David Scott Krueger (formerly: capybaralet) comments on [AN #57] Why we should focus on robustness in AI safety, and the analogous problems in programming