I think a key differentiation between ML and more common statistics is that ML is designed to improve itself based on data. Statistical methods don't do that.

I think a key differentiation between ML and more common statistics is that ML is designed to improve itself based on data. Statistical methods don't do that. So maybe they're trying to do very similar things in a lot of cases, but self improvement is a major differentiator.