Breaking away from the designs of classical and normative structure within sonic and visual art mediums, generative sequencing has grown in discussion since its gain in popularity amongst composers in the 1960’s. Along its progression, the questions pertaining to copyright law have continually been addressed with no definitive answer. Ownership has reached a new complexity now that authorship exceeds the likeness of human intent. Brian Eno, an early pioneer of the generative sound composition, described the technique’s gain in popularity as a decrease in human interest over the replica. The idea in some ways acts as the defiant next-step from the processes of tape-looping in sound recording. Eno’s 1978 breakthrough project, Ambient 1: Music for Airports, utilized a generative composition which ensured that the looped sound recordings were mathematically impossible to sync up in any repetition. Eno stated in regard to this project that, “the thing about pieces like this of course is that they are actually of almost infinite length if the numbers involved are complex enough. They simply don’t ever reconfigure in the same way again. This is music for free in a sense. The considerations that are important, then, become questions of how the system works and most important of all what you feed into the system.”
This generative composition process highlights a new relationship born between the artist/composer and the work before them. These works haven’t been heard or seen before. They are “put into motion” by the artist and then stepped-back and looked-upon. In copyright law this can be a perplexing issue. Jessica Fjeld and Mason Kortz of Havard Law’s Jolt Digest have written about elements of generative composition in their 2017 commentary titled, “A Legal Anatomy of AI-Generated Art: Part I.” In their piece, Fjeld and Kortz break the workings of true generative software into four common traits:
(1) Input; the endless variety of preexisting works that may be fed into the system.
(2) The Learning Algorithm; the system of computer-learning that operates off any mixture of secondary-software and/or codes.
(3) The Trained Algorithm; described by the authors as “the information that the Learning Algorithm has generated from its operation on the Inputs, along with instructions for turning that information back into a work.”
(4) Output; the perceived works produced by running the Trained Algorithm from some form of a starting point otherwise referred to as a “seed.”
Section 1301(b)(1) of the U.S. Copyright Law deems that a design is “original” when it is “the result of the designer’s creative endeavor that provides a distinguishable variation over prior work pertaining to similar articles which is more than merely trivial and has not been copied from another source.” Here, the questions pile up, starting with the concept of royalties. There are many ingredients that contribute to these works, from the artists of Input works to the software developers to the very software itself. Where the credit is amounted for each component is complicated to say the least. Eno, a notorious collaborator himself, suggested that this would lead to an increase in joint works in the future. Fjeld and Kortz asked that if joint ownership pertains to the Trained Algorithm, then does the Trained Algorithm own the outputted work? In all, while the human race’s fascination over replicas finally dies down because of technology, the technological drive for originality has already begun writing it’s own chapter.