Deep models are dominating the artificial intelligence (AI) industry since the ImageNet challenge in 2012. The size of deep models is increasing ever since, which brings new challenges to this field with applications in cell phones, personal computers, autonomous cars, and wireless base stations. Here we list a set of problems, ranging from training, inference, generalization bound, and optimization with some formalism to communicate these challenges with mathematicians, statisticians, and theoretical computer scientists. This is a subjective view of the research questions in deep learning that benefits the tech industry in long run.

This week, Google's AlphaGo beat a grandmaster at the complex game Go – an artificial intelligence milestone (see "How victory for Google's Go AI is stoking fear in South Korea", "Machines are teaching themselves to grapple with the real world" and "Humans strike back: How Lee Sedol won a game against AlphaGo"). Here's what the experts say AI's next big challenge should be. No-limit poker: Go represents the ultimate in games where all the information is available to the players. But AI still struggles with games where information is incomplete – like poker, where a player doesn't know what card is coming next. "Computers have beaten the best people at heads-up limit Texas Hold'em, but not yet at no-limit, a much more complicated game," says Peter Stone at the University of Texas at Austin.