Vanessa Kosoy comments on [Link] Adversarially trained neural representations may already be as robust as corresponding biological neural representations