Inspect, Understand, Overcome: A Survey of Practical Methods for AI Safety