Comment by malshe

Thanks for finding this article. I agree, these are important developments in particular because so many econometric models are now using machine learning without any distributional assumptions. Using correlation coefficients based on linearity is grossly insufficient.