- Songs made with generative AI are infiltrating streaming services.
- One Spotify user was recommended the same song under 49 different names and suspects AI is behind it.
- While the songs sound exactly the same, each has a different title, artist, and art.
That latest earworm you cant get out of your head? It might have been written by a robot.
Turns out, AI-generated songs are taking over conversation in the music industry. Over the past week and a half, mainstream names like Drake, The Weeknd, Beyonce, and Rihanna have all been involved, in some way or another.
Aside from the technology’s ability to, in theory, recreate any song across the history of recorded music, there is another unsettling situation unfolding: Songs, seemingly created by AI, are being uploaded to streaming services and recommended to listeners — but no one can figure out how they got there.
Spotify user Adam Faze took to Twitter to try to find out. Faze noted that “on a Spotify radio this week, 1 song annoyingly kept playing. except every time I looked, it was a different song name and artist entirely.”
He decided to add each of these recommended tracks to a playlist, which now includes 49 “different” songs. These songs each have different titles, artists, writing credits, and cover art — but the audio is the same across the board, a 53-second recording of what sounds like the intro to a prestige television show, ripe with steady piano and stringed instruments, that could be designated as classical music.
Some of the song titles include “Fosticrians,””Thorncutter,” and “Vattio Bud,” while some of the artists listed are Moditarians, Elval Woodridge, and Trey Dawson. All of these artists have little to no presence, with only a handful of tracks that are often slight variations to the same classical number mentioned above.
Faze first discovered the puzzle when he was looking for 1950s orchestral music and began using a Spotify playlist to satisfy the urge, he told Insider.
After listening through this classical collection, Spotify’s algorithm took over, making recommendations based on the playlist, and began playing the track in question.
Faze brushed it off initially, chalking it up to a strange coincidence, but after showing a friend what his Spotify was recommending, the two compiled all of the songs and compared them.
Examining the upload date on the songs reveals that they have all been uploaded to Spotify since June 2022. The only label listed for for some of them is Sky Tech Distribution, while others have no label at all.
In his Twitter thread, Faze pointed to a website, which has compiled all of the music associated with Sky Tech Distribution. There are 1,944 albums tied to the label, and Faze believes all the music is AI-generated, along with its corresponding cover art, considering the generic nature of both.
Sky Tech Distribution could not be reached for comment.
Spotify also did not respond to a request for comment, and considering the nascent field of generative AI, there doesn’t appear to be an official company policy on uploading songs made with AI to its platform. The streaming service does incorporate AI into its DJ tool, which makes personalized recommendations to users based on their listening habits.
Faze’s thread, which went viral, attracted some users in the replies who had similar findings.
“I just found 10 more without even trying,” commented one Twitter user, linking to a playlist of their own, containing 24 versions of the same song, with some sped up or slowed down.
Musician and cultural critic, Jaime Brooks, wrote about AI and music in February, stating that the emergence of AI-generated tracks essentially lowers the value of music from major labels. The more songs uploaded to music platforms — 100,000 a day as of October, per Music Business Worldwide — the more competition there is for user attention.
Last month Universal Music Group issued a request to Spotify and Apple Music to block AI services from training their generative models on its catalogue of copyrighted music, per the Financial Times.
One theory Faze presented is that Spotify is introducing AI-generated music onto its own platform and boosting its discoverability through playlists that focus on passive listening. If users are satisfied with songs that can be generated at small costs and in large quantities, the platform would have less reliance on licensing original material from major labels.
Of course, individuals or groups who recognize an opportunity to create cheap content en masse and earn money off the collective streams could also be behind AI-generated tunes.
Regardless of who or what is behind the mystery of the 49 tracks, it remains to be seen if people will actually be duped into listening to music created by AI.
When asked how long it took for him to question if the tracks were AI-generated, Faze was quick to answer: “Immediately, because the songs were so bad. I didn’t think a human would ever intend to make something like this.”