Which is exactly why we made no progress in the theory of Machine Learning in the past decade. With massive industrial trial and error we identified things that

Which is exactly why we made no progress in the theory of Machine Learning in the past decade. With massive industrial trial and error we identified things that work for specific cases, but we have no clue how to generalize.