What would you call a machine that routinely and predictably produces evil options with marginal differences between them? How reliable of an evil-producer does this machine have to be before you recognize that the production of a certain type of evil is the machine’s purpose?