Chris Leong comments on Paths and waystations in AI safety