I had this vision of the future where record companies would have computers in the basement and manufacture artists. - Trevor Horn, The Buggles
OF ALL THE MALEVOLENT predictions in the novel 1984, one always struck me as implausible and thus a bit silly. In two quick asides, Orwell mentions the ‘versificator,’ a machine for creating songs without human input (perhaps inspired by the ‘musicometer’ of Yevgeny Zamyatin’s 1924 novel We, a forerunner of dystopian literature). In the world inhabited by Winston Smith, a versificator-written tune “had been haunting London for weeks past. It was one of countless similar songs published for the benefit of the proles by a sub-section of the Music Department.” In the 1984 film 1984, Smith, played by a gaunt, haunted John Hurt, peers through a grimy window down at the washerwoman singing as she works. “Listen to that,” he says, with a hushed reverence. “How can she make a song written by a machine sound so beautiful?”
Somehow, incredibly, this prediction was only 39 years off. In less than four months, AI has evolved from being able to produce a novelty fake Drake song (a brief rap about beans in chili) to writing an actual hit (“Heart on My Sleeve,” feat. the Weeknd) with millions of views and streams. If you listen to these two tracks back to back, you’ll catch the jump. The first song, “Make Drake Rap About Beans” (or at least that’s the video title) sounds real enough, meaning you’d never think to question it if you heard it in public, at least if you weren’t listening carefully. And if you did listen carefully, it’d probably sound like what it is: a joke where part of the punchline is that it sort of sounds like the real thing, only minus conviction, or a hook.
“Heart on My Sleeve” is a sumptuous song. It sounds huge, lonely, defiant, expensive. It makes me think back to good times I’ve had, long ago, with people now dead. The song is wistful about itself, wistful with the bittersweet horror of what is happening to us as a species, right now. It forces the listener to acknowledge that this song marks a moment, that you are one of countless humans hearing this song and understanding what it means for humanity.[1] How can a song written by a machine sound so beautiful?
I DON’T EVEN LIKE Drake, or the Weeknd[2]. Their songs have always sounded fake to me. The music I’m into—the first decade of hardcore punk—falls on the opposite end of pretty much any spectrum: production, emotion, profitability, ideology. But these distant poles of musical taste now share one common denominator: availability. I can now listen to a super-obscure hardcore song just as easily as I can hear a hit rap. YouTube has evolved into the new Napster, a place where anyone can hear anything. Only this new incarnation of Napster is free, legal, and doesn’t take four hours per download. This evolution happened gradually, disguised by all the other nifty and horrific things people can now do with instant universal networked video.
In the three weeks since “Heart on My Sleeve” dropped, Napster was everyone’s inevitable comparison. When the peer-to-peer file sharing application debuted in 1999, the music industry grappled with a new set of facts about their product. And this time around, musicians (if not their industry) have one advantage; audiences want to hear songs they know. Some of the same rules that helped performers before streaming music still apply now.[3]
But seriously, c’mon: fan fiction music is not good news for any musician. We can’t even imagine all the implications from this vantage point. Maybe there will be a shift away from the song as the basic unit of modern music, the same way streaming music has already shifted listeners away from the album as a larger unit of music. Perhaps some listeners will demand new music on every listen.
I WAS LATE TO this YouTube-is-the-new-Napster thing, and I’ve been trying to make up for lost time in the last year. Even in the teeny tiny sub-subgenre of human music I like, I’m shocked at how many bands and albums I’m still discovering, some from cities and scenes I’d assumed I knew thoroughly. “Heart on My Sleeve” means one particular thing to me: my days of searching are numbered. Soon, I will be armed with the knowledge that someone somewhere could have created and uploaded forgeries of supposedly real and obscure songs. And that “someone” could be a bot, one able to build a prompt from my viewing history and then write, record, mix and post new “old” songs from vintage 1981 hardcore EPs that never were.
It gets weirder. High resolution 3D printers can currently make playable record albums. A smart counterfeiter could, in the very near future, use AI to forge a great vintage hardcore 7” from 1981, add the band or release to Discogs.com, and then physically manufacture the counterfeit object, which could then command a high price on the collector’s market. The lore of old hardcore bands is full of lost records made by bands whose already small press runs were severely depleted, warped in trunks or destroyed in flooded basements. Online discographers are notoriously imperfect. How would anyone be able to prove that a record never existed? On a long enough timeline, this is inevitable. This actual, literal thing.
PERSONALLY, I DON’T THINK AI music is like Napster. I don’t think there’s any comparison for what this is. We’re in uncharted territory, outside human history. On one of countless postings of “Heart on My Sleeve” that pop up daily, one top commenter summed the horror nicely: I can’t stop singing this I’m sorry they got us.
[1] I suppose that’s another little illicit thrill for this one song: every listen could be your last. Lawyers keep knocking it offline.
[2] The Weeknd apparently doesn’t sound exactly like the Weeknd. I wouldn’t know. Online reviewers have noted the offness of his voice. I pressed my wife, a fan, for details. She agreed, but said it was hard to quantify. “No conviction” was her best take, which was exactly my note on the fake Drake ‘chili’ rap. Drake’s been famous longer than the Weeknd, so it makes sense I’d know more about his voice. As would an AI.
[3] This quirk of consciousness benefits musicians, but not comedians. Emotions attached to songs are retrievable with every replay. In contrast, comedy relies on surprise to elicit involuntary laughter. A few rare performers—Monty Python, Steve Martin—have gotten away with performing their hits, but for nearly every other comedian, audiences demand new material. (Thanks to Matt Ransbottom of Chicago, IL, for pointing this out to me).
You stated above "that someone somewhere could have created and uploaded forgeries of supposedly real and obscure songs." I have a blog & have many times exposed fakes (hard & prog rock, especially. With in insularity of many punks scenes, with the DIY ethic of self-releasing, punk music would seem the perfect fertile ground for forgeries.
I do a music blog & try to share factual information about the music I share & showcase, but recently I have been pondering what is to stop an AI from creating a fake blog with only fake music from fake bands. I don't believe most people would catch on. If people just went to live gigs & supported real musicians, I believe that would be our first volley fired for stopping Artificial Reality.
Will rare-counterfeit proliferation defang the know-it-all collectors, or will it empower the romantics with the ability to make the most unique mix tape of songs ever for their crushes?! (*Playlist?)