Towards_Keeperhood comments on Giving AIs safe motivations