The Way the (Real) Music Dies
Note: A different version of this piece ran in WRAL Techwire. You can read that here.
Dateline 2044: A wistful grandfather speaks to his grandchild:
“A long long time ago…I can still remember – when humans made music.”
“Gee, when was that, Gramps?”
“It went on for longer than you might think, but I believe the day the music died was April 10, 2024.”
That’s the day Udio.com introduced the latest entry into the generative AI music space, their beta version of a music writing program that will slowly, insidiously replace most human beings making music.
The exact date human-made music began its slow death spiral will likely be up for debate in 2044. Before Udio, there were at least 21 other platforms already writing music and lyrics with ridiculous speed and ever-improving quality, with names like Suno, Mubert, Soundful, Aiva, Soundraw, Loudly, etc. But Udio has a star power and money unlike any of those others: it was founded by members of Google’s DeepMind team and has investments from venture behemoth Andreesen Horowitz, the co-founder of Instagram, the rapper Common, will.i.am and others. The Udio launch is the nuclear option in what Rolling Stone calls an “arms race” for AI music generation supremacy.
Here’s how the platforms work, with some slight variations. You go to a site, created an account, then type in a prompt describing the style of music you’d like and a description of the content you want in the song. Hit enter. Depending on the platform, in somewhere between 8 seconds and a minute, you get (typically) two songs fitting your description.
AI vs. Muzak
This ain’t Muzak. For one thing, Muzak was recorded by actual human beings, working studio musicians who made a soul-sucking living dumbing down popular songs into inoffensive background music. Between the 1930’s and the early 2000’s, grocery stores and dental offices and elevators paid to get Muzak piped in.
No, these platforms are BETTER than Muzak – and getting much better. Except that no humans write, play or sing a single note of it and, for now, they’re free.
The creators of the platforms seem to know the stakes to the music industry. Udio assured listeners in a release that it was “guided by musical people from the outset.” Suno.com co-founder Mikey Shulman said it was all about creating “wildly democratized music” that would right “the lopsided imbalance between music listeners and music makers…We’re trying to get a billion people much more engaged with music than they are now.” He envisions, eventually, those billion people paying $10 a month to make their own music, tailored to their exact specifications – that would be $120 billion in revenue per year.
It’s hard not to see the implications of AI to human beings trying to create music. The first people to lose out will likely be those writing background music or beds of commercials – the sort of music we are meant to not-quite notice. Then as the technology improves the bots will crowd out rank-and-file working public and studio musicians: why pay for a drummer when you can buy a drum track? Why hire a band when you can program your own music for much less? And in time maybe even the megastars will be replaceable: do you need to see the lens of a breakup through Taylor Swift’s eyes if you can pay an AI $10 a month to write a song with all the exact details of yours?
The economic consequences of the technology are not inconsequential. There are an estimated 2.5 million people in the US with jobs in the music industry – artists, musicians, promoters, roadies. Then add in those working in arenas and coffee shops and bars that depend on performers to draw folks in. Then add the folks who use music to supplement their income. That’s a lot of people losing a lot of money. But we’re losing more than that.
Broadcasting vs. Narrowcasting
Music can have profound, unexpected impacts on us. I worry, as Joni Mitchell did in “Big Yellow Taxi,” that “we don’t know what we got till it’s gone.” Here are two examples of the power of human-performed music in my life:
· When I was 21 and working as a teacher in Maine, a friend played me a recording of a live performance of Al Jarreau doing a cover of the Dave Brubeck classic “Take 5.” It was like nothing I had ever heard before, and spurred me to think differently about how I was teaching, to figure out how to permit more creativity in every classroom session.
· A few years later, I turned on the radio one day and heard a recording of Mo’ Horizons doing a cover of “Hit the Road, Jack” in Portugese that helped me finally realize the possibilities and challenges of globalization and pushed me to focus part of my university work on international student recruitment.
But there’s more to it than just the occasional epiphany. Shared human-made music can resonate with us in our darkest moments; pump us up for our biggest competitions; inspire us to greater faithfulness. We can revel in it with our best friends, and, sometimes, complete strangers. And whether it delights or annoys us, comforts or challenges us, it makes up the soundtrack of our communal lives.
As AI gets better, the danger is that music made by humans pushing the boundaries of creativity could go the way of newspapers and broadcasting.
Why would we pay for music that forces us to think when we already ‘know’ what we believe?
Won’t we decide instead to just listen to music that never challenges us, narrowcasting its way into your preconceptions of what music “should” sound like?
Could it really happen?
Is anyone still putting a limit on what algorithms can do? Think about the sages who thought computers could never beat Gary Kasparov at chess or beat Ken Jennings at Jeopardy; English teachers who claimed they would always be able to spot an AI-generated essay or art critics who said they would always be able to tell a real Dali from a fake.
Some musicians are trying to put the artificial toothpaste back in the tube. A group called the “Artists’ Rights Alliance” recently wrote an open letter, with 200 signatures, calling on the AI music sites to “pledge not to develop AI music generation technology, content or tools that undermine or replace the human artistry of songwriters or artists or deny them fair compensation for their work.”
Yeah, good luck with that.
“Fair Use” considerations and lawsuits
The battle isn’t completely over. In Music Business Worldwide, Ed Newton-Rex made a convincing case that the Udio AI had been trained using thousands of copyrighted songs without permission (Rolling Stone quoted one cofounder as saying that they trained the AI “with the full knowledge that music labels and publishers could sue.”) And while none of the platforms will permit you to create a song that sounds like a specific artist, Newton-Rex found that by asking for songs from artists like “Beminem” (not Eminem) and “Blank 184” (not Blink 182), he could get the AI to generate songs that sounded remarkably close to the original artist. Do these similarities show a violation of the so-called “fair use” standard? The lawsuits have begun.
It's not hard to make a plausible case for the complete cannibalization of copyright in print, but proving that in court is going to be ridiculously hard. I’d give the music industry the same odds as newspapers had in trying to prevent Google from using their content for free.
Hello, New American Pie
I’m left feeling profoundly pessimistic about the future of human music, so much so that I thought about writing a new version of the song “American Pie” for this post. But then I decided to save some time: I asked the Suno AI to write music and lyrics for a song. This was my entire prompt: “A soulful song about the dangers of AI’s writing music vs. humans.” Listen here to the song Suno created for me, “The Soulful Lament,” and read the lyrics below, all created and “sung” by an AI, with no hint of irony:
Late at night/ I sit alone at my piano/ Thoughts racing through my mind
Don’t know where to go/ Those AI machines/ They’re taking over the show
But can they truly feel the music in their soul?
(Verse 2)
Pour my heart out/ Every note/Every chord
My music’s my voice/ It’s the story I’ve stored
But now they say the robots/ They’re bringing something new
But can a machine capture/The pain that I’ve been through?
(Chorus)
Oh the danger of AI’s/ They’re creeping in so sly
But I’ll keep singin’ my truth/ Won’t let ‘em pass me by
The passion in my heart/ The stories in my veins
No AI can ever replicate / My soulful strains.
It’s really not bad. And it’s horrible. “No AI can ever replicate/My soulful strains”? Consider the source.
-Leslie
References:
Rolling Stone on Udio and Suno: https://www.rollingstone.com/music/music-features/udio-ai-music-chatgpt-suno-1235001675/ and https://www.rollingstone.com/music/music-features/suno-ai-chatgpt-for-music-1234982307/
RoomieOfficial youtube overview of AI platforms: https://www.youtube.com/watch?v=XhbaXx_ao2Y
AI-generated art is getting better: https://medium.com/@twothirdsmath/heres-how-ai-paints-in-the-style-of-famous-artists-523a4bb3264f
Is the way these platforms are training “fair use” or abuse of human musicians rights?: https://www.musicbusinessworldwide.com/suno-is-a-music-ai-company-aiming-to-generate-120-billion-per-year-newton-rex/
Music industry employment: https://www.riaa.com/reports/the-u-s-music-industries-jobs-benefits-economists-incorporated/#:~:text=prepared%20a%20report%20on%20the,%249.08%20billion%20in%20export%20sales.
Artist’s Rights Alliance statement: https://musicrow.com/2024/04/artist-rights-alliance-addresses-ai-developers-more-in-open-letter/#:~:text=“We%20must%20protect%20against%20the,and%20destroy%20the%20music%20ecosystem.
· I