Who will AI save?
Since it will necessarily be humans at the helm of the AI creation process, it makes sense that AI be programmed in such a way to minimise risk of harm to all living things. Many people wonder, especially in the context of self driving cars, who will be killed? a pedestrian crossing the street, or the passengers in a car? (if you could only save one set)
Humans make AI
While the exact answer to this question varies from situation to situation, the ultimate answer depends upon the way the AI is programmed to behave. You see, AI, atleast in 2017 isn’t completely autonomous yet, which leaves some room for humans to decide who that value is offset towards protecting. Afterall, if AI dies, Humans can resurrect it but if the human dies, AI wouldn’t be sophisticated enough in time to prevent eventual catastrophe, unless it fundamentally taps into all of mankind’s minds simultaneously (and still encounter limitations)
My guesstimate is, that the AI would (or should) be programmed to minimise life loss in situations of great peril, where loss of life is completely unavoidable. That has to be a quantitative measure rather than a qualitative one, simply because of the fact that humans are nervous, twitchy, and have a potential to be random, creative, and overwhelmingly intelligent in spurs of time, so a computer should have no way of reliably identifying whose life is more important, even if we live in a totalitarian machine learning-controlled society.