Have Rosalía and Quevedo sung together? The challenges of music in the face of AI

Bad Bunny has been the last artist plagiarized through artificial intelligence (AI), an open door to creating music bypassing the musicians that announces profound changes in the sector industry and, above all, in the way songs will be created.

Oliver Thansan
Oliver Thansan
18 November 2023 Saturday 09:25
3 Reads
Have Rosalía and Quevedo sung together? The challenges of music in the face of AI

Bad Bunny has been the last artist plagiarized through artificial intelligence (AI), an open door to creating music bypassing the musicians that announces profound changes in the sector industry and, above all, in the way songs will be created. of the years to come. “If you like this shitty song,” says the Puerto Rican singer, “get out of this group right now.” It is the visceral response to the publication on TikTok of NostalgIA, a song with the voices of Bad Bunny, Daddy Yankee and Justin Bieber that as of this writing can still be heard on YouTube, the same as Heart on my sleeve, the song that this summer was published with the voices of Drake and The Weeknd, or Ahora te jode, the song that mixes the voices of Quevedo and Rosalía. All of them have in common that they have been made using artificial intelligence without the approval of the artists who star in them.

After suffering firsthand the streaming download revolution, which turned the music sector upside down in the early 2000s, the industry faces a new challenge with the lesson learned, looking for a fit for artificial intelligence, instead of deny the greatest to undertake an all-out fight against progress, technological progress at least.

“AI is a great opportunity because of all the resources it offers musicians and producers, like Photoshop for designers,” says Marc Isern, co-director of the Halley Records label. “In the same way that now everyone records in digital and recording in analog is an almost nostalgic decision, soon everyone will integrate AI resources to make music.” Isern makes this reflection while his company receives proposals to use the voice of his artists in AI-generated birthday songs. Giving a “happy birthday” personalized with the voice of your favorite singer can be a great gift, but it leaves the ethical and copyright debate on the edge. “When a technology company offers us to create products derived from our artists with AI, they do not fully understand that what they consider a product or merchandising object is, for artists, an artistic work.”

With the intention of taking the lead in the face of these new challenges, the Association of Catalan Phonographic and Videographic Producers and Editors (Apecat) organized this Friday a round table with experts from the sector that, among other aspects, will take as a reference the measures that are being taken currently in the arms sector. “The laws will take time to arrive, it is more feasible to think that we should have ethical behavior rather than normative behavior,” comments Eva Faustino, manager of Apecat. “We are interested in ethical reflection on the limits that we can apply, achieving a sector agreement before the norm that is imposed on us.” It is in this sense that they look at the arms sector "because it is a much more sensitive issue and is more advanced."

The objective is to prevent the technological revolution from overwhelming the sector again, as happened in 1999. “Now we know that a disruptive event can change our lives” and remembers how the European Union took 10 years to transpose the operation of the digital market to its regulations. “The last directive that came to solve digital copyright was from 2021, now we have solved what started in 99 and we already have artificial intelligence here.” In this sense, he celebrates that Spotify has removed from its platforms songs that use other people's voices through AI, such as the one made from the voices of Drake and The Weeknd, "it puts us in a better situation, it is establishing a position in that sense".

In the same direction, the Swedish streaming platform removed tens of thousands of songs uploaded by the Boomy platform and generated by AI in May, as reported by the Financial Times. At the same time, the Universal record company is in negotiations with the streaming platform Deezer and Google to establish the rules of the game in the sector. The entertainment giant is seeking a licensing agreement for the use of its artists' voices and melodies in songs generated by artificial intelligence.

“There is concern that the author's imprint, of the originality of human creation, could be lost,” says Cristina Perpiñá, general director of the General Society of Authors and Editors (Sgae). The entity in charge of managing the copyright of creators is favorable to the use of new technology as long as the author is maintained as a central element, otherwise "it loses culture and quality." To this end, the Sgae is introducing a clause in the licenses it grants to exclude the use of its catalogs to generate music using artificial intelligence, unless it is with the approval of the authors and in exchange for remuneration. It is a first step although they are clear that “in the end an agreement between the rights holders and the large platforms will be necessary.”

“I think that with artificial intelligence you can do almost anything,” says Antònia Folguera, curator of Sónar D, the academic branch of the advanced music festival that has been behind this technology for years. “Right now there are many processes in musical production, from composition to production, where AI has been involved for a long time even though we were not aware of it.” The emergence of vocal transposition, which allows singing using another person's voice, is what has triggered interest in this technology, which raises many ethical doubts. “There is no resolved question, but there are creative ways in which we try to respond to these dilemmas,” he comments, and remembers how in 2021 Sónar D already hosted an experiment where the artist Holly Herndon gave a concert using the voices of Maria Arnal and the Tarta Relena.

The singer-songwriter Roger Mas is initially reluctant to use AI in music, but remembers that, for many years, the voice and instruments have been recorded in the studio “two, three, and a thousand times one on top of the other, and We find that normal although it is impossible that it could happen in reality.” That is why he sees it as understandable, and inevitable, that the next generations of artists experiment with this technology, which he downplays. But remember how in the 19th and 20th centuries a lot of importance has been given to the artist, to originality, and it has been forgotten that in previous times the artist was considered a simple craftsman. “Probably one of the good things that AI will bring is to show that artists are not so original or important,” he reflects, adding that this causes him “sympathy” for the new invention.

When facing these changes, the sector must make use of legislation that comes from the last century, where voice issues are not dealt with through intellectual property, but through the organic law of the right to honor of 1981, since The voice is considered to be related to the right to privacy and one's own image. On the other hand, in matters of appropriation of intellectual property, "we are trying to use a law that was created at the end of the 19th century and has been valid during the 20th but generates tensions in the uses of the 21st," laments Eva Faustino. The lawyer recalls that attempts are being made to legislate AI through the “pastiche” concept, which allows the creation of works derived from references to previous works, legislation that allows comedy programs such as Polònia to use songs with changed lyrics. “This was regulated in an analog environment, but I consider that it does not cover artificial intelligence.”

The regulation of AI involves the use of technologies that allow authorship control. “Blockchain technologies should be able to solve these types of problems,” says Antònia Folguera, “if the song carries a trademark and registers that the voice belongs to a person, the blockchain would allow that every time this voice is used or played, the artist would be paid for his work and the artist's identity would not be broken, that it would arrive intact or that it would not be violated.” This is the path that the Sgae is betting on regarding the intelligence law that is being negotiated in the European Parliament, and which is considering including an obligation of transparency. Traceability in the authorship of the pieces “so that rights holders know that their catalogs have been used, and also so that users who consume AI music know that it has been generated this way and not with an author.”

YouTube, owned by Google, is one of the companies that is taking steps in this direction. It recently announced that starting next year it will force video creators to report the use of synthetic or manipulated content that looks real. The measure would apply to videos that use AI to compose realistic-looking situations that have not happened in reality, or show people saying things that they have not actually said. A regulation designed mainly to avoid conflicts on sensitive issues such as elections or a conflict, although it could also be applied in other fields such as music.

“As a record company, companies have already come to us that ask us for a price for giving us the voice of our artists and they even give us the authorship of what comes out,” says Marc Isern, “but we are generating songs from songs, and in an artistic work. What role does the derivative work play?” he asks, and remembers that there is a responsibility on the part of the industry itself “to not want to fill the market with absolutely absurd competition.”

The danger of promoting harmful uses is contemplated by the Sgae, "the opening of a music market generated by AI could have negative consequences for artists even if it were regulated," warns Cristina Perpiñá, and recalls that currently works generated independent and autonomous by an AI are not protected by copyright, for this the intervention of a natural person is necessary. “What is being debated is to what extent not giving any protection to these works can generate a parallel market for machine-generated works at a much lower price, because they would not have to pay copyright,” a situation that “could completely displace the author.” From the market".

Beyond the doubts it raises, AI opens a vast field of musical exploration on a creative and expressive level. “The voice of a human has the limitations of the human body, but if you release it it can overcome many limits of volume, of pitch, if it is lower, higher, also cosmetic forms, so to speak, that you cannot do with your natural voice”, proposes Antònia Folguera. A path that has already been opened by tools such as autotune, designed to tune the voice. "But there comes a time when, by exaggerating the possibilities of this plug-in, you achieve voice effects that are not natural, and that have greatly defined the voice in certain musical genres today."

The pianist and composer Meritxell Neddermann is passionate about everything that can bring novelty to music, “I like lack of control,” she says, although she thinks that AI has a “monstrous” point when it is used to appropriate the voices of others artists, so he would not give up his, any more than he would with his piano chords, “it's like giving up my voice, it would make me sick.” He closes the door to using the voice of artists, “something that already exists and has its own spiritual entity”, but he would love to work with voices generated entirely by AI, “it is the same path as synthesizers, but extended to anything else, "That they can speak, create lyrics, make a vocoder choir, all that makes me very curious."

At the heart of the debate lies the question of what is considered art. “With culture, societies are built, there is an adaptation of the ethical values ​​of society,” recalls Eva Faustino, concerned that “if this cultural self-regulation begins to be created through AI, it is somehow lost.” “For something made by an AI to be considered art and have a soul, the person controlling it must be a human, it is an artist,” says Antònia Folguera. “As long as it is this artist who puts the spirit, the soul or the intention, it will be artistic”, and remember that AI by itself does nothing, “a computer has no concern or vital need to express itself, to create, Humans do.” In her opinion, AI is not creative “because it does not create anything new, it always samples the past. It is a wilder, more modern, more careful and more refined form of sampling than we knew, which has been used in the world of music since the time of hip-hop. But basically it expands what already existed, it does not move forward.”