Pattern comments on What other problems would a successful AI safety algorithm solve?