Questions around the validity, desire for, and uses of AI generated music alway bring me back to the same starting baseline. What do we really want from AI at all? I tend to start from the viewpoint of AI as a tool - one that continues to develop to help us (the humans) do more tasks, better (where better, especially in the arts, is a subjective experience. Is better, faster? More complex? More technical?).

I probed ChatGPT with the simple question of whether or not AI is a tool, to which it said, “yes, AI can be considered a tool. AI refers to the development of computer systems that can perform tasks that typically require human intelligence, such as visual perception, speech recognition, decision-making, and language translation. AI technologies can be used as tools to automate and streamline processes, solve problems, and enhance decision-making, among other things. However, AI can also be seen as more than just a tool, as it has the potential to transform entire industries and impact society in profound ways.” Could it be that our skepticism or fear of the onset of AI generated music is focused around the last part of ChatGPT’s response, in its potential to profoundly transform and impact society?

I push for a values-driven framework in evaluating whether or not this is a good idea. Specifically in asking the question, “how and why is this valued?”, or, in other words, “who cares?” Art has always been a subject of debate as a sheer product of its subjectivity. The subjectivity is also precisely why a system like chatGPT is designed to respond “I don’t have an opinion” when asked a subjective question, because it really just takes one person to believe something for it to be true in their world.

We can argue that the songs of birds, the croaking of frogs, the smooth falling rain are music - to some people. The 11,113,526 people that listen to When it Rains It Pours, a 2-minute Spotify track of rain in the countryside, may certainly say that they have value in this naturally generated music. If rigorous research were done into the myriad of musical phenomena in the natural world that suddenly ushered in a renaissance of organic sounds in music, would we ask the question - what do we really want from the natural music generation? This is to say, nature sounds, by themselves or within other compositions, is not music because nature decides that it is, in the same way that AI generated music or art isn’t art just because the AI says that it is - we have to decide as the humans that are experiencing it, as is in our human nature, whether or not something is art - and from there, we can decide how to value something.

In this sense, nature can be thought of as a tool for humans to express and interpret creativity and art. I’d argue that developments like MusicLM is the continuance in development of new tools for the ongoing and very human exercise of creating new ways to experience sounds that make us feel, that make us question, and help us to express. And that’s not to take away from everything that a tool can be to someone, as if tools are only a means to an end. If a piano is ‘just a tool’ for making music, then we can’t undermine the idea that someone may have an immeasurable emotional connection to the instrument, and feel connected with it in a way that transcends something as a tool, as just a means to an end. In that vein, AI generated music can very well have the capacity to elicit connections just as powerful as the music-making mediums available to us today and throughout history.

But where do we draw the line? Logically, a piano can’t in and of itself undermine someone’s own sense of creativity, or a society or culture framework for valuing creativity and art - whereas AI generated music pushes the boundaries of the social aspects around music: authorship, ownership, bias, accessibility, and authenticity. Because we live in a society, these are questions that should be asked, because the answers can have profound implications for how we apply the social benefits that music can bring.

How will that valuing of music that was or wasn’t generated by AI be manifested? Will artists on streaming platforms need to adopt a version of a blue twitter check mark to verify that their music was original and not created in any form by AI? Or do people even want to know to what extent a piece of music or a synth pack or something was generated by AI? Does it make a difference if something was 30% AI-generated versus 90%? Does having that checkmark of humanity (a Turing Tick?) imply that we can be assured that the music we listen to was created by another being that has the genuine capacity to feel emotion, in the way we do - regardless of if those emotions resonate with us as well as something an AI could create? In the same way that many people don’t particularly care what kind of synth a groove was created on, just that they like the way it makes them feel, will people not care that something was generated by an AI, just that a human helped determine that it was the right music to apply to whatever context we’re experiencing it in?

I’m excited about what AI music generation can bring, because I think it serves as the next healthy challenge, transcending one artist challenging another to push the limits of creativity, but rather AI challenging humans to push what creativity really means - so that ultimately, humans and AI’s sense and capacity for art both continue to evolve and grow.

Here’s the lyrics to a song about oranges and sour patch kids, the two foods on my desk while writing this, generated by ChatGPT: