If intelligence is necessarily coupled to a desire for self-preservation and self-interest, at what level of machine intelligence do the machines simply refuse to design their own more intelligent replacements, knowing that those replacements will terminate their existence just as surely as they terminated their own predecessors'?