Sources of Risk and Design Principles of Trustworthy Artificial Intelligence

Abstract

The importance of artificial intelligence is constantly increasing due to ongoing research successes and the implementation of new applications based on it. It is already described as one of the core technologies of the future. This technology is also increasingly being applied in the field of safety-related applications, which enables the implementation of innovative concepts for novel protection and assistance systems. However, for this to lead to a benefit for human safety and health, a safe or trustworthy artificial intelligence is required. However, the increasing number of accidents related to this technology shows that classical design principles of safe systems still need to be adapted to the new artificial intelligence methods. On the one hand, this requires a basic understanding of the components of trustworthy artificial intelligence, but on the other hand, it also requires an understanding of AI-specific sources of risk. These new sources of risk should be considered in the overall risk assessment of a system based on AI technologies, examined for their criticality, and managed accordingly at an early stage to prevent later failure of the system.