I do not understand music theory
I can barely read a score and I do not know music theory. I write entirely by feel. Given my temperament, I think that if knowledge came first I would end up trapped in a form.
When I compose I sound out notes one at a time and hunt for the ones that feel good. It is not an efficient method, so a single song takes me a long while.
The only instrument I can play is guitar. I cannot play piano, and with guitar I really just know how to hold the chords.
I once had someone with perfect pitch, a music-conservatory graduate, look at the score of a song I had written. Their entire verdict was: "unusual." I have chosen to take that as a compliment.
Where does "an AI work" begin?
Music generation tools such as Suno AI have arrived recently. With them, I became curious about how much delegation to AI it takes before people start to object.
There are many possible patterns. Handing everything to AI, without touching a DAW like Studio One. Not playing or recording anything, and leaving it all to AI. Feeding in your own melody and letting AI handle the whole performance. Building a rough track yourself and letting AI arrange it. Using AI only for a missing instrument part, or to fill a gap in an idea. I honestly do not know where the line should sit.
Go further and you find cases where a track built in a DAW is finished with the AI assistant in iZotope Ozone (Master Assistant). How many people would call that an AI work is another hard call.
If I sort it out for myself, it comes to this. If the lyrics and the music were both generated, it is an AI work. If only the lyrics were mine, the writing is mine and the music is AI's. And if almost every part is mine but the drums are AI's? AI was certainly involved, so for now "an original song made in collaboration with AI", or "a song that exists because AI existed", feels closest to the truth.
Whether to accept it is up to each listener
In the end, whether you accept work that AI had a hand in is a judgement each listener makes.
And if AI quality keeps improving, it is entirely plausible that telling human work from AI work will only be possible through embedded digital information, like a watermark.
How the music gets used
For me personally, what matters is the form in which a finished song gets used.
If it only ever plays on a streaming service such as Spotify or Amazon Music, I think it is easy to receive it as a work in which AI weighed heavily. But if people take that recording, perform it live and play it in front of an audience, then it becomes music actually sounded by human hands.
As things stand, AI output tends to converge on the average, and I think it will be a while before it can make work that lodges deep in a person. Even if that becomes possible and AI music becomes the mainstream, as long as the listener is human I suspect what will finally be asked for is humanness.
So I feel that how a song is put to use matters more than how much of it AI made. Put another way: who sounds the notes connects directly to what the song is worth.
How work reaches people
Not only in music but in images and video, we now live in an age where people weigh whether a thing was made by a person or by AI. Precisely because of that, I think occasions where people meet each other directly will become more valuable than ever.
Rather than how far you keep AI at arm's length while making something, the real question may be not to misjudge how you deliver it. That is what I spent today thinking about.