[rob-p]

I think that one of the problems I have with the description in the post is that it draws a line where, in reality, no real line (or at the very least, only a blurry mess) exists. For example, where do hierarchical models and Bayesian nonparametric models fit? Where does model selection fit? These notions have existed in statistics for some time. I'm not one of these particularly dogmatic people who believe that "it's all just statistics" or, conversely "machine learning is entirely new / different". In fact, I think it's the deep (and continuing) connection between these fields that make them both so interesting and powerful. However, I do tend to agree that the type of hype used in this post massively oversells ML while simultaneously underselling Stats, based, partly, on the false dichotomy drawn between them.