There’s a band you’ve probably never listened to which has recently set the internet ablaze with debate about the future of music.
Actually, the Velvet Sundown isn’t a band at all, more a fever dream created by a computer. When the “artist” racked up over one million monthly listeners on Spotify with an album of guitar-driven soft rock tunes, it had many scratching their heads over the identity of this slick new act.
But the giveaway was the AI-generated photo of the band, complete with tell-tale weird fingers and guitar strings in the wrong places. The truth was that this four-piece indie act was generated entirely using AI.
To find out how simple that was to do, I created a song myself on the platform Suno, which is being sued by the Recording Industry Association of America for allegedly pillaging a vast number of songs to train its AI machine on.
The prompt I gave Suno for my song was very simple: “modern progressive rock, soaring guitar solo in the middle, with the lyrics inspired by Dark Side of the Moon about the pressures of modern life”.
I’d initially asked for it to be in the style of Pink Floyd, but Suno rejected the prompt as users are not allowed to refer to existing artists, no doubt on the advice of Suno’s lawyers. Within 30 seconds, my fully formed, three-minute song had arrived.
Shadow Machine is more like nu metal band Linkin Park than Pink Floyd. In fact, it sounds very similar to the recent Linkin Park hit The Emptiness Machine. In the soaring chorus of my song, a male American-sounding voice sings:
Turn it off, turn it off, the shadow machine
Grinding gears in the dream unseen
Turn it off, turn it off, the shadow machine
Who am I when I can’t breathe clean?
Not bad. The guitar solo is great too. I’ve been humming the tune to Shadow Machine all week.
Is this the first time a musical act has owed its existence to the soulless march of technology? I’d wager that most of the people losing sleep over AI composition have happily grooved along to music that owed as much to silicon chips as to human creative angst.
Remember when synthesisers weren’t quaint retro set pieces but the harbinger of creative doom? The 1980s and 90s saw entire genres emerge from the cauldron of electronic sound. Fans of Kraftwerk, New Order or The Chemical Brothers didn’t complain. Instead, we danced to their music in crowded clubs.
The reality is that music’s “soul” has always been smuggled in through technological back doors. The Pink Floyd track On the Run, off 1973’s Dark Side of the Moon, is often cited as one of the earliest techno songs. The band created it using an EMS Synthi AKS synthesiser.
Electric guitars scandalised the 1950s. Drum machines made percussionists weep. No one seems convinced auto-tune improved vocal performances, but Cher’s Believe was a banger nonetheless.
AI is no Rubicon-crossing moment in music. I’ll happily listen to AI-generated content, but there’s nothing wrong with wanting transparency.
The likes of Spotify and Apple Music should clearly label AI songs and albums, and allow you to filter them out of their streaming libraries – “100% human-made” will soon be a powerful marketing tagline.
The bigger issue is the grey area in US copyright law around whether new works generated from existing music libraries constitute copyright infringement. Numerous lawsuits in the US are litigating this and will establish vital precedents over the next few years. Until this is clarified, the safe and ethical position for AI music developers is to license what they train on, credit what they use, and ensure artists are part of the conversation and reward.
