Third, and most importantly, recent advances in machine learning and information retrieval let us convert previously useless data into knowledge. For example, Netflix’s very profitable
streaming service offers a large but rather low quality and slightly outdated movie and TV show collection. However, its movie recommendation system helps users find the few movies in this huge collection that they might actually enjoy watching. Key to this system is a rather new machinelearning technique called alternating least squares, a form of collaborative filtering that lets Netflix compare users with each other at scale to make individual recommendations.