As AI advancement opens new possibilities for music creation, it also introduces challenges regarding intellectual property and compensation. Systems like ChatGPT4 can transform existing songs into different genres remarkably well. When asked how it learned to do this, one system replied it had studied various forms of literature and music extensively.
Three systems were able to create a song in different styles, while another refused due to copyright. They show skill in composition, adapting melody, harmony, and lyrics skillfully according to genre conventions. However, questions arise around who provides the training data driving such abilities, and whether they should receive attribution or payment for derivative works.
Music AI training requires large datasets spanning notation, lyrics, theory, and recordings. Major companies have acquired these rights from record labels and publishers. But do these blanket deals sufficiently consider songwriters as primary content creators? Systems demonstrated reworking famous songs, but did songwriters approve this use? As blackstone invested billions in certain catalogs, such applications could devalue those assets without consent.
Royalty structures also need examination. Early mobile tone deals fairly split revenues between labels and writers. But subsequent platforms saw labels retain more. Now streaming pays labels and performers well while shortchanging writers. If AI expands monetization further, all stakeholders deserve a seat at the table to negotiate fair terms.
To address these issues, industry leaders must have open discussions. Questions around attribution, approval of training data use and royalty distribution require answers. Collective management organizations could unify support for writers globally. As equity funds invest heavily in their interests, now is the chance to build cooperative solutions before revenue streams solidify without them.
