About the project
M&M is both an artistic and a research exercise. We want to know more about the legal implications that using copyrighted music to “teach” AI machines to create music and the preparedness of the UK legal framework to accommodate AI-generated works with all its possible nuances (e.g., pieces created only by the machine or in collaboration with a person). In a nutshell, this work aims to test whether the music industry’s current legal and economic frameworks are prepared for this emerging phenomenon and to offer recommendations for the future. This is the research bit …
In order to make this possible, our team decided to design a potential future scenario (perhaps even already present): we asked a composer to create new music with two AI music machines (using the same technology used in the viral OpenAI ChatGPT) created for this project. One of the machines was trained with (out-of-copyright) Romantic music and another with (in-copyright) Pop and Rock music. By trained we mean that the model “heard” a lot of music pieces and asked to infer the rules that allow it to create music in the style of the music it heard (in essence not very different from what humans do). By following these steps, we were able to know exactly which songs the machine used to learn how to compose and therefore we have a very transparent and clear platform to explore the legal implications that arise from the whole process (something that would not be possible if we used an already available commercial AI music machine).
The outcome of this exercise are four new pieces that will be performed in live concert at the Tung Auditorium (University of Liverpool): two inspired by Schubert’s Die Forelle and the other two by a new piece commissioned to the Liverpool-based band Steeling Sheep: Just Data.
The research
The research behind this project intends to answer a set of key questions:
- Is there currently legal capacity in the UK to offer copyright protection to AI-generated musical works?
- Who is considered the author of an AI-generated work (and therefore the first owner of the copyright)?
- Is there a framework in place in the UK to remunerate the copyright owner of AI-generated works?
To answer them, we will to attempt registering the four new AI-generated songs with the Performing Rights Society (PRS) with the AI system as the main author/composer. This process aims to test:
- whether a machine can be considered an author and therefore owner of the copyright;
- whether AI-generated music can be exploited to generate income; and
- how and to whom royalties can be paid.
The concert
The concert will start with a short introduction to AI, how AI machines are being used to create new art (including music) and an overview of how we used AI to create a new set of songs over the past few months.
There will be five world premieres: four new songs composed with two AI machine developed by Rui Guo in collaboration with the Applied Music Research Lab and a song commissioned to the Liverpool-based band Stealing Sheep. The lyrics for all songs were created with the now infamous ChatGTP.
After the concert, Rachel Drury (Ph.D. research at the University of Liverpool) will talk about the implications of AI to copyright law and the research behind this project.
At the end, the public will have the chance to ask questions about the concert and the project.
The concert of FREE (but you will need to book tickets in advance).
