The Frontiers of Fairness in Machine Learning