Exactly. And to add to what you say, adversarial problems in data science are much harder than the regular ones which data scientists deal with and the hubris that I often see with data scientists won't cut it here. You need a different mindset to tackle these problems and they must be handled incrementally with the real world in the loop.