The Generative AI: Revolution or Threat to the Music Industry ?

Generative AI tends to replace many automatable tasks in the execution of a job or the application of a process. In companies, the biggest changes are observed in sales, marketing, and IT with productivity gains of around 30%. In these fields, the impact is significant, and companies have realized the potential of the technology, but we are not witnessing a restructuring of the value chain. However, generative AI has a fundamentally transformative power in certain markets, such as the creative industries. In particular, generative AI is profoundly transforming the music industry. Historically, music creation was perceived as a domain reserved for humans, a bastion of artistic expression that artificial intelligence could never conquer. The emergence of platforms like Boomy, Suno, and Udio, capable of generating music from simple prompts, has upended this perception. These platforms rely on generative AI to create music tracks in response to textual descriptions, a process known as "text to song." For example, it is now possible to ask an AI to produce a Beatles song with a reggae melody performed by North American Afro voices in their fifties. The result is surprisingly faithful to the request, enough to raise questions about the redistribution of value.

The creative process with generative AI differs radically from traditional musical composition. Rather than relying on individual inspiration and talent, AI uses billions of musical styles ingested as inputs during the model's learning phase to generate new compositions. This capability allows individuals with no musical training to produce their own creations, thus opening the door to a new generation of potential creators. This democratization of music creation raises questions about the impact on traditional artists, whose works risk being drowned in a mass of AI-generated productions.

The applications and impact of generative AI vary depending on the music market segment. We mainly distinguish between listening music, intended for streaming platforms, and background music, used to enrich video content on platforms like TikTok or YouTube. AI is particularly effective at producing background music in bulk, where quality is not paramount, and the market is experiencing strong growth driven by the increase in digital content production. For listening music, even though the embodiment of the performer still retains a dimension that protects artists from the tide of AI production, some genres like hip hop or techno, which are already very electronic, are more easily imitable by AI.

In the past year, we saw numerous initiatives for new creation models:

A song titled "Heart on My Sleeve" was created by an anonymous TikTok user, "Ghostwriter977," imitating the voices of Drake and The Weeknd. The song quickly went viral, listened to and shared millions of times, then removed from streaming platforms at Universal Music's request.

In April 2023, an independent group from Hastings, UK, called Breezer, created an AI-generated album, tribute to Oasis and titled "AISIS - The Lost Tapes Vol. 1." The album received positive reactions from listeners and even from Liam Gallagher himself, who praised the AI-generated album on X.

AI has also been used to resurrect the voices of deceased artists, as in the case of Paul McCartney, who used AI to extract John Lennon's voice from an old demo and thus created a new Beatles song.

Other artists, like Grimes, have taken a different approach by allowing the use of AI-generated versions of their voices, with a 50% royalty share. This initiative aims to "open source" her voice and challenge traditional copyright concepts.

Faced with fully automated production, artists also take advantage of generative AI to accelerate their creative process. By using AI as a creative assistance tool, they can produce more quickly and with better quality, especially in the inspiration, mixing, and mastering phases, while respecting the essence of human creation. As in many fields, AI then becomes a production companion. This use of AI as an assistant rather than a substitute helps preserve artistic integrity while increasing productivity.

The economic impact of generative AI on the music market is already significant. In 2023, the market for AI-generated music was estimated at $300 million, with projections reaching $3.2 billion by 2028, an annual growth of 28% (source : market.us). This rapid growth could lead to a decrease in traditional artists' revenues. A study by SACEM and GEMA predicts a 27% decrease in revenue for some artists due to increased competition from AI-generated works. Streaming platforms, which represent 70% of music sales, could reduce their copyright costs by producing their own tracks via AI, further exacerbating the pressure on artists.

The challenges and issues of generative AI in music are numerous. Copyright protection is a major concern, as AI needs to ingest existing works, largely protected by copyrights, to "learn" to create music. Transparency of AI model inputs is therefore crucial to ensure that artists whose works have been used are fairly compensated. Moreover, the question of remuneration for creations generated by non-musicians raises ethical and economic debates. Finally, the quality and originality of AI productions remain subjects of discussion: is AI capable of producing truly innovative works, or does it merely copy existing styles, leading to a homogenization loop in a market whose driving force is, on the contrary, that "little something" that makes a hit's quality?

In conclusion, generative AI represents both an opportunity and a threat to artists. It offers the possibility to create more quickly and diversely, but it could also reduce the revenue and recognition of traditional artists if generative AI startups aim for revenue at all costs, disregarding social and legal aspects. Ongoing lawsuits against AI music creation platforms also highlight the need to regulate this technology to protect artists' rights. Legislative intervention is essential to structure this rapidly evolving market. Laws like the European AI Act, which does not recognize 100% AI-generated works as eligible for copyright, are probably a step in the right direction. CMOs are activating their opt-out capability (explicit refusal to use works to train AI models) to prevent the abusive use of their catalog. These actions represent a first defense but will not be sufficient. It is crucial that policymakers, companies, and artists collaborate to maximize the benefits of generative AI while minimizing its risks. Without questioning the creative novelties enabled by technology, the exponential speed of its application shakes the market, which is undergoing a forced learning process, as was the case with the end of CDs 20 years ago and the emergence of peer-to-peer and then streaming.