Here’s how I have used AI (or something like it) and why I don’t feel like I am compromising my human creativity.

As mentioned elsewhere, I have this software that generates music using fractal mathematics. You configure a bunch of settings and request a certain amount of musical matter and voila, you get a 4-track MIDI file. In its raw form, this MIDI music sounds pretty boring as there’s no instrumentation specified, just notes, but I then take this raw material and explode tracks into multiple parts, assign instruments and processes and effects, and edit like mad, doing anything at all that sounds good to me, and in the end I get a finished piece that sounds pretty good.

I do, however, wonder, “Who wrote it?” I hear things in this music’s complex harmonies that I like a lot but do not understand at all, and so hesitate to take composer credit for. It’s not like the software was a tool of convenience; there’s no way I would have been able to create that quartet on my own at all, not one single bar.

BUT… the raw MIDI played back using a default piano sound sounds dinky and boring and old-fashioned video-gamey while my finished piece sounds, I think, like real music. And I would like to claim some major authorship credit for that real music, though it would be really best characterized as a collaboration with the software. I will happily call this music “mine”, and tag it with my name, as its creator — but I would never think to publish it without explaining its mechanical origin, so the listener/reader can think what they will. *

Of course, there’s a very different level of AI engagement that appears to be dawning all around us: prompt-based generation of finished material given nothing but a few words. “Sad song about railroads in the style of Mozart”, that sort of thing. Taken to its extreme you can even just say, “Give me a song” and get something. That’s what is freaking people out, the idea of the complete disappearance of a human role that’s something other than product consumer.

BUT… with good AI tools, you won’t just say, “Give me a song”, you’ll engage in a long iterative process where the AI offers material and you request it to refine or expand it in some way, and this gets repeated over and over and built upon until you have a robust product with your human creativity all through it. Like: in the middle of a sax solo, you direct the AI to modify a previously-generated phrase to be “Irish-sounding” and add a lyric line about sheep. You do that for hours and hours, for several days, adding and subtracting and refining material until in the end you have something you’re willing to put your name on and which, I think, you will be 100% humanly entitled to.

Advances in AI will result, not in elimination of the human role, but in advances and innovations in the human artistic processes that use it. And I for one can’t wait to play with MusicLM.

Mark

——

* Interestingly (to me) I’ve used the exact same approach to music creation with stuff like farm recordings of chickens and roosters as raw material, using software to convert crowing and clucking and water bowl banging into MIDI melodies and going to town from there. In that case, there’s no AI to share credit with, so I don’t see who else there is to blame this stuff for but me. =8^)