I have yet to write down a o longer post on this, but honestly I think AI have little chance to operate on purely logical terms. This thinking about paradoxes comes from the time when people approached AI with Prolog etc. They didn't get really far.