An AI that was smarter than us might deduce that we were not applying the Deep Structure of our morality properly because of bias or limited intelligence. It might conclude that human morality requires humans to greatly reduce their numbers in order to lessen the impact on other species, for instance.