Medical artificial intelligence and the black box problem: a view based on the ethical principle of “do no harm”