Very nice and quite interesting work, and also an excellent demonstration of https://xkcd.com/1838/ . I’ve seen the same general approach from baby grad students several times in my time working in an ML lab, and honestly the cool thing about ML is that you can generally make this approach pretty well.
Also hilarious, for occasional nuggets of wisdom such as “if you are considering the use of a library function with tricks in the name, your life may be leading you to a place that you don’t really want to go. “
Cool, I wonder if a similar technique would help to find hash collisions. Would that be as useful as a rainbow table?