While we can all appreciate the benefits of AI, it is developing faster than anyone imagined, with no consensus on what constitutes acceptable risk