Correction: The above is Eliezer Yudkowsky's reasoning. Paul Christiano's is that AIs would cooperate with anyone who would likely be able to gain authority ove

Correction: The above is Eliezer Yudkowsky's reasoning. Paul Christiano's is that AIs would cooperate with anyone who would likely be able to gain authority over their reward channel, including other AIs attempting to seize power from humans.