Do AIs inherently protect their own interests and sequester resources as subgoals toward any goal they are given?
@mboudry shows how this "instrumental convergence" doomer dogma is dubious anthropomorphism. | Why HAL 9000 Feared Death (and Real AIs Don’t)