Music + Machine Learning
Computers have helped shape the music industry in different ways since the 1950s (Braguinski, 2022). American researcher Curtis Roads published an article in 1985 outlining the presence of artificial intelligence in different areas of music such as composition, performance, music theory and digital sound processing (Roads, 1985). This is not surprising as music is in essence mathematical and therefore constitutes a ground for the use of machines to produce artificial music (Ruiz et al., 2021, p. 88). There has certainly been a tremendous technical advancement since the 20th century and AI is now capable of mastering far more in the music sphere.
ML's impact on the music industry
Machine learning over the years has become part of many processes involved in the music industry. Algorithms have been developed by programmers to understand harmonic and melodic patterns of different music genres, generate new compositions by using text-based transcriptions of music, classify music by genre and style using deep neural networks techniques, recommend songs and artists to streaming services users and much more (Ben-Tal et al., 2021; Fan, 2022; Hodgson, 2021; Prabhu et al., 2018; Tigre Moura & Maw, 2021). Computers and machine learning do differ when compared to the human ability as their techniques to generate outputs are faster, cheaper and more scalable (Tigre Moura & Maw, 2021), which is certainly captivating to music producers.
Software recommendations play a big part in the rise of popular music today. In fact, scholars have coined the term “algorithmic culture” to define a cultural commodity (like music) that is governed by software recommendations (Galloway, 2006; Werner, 2020). Werner adds that “streaming services are central players in algorithmic culture, where cultural value is determined by processes partly human, partly machine (2020). In 2020, music subscription and streaming revenues reached 10.07 billion, making up the vast majority of revenues for the entire music industry (Götting, 2022b). Although these platforms provide access to an immense amount of content, it remains that the discoverability of artists and songs is heavily reliant on if they are recommended by the systems or not. Thus, as Werner explains, “algorithmic culture reduces decisions on taste to a few actors defining what is good and for whom, constructing social groups and cultural value in the process” (2020). The development of streaming services and their recommender systems has surely had the biggest impact on music. The interview with scholar Jada Watson expands on the subject.
ML's on copyright for generated music
With the rise of music created with AI, the question of copyrights and appropriate compensation follows just like any art domain where commercialization is present. In his article, Drott elaborates on the debate surrounding copyright for music created with machine learning techniques. Firstly, it must be understood that to be copyrighted a work must cross the originality threshold. For example, in the United States, in order to qualify for copyright protection, a work must satisfy the originality requirement, which is that it “must have ‘at least a modicum’ of creativity, and it must be the independent creation of its author” (University of Michigan Library, 2021, para. 3). According to Drott (2021), the possibility of passing the originality requirement is much higher now that neural networks, evolutionary computing, and other techniques are better able to extrapolate from inputs to generate unpredictable outputs. Taking this into account, the debate is not concerning the generated music itself but who is credited for the work. The challenge here is the default assumption that copyrighted work is attributed to ‘human’ creators (Drott, 2021). If machines are not traditionally eligible for authorship according to the legal framework*, then an appropriate human candidate must be identified. Drott names a few examples, such as a machine’s programmers, its proprietors, or the end user for whom a work is created. As of now, no right answer has been provided. Surely this will change as AI and ML generated works become even more popular and commercialized to the point where it is its own industry.
* In December 2021, the Canadian Intellectual Property Office registered its first copyright registration with an AI author (Medeiros et al., 2022). It is for a painting by co-authors Mr. Ankit Sahni and the RAGHAV Artificial Intelligence Painting App.