Music generation based on surprise optimization

Use neuroscience and deep learning to generate music that pushes the right buttons in our brains.

Start date: October 2016
Category: Fundamental Research
Contact point: Francois Chollet - fchollet@google.com

Problem description

It has been suggested that part of our appreciation for music comes from dopamine responses triggered by building-up then breaking sensory expectations, i.e. that music manufactures continuous surprise during listening, an effect that appears sustainable even after repeated exposure to a given musical pattern.

At the same time, techniques for music generation have recently made great progress, with WaveNet [1] in particular appearing to be a promising avenue.

We propose to combine such music generation techniques with a “critic” component that would analyze the degree of sustained surprise (i.e. enjoyment) that a musical track generates during listening. This critic may be a model attempting to predict the short-term evolution of the musical track; the error signal of such a model would provide visibility into listening surprise. This component would guide the music generation process so as to produce tracks that not only sound “like music” but also sound enjoyable.

Datasets

References