Nicole Dieker is struggling with her composition work — it’s all guessing and fumbling, right now — but she knows that once she identifies the specific problems with her current draft, she’ll be able to start solving them.

Last night, L and I were talking about a hypothetical computer program that could use neural nets and machine learning to compose music that would sound both “organic” (L’s word, I take numerous issues) and satisfying.

The kind of music that keeps you engaged the entire time, because it’s got the perfect balance of tension and release, exposition and development and recapitulation.

The kind of music that could be customized to your personal tastes and previous music-listening experience, since it’s fair to say that people who have spent years listening to a lot of different kinds of music occasionally prefer more complicated compositions than people who haven’t listened to as much music (or as many varieties of music).

The kind of music that could be programmed, if we so chose, to amplify or quell whatever emotion you are currently feeling.

So, all right, let’s create ComposerBot.

(Let’s also acknowledge that this is in no way an original idea; the machine that can generate original music to please every unique listener has appeared in more than one book, and I’m trying to remember if it appears in the Magicians trilogy.)

If ComposerBot exists, what becomes of human composers?

The most obvious answer is that any humans who want to take up music composition do so on a strictly amateur and home-based basis; it’d sort of be like painting a picture for your sweetheart, even though you know there are hundreds of professional artists (and, if we’re taking this sci-fi future to its obvious derivation, an ArtistBot) who can do what you’re doing way better than you can.

Or it might be like baking a loaf of homemade bread.

Or learning how to play chess, even though the computers will always play chess better than us.

The point is — as L and I argued — that there would be no need for professional (that is, hired-and-paid) human composers if ComposerBot existed. Pop stars and symphony orchestras and film directors could all give their specs to ComposerBot and immediately receive the perfect composition for their needs — and yes, there would still be pop stars, there’s still value in “everyone singing the same summer jam” even if you can simultaneously ask ComposerBot to create your own customized summer jam, plus pop stars aren’t just popular for their musical talent.

(Arguably there’s a sci-fi-future in which PopStarBot would create some kind of AI that was, like, a customized pop star who met every single one of your emotional and parasocial needs, and also you could hang out with that pop star in your house the way you’ve always dreamed about hanging out with Taylor Swift or whatever, but if everyone has their own pop star then there’s no value in having access to a popular person.)

So we kept talking about this, what would exist and what wouldn’t exist if ComposerBot existed, and then I said something that I’ve been thinking about ever since:

“So this computer program would, like, know everything. It would have the entire history of music, from all eras and cultures, at its disposal — I mean, if you did it right, it wouldn’t just know music, it would know everything there is to know about people, too. And the world.”

And then I literally shouted, in the car:


I mean, people have always created art at the edges of their abilities, and art that pushes at the edges of what we have seen and experienced before, and art that attempts to try to communicate a new idea.

People have also created “art” that is well within their abilities, and we often call that art “tired” or “hackneyed” or “derivative” or “not as good as their previous work.”

Art, perhaps, is where the knowing pushes up against the not-knowing.

Where the cliff meets the leap.

Where the new idea is shared, because it’s so new and exciting and never-thought-by-you-before (or by-anyone-else-you-can-think-of) that you can’t not share it.

Where a new problem is created and then solved, with the acknowledgment (as is always the case, when you make something new and very close to the limit of what you can do) that it isn’t necessarily the definitive solution; that there’s still something left for the artist and the audience to consider together.

The art that pops out like “oh, here it is, here’s exactly what you were expecting,” is closer to what ComposerBot might make.

Because ComposerBot knows everything already, so it doesn’t have to create anything. ❤️

One thought on “Art Is Where Knowing Meets Not-Knowing

  1. I could feel my emotional opposition to this entire idea, rising like a flashy winter river about to flood with just ONE more drop of rain. And then you flipped the entire discussion neatly on its head, and it stopped raining but I was ALSO left surprised and thoughtful.

Leave a Reply