The paper addresses an approach to combat corporate tax evasion by using the possibility of indirect accounting of latent factors in taxation processes. Value chains formed in the digital environment are increasingly falling out of government control, and digital platforms allow the building of relationships that are opaque for tax authorities. The digital market itself is becoming a factor that creates new value and possibilities. The sharing society is self-learning to reduce the tax base: business is getting rid of intermediaries, eliminating the regulator, and avoiding double taxation and taxation in general. This is happening against the backdrop of the chaotic development of the global financial system. The paper’s idea is based on creating a decision-making system for predicting tax evasion events by forming a special information framework of potentially suspicious events, using methods of deep learning of neural networks, and cognitive modelling. The latter was enriched with non-formalizable cognitive semantics, which increases the transparency of tax events. In real practice, the approach has helped to diagnose suspicious tax evasion cases during supplying products for the Arctic zone of Russia

The paper is devoted to finding the time optimal route of an agent travelling across a region from a given source point to a given target point. At each point of this region, a maximum allowed speed is specified. This speed limit may vary in time. The continuous statement of this problem and the case when the agent travels on a grid with square cells are considered. In the latter case, the time is also discrete, and the number of admissible directions of motion at each point in time is eight. The existence of an optimal solution of this problem is proved, and estimates of the approximate solution obtained on the grid are obtained. It is found that decreasing the size of cells below a certain limit does not further improve the approximation. These results can be used to estimate the quasi-optimal trajectory of the agent motion across the rugged terrain produced by an algorithm based on a cellular automaton that was earlier developed by the author.

Using the structural-parametric approach, of which the three-dimensional method of S ‒ approximations is a part, we carried out a series of experiments where we approximated the known measured components of the magnetic field by the sum of simple and double layers distributed on two carriers of magnetic masses equivalent in the external field. We obtain data of measurements of the magnetic field of Mars by the magnetometer of the MAVEN station from the PPI PDS public website. Based on samples of satellite measurements from large data arrays on arbitrary grids according to various criteria, models of the Martian magnetic field were constructed using a modified method of S‒approximations at satellite heights and on the planet’s surface in the areas of operation of the InSight and the Tianwen‒1 missions.

This paper investigates the integration issues of new Hybrid Intelligence components implemented in the Data Envelopment Analysis technology based on the Quadrant Enabled Delphi method. Monitoring of activity and control of selective interaction of their multilateral communication with each other is provided by the Blackboard System based on Commonsense Knowledge to create a Virtual Semantic Environment. The report is illustrated by some examples of the consideration of these issues. In particular, attention is focused on the context of proactive protection of Hybrid Intellectual Property objects. The conditions for the emergence of a conditioned technological gap with the potential possession of such breakthrough technologies with reliable signs of Artificial Intelligence are also indicated. Keywords—Quadrant Enabled Delphi (QED) method, Predictive Model Markup Language (PMML), commonsense knowledge, blackboard system, monitoring, system integration, hybrid intelligence.

The purpose – what is the components of Artificial Intelligence (AI) of Smart Systems (SmS)? Design / methodology / approach – like natural intelligence, two conditions should be distinguished in AI: 1) the necessary condition is the presence of the first signal system as an elementary sensorium associated with reflexes, sensations, terminal sensors and performers; 2) a sufficient condition is the presence of a second signal system as a language associated with words, texts, and speech. Findings – the thesis about the necessary and sufficient requirement for AI of SmS consists of two requirements: 1) the existence of an information-system redundant reflexive model of SmS and its theater of action. This model is presented in the form of an “Environment of Radicals" (ER). That is, the presence of an ER-based reflexive model of SmS, and its theater of action (elementary sensorium); 2) the existence of a language of SmS as a semantic sensory superstructure of a ER-based reflexive model of SmS. That is the presence of semantic sensory structure of ER-based model of SmS with the sensory operation system of ER-based reflective model of SmS (the linguistic sensorium of SmS). Originality / value – Redundancy of the ER-based reflexive model of an SmS and its semantic sensory superstructure (a structure with the operating system as the language of SmS). Research / Practical / Social / Environment implications – creating a unified information and system space Artificial General Intelligence of SmS and natural human intelligence. Research limitations – the existing element base of computer science. Keywords: Smart system, artificial intelligence, redundancy [https://www.researchgate.net/publication/354872469_Artificial_Intelligence_and_Smart_Systems = URL: https://www.youtube.com/watch?v=W3gKOUq9b94 ].