If 2020 is going to leave one big question, it’s what is real and what is not. The fake news accusation is looser than ever. Conspiracy theorists doubt even those truths that were agreed upon centuries ago, and anyway you can no longer be sure whether what you see on the Internet actually happened or was falsified by AI programs. So far, this suspicion has mainly been based on photos and videos. But artificial intelligence is also increasingly penetrating the world of music. One is not quite sure yet: is she a competitor or a collaborator?

The research institute OpenAI has been experimenting with a program called the jukebox for a while. This automatically generates new songs in the style of pop stars, complete with melodies and verses. Ella Fitzgerald, Bob Marley and Elvis Presley are just a few of the greats with whose songs the researchers trained their software.

Advertisement

Sounds good? Some find: rather demonic

From the computer scientist’s point of view, this is certainly an impressive undertaking. The audience is definitely more critical of the AI ​​project. “Screams of the damned” commented about one user under the audio samples, “sounds demonic” another Guardian speaks against it of “sad spirits, lost in the machine”. When the computer-generated Frank Sinatra sings about Christmas fun in the whirlpool, many have crossed the line to sacrilege.

But why so much outrage? Is someone watching over the memory of the former idols all too jealously? But isn’t it the continuous reproduction and rededication and reinterpretation of pop music in the first place? To put it more simply: What is the difference to the remix and the sampling, aren’t there also clever copies that also enrich the original?

In the end, it’s also about what, the money

But OpenAI is not alone. Start-ups are already using AI programs to produce pleasant jingles on the assembly line. Meanwhile, a Dutch radio station launched the “AI Song Contest”. The participating research groups – or rather their software – produced pounding beats and synth sounds. If you just listen to it, there is hardly any difference to the material that is rumbling around in the Eurovision Grand Prix model. Finally, arbitrariness can also be a quality feature.

Advertisement

But that in turn leads directly to more tangible concerns. How could it be otherwise, it’s about money. Wouldn’t it be conceivable, according to the critics, that streaming portals would soon be playing AI songs and thus save the royalties on human artists? Spotify also has its own research program for artificial intelligence. And what about allegations of plagiarism if the real voices and melodies are not used, but almost perfect copies? Either way, it can be said: Anyone who believes that digital technology has already made copyright issues even more complicated will still be amazed.

Fortunately, there is also an optimistic take on AI pop. If you feed millions of songs into a computer program, you could then also have it look for that one pattern, that one hook, that one refrain, which is still missing in human discography. And use it to produce something new and previously unheard of.

Advertisement