Let us know when polynomial regression succeeds at any machine learning task.

A lot of people publish results that say deep learning is "just" something else, where the something else doesn't work.