The Big Problems Between Creativity and Copyright

The Internet is welcoming an ever-increasing number of devices artificial intelligence for each activity. Automated video creation has also been greatly simplified through AI thanks to applications such as Static Diffusion Video or FlexClip, which take advantage of massive databases and royalty-free resources to allow anyone to create animated graphic works in minutes. Could The same goes for music, as Google has shown over the past few weeks by bringing MusicLM to life. AI that creates songs from scratch Starting with a simple text description.

In most of the cases mentioned, we are talking about services that are still in the testing phase, access to which is limited behind paywalls or after registration on a waiting list for an unknown amount of time. In other words, access isn’t free or, if it is, it takes a lot of patience to get permission to use it. And yet, the first fruits are already showing up on streaming platforms and inspiring Discography creators and veterans fear,

between April and May 2023, in fact, even Spotify It was forced to eliminate thousands of music tracks generated by startup Boomi’s artificial intelligence, which was also granted a request for ad hoc measures by the Universal Music Group, or UMG, label. It’s common these days to have fun with the generation of original music that appears to be created by AI with the intent of simple entertainment. between creativity and copyright

However, several problems arise.

AI for Music Experiments

the use of artificial intelligence is creating rapid and somewhat noticeable change of art, the way it is made and perceived. This applies to paintings and drawings, but also to music and literary works. The more AI models are trained, the more the public has to ask questions about the nature of an art project: Was it born of human effort or complex data crossing by AI?

Once you understand the key characteristics of the second type of product, it’s easy to distinguish a human producer from an artificial one: they often use highly dissimilar sounds, in an unnatural or too complex order that a human can hear. related to the mind. Then there are those who are also pushing AI solutions for music on social media, for example showing up between TikTok and YouTube. Songs from the cast “Clone” A change of voice of a song, especially one of fame, or with a different singer than the original.

Experiments are the order of the day, using tools such as ChatGPT and Google Bard to generate lyrics, or UberDuck to sample audio files based on vocal data from selected artists. You can listen to some of the results by searching on YouTube”ai cover”, And honestly they are worried: Freddie Mercury sings Hotel California, Kanye West sings Viva la Vida by Coldplay, and The Beatles are having fun on the notes of Bohemian Rhapsody. There really is something for all tastes!

an endless endless debate

so the need to discuss it is obvious Relationship between AI and Music, because neutral legal dilemmas do not arise. First, we need to talk about the data on the actors’ voices: what files is the artificial intelligence trained on? Who owns the copyright on the songs sung by the artificial correspondent of the world’s best singers? One can attribute copyright to the creator of the algorithm, while record labels and artists can claim it, as well as the person who commissioned the song from the AI.

In the United States, the concept of “hybrid copyright”, according to which intellectual property is divided into degrees depending on the contribution made by the artist, the owner of the copyright of the song used as the basis or inspiration, and the creator of the artificial intelligence.

It is very difficult to find the square in this confusion, but there are also proposers individual solutionsFor example, Canadian singer Grimes has offered the creators of an AI system to use her voice to train the software, on the condition that she receives 50% of the royalties. The idea behind this solution is that the AI ​​model already has the data it needs to carry out the creative process, regardless of the song creators’ wishes. Furthermore, it is clear that music available on the Internet is accessible to anyone with a connection and can use it privately, freely.

However, if it is used “behind the scenes”, it is difficult to determine the exact approach of the developers to this material: how is the song used to train AI? It should be the manufacturer who makes this known, which implies a certain transparency towards the owner. copyright And who wants to use its service.

Ready for revolution?

The discussion should therefore involve singers, musicians, record labels and developers, with the possible aim of proposing a centralized solution that “allows the generation ofhybrid music”, or covers in which AI combines the premise of a song with the voice of another singer, perhaps usually placed at the antipode of the chosen musical genre.

The future of music is highly personal And from every perspective of the music industry, from album production to marketing, the opportunities accessible to the masses appear to be truly immense. However, this revolution should not compromise creativity, it should not replace the producers, singers and writers for whom AI can now ingratiate itself with those who use them.