Limitations of Deep Neural Networks: a discussion of G. Marcus' critical appraisal of deep learning

Deep neural networks have triggered a revolution in artificial intelligence,\nhaving been applied with great results in medical imaging, semi-autonomous\nvehicles, ecommerce, genetics research, speech recognition, particle physics,\nexperimental art, economic forecasting, environmental science, industrial\nmanufacturing, and a wide variety of applications in nearly every field. This\nsudden success, though, may have intoxicated the research community and blinded\nthem to the potential pitfalls of assigning deep learning a higher status than\nwarranted. Also, research directed at alleviating the weaknesses of deep\nlearning may seem less attractive to scientists and engineers, who focus on the\nlow-hanging fruit of finding more and more applications for deep learning\nmodels, thus letting short-term benefits hamper long-term scientific progress.\nGary Marcus wrote a paper entitled Deep Learning: A Critical Appraisal, and\nhere we discuss Marcus' core ideas, as well as attempt a general assessment of\nthe subject. This study examines some of the limitations of deep neural\nnetworks, with the intention of pointing towards potential paths for future\nresearch, and of clearing up some metaphysical misconceptions, held by numerous\nresearchers, that may misdirect them.\n