Gunnar_Zarncke comments on Giving AIs safe motivations