An interesting article on deep learning and symbolic logic:
http://nautil.us/issue/67/reboot/why-robot-brains-need-symbols