Three policies of AI-safety in viewpoint of functional safety and asymmetric classification methods for judgement of safety

Constructing way of functional safe systems using artificial intelligence has not been sufficiently discussed. In order to develop and understand safe systems using artificial intelligent technology, we should first clarify the role of artificial intelligence technology in safety systems. We propose to classify the role of artificial intelligence in safe systems into following three policies. (1) Use AI independently from safety related system. (2) Use AI as part of safety related system. (3) AI is used as a human substitute. Generally, the reliability required for safety related systems is extremely high, and there is a gap with the reliability of artificial intelligence technology. As a measure against this problem, we propose asymmetric classification methods for improving reliability on a viewpoint of functional safety. These are to bias decision on non-safety side, are to prevent ambiguous cases mis-recognizes as safe state while possessing the judgment function on the safety side. By these methods, we expect that these techniques will lead to the utilization of artificial intelligence technology in functional safety systems.