The moment you create your first AI music track can feel quite magical, especially if you’re as completely unmusical as I am.
I can’t hold a note or even reliably keep a beat, but now I can compose an entire pop song, lyrics included, in 30 seconds.
Or a soul song. Or maybe a metal track. Whatever comes to mind – you just say it A.I. You put what you want into the engine and it does the rest. The output may be a little generic, but as far as the listener is concerned, it’s indistinguishable from the real thing.
Even people who know much more about music than I do can’t tell the difference between a human song and an artificial song. The musical’s Turing test has been passed with flying colors.
AI music is a great example of the power of artificial intelligence to handle complex tasks and automate them to levels unimaginable just a few years ago.
But as with all technology, removing friction comes at a cost.
I’ve been researching AI music for the past few months. The full story has been revealed of a massive fraud attempt in which technologically-enabled criminals used AI tools to steal billions of pounds from real-life musicians.
This scam, which takes place in two stages, sounds like something out of a science fiction novel, but is now part of everyday life in the hidden world of the internet economy.
AI-generated music is trending – and it’s easy to get it out there
First, scammers create a large amount of AI music. They then build a bot that streams that music over and over again, thereby earning royalties.
Yes, this is a case of robots listening to robot music. What about defense? More robots. Let me explain.
AI music is so easy to create that its production has already reached industrial scale.
The highest figures regarding the number of AI tracks being released come from the French equivalent streaming site Deezer. spotify Or Apple Music. It is estimated that 60,000 complete AI tracks, more than a third of total production, are uploaded to the company’s site every day.
To put this into context, approximately 57,000 songs were created across the U.S. music industry in 2015.
In 10 years, Deezer will receive 21 million AI tracks per year. This is a conservative estimate, as the scale of AI music production is growing by the month.
“This is a way to completely flood music streaming services,” said Romain Henequin, head of research at Deezer, which has developed an algorithm that detects AI music being uploaded to the platform and detects small features in music that are inaudible to the human ear.
Deezer’s Romain Henequin (right) developed an algorithm to detect AI tracking
The truck itself isn’t actually cheating, but the actions around it are.
Someone would upload an AI track, use an automated system (bot) to listen to it over and over, and earn royalties from it.
This is no small aspect of AI music perpetrated by a few bad apples.
According to Thibault Roucou, head of royalties at Deezer, “the vast majority of listeners to this content are actually people who engage in so-called stream manipulation or fraud.”
Deezer highlights AI content and is the only site to flag if a track is entirely generated by AI
His algorithm detects unusual activity on the platform in the same way that banks look for unusual activity around payments. The algorithm suggests that as much as 85% of all fully AI music listening is fraudulent.
Because of the way streaming services pay, this isn’t just an issue for Deezer, it’s also an issue for artists.
There is no set price for a single stream. Instead, artists receive compensation from a common copyright pool based on the percentage of streams they get.
This means that when someone generates a huge number of streams, they take money away from others by reducing the amount in the common pool.
The amounts involved are very large. Deezer CEO Alexis Lanternier said, “We find that 8-9% of streams are fraudulent.”
“If you take that 8% and bring it into the world of music, it’s roughly billions of dollars, two to three dollars.”
“The battle continues”
Deezer says it uses its own automated system to identify bots to prevent it from accruing royalties from the tracks it streams, but scammers are always looking for new ways to get past its defenses.
“The battle continues,” said Locoux, head of loyalty.
“I don’t think we’re going to lose, but we’re not going to win anyway. They’re going to continue to improve, and so will we. And I hope we can … prevent them from taking too much money from other artists.”
When I told human artists about this situation, they were shocked.
Musician Lyra Tristram says her ‘blood boils’ after losing money to bots consuming AI tracks
“As artists, thanks to the streaming system, we only receive a fraction of the money we could actually earn, and to see it cut short and short by robots… makes my blood boil,” says folk musician Lyra Tristram.
“The music industry needs to do some work on this, or it could quickly spiral out of control,” said Aidan Grant, founder of music production company Different Source.
But what can you do? The AI music genie is now out of the bottle, and discussions are now taking place about how best to warn consumers and musicians.
Deezer has decided to label its fully AI tracks as AI, and is currently the only streaming site to take this approach.
Spotify, the world’s largest streaming service, opposed the move due to concerns that musicians using AI could be stigmatized. This is a potential problem in the future where all songs are created using some kind of artificial aid.
Read more from Sky News:
Driverless cars to arrive in London ‘this year’
Man given Musk’s AI brain chip says it’s like magic
AI-generated music is being uploaded to major streaming sites in droves
Neither YouTube nor Apple Music labels their AI tracks. YouTube said it is asking creators to mark their AI as AI if it looks realistic. Apple did not respond to a request for comment.
Spotify joins Deezer in an effort to stop the flood of AI music and fraudulent streams.
Last year we removed 75 million spam tracks, many of which can be attributed to AI. For comparison, Spotify’s entire catalog has 100 million songs.
What is the only hope for musicians? Some AI tracks have garnered a lot of hype and generated millions of streams, but so far there hasn’t been much demand other than the occasional one-off viral hit, and of course, scammers.
It seems that for music to gain a meaningful audience, it still needs to be performed and promoted by someone with genuine personal appeal. Relationships are important…for now anyway.
