Towards explainable A.I.?

Does the advent of machine learning mean the classic methodology of hypothesise, predict and test has had its day? I loved reading this interesting article in The Guardian written by Laura Spinney.