Play It Anyway

Before they are published, these Obiter pieces exist as essays on my hard drive. I have a backlog of concepts, and few that I’ve developed a bit beyond that. This past week, I started to do the deeper research into one, which I killed part-way in because I stopped believing in the underlying thesis. I finished a first draft of another one, which would have been published today were it not for the one you are reading now. It was fully researched and verified, and I might publish it next week. And then there’s this one, which jumped the queue when, on Sunday, as I was drawing with my two year old son, he turned to me and asked if I would play “some piano music.”

The request was not a surprise. Classical piano sits as an eye-level option in my home, right beside The Wiggles. The surprise was the sudden realisation that I could not ever recall playing the piece that immediately jumped to my mind that day. And upon hitting the play button, the idea for this essay struck me in a way where I knew I wouldn’t shake it until it was out. So here it is, now, ahead of the piece you might otherwise have read.

I want to take a moment to point out the cost of the sequence that I’ve described. For most of my life, drafting something required commitment to a thesis. There’s a deadline that can be real, imagined, or resultant only from one’s ability to stick with a task long enough to grind towards completion. The commitment was a necessary component because you had to do the research and develop the idea in pursuit of the deadline. By the time you had an opportunity to realise that you didn’t believe what you had started with, or that you didn’t really care all that much about the work, you were too close to the deadline to do anything about it. You published anyway because the alternative was missing it. What changed, for me at least, is not my discipline. What changed is what it costs.

The “piano music” that I played for my son was recorded on 24 January 1975. By the date alone, some readers will already be able to recite the story that follows. It takes place in Cologne, Germany, where on a stormy night, the Opera House is hosting its first ever jazz concert in front of a sold-out 1,400-person audience. The show is starting at 11.30pm, because that is the only slot the Opera House would offer to the 18-year-old promoter, Vera Brandes.

The performer is Keith Jarrett. A brilliant musician, and a particularist for his craft. Jarrett had demanded a Bösendorfer 290 Imperial concert grand piano. Due to a combination of errors, bad luck, and inclement weather, the piano never reached the stage and the only available alternative was a small rehearsal Bösendorfer found backstage. It is out of tune, has sticky pedals and notes that sound like there are nails on the strings, and is acceptably musical only in the middle register. A piano tuner has done what they can, but the piano remains effectively broken.

Jarrett arrives exhausted, in pain, and in a back brace. He sees on his stage a broken piano, insultingly different from the Imperial Grand he’d ordered, and he removes himself from the Opera House and returns to his producer’s Renault 4 in which he’d arrived. As Jarrett prepares to be driven off, Brandes rushes out and, standing at his window in the pouring rain, pleads with Jarrett to reconsider. Jarrett sees her - an 18-year-old jazz enthusiast with her career resting on this show. He relents. “Never forget. Only for you,” he tells her.

The recording gear is set up and ready. Were it not for that, what happens next would be retold but not understood. The audience takes their seats and Jarrett returns to the stage. He sits, rests his hands on the keys of the broken piano, and begins to play.

The first six notes ring from Jarrett’s fingertips. It is disputed whether they are an imitation of the Opera House’s signal bell or of the end melody of the house 4711 chiming clock, but what follows is a 26-minute improvisation that grows out of that tune. From the first moments it is undeniable that the music is being pulled from a place of unknown beauty.

Jarrett played in total for a bit over an hour, and the recording of that night, The Köln Concert, remains the best-selling solo album in jazz history, and the best-selling piano recording of any genre. And last weekend I played it while drawing with my son.

The classic telling of this story credits the broken piano as the hero. It posits that, with his options limited, Jarrett was forced into a register, and into techniques, he would never ordinarily have chosen, and that this drove a type of creativity that would not otherwise have been realised. The masterpiece is credited to the obstacle.

As a lesson, that telling of the story is fair. But what it doesn’t confront is the simple fact that the piano could contribute nothing on its own. It plays no music and enjoys no imagination. The piano was an inert object whose limitations required acceptance and skill to overcome. The masterpiece was not produced by the piano. It was produced by the insulted, exhausted, and agonised human being who sat down in front of it. And he sat down in front of it because of an 18-year-old jazz promoter named Vera Brandes, who removed the excuses not to sit. The tickets were sold. The Opera House was booked. The tuner had done all that could be done.

This essay is about AI, but it is not about the creativity of AI, and it is not about how AI drives the creativity of people. It is about how AI strips away the excuses not to be creative. It is a claim about the price of your creativity. Once the excuses are removed, the only thing left to decide the outcome is the willingness of the operator. The AI machine is not the broken piano, though the temptation is to pull in that direction.

The AI machine is Vera Brandes.

There are three angles from which the machine can attack the excuses. I’ll address them each in turn, and I’ll say up-front that I have seen all of them die at my feet as I conduct my work.

The first excuse was deadlines. One of the most difficult things to do as a creator of any sort (be it a writer, developer, songwriter, artist) is to break your own work. It’s a practice that is praised, but creative destruction has historically been both painful and expensive. A writer killing a thesis was an article that didn’t make it into that month’s newsletter. A musician concluding that a motif was generic or uninspired was an album with a shorter track list. It was a rational bias, conscious or not, to treat a piece of work as being sound even when it didn’t meet the standard. AI has eliminated the need to maintain the bias to the same degree. When I’m writing these essays, I can think of the concept, attack it from multiple angles, reshape it a dozen ways, abandon it entirely, and start again five times without worrying that I’ll miss the publish date. As I type this paragraph, I am working on the sixth version of this essay, and it is materially different from what I started with and entirely different from what I was going to publish in its place. The entire thing is written, above and below, but I simply have no need to commit to what currently exists as a final version, or even a final thesis. If I needed to start it again, I could.

But the fact that it is cheap to destroy a creative work does not mean that there is a corresponding willingness to do it. A constraint that is imposed, like a broken piano, creates a necessity to diverge from comfortable defaults and explore new ideas. A writer with the ability to attempt an essay six times is more likely to attempt to refine the starting thesis six times over. Although the excuse of the deadline has been beaten into submission, the risk that the starting thesis remains intact because of our own unwillingness to have it be otherwise remains unscathed.

The second excuse was the cost of feedback. Creative work has never had an issue with the ability to obtain feedback, but feedback that is both valuable and inexhaustible was expensive. It is not a great relationship-building strategy to express to a colleague that their ideas regarding your work are entirely flawed, and reflect a complete misunderstanding of both your message and your tone. Irritation avoidance is a legitimate factor to weigh in interactions with people. AI is an emotionless coworker. If the feedback, suggestions, editing, or anything else is inadequate, then, to draw again from the Köln Concert analogy, you can just keep bashing the piano keys until the music is beautiful. You cost yourself no goodwill, and you diminish no one. It is no longer a cost to ask for feedback ten times on the same three sentences. The machine cannot be exhausted by it, and will not take offence at your rejection of its feedback.

A study that at first blush appears to run contrary is, I think, strong evidence for it. In 2024, the journal Science Advances published an article by Anil Doshi and Oliver Hauser. I’ll spare you the details of the study, but it concluded that access to AI made writers measurably better on the whole, but also more similar to one another. The gains were mostly concentrated among less creative writers. The authors’ summary was this: “With generative AI, writers are individually better off, but collectively a narrower scope of novel content is produced.” I believe that this article draws us again to a conclusion about operator willingness. When the operator makes use of AI to fall into a comfortable seat at the keyboard, the output naturally becomes homogenised. When the operator is willing to make use of AI to test the seat in which they find themselves sitting, the instrument is fundamentally serving a different function. The same tool that is the Imperial Grand to an operator who accepts the output as written is the broken piano to the operator who is willing to take inadequacy as redirection.

The final excuse was the social price of being seen to be wrong. The witness cost. Correction from others has always carried a price, be it against personal ego or social perception. When an idea is rejected by the meeting room; when an answer is rejected by a teacher; when a riff is rejected by the band, you pay. Show me a person who claims never to have defended an idea past its death, and I will show you a liar. On at least two occasions, while writing this essay, I was explicitly told by the AI agent that proofreads my work that something I wrote was beneath my own standards. On both occasions it was right, and I had the luxury of being wrong in front of my own screen and correcting it before you had to read it. It’s like eating a canapé at a wedding on the way to the bathroom, then catching the poppy seed in your teeth in the mirror before anyone sees you again. There’s no need to get defensive about a fair catch. We can afford to be wrong in private as many times as the work requires, but all of the choices about what is right or wrong remain ours to make. The cost is nil. What remains is whether you are willing to work as though the social price is gone — because it is.

And so I won’t conclude that the lessons taught about the broken piano are wrong. But I will conclude that the most seductive metaphor about those lessons when you are applying them to AI is the wrong one. The broken piano does not map cleanly onto the AI machine. AI is not an obstacle whose constraining force will draw out genius. And the people who decide that it is the perfect instrument — that it will make the work effortless — are, I think, falling into the trap that will produce infinite copies of the same thing. The hero in this telling of the story is not the broken piano. The hero is the person who, when the excuses for not playing the piano piled up in front of Jarrett, removed them one by one until all that was left was a question of whether he was willing to sit down and play. When there is no longer a reason not to be destructive with your work; no longer a reason not to ask for feedback; and no longer a cost to being wrong, there is nothing left to hold as a reason not to do the creative work - and to do it completely.

The instrument in front of you is not a perfect one. It never will be, and if you expect otherwise then you are sacrificing your own creativity for that expectation. But the audience is seated, the tape is rolling, and the tuner has done all they can. The question is whether you are willing to sit down and play.

The piano is broken. Play it anyway.