Volume 14, Nº 4 (2024)

The first part of the article discusses the book The Myth of Artificial Intelligence by American scientist and entrepreneur E. Larson, which focuses on debunking some myths about artificial intelligence. These myths, which have persisted for over half a century, suggest that the emergence of human-like ("general") AI and eventually superintelligence is inevitable, occurring naturally as AI systems evolve. The book criticizes these myths in two ways: scientific and social. It is shown that machine learning does not lead to general AI, and the myth of AI makes human potential look weaker. The second part of the article considers the problem of understanding. The concept of cognitive semantics is proposed, based on the ideas of J. Lakoff, S. Pinker, A. Damasio and A. Seth. In particular, it is noted that: understanding is an interpretation in terms of a person’s picture of the world; the picture of the world is constructed by our brain, and it is structured through the categorization of human experience; meanings (senses) are formed earlier than conceptual structures are formed; biological goals underlie meanings; not only the brain but also the body participates in cognitive processes, and understanding is associated with actions in the environment, knowledge of which is contained in the picture of the world. The article concludes by pointing out dead ends, difficulties and dangers on the path to general AI.

The design of information models in the development of automated systems involves constructing entity-relationship models, which define classes of entities and interclass relationships of the ancestor-descendant type, serving as a conceptual and ontological foundation for the subsequent creation of databases. This article explores aspects of representing the semantic constraints of the subject area imposed on relationships between entities in entity-relationship models. A special type of relationship between classes of entities, termed multiple ancestor (MA) relations, is highlighted, where several instances of the ancestor entity may exist for a single instance of the descendant entity. The article analyses the possible semantic constraints that arise in such cases and the anomalies they generate. Based on the introduced concepts of ascending kinship lines and the MA predicate, a formalisation of MA integrity is presented. Examples of formal MA constraints are provided, including positive (requiring the coincidence of ancestor instances), negative (requiring the mismatch of ancestor instances), and mixed. The mutual influence of several MA constraints with partially intersecting lines of ascending kinship is also examined. Finally, the feasibility of implementing MA constraints in relational database models is evaluated.

Patient condition monitoring is essential at every stage of the treatment and diagnostic process. During treatment, the key tasks of monitoring include evaluating the effectiveness of treatment strategies and ensuring safety, specifically by avoiding complications and adverse effects from medications, procedures, or their combinations. To enhance doctor-patient interactions and improve treatment quality, intelligent monitoring tools are needed. These tools must use patient data and formalized knowledge of the subject area to identify which observations, at specific times, can indicate any deviation of the patient's condition from expected outcomes.This study aims to develop an ontology that formalizes the knowledge necessary for selecting and explaining monitoring parameters throughout treatment. Key concept relationships that address the challenges of monitoring patient conditions during treatment are identified, leading to an ontological graph for a specific class of medical monitoring issues. The study describes a method for building knowledge graphs applicable to various diseases and a reasoning process to identify unexpected conditions. A process of reasoning and the resulting recommendation is presented. This approach establishes a foundation for a decision support system tailored to this class of problems, where monitoring parameters can be adapted according to treatment stage, patient condition, and individual characteristics.

To manage pedestrian and vehicle flows at intersections, systems are implemented that use adaptive traffic light models, adjusting time intervals based on the volume of pedestrians and vehicles. These systems include video cameras that monitor road user movements, enhancing real-time traffic control. This paper introduces an information and analytical system for managing transport and pedestrian flows using the YOLO neural model, which enables object recognition. The system performs several operations: converting the original image to gray scale, applying Gaussian blurring, detecting object boundaries through the Canny filter and fuzzy logic for edge detection, and contour processing, where each identified contour is assigned a unique number. The neural network then compares detected contours to the training sample data, determining whether the object is a person or a vehicle. The paper presents experimental results for these algorithms in object recognition. Modified software and images of intersections with pedestrian crossings captured from street video cameras in Kursk were used in the experiments. The recognition accuracy rate from the experiments was 72.4%.

The article outlines an approach and software tool for the automated enrichment of domain-oriented knowledge graphs with new facts derived from semantically annotated tabular data. For semantic annotation of table columns, a combination of three heuristic methods is proposed, leveraging named entity recognition in cells, lexical matching, and feature grouping. This approach is implemented as a dedicated handler within the Talisman software platform. An example and experimental evaluation of the approach during the semantic annotation of columns are provided using a test set of tabular data across six thematic categories: "organization employees," "open vacancies," "car model market," "famous scientists," "book sales," and "tennis player rankings." Evaluation metrics included precision, recall, and F-measure, with final results across all six categories as follows: precision - 79%, recall- 63%, F-measure - 70%. These results highlight the potential of the developed approach for enriching domain-oriented knowledge graphs with new facts from semantically annotated tabular data. The limitations of the proposed approach are also discussedfrom semantically annotated tabular data. The paper also provides a number of limitations of the proposed approach.

The article examines the functionality and architecture of an interactive application designed to support multi-criteria decision-making using S.A. Piyavsky's confident judgment method. It provides a description of key components of this method essential for presenting the topic in a computer-oriented format for data and knowledge representation. The application stores and provides access to a bank of universal importance coefficients for specific criteria and serves as the core of an information methodology for decision-making based on the confident judgment method. This methodology includes the following stages: compiling a list of alternative solutions with evaluations by a set of specific quantitative criteria; establishing a selection policy that orders criteria by importance for the decision-maker; normalizing the multi-criteria problem statement in line with a foundational optimization model; evaluating each alternative solution within this model; and selecting the optimal alternatives. The application automatically coordinates the creation and editing of the list of alternative solutions according to variation boundaries set by the selection policy, integrating all solution stages with user feedback. The application’s functionality is represented by a diagram linking the main activities and their outcomes, along with a UML diagram of application use cases. The architecture of the application is presented as a mind map. The application is implemented on the Excel platform using VBA programming language, chosen to ensure accessibility and support the development of advanced multi-criteria decision-making methods for a broad user base who use Excel in their professional work.

Decision-making requires in-depth analysis through the evaluation and comparison of various alternative solutions. These tasks are characterized by multiple criteria used for comparison. The variability of parameters and conditions in decision-making has led to the development of numerous methods aimed at facilitating this process effectively. Currently, several hundred multi-criteria decision-making methods are known, making it challenging to select the most appropriate one for practical use. Therefore, research on the analysis, classification, and systematization of multi-criteria decision-making methods is relevant and essential for decision-makers and developers of decision support systems. Most classifications emphasize the methodologies and purposes of these methods, often overlooking their inherent limitations. This article attempts to systematize the main multi-criteria decision-making methods, considering their limitations. It analyzes the most widely used methods and their classifications. Based on identified limitations, the paper proposes a classification of multi-criteria decision-making methods and guidelines for selecting them. Applying this approach can enhance the validity of method selection and, consequently, the effectiveness of the decisions made.