ENSURING RELIABILITY IN THE LIFE CYCLE OF RESPONSIBLE ARTIFICIAL INTELLIGENCE SYSTEMS

Abstract

Background. Over the past few years, there has been a dramatic expansion in the capabilities of artificial intelligence (AI) systems, which has simultaneously led to new risks and potential benefits. The autonomy of systems, which allows moving from decision support to self-development of a decision, provides conflicting results in areas of responsible application. Such sensitive areas of application as human health, its ecological living environment, economic and social status. In the field of armaments, the means of creating autonomous systems of a new generation and the concept of a future «hyperwar» associated with them are discussed. Materials and methods. Due to the growing use of AI around the world in these sensitive areas, there is a demand for reliability when using such autonomous systems. It is necessary to formulate the risks and benefits of this technology, including compliance with fundamental ethical principles. The application of critical decisions must be controlled by the person in charge. Measures to ensure the reliability of an autonomous system with AI must be provided at all stages of the life cycle, and only in this way it is possible to control risks and create explainable and manageable AI. Results and conclusions. The authors outlined the concept of the concept of «reliable» AI and described the implementation of its individual provisions at the stages of the life cycle of an autonomous system for responsible purposes.

Keywords

• reliability of the artificial intelligence system
• responsible assignment system
• decision-making by an autonomous system