Song is a universal form of expression that resonates across cultures and time. It serves as the soundtrack to our lives, from wrapping up the working week to navigating break-ups. A song can have profound impact, as do the artists behind the music.
But with the advent of artificial intelligence (AI), music as we know it is changing, posing the question: Does it matter if the music we listen to was created by a human?
AI continues to advance at an exponential rate, integrating across platforms and applications to the point where all of us are likely interacting with some form of AI on a daily basis whether we realise it or not.
Early iterations of AI-generated music produced songs that sounded tinny and lacked proper inflection, structure and emotion.
As the tech matured, generative music began to take on a more uncanny quality; a faux studio sheen punctuated by a vocalist just unnatural enough to unsettle tech-literate listeners capable of distinguishing a human voice from a synthetic one.
All this is set to change with the release of industry mainstay Elevenlabs’ generative music technology. After cornering the market with voice synthesis, Elevenlabs is entering the AI music market, competing with new players Suno and Udio.
While the latter two services still stumble with cohesion, fidelity and myriad other kinks that tend to steer their compositions toward the uncanny valley, Elevenlabs appears capable of producing AI music that could fool even the more discerning listeners into believing a human being with agency had a significant role in the composition, performance and recording process.
By training their tech on the copyrighted works of countless recording artists, these generative AI music companies can create a full song in seconds – Udio can create 10 songs a second, based purely on a single prompt. By inputting words like “80s pop synth song on partying”, the service can generate an entire track complete with lyrics, vocals and all other replicated instruments and production elements with a few button presses.
Presenting anyone with a smartphone, tablet or computer the ability to materialise an artificially assembled album without any musical skill, musical talent or musical sense risks devaluing music as both skillset and art form and, by ease and sheer volume of output, will eclipse human-created musical content in a range of fields. Germany was just host to what was quite possibly the first fully AI-generated track to reach the pop music charts.
Why approach an indie band for placement in a TV commercial or hire a composer to write a jingle or musical cue when anyone with the right software and a few “tokens” – one of the central currencies of AI-generated content across platforms – can simply click a button and populate their project with passable music in seconds at a minuscule fraction of the cost or effort?
The influx of generative AI music tools could also significantly disrupt the future of musical talent: if young people are raised to normalise and rely on AI tools to write their lyrics for them, their guitar riffs, their song structures, their production styles, rather than learn these instruments organically and form their own thoughts into words and melodies, how will their creativity be stimulated and how will these skills grow? When we outsource human creativity to machines for ease and instant gratification, we are left with only a superficial approximation of personal expression.
The use of AI-generated music is raising significant ethical and copyright questions. In order to create content, the AI programs have to be fed music composed, performed and recorded by humans; this is how the algorithm discerns what jazz music is, or what a Motown singer sounds like. If you were to train an AI music generator purely using Taylor Swift’s music catalogue, then all the music created by that AI platform would sound similar to Taylor Swift.
The challenge for the music industry, and for artists in particular, is the speed at which generative AI is advancing, the lack of legally enforceable frameworks, and the insatiable greed of companies who will happily forgo human-created content for cost-effective machine content
This raises several points of concern: what content has been used? Was the music copyrighted? Have the artists whose work was harvested to train generative music models even been made aware that their works have been disassembled and reduced into data points for the profit of wealthy tech companies?
The challenge for the music industry, and for artists in particular, is the speed at which generative AI is advancing, the lack of legally enforceable frameworks, and the insatiable greed of companies who will happily forgo human-created content for cost-effective machine content.
If we want to protect the arts and safeguard human-created content from irrelevance or obsoletion, a global, enforceable, legal framework is required.
For the rest of us, we’re left to wonder if the next song we hear on the radio or on a Spotify playlist was created by a human or a machine.
