Artificial intelligence (AI) has long existed on the edge of our imaginations as a threat, a challenge and, perhaps, a promise. Indeed, the idea of robots making music, via the likes of Kraftwerk and Daft Punk, has over the years become almost comfortingly retro. But it is here now. Only not in the manner those electronic icons once presented.
If you count the algorithms that dictate our use of streaming services as AI then it has already completely re-shaped pop music over the last decade. The ‘Like this? Try this!’ culture that algorithms have engendered has overturned the way we are introduced to music. It’s proved a double edge sword. Whilst they’ve undoubtedly been useful to consumers there is an argument that their presence – alongside Spotify playlists – has homogenised pop to the extent that these days anything daring and different doesn’t stand much of a chance of making a commercial impact.
From top line to live: how AI tools are shaping new music
But aside from that they are also shaping the creation of music. The last few years have seen a proliferation of AI tools for musicians, from ones that aid live performance such as Rhythmiq and NotePerformer, to those that arrange samples (Atlas), to compositional software such as Aiva and Amadeus Code. The latter is a songwriting aid into which you input the type of music you intend to create. It then makes suggestions about which chord you could use and might recommend a melody that’s ‘inspired by’ an existing track.
Of course, you don’t have to follow its lead, but suppose you do allow Amadeus Code to chose the chords and top-line melody – who should get the copyright? Yourself? The company that developed the software? Or the writer whose track ‘inspired’ the melody? It’s a potential minefield.
The impact of AI on MU members’ livelihoods
AI has long been talked up by cutting edge figures – Brian Eno was endorsing ‘generative’ music back in the 90s. More recently Berlin- based artist Holly Heardon released her 2019 AI-assisted album, Proto. But established electronic auteurs incorporating AI is one thing, more concerning is the implication it has for other musicians. On library music composers, for example. Universal Music has apparently stated that within five years all their library music could be AI-generated.
Michael Sweeney, the MU’s Recording and Broadcasting Official fears the impact this may have on members’ livelihoods, and the implications AI has in terms of copyright. “While AI is unlikely to ever replace human songwriters and composers, there are dangers for cultural diversity when it comes to what they might produce and what people’s search terms might direct them to. We should not be afraid of the technology, but we do need to understand it, and in particular be alert to the fact that human beings, with all their potential for inherent biases, are the ones creating the software.”
Why the Act needs updating
Copyright in the UK is still covered by the Copyright, Design and Patents Act of 1988, passed long before mass usage of the internet. Many industry observers have suggested that the Act badly needs updating to reflect the rapid growth of computing power since. Others are more relaxed about the implications of AI. “I think a lot of the sturm and drang we have about AI and copyright at the moment is a bit overstated,” insists Trevor Cook, a solicitor who specialises in intellectual property. “There are people making a name for themselves on the conference circuit talking about it. One does question how
close to reality the development of music or literature without ANY human intervention is? Or, in fact, are all these machines tools, in the same way as a word processor or computer programme is?”
Think of algorithms as music instruments
Cook suggests that algorithms should be treated as just another instrument. “Nobody is suggesting because Microsoft Word is provided by Microsoft that they own the copyright to everything that you write. I mean, it could be a... remote possibility,” he concedes.
“You know these licences that you click through and never read? It’s entirely possible that one day somebody might slip in some sort of clause like that. The question is the enforceability of that. I think if they did start trying to do that then people would come down on them pretty hard. It’s dangerous to try and anticipate problems and introduce new laws. Only once you identify the problem and see how it is happening in practice can you judge whether the existing law is inadequate to address the issue.”
Fabs by numbers: why flair and spirit matter
Florian Koempel is a copyright consultant who has advised UK Music and the Music Publishers Association. “First of all the algorithm itself. It belongs to the company or the developer of the algorithm. That is the neutral part. Then there’s the secondary level – what happens to the ideas generated by the algorithm? On the legal side it is pretty clear that under EU, UK and US law copyright is the remuneration for human endeavour. Humans take decisions. What interests me is the situation with the material that the computer uses to learn from.”
He is scathing about the examples of purely AI-generated music thus far, including Iamus, the classical composer computer based at the University of Malaga whose work has been performed by the London Symphony Orchestra, and Daddy’s Car, which is what you get if you feed The Beatles’ back catalogue into an algorithm – a harmony-heavy track with none of the spirit or melodic flair of a Lennon-McCartney composition. Fabs by numbers, literally.
But Koempel does feel that the industry needs to get to grips with the copyright issue regarding collaboratively-used AI. “At the front of our conversations need to be the creativity of the human composer and performances. But to tell you the truth, UK Music doesn’t have a policy statement on it, and the publishers don’t have one.”
Artificial intelligence: the philosophical and legal perspective
The Ivors Academy doesn’t either, although it is working towards one, according to Helienne Lindvall, the Chair of its Songwriting Committee. As both a songwriter and the founder of LANDR (a cloud-based application that supplies AI mastering services) she has a broad perspective on this.
“A lot of this pivots around the philosophical/legal question of where you think the creative process lies,” she suggests. She refers to the 2015 case involving the British wildlife photographer David Slater. Slater was in Indonesia taking pictures of macaques when one approached the camera and took a selfie. When the photographer selected the snap for publication PETA (People For the Ethical Treatment of Animals) took him to court, claiming the picture’s copyright lay with the macaque. Fortunately (for humans, anyway) a US court held that animals cannot hold copyright.
Defining copyright and rewarding creators
“It was an interesting case because where does the copyright lie? In the person who actually presses the button? Or is it in the person who makes the selection? This is similar to the way AI works now – in the end, someone somewhere is making choices. Yet at the same time, whatever is put into that machine are copyrights that belong to other people.”
LIndvall explains that she supports a system where creators are properly rewarded if it is clear their work has been used in AI. Where multiple recordings have been used and it is unclear what they are she suggests the copyright proceeds should be collected and distributed collectively.
We need to be on the front foot for this because that train is coming down the track at full speed whether we want it or not. It’s hitting media and library writers first – they will be the first to be replaced by machines. The question is then ‘Whose work is being fed into those machines?’ It feels similar to when you’re trying to unravel a yarn that is so tangled up by the time it gets to you it’s impossible to separate.
The human touch: from composing to copyright
“It’s complicated. If it’s a purely AI creation based on, say, The Beatles, then obviously copyright should go to Lennon and McCartney. But what if another human creator makes choices and changes? We only have to look at things that are not AI where you have legal cases – Marvin Gaye and Pharrell, for example. Where does the copyright lie? What is inspiration, what is production and what is composing? At least in that case there were humans on both sides rather than a robot.”
Deep questions about the nature of human creativity lie ahead, and it’s high time all of us – musicians, writers, industry stakeholders and government – began to get our heads around them.