Music from the machine (1557)
I’ve not written a lot. The timestamps don’t lie. There’s a Malay phrase, hangat-hangat tahi ayam or “as warm as chickenshit”, that came to mind as I reopened my profile and skimmed through what I posted in 2023, 2024. I am not yet improved enough to disapprove of the style, I actually like the flavour in those; I am however removed enough that I forgot the points I had wanted to make, and noted those with amusement. Clearly when I wrote that it was in a fit, and, as the chickenshit goes, the warmth faded as soon as it left the orifice. The remaining 998,000+ words will have to be squeezed out.
I have, however, remained creative. I’ve been composing piano music for just over a year now, bumping headlong into awkward passages, rueing a lack of access to formal training, sipping from the waters in the troughs of self-doubt as bitter as the air from the peaks of inspiration is sweet.
Composing is a wonderful thing. Plenty of us carry some sort of music within us, a catchy tune, a latest obsession, an infection from a glancing earworm, with no real conscious effort at all. In other words there is a natural, always-on gear and a muscle in the brain that retains and digests sequences of sound. It is a phenomena of low demand and low stakes, for the most part.
And if you let it, this phenomena can be bidirectional, and that bit of fatty tissue will also produce strings of notes on its own, and as the establishment has shown us from its (wary but net) validation of Schoenberg, any string of notes is a song no matter how dry or long or jaunty. This is what we mean when we say “Everyone has a song in them”.
You can adorn the strings with baubles, cut and twist and paste them to make repetitions and themes, elaborate on them with more of that music that comes from the opaque dark within you — and you can certainly also take that material and work it into structures, be it the fugue or the passacaglia as the master craftsmen Bach and Handel and Ravel did, and do that working in many faithful or subversive ways. This section is the bit that you can ape.
But before this is the listening.
Some many millennia ago we were crouching in bushes, girded by loincloths, sneaking up on waterfowl and buffalo by day and keeping a wary eye out for being sneaked up on by tigers and snakes by night. The lizard brain that remembers when the gaps in the rungs of the food pyramid, between us true apex creatures and the other non-bipedals, were smaller, is still in us, awake, alert, Sensing, for threats and opportunities.
Sensing is a thing you do between yourself and the outside for survival.
Someone lost in their own head is allocating mental resource to something at the cost of that sensing.
The listening — the deliberate dragging away of the bottle from the mouth of the lizard brain to your unconscious — is the first step to composing. Boredom, the pursuit of minute-by-minute productivity, the unsettling, the scraping inside when you stare at a wall for more than 30 seconds: if the muses are merciful, or perhaps lucky, the music will not need to battle its way through all this to make it to your conscious, and doubly so if you latch on and retain it in your memory before it fades.
You could, of course, make arrangements of other fragments of music and that is fine too, but I am talking about being a lightning rod and turning that lightning into art.
In the Tom Service book about great conductors of orchestras, there is reference made many times to how, when an orchestra really does great, the music seems to come in “from above” or “from outside”, as an external force rather than a consequence of internal mechanicking. This I now believe after composing, until I find a better set of words to describe it.
I have tried sitting down with the primary purpose of receiving such music, often to frustrating ends. All of the fragments that I started working with all came to me as if “from above”. I know it must be that it’s some result of the fatty tissue creating some sequence on its own from all the other sequences it files away, but why fight the feeling when it so much less resembles consequence and more so resembles being visited?
It’s after I am visited that I need to get to work. Here the thinking part of my brain kicks in.
Is the fragment interesting? Could I continue it? Sometimes it’s just not very compelling, and sometimes it’s a tune that I just unconditionally and firmly believe is worth the adorning.
Continuing the fragment is a mix of conscious and unconscious. I used to think this of all the great classical music: what makes this ‘answer’ any better to any other ‘answer’ to the ‘question’? What makes this ‘question’ any better than another ‘question’? Now, as I compose, the answer I have found is simply: because it came from the opaque within me.
Get 100,000 burner words’s stories in your inbox
Join Medium for free to get updates from this writer.
Take a simple fragment of a waltz and you will see how the opaque will carry on with a countertheme, or a second voice, as your thinking — your taste, your preferences, your tools — set accompaniment, key, rhythm. I am fortunate that while I don’t get first-fragments that often, I don’t seem to find it hard to continue them, though sometimes I look at the continuation I wrote, lose faith in it a week after, and rewrite that B section again and again.
More formal education and training would certainly help with the discipline and the craft, but there is something deeply intuitive about the process for me that is critical in it being as enjoyable as it is.
The parallel to generative AI is obvious but let’s delay delving into that for a bit.
When I listen, I receive fragments. I work those fragments into music by adding length and material, and strengthening that material through structure and harmony.
So far this has been a process fueled by unexplainable, base conviction; that is, I keep working the fragments because I believe they’re workable, and I strengthen the material through what I believe is additive rather than subtractive. If I put in lots of chromatic descents it’s because I like it that way, and if I demonstrate harmonic restraint (not often) it’s because I think it’d be good taste to.
Then I will test the piece by playing it on a piano, off the digital workspace, and test that conviction as a different being, one who did not write the music but one who is deciding whether it’s worth learning to play. Music written remains dead unless someone plays it, after all; unless the composer is fine having a collection of ink, it is on them to write a piece that is musically worth the effort to learn to play. And I like the idea of writing music accessible for someone of my hobbyist capability.
Between these two personas, I negotiate, tweak and at some point draw a line to say it’s done.
Composing gives me joy because it lets me exercise faith.
There are other quirky things like pastiching other music I like, or fancying myself a very late apostle to a favourite composer, but fundamentally I am writing music because I like to, which feels different to how I feel about writing — that I write because I want to have written.
Here, then, is the basis of my argument against AI music, as a composer and not as a consumer: it’s only half the persona, that of the editor.
Sure, one could say that the machine generates the strings and that’s effectively no different to being inspired, but I would counter that this is something said by someone who clearly hasn’t really listened or felt that before. Who does the receiving from above matters.
When you put your conviction in the song the machine produces, you are not putting your conviction in the song that came from within you.
When you have the machine craft the structure, it is the one bringing its algorithmic average of everything it’s digested — it’s not coming from the toolbox and archive you collected through your childhood, your cultural musical touchstones, your formative memories and the music backdrop to them.
When you tweak and edit, sure, that’s you and your taste. You get that one. But you’re not navigating that tension within yourself to have the heart of the music reveal itself.
There’s also my personal favourite general argument that the outputs of an AI tool will be mediocre by definition, because it’s trained on a general population of data to produce the best-by-average, but this doesn’t matter much to me in comparison to the next bit:
Using AI tools to compose serves having composed more than composing.
It’s good if you want end products in volume and very deficient if you want a creative experience. I suppose if you really enjoyed curating you could maybe get as much creative flex as one might writing a restaurant review on Google? That’s not nothing, but you’d get far, far more doing it my way.
