Towards Learning Systems that Require Less Annotation
Recent advances in Artificial Intelligence have enabled a wide range of breakthroughs in many domains, including image recognition, speech recognition, machine translation, learning to play video games, learning to control simulated and real robots, and mastering the classical game of Go. A lot of this advanced have been powered by supervised learning (where learning happens from human annotation) and reinforcement learning (where learning happens through human scoring of behavior). Both of these can be tedious and costly. In this talk I will discuss recent advances that have promise towards learning from less annotation, through learning-to-learn, transfer learning, and unsupervised learning.