(AFP) – “Artificial intelligence: do I really need it?”, sings Lulu Gainsbourg in her title “Elle”, which echoes the questions and even the anxiety of Music creators, jostled by the rise of generative AI.
The son of Serge Gainsbourg, who humorously hums “AI, dominate me”, created this piece after a friend showed him a song entirely written by artificial intelligence and told it on the social network Instagram (Meta group ) to have been “really shocked”.
Other artists have chosen to take advantage of this new technology, capable of generating all kinds of content on a simple request in everyday language.
– “Surprising things” –
This is the case of the British singer Imogen Heap, known for her song “Hide and Seek” (2005). She created her own AI model from her voice, which is used on her latest musical project.
She also revealed during the Web Summit, a major tech gathering held in mid-November in Lisbon, the creation of a filter allowing anyone to generate sound from their works thanks to a partnership with the start -up American Jen.
Canadian composer and singer Grimes also designed a vocal clone last year and invited Internet users to use it to create new sounds, while American singer Taryn Southern and the group YACHT released entire albums composed and produced with the help of artificial intelligence.
This new tool “allows us to do surprising things that we wouldn’t have thought of,” François Pachet, an artificial intelligence researcher who worked at Spotify and Sony, told AFP.
In 2018, he produced the album “Hello World”, where the Belgian singers Stromae and Canadian Kiesza collaborated with AI.
Many music creation platforms with artificial intelligence have also emerged, such as Aiva, Suno or Udio, while giants like Meta and Google also offer this type of service.
“People who are not musicians at all can suddenly make whole songs,” Mr. Pachet enthuses.
On the Web Summit stage, the co-founder of the Moises platform, Eddie Hsu, for example, showed how, in a fraction of seconds, AI could add drums to a few lyrics and transform them into a bossa nova sound.
– Transparency –
But many voices are being raised in the music industry to challenge the way in which some of these models are trained.
In the United States, the Recording Industry Association of America (RIAA), which represents the largest record labels, filed a complaint in June against the start-ups Suno and Udio, accused of “copying the work of an artist and exploit it for their benefit without consent or remuneration.”
More than 35,000 artists, including Thom Yorke from Radiohead, Björn Ulvaeus from ABBA and Robert Smith from The Cure, also signed a petition at the end of October, denouncing the unauthorized use of their creations to feed algorithms.
With generative AI, “there are entire sections of musical creation that can be replaced, such as the sound system in certain stores, the dressing of certain channels”, details for AFP, Cécile Rap-Veber, general director in France of Sacem (Society of Authors, Composers and Music Publishers), signatory of the petition.
“Unfortunately, there is a whole middle section of creators who have until now made a living from their art and who tomorrow risk no longer being able to,” she adds.
Professionals in the sector are demanding more transparency from technology companies, remuneration for the use of works by AI and the creation of new content when they are used to replace human works, specifies Ms. Rap-Veber.
For François Pachet, this technological advance will above all force creators “to do more original things” since “conventional things will be able to be done, in fact, more or less automatically”.
As for the replacement of humans by machines, he hardly believes in it.
“There is yet no song composed by artificial intelligence that would be so good that everyone would want to cover it,” he argues.
