[version_five]

> Chances are a lot of machine learning implementations can be replaced with just "boring" statistical models and achieve more power.

This is of course true, but the innovation in ML is not that a neural network (or whatever model) is equivalent to something else, it's finding the weights in the first place.