Suno AI, the Cambridge, Mass.-based startup that Rolling Stone described this week as “ChatGPT for music,” is pitching its app as a powerful new tool for non-musicians to conjure exactly the music they’d like to hear, no skills necessary. At Suno’s website, which offers a free trial of the app, its magic is apparent instantly. Simply type a prompt into the box, and moments later a clip of remarkably high-fidelity, realistic-sounding music to fit your specifications will begin playing. One breathless CEO called it on X “the next AI unicorn,” proudly sharing his creation of a song from a “metal opera about Mondays.” An AI app that creates customized music on demand raises obvious alarms for human musicians, who fear they might be supplanted — as well as for streaming providers like Spotify whose platforms could be overrun (even more than they are now) with spam uploads. (Mikey Shulman, one of Suno’s co-founders, mused ambivalently to Rolling Stone that “Spotify may one day say ‘You can’t do that.’”) Damon Krukowski, the indie rock veteran, organizer, and music blogger, wrote to me that the project was a “NIGHTMARE” (capitalization his), and pointed to a blog post of his about the labor-rights tug of war between artists and AI. Vernon Reid, the Living Colour guitarist, wrote to Rolling Stone that the “long-running dystopian ideal of separating difficult, messy, undesirable, and despised humanity from its creative output is at hand.” Their issue isn’t that technology and music are somehow incompatible. The two fields have nourished each other through the years, from Glenn Gould’s experiments with the recording studio, to Tim Follin creating digital symphonies on a primitive computer with a single “BEEP” command in BASIC, to modern digital audio workstations that offer high-end studio quality to any bedroom musician. That’s not the situation with Suno. Suno’s creators present their app as a tool meant to empower the non-musician, but in taking any kind of human effort out of the music-creation process and reducing it to DALL-E-style “prompt engineering,” its critics worry that it erases the expression, spontaneity, and social context that make music a world-changing force. Other venture-funded businesses that have transformed the world, like DoorDash or AirBnB, have also taken some of the “humanity” out of it. Still, at the end of the day you get the food; you get to stay somewhere (well, most of the time). With Suno what you “get” is technically a piece of music, but even if it did consistently and accurately fulfill users’ prompts, that music shares none of the human intent or meaning that even the most purposely cybernetic artists working today stubbornly retain. Of course, I had to try it. As a longtime amateur musician myself, I was naturally sympathetic to Krukowski and Reid. But I came out of the experience thinking differently about Suno than I expected to — becoming maybe less worried about the app, and more worried about what AI is doing to human effort in general. To be blunt, the songs it creates are excruciating. “In the land of the Horn, a story untold / A nation's journey, from young to old,” went the lyrical clunker that introduced one song I created this morning with the (arbitrary) prompt “Write a jazz-inflected 1970s classic rock song explaining the history of Somalian nationhood.” The app simply doesn’t work very well yet: This song, meant to be a Steely Dan soundalike, instead featured an eerie, generically macho voice belting the ChatGPT-generated lyrics over a bed of synthesized strings and percussion that was redolent of the most generic 1990s “adult contemporary,” or a particularly mediocre Disney musical ballad. One “downtempo punk dirge” came out sounding like a mid-tempo, minor-key pop-R&B anthem. A “commercial jingle that recites some of Tayshaun Prince's statistics from the 2007-2008 NBA season” failed to include any statistics but did, for some reason, call the retired forward “the ultimate basketball cat” (complete with a “meow!” for punctuation). I’m not the only one who had this visceral reaction. Ben Recht, a computer scientist and engineer at the University of California Berkeley, emailed me in a fit about the app, writing “You can buy a cheap royalty-free sample pack and make infinite background music in GarageBand. It takes literally zero expertise to be up and running in an hour. And the music sounds way better than whatever this crap is. I will never understand why people don't appreciate that music is inherently derivative and what makes a song great is how it ties to current social experience.” It’s fun to tweak the app’s misfires and shortcomings — it did cause me to legitimately laugh out loud several times. But its basic failure points to what its promoters either don’t understand, or seek to elide or erase, about human-generated music: The social context and human effort that go into producing music are not just crucial to its value, but comprise the entirety of that value. It’s impossible to imagine the current incarnation of Suno producing pop music someone might actively want to listen to, much less anything with the influence, feeling, and social impact of the art form’s paragons like the Beatles, or Prince, or Beyoncé. But it’s very easy to imagine a world where our mechanisms for delivering culture are defined by cost-cutting and financial speculation, leading to a bottomless supply and demand for AI-generated library music. If apps like Suno change the future of music it might not be by replacing today’s pop stars, but by carving away most of the wonder, weirdness and auteurism that make music matter to us at all.
|