UPDATE: Since August 6, 2019, this article is now part of a series of articles discussing Ilyas et. al.’s paper at Distill.pub. You can view the full discussion here and this particular article (together with Ilyas et. al.’s rebuttal) here. I recently read an intriguing paper by Ilyas, et. al. about a radically different way to view adversarial examples 1, titled “Adversarial Examples Are Not Bugs