Gunnar_Zarncke comments on why assume AGIs will optimize for fixed goals?