Quantum machine learning has garnered an exceptional amount of attention over the last five years: the promise of exponential speedup in the most hyped application area of computer science is proving irresistible to physicists, but less appealing to the machine learning community. We look at why this asymmetric relationship arose and contemplate whether a quantum AI winter is coming. As quantum technologies are maturing, but remain imperfect, one promising avenue forward is to develop practical learning protocols on contemporary quantum computers. On a more theoretical level, the nature of learning as bound by quantum physics is barely understood. These directions offer a rich, largely unexplored area at the intersection of many fields of quantum physics and computer science that welcomes newcomers.