romeostevensit comments on Disentangling arguments for the importance of AI safety