"In my opinion the hard problem in machine learning has moved from prediction to characterization."
Is this because prediction has been "solved" or because characterization is currently more commercially useful?
Is this because prediction has been "solved" or because characterization is currently more commercially useful?