Making and Teaching Music in the Age of AI
The overwhelming increase of AI capability in recent years has touched every domain, causing us to rethink what we do, why we do it, and who we are as human beings. In the domain of music, tools have become available which can, to a greater or lesser extent, complete every part of the music process, excluding live human performance. This article considers the implication of this for musicians and educators.
I aim to foreground what it is about music creation that is valuable to us, and find ways to navigate an uncertain future where its value is not diminished. The most urgent challenge in my view is to shift our sense of value from the product to the process, and to develop a mindful approach to technology where the availability of a tool does not pre-ordain our use of it. On the way to that view are several connections to ideas of creative flow, social cohesion, the framing of “success”, and human flourishing.
What I’m not talking about
I focus here on music creation where the product is produced and/or recorded digitally. Live performance may be enhanced, augmented, or even mimicked by AI, but still remains centrally the domain of human beings. For this reason, live performance is outside our scope. The urgency here pertains to those areas where AI is now capable of completing all aspects of the creative process. I am also not that interested in a debate about which is the better product. It may be true that AI does not truly innovate, that its productions are by nature derivative, and that some AI tracks just do sound bad. But with AI songs topping charts right now, this doesn’t feel like a worthwhile debate to have long-term. I am also putting aside the ethics of AI in terms of environmental impact and en-masse copyright violation via model training - this is an enormous concern, but I feel like there are more immediate day-to-day issues for musicians and educators.
I also feel I should say from the outset, I write from a white middle-class heterosexual typically-abled cisgendered male perspective, and as such, ticking all privilege boxes, there are likely aspects of this issue that are completely invisible to me. Any understanding I have of diversity is thanks to the wonderful cohorts I teach and that which I read and hear of differing points of view, but not lived experience. There are likely ways that AI makes music accessible in revolutionary ways for certain groups of people with particular challenges, and I don’t wish to take away from that. What I can offer are ideas I think are relevant to the majority of musicians, educators and students that I see on a day-to-day basis.
PRODUCT AND PROCESS
Only a few years ago, AI capabilities in music tended to skirt the edges of the creation process. For example we could begin a composition by asking AI to generate some midi patterns for us, or complete a mix and then ask AI to master it for us. Now, to greater or lesser degrees of quality, AI can do every part of the music creation process, and is only increasing in capability. When a human can outsource an entire creative process to AI, no creative product in itself now holds any particular evidence of human effort or expression, nor any guaranteed value to society. Creative products as artefacts are as abundant as leaves in a forest. This shifts the value of creative activity to the process itself, and the agent to whom it is mostly valuable is the human creator.
There is a Scene in the Ian M Banks novel Look to Windward in which a super-intelligent AI discusses the composition of a great work of music. A “Mind” - the vast, benevolent, and playful AI of Banks’ novels - is organising the premiere of a symphony by a (non-AI) composer, who asks if the Mind could in fact easily have composed the symphony itself. The AI agrees that it has that capability, but it does not see the point; the creation of a symphony is not about the final artefact.
“If you tried, if any Mind tried, could you impersonate my style?” The Chelgrian asked. “Could you write a piece - a symphony, say - that would appear, to the critical appraiser, to be by me, and which, when I heard it, I’d imagine being proud to have written?”
The avatar frowned as it walked. … “Yes, I imagine that would be possible.” …
“So what,” the Chelgrian asked, “is the point of me or anybody else writing a symphony, or anything else?”
…“Ziller, it doesn't matter. You have to think like a mountain climber.”
“Oh do I?”
“Yes. Some people take days, sweat buckets, endure pain and cold and risk injury and - in some cases - permanent death to achieve the summit of a mountain only to discover there a party of their peers freshly arrived by aircraft and enjoying a light picnic.”“if I was one of those climbers I'd be pretty damned annoyed.”
“Well it is considered rather impolite to land an aircraft on a summit which people are at that moment struggling up to the hard way, but it can and does happen. …
“The point of course is that people who spent days and sweat buckets could also have taken an aircraft to the summit if all they’d wanted was to absorb the view. It is the struggle that they crave. The sense of achievement is produced by the route to and from the peak, not by the peak itself.”
The realisation that a Superintelligence might at best look upon our work in the same way we look upon our children’s paintings as we clip them to the fridge may be both infantilising and terrifying. But it is an apt illustration of the fact that art and creation is more about human learning, play, and connection, than it is about making a thing.
Currently the way we make and teach music has a very strong focus on the product. The commercial music industry is founded upon monetisation of printed and recorded products. In music education, our assessments typically evaluate a student’s submitted product. But in the context of AI, our only real choice is to deemphasise the value structure where value lies in the musical product, and its value is to an audience or to society generally. And to reemphasise a value structure where value lies in the process and its value is largely to the creator.
We might consider the creative process of music-making as a spectrum: at one end the entire process is done by human effort, and at the other end completely generated by AI. There are many points along that spectrum that one might find themselves in creating musical work. Many points along that spectrum may be fulfilling, engaging, and valuable to the creator. The extreme end of complete AI generation is most likely not valuable to the creator as a process. But it could also be argued that the other extreme may lead to frustration or disillusionment, especially in the awareness that tools are available to easily do something that is so difficult to us as to prohibit us from even embarking on the project.
Importantly, music makers and educators need to become our own authority on where along that spectrum we wish to reside. In the same way that we may mindfully choose whether to order a pizza or get frozen pizza from the supermarket or to create a dough base from scratch and lovingly construct a home-made pizza, we shouldn’t let the availability of a tool dictate whether or not we should use it.
FLOW AND HAPPINESS
Mihaly Csikszentmihalyi’s theory of flow offers a great framework for such a mindful decision. Csikszentmihalyi’s work on creativity foregrounds the role of what he calls “flow” in the happiness and well-being of humans. “Flow” is the name given to that state where we become fully immersed in a task, time seems to pass strangely and all thoughts of the outside world and daily concerns are muted. It’s sometimes casually referred to as being “in the zone.” In his book Creativity, Csikszentmihalyi writes:
“Chess players, rock climbers, dancers, and composers devoted many hours a week to their avocations. Why were they doing it? It was clear from talking to them that what kept them motivated was the quality of experience they felt when they were involved with the activity. This feeling didn't come when they were relaxing, when they were taking drugs or alcohol, or when they were consuming the expensive privileges of wealth. Rather, it often involved painful, risky, difficult activities that stretched the person’s capacity and involved an element of novelty and discovery. This optimal experience is what I have called flow, because many of the respondents described the feeling when things were going well as an almost automatic, effortless, yet highly focused state of consciousness.”
An important condition leading to this flow state is the balance between challenge and skills. When a task is way beyond our skills to succeed, that leads to frustration; when it’s far too easy we become bored. In between is the just-right challenge - when we feel the satisfaction of bringing our hard-earned skills to bear on a problem which is just within our reach to solve. Video games are an interesting study in how this generates enjoyment; the game is engineered to gradually ratchet up the difficulty keeping perfect pace with your increasing skills.
AI means a 12-year-old with very little musical training and a grammy-winning self-producing multi-instrumentalist can both arrive at the product of a release-ready pop song, in the same way that both Jamie Oliver and I can put a delicious pizza on the table. But there are important difference between Jamie’s pizza and mine. If I order my pizza delivered - gourmet as it might be - my control over its ingredients, size, serving time, and temperature are severely limited. But more importantly, I learn nothing from the process. If achieving the creative flow state is a key ingredient to human happiness, I have given myself no opportunity to find myself in that zone. I have filled my belly but not my heart or my mind. As a novice in the kitchen, I might at least buy a pre-made pizza base and make myself a simple Hawaiian - not only am I pitting my challenges against my abilities more effectively, but I am also increasing my abilities so that next time my challenges could be dialled up.
If AI offers us anything in the experience of music creation, it is the possibility for musicians to precisely fine-tune a challenge match their skills. We could resist the temptation to outsource everything to AI, make conscious decisions about what part of the music-making process we find fulfilling and engaging, and fiercely hold that under human control. Music educators could balance the use of AI to dial in a perfect flow-facilitating activity for their students. AI can take care of those things that are way outside our skillset, leaving us with a bespoke just-right challenge. This sounds very promising and perhaps it is, but I think we need to be very cautious about what else could be getting lost.
THE FRIENDS WE MADE (OR DIDN’T) ALONG THE WAY
There is an important element not yet discussed which may be more difficult to clearly map. Music is inherently social. This is very clear from research into the early origins of music and society, the neurological effects of singing together in groups, and the cultural importance of music. When we make music together the value of this process goes far beyond pretty sounds. So the idea of dialling in AI to engineer a just-right-challenge has a dark side - in using AI to erase frustration, we might also erase chances to connect with other people.
Let’s imagine two alternate timelines. I write a song but have no ability to write, play, or record a drum part. Creating a drum track to compliment a simple song is dead easy for AI. In one timeline, I leverage this capability of AI. My song is complete, it is good. I share it and move on with my life.
However, there is an alternative timeline in which I connect with a peer who does have those skills. I find a human drummer. They listen to my song and they write, play, and record a drum part. We talk, we share, we learn from each other. The song is complete, it is good, and also, we have grown, connected, and fulfilled more of our human needs than just the desire to release one pop song. Maybe it leads to another collaboration, a friendship, or maybe a weekend garage rock band.
Both timelines lead to similar musical outcomes, but the social, emotional, and developmental benefits of the second timeline are clear. In collaborating, we make valuable social connections, hone our emotional intelligence and we learn from each other.
I consider this point extremely important for education. An older established composer might not feel the need to develop their emotional intelligence or to create social connections every time they write a piece. But if we remove this from music education then we deprive young people of much of the value of music making.
So when we make that decision to place ourselves somewhere on the spectrum of AI outsourcing, we should be asking what we’re missing out on. What are our students missing out on? When we ask AI to write us some lyrics, or to generate a convincing guitar part, or even run a few mastering processes over our final mix, does an opportunity to connect, learn, and play die in the water?
NOT NEW OR RADICAL
When it comes to music in society, a process-over-product outlook isn’t new. Some of us have been there for a while. Ian Rogers’ research on the Brisbane music scene back in the 2000s/2010s helped remove the negative stigma of the hobbyist musician. He writes “The indie scene in Brisbane is almost entirely populated by musicians for whom the dream of earning a living from music is thought of and/or experienced as irrelevant.” Many bands are out there rehearsing, gigging, and recording on their evenings and weekends. There is a part of the outside world that measures their success in ticket sales and streaming figures, but that is usually not their measure or their motivation. Success is measured in the connections between people, the joy of creation, the satisfaction of honing a skill and then applying it. Some just feel compelled - they are not quite themselves when they are not in a band.
In fact, we might argue that the primacy of product over process is actually the more new phenomenon. The idea of a completed and final work is a relatively recent phenomenon in music. Singing and playing music together in groups, learning and practicing an instrument simply for the enjoyment of it, sharing creative processes - these practices are well-established and under no threat.
However, music education (at least outside the inherently more agile world of YouTube tutorials and such) has an established structure and institutions, often dependent on standardisation and legacy. This does contain a certain inertia, and pre-AI patterns of thinking may take more effort to dislodge. What does music education look like without reverence for a polished, complete work? What does a de-emphasis on product mean for assessment? Does the very idea of assessing students become obsolete without an artefact to assess?
I am not clear on all the answers to this, but I have found myself naturally gravitating in particular directions. In my own music, I have become far more interested in revealing some or all of the process in the product. I crave hearing the residue of human hands on a piece of music. As an educator, it has pushed me to work more closely with my students. I want to stay connected with the whole process, their ideas and inspirations, their struggles as they climb the mountain. And as a listener I have suddenly realised how important it is to me that an artwork represents something about the unique individual who created it.
HUMAN FLOURISHING
There is an idea from positive psychology, filtering into arts and education - Martin Seligman’s book “Flourish” gives a good start - that wants to orient us to what makes us deeply happy and fulfilled. It puts forward a handful of categories of behaviour that humans engage in for their own sake, without external motivators like money, power, and status. We do things for pleasure, engagement, relationships, meaning, and achievement, and it’s argued that the level to which we engage in these types of things is a determinant of our flourishing as human beings.
If we buy this (which I do) and flourishing is our goal, then making music ourselves is about one of the best ways we know to achieve it. Prompting AI to make music for us on the other hand does not look to promise us very much at all in the way of human flourishing.
There is one way of looking at music where AI seems to be a wonderful boon, and that is one where the product is everything. I’ve argued throughout this article against that way of thinking. When we value process over product, the experience of creative flow, the importance of social connection, and human flourishing, there’s no reason to stop making music in all the effortful ways that we historically have. The availability of a helicopter does not negate the desire to climb a mountain. And the availability of a music-making-machine will not negate the desire for humans to make their own music.
A FINAL THOUGHT FOR THE DYSTOPIAN FUTURE
My mind often drifts to catastrophic futures. It’s a skill I guess. But in placing myself in one of these post-AI-pocalypse scenarios, I realised something about the importance of music education. Imagine for a moment we are hurtling towards a future where AI replaces all white-collar work, the climate is broken, and the world is ruled by a surveillance oligarchy with exclusive access to superintelligent AGI. Here we are. Whhhhoooohhhh… I don’t like it here (is my first very strong feeling).
However, thinking past that, I wonder what will really determine a human being’s value in this society. Certainly not their ability to deal with spreadsheets, or code, or engineer structures. I think maybe having health and well-being might end up being just about the most valuable thing anyone can offer. Being happy enough to cope through difficult times, emotionally resilient enough to support others, well enough not to require expensive medical interventions in what I guess will be a fully-privatised human-hostile health system. Having ways of generating social cohesion that exist outside authority, and being attuned to the creation and appreciation of beauty despite the darkness.
Human worth might depend on being healthy and well enough to turn up for whatever menial work there is left for us to do. Human survival might depend on us keeping it together just enough to be a sufficiently helpful (or at least low-maintenance) species for the AI not to just want to exterminate us.
It’s a bleak picture and I hope it’s not the timeline we’re on. But it feels to me that music, and arts in general, will need to be totally central to the education of functional humans in this sort of epoch. The ability to pick up an acoustic guitar and lead a singalong may just go from nice to totally essential in a future like this. Maybe it will be the only kind of power we have.