Two remarks on the new AI control problem

This paper examines the new AI control problem and the control dilemma recently formulated by Sven Nyholm. It puts forth two remarks that may be of help in (dis)solving the problem and resolving the corresponding dilemma. First, the paper suggests that the idea of complete control should be replaced with the notion of considerable control. Second, the paper casts doubt on what seems to be assumed by the dilemma, namely that control over another human being is, by default, morally problematic. I suggest that there are some contexts (namely, relations of vicarious responsibility and vicarious agency) where having considerable control over another human being is morally unproblematic, if not desirable. If this is the case, control over advanced humanoid robots could well be another instance of morally unproblematic control. Alternatively, what makes it a problematic instance remains an open question insofar as the representation of control over another human being is not sufficient for wrongness, since even considerable control over another human being is often not wrong.