The report discusses the successes of reinforcement learning (RL, which can be applied both to DL and other paradigms); graphical and Bayes models, especially with probabilistic programming languages; generative models that may allow training with much smaller data sets; and other kinds of probabilistic models such as those that have shown remarkable successes in question answering (e.g., IBM’s Watson), machine translation, and robotics. While DL will certainly affect all of these fields, it is not the only or final answer. More likely, DL will become an essential building block in more complicated, hybrid AI architectures.