Or, more succinctly, if we don't solve coherent extraoplated volition, we are screwed regardless of whether Kruel or Yudkowski is right about the specific threat of unfriendly AI.