When AI Takes the Conductor's Baton

Music | par Franskaya

Franskaya's Blog

In this article, I explore the impact of Lyria and MusicLM, Google's latest innovations in musical artificial intelligence, on artistic creation. While MusicLM turns textual descriptions into high-fidelity melodies, Lyria goes a step further by generating complete songs with lyrics, music, and vocals. These advancements raise crucial questions about the future of music creators in the face of rising AI, particularly regarding copyright and compensation. The text raises ethical concerns and explores the reactions of the artistic community and legislative responses, such as the public consultations initiated by the Canadian government and the positions of the SPACQ, while acknowledging the innovative potential of these technologies.

Lyria and MusicLM: At the Dawn of an AI-Revolutionized Musical Era

Whether they like it or not, composers will soon have to contend with AI

As a passionate electronic music composer and avid technology enthusiast, I find myself at a fascinating yet unsettling crossroads with the advent of Lyria, Google's latest innovation in the field of musical artificial intelligence (AI).

In this series of articles on music, I address various themes of interest to musicians, artists, or anyone attuned to the magic of music.

Google had already shaken the artistic community last January by introducing MusicLM, which allows the creation of melodies from simple textual requests. MusicLM can create high-fidelity music, in 24 kHz, that remains coherent for several minutes, enabling a user to design an instrumental track without touching a synthesizer keyboard, simply by specifying the musical genre, type of instrument, mood, location, or era, for example.

AI-generated image by Dall-E of robots playing music on an electrified stage

For now, some commentators indicate that we are still far from wanting to add these sounds to our playlists, and even that most of the music created by this software can quickly become unpleasant to our human ears. But obviously, these are just the initial milestones.

In its latest announcement, Google reveals Lyria, an AI developed by its subsidiary DeepMind, which goes even further than MusicLM. This AI not only generates melodies but also assembles lyrics and music, like a virtual composer and lyricist. Lyria also sings, allowing the generation of a complete song! Soon, anyone will be able to ask the program to create a personalized piece, based on a simple written description! In a promotional video, someone hums something, then asks the AI to turn it into a saxophone score. Then the AI adds other tracks, and the final arrangement is particularly convincing!

These developments, while fascinating, raise important questions. As a composer, I wonder how these technologies will transform the industry. Will they complement or compete with human creativity? Imagine Spotify offering stations where music is created live by an AI, inventing new melodies on the fly! Knowing the immense competition already existing among humans for listening time, will the addition of this player result in reducing the share of listening time allocated to humans? We already see the scenario looming where giants like Google won't have to pay royalties to creators by playing auto-generated music… With the giant's already colossal profits, these contents will further reduce the revenues of musicians and creators. Will we witness a gradual decline of human musical creation, eclipsed by these artificial geniuses? Will this new digital revolution put more and more musicians and creators out of work?

AI-generated image by Dall-E of a robot violinist playing music on an electrified stage

We are currently witnessing the complete upheaval of the traditional media business model, based on advertising. The TVA group is already suffering the consequences of the changing rules of the game and has recently been forced to cut more than 500 jobs. The Canadian government recently asked giants like Google and Facebook to share their immense advertising revenues with Canadian content creators, journalists, publishers, but they refuse. Will music creators face the same fate? There is cause for concern! If one no longer needs to undergo lengthy musical studies to create music for films and any producer can simply ask an AI to propose a free soundtrack, we risk seeing the disappearance of certain professions. Indeed, if potential clients can do everything themselves thanks to AIs like Lyria, MusicLM, or OpenAI's Jukebox, for example, who will seek the costly services of a human for mixing, arranging, or composing?

Google responds to these concerns by stating that it has adopted a code of conduct to ensure that the impacts of AI are positive for the future of music. It promises that music tracks created using Lyria will be encoded with a watermark technology called SynthID, which will allow the identification of content generated by an artificial source. Google says its goal is to provide artists with additional creative means to those already existing, allowing them to explore new soundscapes. It will therefore offer certain tools, for example, the possibility to create an instrument track from a few hummed or played notes, or to transform a few chords into vocal harmonies.

AI, undoubtedly, opens doors to unprecedented creativity. It will allow everyone to become a journalist, composer, writer, graphic designer, and more. But it also raises serious concerns. Google used several thousand hours of human works protected by copyright to train MusicLM, without consent, credit, or compensation for the original creators. This situation raises important ethical and legal questions.

Fortunately, in response to these challenges, the Canadian government has initiated public consultations on the impact of generative AI on copyright. These consultations, open until December 4, 2023, aim to gather feedback for the development of copyright policies, particularly regarding the use of protected works for AI training, the ownership of rights on AI-produced content, and liability in case of copyright infringement.

AI-generated image by Dall-E of robots playing music on an electrified stage

The Société professionnelle des auteurs et des compositeurs du Québec (SPACQ) has also taken a stance on this issue, and is making representations in the context of this governmental consultation. The SPACQ's submission highlights the impact of generative AI on the cultural sector and the risk of economic harm to rights holders. In its submission, the organization supports the implementation of safeguard regulations.

One thing is certain: the era of AI-assisted music has begun. It will offer incredible opportunities and possibilities, no doubt! But it also confronts us with new complex ethical challenges and consequences yet to be discovered.

AI-generated image by Dall-E of a robot playing a synthesizer on an electrified stage

As a tech-savvy artist, I am both excited by the prospect of playing with new futuristic tools and concerned about having to do so in a context devoid of revenue.

Will AI one day monopolize complex jobs to the detriment of humans? What will music sound like in a hundred years? What do you think Karl Tremblay of the recently deceased Cowboys Fringuants would think of all this?

Notee: The illustrations for this article were created with OpenAI's Dall-E. This concretely illustrates the points of my text: no human photographer or illustrator was solicited for this article, highlighting the absence of revenue for these creators in the context of the emergence of AI.

Main references:

Portrait of Franskaya, music producer and composer from Quebec, Canada


Blogger, author, artist, Franskaya is also a songwriter, sound technician, and music producer from Quebec, Canada. Don't miss the next post by following him on his Facebook page or click here to learn more about the author of this article.

Image of the Apotheo music single cover by Franskaya


Listen on Streaming

Spotify icon
Apple Music icon
YouTube icon
Tidal icon
Deezer icon

Other blog articles