Fascination, Challenges and Constraints
In recent years, there has been a great deal of academic interest in applying Deep Learning to creative tasks such as generating texts, images or music, with fascinating results. The research interest in these generative models is based on the assumption that producing new, similar data is only possible by learning some essential understanding of the nature of the input data.
Technically speaking, Deep Learning models can only learn the statistics of the data. Thus, they often can learn relationships in the data that human observers have not been aware of, and can therefore serve as a new source of inspiration for human creativity.
This workshop focuses on current technical approaches for automatic music generation. We will also discuss questions like: What makes musical data so special? What are the structures that only occur in musical data? What can artists, scientists or music enthusiasts expect from these models? To what extent do listeners accept music composed by AI?
Biography:
Ali Nikrang is a key researcher & artist at Ars Electronica Futurelab. His background is in both computer science (Johannes Kepler University in Linz) and classical music (composition and piano performance, Mozarteum in Salzburg). His research interests include creative and interactive applications of artificial intelligence systems.
https://ars.electronica.art/futurelab/team/ali-nikrang/
