I think the appeal of AI, other than saving money on staff, is that it’s a non-deterministic black box without ethics. Having no-one on staff who understands exactly why the AI made a decision is great for plausible deniability in court.
Not without ethics. With programmable ethics. Programming null ethics is still a choice.
Correct. AIs do not absolve people of responsibility. You can do awful things with the knowledge in the library, but we all know humans are the ones putting things into action, not books.
“Your honor, it is impossible to say why the sentient D20 chose to roll <5 for this patient. It is omniscient, and has been weighted carefully by our alchemists to produce decisions as cold and impersonally as the best of our insurance drones.”