I'm about to embark on a project that I suspect will earn me the enduring, red-hot enmity of a great many people, including some whose opinions I respect.
I'm going to use artificial intelligence to make music.
More specifically, I've been experimenting with Suno, an AI music-generation platform, to create "recordings" of songs I've written. And I'm planning to take those "recordings" and distribute them through DistroKid, which will make them available on Spotify, Apple Music, YouTube, and other streaming services.
I realize that, for some people, this puts me somewhere on the moral spectrum between a scab and a puppy strangler.
But hear me out.
First, a little background.
I'm a songwriter. I'm not a musician.
Well, that's not entirely true. I play guitar (badly), and I can sing (badly). Whether anyone would want to listen to me doing either is another question. My talent, such as it is, lies primarily in writing songs: coming up with lyrics, melodies, chord progressions, and the occasional musical idea that I think is worth sharing with the world.
But there's a problem.
Writing a song and making a good recording of that song are two very different things.
If I want a professional-quality recording, I generally need to hire musicians, singers, producers, and engineers. And those people, quite reasonably, expect to be paid for their work.
I've done this before. In 2023, I released an album called Music No One Else Can Hear, consisting of fourteen songs (seventeen tracks, as there were a few alternate versions) that I wrote or co-wrote. I hired people to turn those songs into recordings I could be proud of.
It was a wonderful experience. It was also expensive.
And there's another complication. Sometimes I want to hear a song performed in different ways. Maybe a song I originally imagined as folk-rock would work better as country. Or punk. Or a big, overblown theatrical production featuring a singer who sounds as though he's trying to cough up a carburetor.
With human musicians, every one of those experiments potentially costs money. Sometimes quite a lot of money.
With Suno, I can experiment to my heart's content, or at least until I exhaust my monthly allotment of credits.
I can provide a song I've written, complete with its melody and chord structure, and ask Suno to produce a "recording" in a particular style. I can revise the instructions, change the arrangement, try different vocal approaches, and generate multiple versions until I find something I like.
Is this creative?
I've heard the argument that AI-generated music involves no creativity. And I think that's wrong -- at least in my particular case.
The creative work I'm primarily interested in is songwriting.
When I hire a musician to record one of my songs, I don't cease to be creative because somebody else is performing it. I wrote the song. I made decisions about how I wanted it to sound. The musician contributes his or her own creativity, skill, and interpretation, and the result is a collaboration.
When I use Suno, I still wrote the song. I still make decisions about how I want it to sound. I still listen critically, reject versions that don't work, revise my instructions, and try again.
Obviously, there's a difference. A human musician brings judgment, experience, imagination, and an individual artistic sensibility to the project. Suno is software. I'm not suggesting that those contributions are equivalent.
But I don't see why using software to realize my musical ideas somehow invalidates the creativity that went into writing those songs in the first place.
And, frankly, I don't think anyone would object if I used Suno strictly to produce demos.
That's one of my motivations. I'd love to have good-quality demos of my songs that I could send to performers, producers, or publishers who might be interested in recording them. A good demo can help someone hear the potential in a song, and Suno makes it possible to create one without spending hundreds or thousands of dollars. In fact, I have cowritten a couple songs for movies that are now in preproduction, and used Suno to generate "recordings" that I could use as proofs of concept.
But here's where things get more controversial.
I'm not planning to stop at demos.
I'm planning to distribute some of these "recordings" commercially.
I know. Fetch the pitchforks.
Why?
Because I like my songs, and I'd like other people to hear them. And if I can produce a version that sounds good to me, why shouldn't I make it available?
I'm not under any illusion that I'm about to become a streaming sensation. I have no expectation of making a fortune. Indeed, based on my experience with streaming royalties, I suspect that the financial rewards will be sufficient to purchase a moderately priced sandwich, provided I don't insist on cheese.
But that's not really the point.
The point is to get the songs out there.
Now, I understand that there are legitimate objections to AI-generated music. There are serious questions about the material used to train these systems, the rights of musicians and other creators, and the potential for AI to displace people who make their living creating and performing music.
I don't dismiss those concerns.
In fact, there's an obvious tension here. I'm a songwriter. I believe songwriters deserve recognition and compensation for their work. I would be understandably unhappy if someone used my songs without permission or compensation. So I can't simply wave away the concerns of other creators because this particular technology happens to be useful to me.
But I also don't think that every use of AI in music is inherently illegitimate.
I'm not asking Suno to write my songs for me. I'm using it, primarily, to create performances of songs I've already written. There may be occasional exceptions where I provide lyrics and experiment with AI-generated melodies, but the project is fundamentally about getting my own songwriting heard.
And I'm not pretending that I assembled a band, hired a singer, and recorded these tracks in a studio. I'm perfectly happy to tell people exactly how the "recordings" were made.
Which brings me to the quotation marks.
You'll notice that I've been putting the word "recordings" in quotes throughout this post. That's deliberate. These are not recordings of actual musicians performing actual music in the traditional sense. They are computer-generated audio tracks that simulate such performances.
They may sound like recordings. They may serve the same purpose as recordings. But I think the distinction is worth preserving.
At least for now.
Ultimately, I don't know where this experiment will lead. Perhaps I'll create a few "recordings," distribute them, and discover that nobody cares. Perhaps I'll discover that some of my songs work much better in genres I'd never previously considered. Perhaps a performer will hear one and decide to make a real recording of it.
Or perhaps I'll simply have a collection of versions of my songs that I enjoy listening to.
Any of those outcomes would be fine with me.
And if the whole project earns me the enduring, red-hot enmity of musicians everywhere?
Well, at least I'll have a soundtrack for it.
Even if nobody actually played on it.
No comments:
Post a Comment