ABSTRACT
It is useful to consider AI safety within the broader context of safety engineering. This provides fundamental principles for identifying and managing risks, drawing general lessons from the management of other systems such as airplanes or nuclear power plants. Principles of safe design, crucial for improving a system's safety and controllability, are discussed in this chapter, particularly in relation to AI systems. Any competent form of risk management needs to consider tail events with low probability but high impact. The concepts of tail events and black swans can be applied as part of strategies to mitigate unforeseeable risks from AI.
