Reliability of artificial intelligence in civilian and military applications

Introduction/purpose: The goal of this paper is to show the importance of reliability and availability regarding artificial intelligence (AI) which is nowadays applied in almost every area of human life, both civilian and military. Everything can fail, including AI as any other product or device. AI also needs to be reliable. It is mainly realized as software, but also involves hardware, data and the human factor, so it is complex for building in and for reliability assessment. Methods: This study has used the published articles of this author as well as some other papers and documents. Results: The research has resulted in brief definitions of reliability, availability, artificial intelligence, reliability of hardware, software, data, the human factor, and AI as well as in a brief analysis of some reasons for the unreliability of AI, such as data quality, a small or insufficient amount of training and test data, training on unreal data, the nature of learning algorithms and user incompetence. Conclusion: The reliability of artificial intelligence is very complex and due to its importance it must be considered during AI design and application.