What do you consider to be some of the major open problems in machine learning and its associated fields? Both practical and theoretical problems are welcome, but for the sake of conciseness leave out vague problems such as general intelligence, etc.
[link][12 comments]