Are There Deep Reasons Underlying the Pathologies of Today's Deep Learning Algorithms?