In mid-November, Shazam, the SMS service launched in London in 2002, reached the new record of 100bn song identifications, i.e. 12 songs identified for each person on Earth. To achieve this, a single person would have to search for a track on Shazam every second for 3,168 years.

This is more than 2,200 times the number of identifications of the most searched track on Shazam, "Dance Monkey", which has more than 45m tags. The track's track record  is just insane: number 1 in the charts in 18 countries, 27 platinum discs and one diamond disc, and three awards at the 2019 Music Awards!

Benson Boone's 'Beautiful Things', which was one of the Shazam 2023 predictions, was the first track released this year to reach 10m IDs, in a record time of 178 days. At this rate, it would take more than 4,800 years for it to reach 100bn.

More than 300m users connect every month to identify a piece of music they hear nearby.

History of technological fears

And as at every stage in the history of music, a new fear has arisen.

The end of the concert and printed scores with the phonograph (late nineteenth - early twentieth century): the appearance of the first sound recordings on cylinders, then on discs, gave rise to serious concerns. Musicians feared that "mechanical" listening to works would replace the living experience of a concert. Sheet music publishers, for their part, saw these new media as a threat to their revenues, as the public might prefer to listen to a recording rather than buy a score to play at home.

Radio and the fear of loss of control (1920s-1930s): when radio became a mass medium, labels and composers wondered whether it would discourage people from buying records. Critics also feared a decline in artistic quality, as radio was perceived as a medium that favoured immediate access to music, without regard for accuracy of execution or subtlety of interpretation. Musicians and unions are also concerned about the unfair competition that free broadcasting of recorded music represents, jeopardising the fees paid to performing artists.

The tape recorder and private copying (1950s-1960s): access to the domestic tape recorder worried the record industry, as it became possible to copy music broadcast on the radio or to duplicate albums between private individuals. It is feared that sales will fall if the originals are not purchased. Similarly, some are worried about the devaluation of the musical object, which can now be reproduced ad infinitum, with no immediate loss of quality.

Cassette and the start of home tapping (1970s-1980s): the arrival of the cassette made home tapping easy, i.e. the personal recording of music from the radio or duplication between individuals. This phenomenon also contributed to the emergence of the notion of private copying rights and compensatory levies.

The CD and digitisation (1980s-1990s): the transition from analogue to digital raised fears about artistic quality (sound that was "too clean"), but above all, in the long term, about ease of reproduction. Although the CD initially boosted the market thanks to its quality and durability, the possibility of ripping and sharing files on computers loomed on the horizon, creating a climate of suspicion about future computer piracy.

MP3, peer-to-peer and the fear of mass piracy (late 1990s - early 2000s): the advent of the MP3 format and file sharing via peer-to-peer networks represented the culmination of the industry's fears. Labels and artists feared a sharp drop in revenue, the distribution of music without control or remuneration, and the possible disappearance of the traditional business model of the record industry. The result is a legal and media war on piracy.

Streaming and the fear that works will be devalued (2010s - present): with streaming, unlimited access to huge music catalogues for a modest monthly subscription is seen as a new upheaval. Artists worry about the remuneration they receive per listening session, which is often deemed insufficient, and about the fact that music is becoming a simple flow, dematerialised and therefore less valued. Fears are crystallising around the role of algorithms in discovery and promotion, and the loss of artistic identity in favour of catalogue performance.

A worldwide warning

This week, Luxembourg's Sacem music licensing agency relayed a worldwide campaign to express its fears about the emergence of music generated by artificial intelligence.

According to a study commissioned from PMP Strategy, the market value of GenAI music will reach €16bn by 2028, and its audiovisual counterpart €48bn. To reach these figures, service providers will have to evolve, who running the music of yesterday and today on their artificial intelligence models and are expected to attract €4bn for music and €5bn for audiovisuals.

According to consultants commissioned by music rights organisations, this cannibalisation could affect up to 24% of the revenues of 'traditional' music creators and 21% of audiovisual creators, or €22bn over the next five years.

In addition to the creation of music itself, PMP anticipates the massive creation of background music for audiovisuals or public places instead of having to pay the industry's music rights, or the use of generative AI to replace translators, scriptwriters or even conductors and directors.

"In the absence of a change in the regulatory framework, creators will endure losses on two fronts: a loss of revenue due to the unauthorised use of their works by generative AI models without any form of remuneration, and the replacement of their traditional sources of income due to the substitution effect of AI-generated products, which will compete with the works of human creators," said the study.

"For creators of all repertoires, from songwriters to directors and screenwriters to film composers, AI can open the door to very promising new opportunities - but we must recognise that, if not properly regulated, generative AI also has the power to do great harm to human creators and damage their careers and livelihoods. Which of these two scenarios will win out?" asked Björn Ulvaeus, president of the International Confederation of Societies of Authors and Composers (Cisac), which represents more than five million creators. "This will largely depend on the choices made by political decision-makers when they revise the legislation currently in force in many countries. It's essential that we get the regulations right, protect creators' rights and help create an environment that preserves human creativity and culture."

AI and music, already an old story

Except that attempts to introduce AI into music are not exactly new.

In the 1980s-1990s, composer and researcher David Cope developed algorithms capable of analysing vast quantities of scores by composers such as Bach, Mozart and Chopin, then generating original pieces in their respective styles. The results were sometimes so convincing that some listeners found it hard to distinguish the work generated by the machine from that of the human composer.

In 2016, Sony Computer Science Laboratories presented "Daddy's Car", a song composed by an AI trained on a large corpus of pop music, in particular that of the Beatles. The aim was to demonstrate the algorithm's ability to assimilate stylistic characteristics and produce a novel but familiar 'Beatlesque' song.

OpenAI developed a model called Jukebox, capable of generating music (including vocals) in the style of many well-known artists, from Frank Sinatra and Elvis Presley to rock bands and contemporary pop. By analysing the repertoire of these artists, the AI can produce previously unreleased recordings that evoke their signature sounds.

AIVA (Artificial Intelligence Virtual Artist) is an AI specialising in the creation of music, particularly in the style of classical and illustrative music. By studying the scores of great composers (Beethoven, Mozart, etc.), AIVA can generate symphonies, piano pieces or orchestral works reminiscent of the great masters.

Google's Magenta project, dedicated to artistic creation using AI, has produced examples of music generated from existing audio data. These experiments use machine learning to produce new material inspired by well-known music, exploring the boundaries between reinterpretation, pastiche and true originality.

In Japan, Hatsune Miku is an icon of the phenomenon: she is a virtual singer, the product of voice synthesis software. She doesn't exist physically, but she has given concerts (in the form of holograms), sold albums and inspired fans to create numerous songs and works of art. She has become a true digital celebrity.

Read the original French-language version of this report /