AI Making Music For AI

Songs generated by AI are flooding the market as bots coded to “listen” to them push up their popularity rankings.

Higher visibility means more robots will want to check out the songs. And maybe people.

The set-up would be funny if it weren’t so frightening.

Late last month, songs appeared on streaming services that used AI that emulated  Drake and The Weeknd. Faux Oasis songs exploited AI to create lead vocals that sounded like Liam Gallagher. 

You can bet there are more such songs being produced and posted as you read this essay.

For instance, a service called Boomy matches human users with no musical abilities whatsoever with an AI and sound artifacts that can get patched together into songs and posted. Then it unleashes bots to click on the songs and push up their visibility (it’s called “artificial streaming”).

Granted, talent was never a prerequisite for rock stardom, and many bands have hired new lead singers who’re indistinguishable from older ones (Yes and Journey come to mind).

And, on the tech front, it’s already hard to separate the human and automated components of popular music. Sounds are generated and modified by computers, and that’s usually after advanced tech research tools have identified what they should strive to sound like. Some to much of what you might hear from a band performing on stage is prerecorded or manipulated by tech.

A recording studio has more in common with NASA’s Mission Control than the laid-back jamming space you might imagine. Same goes for the way songs are initially conceived (i.e. even if a tune emerges from a lone artist on her guitar or piano, the literal machine of the music industry takes over soon thereafter).

It’s why so much popular music sounds so perfect, like it has been created by robots.

It has.

Add to that the likelihood that many if not most of the “likes” registered on social platforms involve little to no conscious attention from the likers. Likes tend to prompt likes. 

It’s conceptually no different than the way music gained popularity in the old days of analog radio, when sales reps plied DJs with cocaine and cash to up the plays for their songs. There was nothing “organic” about it. 

So, what’s the big deal now? It’s not a fight over artistic merit or some assertion of humanness. Popular music has depended heavily on tech for years, just as the music labels have depending on artificial means to achieve their awareness and sales goals.

Is using AI the problem?

No. The music labels don’t want people using someone else’s AI.

It’s all about who makes money off of it.

I’m all for defending ownership rights, and artists have an added moral right to get paid for their work. Imitation might be the highest form of flattery but it robs artists of the fruits of their labor, however indirectly.

Awareness is a precursor to financial reward, but it’s a poor substitute. Literally.

There’s talk of insisting on some form of disclosure on music generated by AI, like a label or something. I think that’s the wrong way to go; rather, why not create a reliable system for asserting and even ranking the human elements in songs?

You could imagine a higher number (or letter, or whatever) correlating with how many components (or overall percentage) of a song didn’t just originate with a person but kept its form through the end of the production process. Kinda like listing organic ingredients, or asserting that certain elements are excluded.

There’s a campaign among artists and agents called The Human Artistry Campaign that seems to amount to nothing more than yet another manifesto of principles. It claims that art created by human beings is special and that AI could never replace it.

Only it already did so when the first code corrected a singer’s pitch.

Most listeners would be shocked when my labeling scheme told them they were listening primarily to Auto-Tune, or that so-and-so percent of a song’s structure had been determined by a big data analysis of the artist’s last hit (which was really big with the bots).

If the difference separating music that’s modified versus generated by AI is little more than a question of who pushes the record, send, or like buttons, the distinction is pretty meaningless.

I think the opportunity isn’t to chase after AI and presume to take it out of music, but rather find ways to put human beings back into it.

People making music for people. What a concept.