayesian regularization approach. In particular, she listed the methods that her team used to develop intelligent sensor networks under conditions of information uncertainty and showed a structural diagram of the resulting network.

The developed network was tested by measuring moisture levels in wells. Experiments have shown that for greater efficiency, it is necessary to classify the conditions that can affect the measurement results.

“The choice of the measurement scale of a thermal imager is highly dependent on the measurement conditions: snow height, wind, surface layer temperature, pipe depths – a whole set of factors that create absolute uncertainty. It seemed to be a big challenge to dial in a dataset to train the network under such conditions. I proposed to classify these conditions and make a measurement scale for them to lower the size of the dataset and build a series of interconnected neural networks. This solution covers many changing conditions and dramatically reduces the need for data, which is quite difficult to collect,” explained Svetlana Prokopchina.

Vadim Borisov, President of the Artificial Intelligence Association, devoted his speech to the expectations from Industry 5.0. The speaker talked about the developed method of analysis and composite hybrid modeling of the complex technical systems: centrifugal compressors, pumps, fans, and turbines, and about the created method of fuzzy situational control of complex technical systems based on hybrid models.

These methods were used to solve compressor unit control problems in the research laboratory “Modeling of Technological Processes and Design of Power Equipment” at St. Petersburg Polytechnic University Peter the Great. The experiment allowed increasing the modeling accuracy and control efficiency of compressor units following the criteria of safety, economic efficiency, and regulation accuracy.

Vladimir Gorodetsky, an Honored Scientist of the Russian Federation, delivered his report “Artificial Intelligence and Data Science: Together or Separately?” He devoted his speech to acute problems and interconnections between the two fields.

Sergey Jurish, President of the International Association of Sensor Systems (Spain), made a presentation titled “Industry 4.0: Market and Trends”, during which he talked about the transition from Industry 4.0 to Industry 5.0. The scientist described aspects of the past industrial revolutions: mechanization, electrification, automation, and smart production, and also raised the issue of the peculiarities of Industry 5.0.

Roald Taimanov, Head of the Laboratory of Mendeleev Institute for Metrology, and his Deputy Ksenia Sapozhnikova highlighted the problems of trust in the measurements made by artificial intelligence. The scientists considered the issues of training both weak AI, which is increasingly used in various hardware, and strong AI, a feature of Industry 5.0.

The speaker noted that the level of trust in AI depends on the reliability of measurement results. However, any AI can make mistakes due to software failures and other malfunctions. Experts suggest solving this problem through “human” education. “To build trust in a strong artificial intelligence, we should take advantage of the experience of raising a person who can be trusted,” he said.

Alexey Averkin, a leading researcher of the Educational and Research Laboratory of Artificial Intelligence, Neurotechnologies and Business Analytics at the Plekhanov Russian University of Economics, presented a paper on the methods of explainable AI in the works of Lotfi Zadeh. The speaker pointed out that there is a contradiction between the accuracy of AI and explainability. The most precise methods, such as convolutional neural networks, do not provide any explanations, and explainable rule-based methods are often less precise.

The conference continued with breakout sessions on 13 areas devoted to the general theory of measurement, metrology, probabilistic methods in information processing, Bayesian approach, system modeling, neurocomputer networks and neurotechnologies, AI cognitive systems, fuzzy systems, new approaches in measurement, intelligent measurement systems, Big Data technologies, Internet of Things and other areas. A total of 86 papers were presented in Russian and English.

The 25th International Conference on Soft Computing and Measurement SCM’2022 was organized by the Ministry of Science and Higher Education of the Russian Federation, Institute of Electrical and Electronics Engineers (IEEE), ETU “LETI,” Lomonosov Moscow State University, International Fuzzy Systems Association (IFSA), Association for Artificial Intelligence of Russia and Russian Fuzzy Systems Association.