Towards_Keeperhood comments on Can we safely automate alignment research?