Using a machine-learning system known as a deep neural network, MIT researchers have created the first model that can replicate human performance on auditory tasks such as identifying a musical gen