Gary Marcus and Ernest Davis argue deep learning alone can’t deliver trustworthy AI — and lay out what hybrid, common-sense-grounded systems would need to look like instead.