In this talk I will describe the developments in music discovery technologies over the last 20 years. This journey will depart from the question of how music can be algorithmically modeled for content-based search and recommendation. Then I will detail how additional data sources representing diferent types of user interactions have gained importance for these tasks. Ultimately, we will arrive at a discussion of what the objectives of recommender systems are and-most importantly-could and should be in a world where music is virtually freely available everywhere and at all times. This will bring us to the
W open question of fairness in music recommendation and how future technology could contribute to this.