Music: man or machine? Confusion about AI band on Spotify

Three published albums with 13 songs each within just one and a half months: The band The Velvet Sundown is currently producing one more or less shallow indie ballad after the other. How do you get it quickly? Quite simply: not people should write their songs – but apparently an artificial intelligence.

“Not quite human. Not quite a machine. The Velvet Sundown lives somewhere in between”: This is how the alleged AI band is described on its official accounts on social media and on the streaming platform Spotify. Who is behind it? That cannot be clearly stated.

The four supposed band members have never appeared publicly, pictures of the group were obviously created by a AI. It was only at the beginning of June that the band’s first music appeared on various streaming platforms. Most of the songs with their rock, country and folk elements are relatively interchangeable, neither great nor abysmally bad- and yet in a few weeks the group has managed to trigger confusion and even worries about the future of the music industry.

“The Velvet Sundown is a synthetic music project that is led by human creative leadership and composed, set to music and visualized with the support of artificial intelligence,” says the said social media accounts. With everything, including music, voices and texts, these are original creations that were created with the help of AI tools. There is talk of an “artificial provocation”, which should question “the limits of authorship, identity and the future of music in the age of AI”.

This provocation seems to have been successful for now: Velvet Sundown currently has around 1.4 million monthly listeners on Spotify, the single “Dust on the Wind” has been heard almost two million times so far. And not only that: The phenomenon has also led to a debate that the Swedish industry leader has not yet explicitly characterized the music of the band – contrary to streaming rival Deezer – as an AI generated.

Spotify: So far no industry-wide standard for AI use

According to Spotify, there is still no industry-wide metadata standard on how the use of artificial intelligence has to be disclosed. “As soon as there is a uniform approach, we will carefully check how we can implement it,” said the group on request. With each track delivered, streaming services are ultimately dependent on the rights holders in order to obtain the necessary information for the identification of the composition and the contributors.

Some social media users are already going to the barricades against Velvet Sundown. “Death of AI music” is even one of the more civilized reactions to the band.

Also (previously) people behind the machine

It is clear that a person must also be behind the machine. “The AI does not just wake up in the morning and decides to put music on Spotify,” says Virginia Dignum, who is a professor of responsible artificial intelligence at the University of Umeå. “So people are involved and there are people who make profits with it, even if it is only about marketing or attention.”

One problem is that the AI feeds their musical knowledge of songs and lyrics of real authors. “The data on which all this music is based are originally generated by people,” says Dignum. The claim that AI create music, texts or pictures is therefore somewhat misleading, because the underlying data would ultimately come from human authors. “And most of it was copied and stolen, regardless of copyrights or the authorship of those involved,” says Dignum.

Effects of AI on artists

In the long run, all of this could become a major problem for songwriters and authors in general: As the Swedish newspaper “Dagens etc.” recently reported on a report by the International Copyright Association of CISAC, Ki-Musik could give real musicians in the coming four years of income losses of up to ten billion euros.

Among other things, the demand for Spotify, among other things, also stems to clearly name music as such. “Every information should be characterized with the correct authorship. And if this authorship was supported by AI, it is important that this is known,” says Dignum.

Like other AI experts, the professor draws attention to the fact that a strong regulation must be a central part in dealing with AI. At the same time, people would also have to be trained more to deal with it. The right rules should be set up as to say that a car must have brakes – but you also have to have a driver’s license to understand how to use these brakes.

What remains for the listener?

For streaming users, it is ultimately a matter of better distinguishing between human and machine-produced music. As the example of Velvet Sundown shows, the borders are increasingly blurring. For the music industry, the big consequences could have – according to the buggles, who sang in “Video Killed the Radio Star” 46 years ago: “They Took The Credit for Your Second Symphony, Rewritten by Machine and New Technology.” This can be translated like this: “You have recruited the fame for your second symphony, which was rewritten by machines and new technology.”

Can – to stay in the picture of the song – so that Ki mean the end of the streaming star? Probably not. At least however, she can compete with many musicians for revenue that smaller artists are particularly dependent on.