The AI goldrush has begun. Across the world, tech companies are sprinting to integrate artificial intelligence into their products – the music business is no exception. Out in front is Spotify, which has recently partnered with OpenAI to bring the company’s ubiquitous ChatGPT to the platform.
What does the world’s largest music streaming service want with an AI chatbot? Well, the answer is actually ingenious. Take the world’s hottest AI, mix in Spotify’s formidable music recommendation algorithm, slap on an AI voice generator, and you get Spotify DJ: a tireless artificial disk jockey that never runs out of things to say and always knows what you want to hear.
This new feature, currently only available in the US and Canada, is sure to please subscribers and company investors. However it may also inadvertently pour fuel on a long-smouldering fight that has dogged Spotify and major labels throughout the streaming era. We’re talking, of course, about royalty payments.
To understand the issue, you have to look back to the great grandpa of telecommunications: radio. In the UK, when a song is played on the radio the resulting royalties are split 50/50 between the artist and their record label. This is due to a concept in UK law known as ‘equitable remuneration’. By contrast, if that same artist sells a physical album, the money they receive will be dictated by their recording contract which, most commonly, gives 80 per cent of royalties to the label and only 20 per cent to the artist.
With this financial incentive in mind, it should come as no surprise that when streaming platforms first hit the scene, many major record labels argued that they were not the same as radio. Warner Music, Universal Music Group, and Sony lobbied for a legal framework that classed each individual stream as a sale, just the same as when you purchase a vinyl or CD. This neat trick entitled them to the lion’s share of the profits.
Which brings us to Spotify’s DJ, a feature that mimics the format of radio almost exactly. For a start, the songs are selected for you, and while you can certainly skip an unwanted track – something you can’t do on radio – the ultimate goal is that you won’t have to. The principal drawcard of algorithmically generated playlists is that, as they learn more about you, their track selections become increasingly seamless and require less and less input from the user.
Then there’s our digital host, who serves up a stream of AI generated, but believably human, dialogue. Modelled from the voice of Spotify’s Head of Cultural Partnerships, Xavier Jernigan, DJ breaks up the listening experience by offering biographical facts about the artist and trivia about the music.
All of which raises the following conundrum: why should artists earn a 50 per cent royalty split when a flesh and blood DJ plays their song, but a 20 per cent split when an artificial DJ plays it?
And how about infinite radio stations that use machine learning to generate endlessly evolving tunes on the fly? Aimi is already offering just that. The company recently launched a series of live Youtube streams, each one suiting a particular mood. You can tune in and vibe out with no fear of ever hearing the same song twice. The company’s mobile app, currently in beta, takes things even further – allowing users to mix, match, and modify the sounds in real-time.
Thankfully, the team at Aimi makes clear that its machine learning algorithms have been “trained ethically and legally”. The company worked with a group of artists including Max Cooper, Ian Pooley, I. Jordan, Hammer, Dam Swindle, DJ Mix, Object Blue and more. All were tasked with creating the raw materials needed to produce this musical marvel.
However, other companies may not see the need for such ethical standards. Both the publishing industry and the art world have been rocked in recent months as countless copyrighted images and texts have been scraped from the internet and used as training data for generative AI applications such as Midjourney and ChatGPT.
Will musicians also have their tunes sucked into algorithms without permission and used to generate endless playback? Almost certainly. What, if any, royalties are they entitled to? No one knows. Our legal systems are still catching up with the technological breakthroughs of the 2010s, let alone generative AI of the 2020s.
But the problem of payments is not trivial or speculative; musicians are one of the most financially vulnerable professions in society. UK based charity Help Musicians found in its November 2022 study that a staggering 90 per cent of surveyed musicians felt concerned about their ability to afford food over the next six months, while 84 per cent had similar worries about paying rent or mortgages.
At the same time, the music industry as a whole has enjoyed seven-straight years of economic growth, with the majority of that increase coming from streaming. Universal Music Group, the world’s largest music company, saw revenue of $10.96 billion in 2022 – a 21.6 per cent increase on the previous year.
Of course, radio and music streaming are not identical – there are any number of structural, technological, and financial differences that make this obvious. However, it’s equally clear that playing a song on Apple Music or Tidal is categorically different from purchasing a record. The fact that we’re even trying to tackle this problem by applying legal concepts designed for terrestrial radio and physical album sales says a lot about how outdated the regulations for music royalties are.
Campaigns such as Broken Record have been drawing attention to this absurd state of affairs and are lobbying for significant reforms. Among a number of suggested changes, the group has pushed for streaming to be included in the equitable remuneration laws which already cover radio and other public performances. It’s debatable how far this would go in solving the music industry’s economic polarisation, but it would surely be a step in the right direction.
One thing that isn’t up for debate is AI’s potential to reshape the creative industries. How music is made, how it’s consumed, and who gets paid may all become open questions in the years to come. Sooner or later, creators, lawyers, and governments are going to want some answers.
For more features, click here.