AI will change music, but how? [Mark Mulligan of MIDiA]

Mark Mulligan of entertainment consultancy MIDiA discusses the possibilities AI makes possible and how these changes can be good and bad.

by Mark Mulligan of Music Industry Blog

Every new technology goes through a period of being overhyped before the dust settles, and that technology either fades or builds steadily thereafter. Think 3D printing, VR, NFTs. In my 20+ years as a media and tech analyst, only three technologies have had a level of hype that felt like it was going to live up to expectations: 1) the internet (which was already in full swing by the time I started out – I’m not that old); 2) smartphones / apps; and 3) AI. Those technologies have one big thing in common: what they could become is ungovernable by its originators. But while it was human-power that unlocked that potential of the first two, it is the technology itself that is the accelerant for AI. Of course, people will amplify it as well, but AI itself is already creating many of the new pathways. The business, societal and even humanity implications are so vast that the implications for music are small in comparison. This, however, does not mean that they will not be equally transformational and disruptive within the confines of the music business. Which brings us nicely onto ‘heart on my sleeve’.

For those of you that have been on Mars for the last few days, this AI-generated track mimics the musical style and voices of Drake and The Weeknd. As Trapital’s Dan Runcie observed “[It] isn’t that good, but it’s an improvement from 2020’s TravisBott and other generative music attempts in recent years”. UMG’s response was to encourage DSPs not to host generative-AI tracks, and Drake himself was not happy with the last time a ‘fake Drake’ track did the rounds. Drake will probably be even less happy with this latest AI addition to the Fake Drake roster, which raises the question: will Fake Drake Break Drake?! While there are valid concerns from both parties, there is a real risk of this becoming an old world versus new world conflict, and in such scenarios, the new most often comes out on top.

AI is going to change the future of music. That genie is well and truly out of its bottle. Should more have been done by the traditional music industry to work with music AI companies earlier on? Of course, but we are where we are. So the focus now should be on trying to work out how to influence and shape what the future might be, through collaboration as much as (perhaps more than) enforcement:

We have been here before: The music industry was vehemently against P2P piracy (and I am old enough to remember that). After more than a decade of trying to fight it, the music business finally built an entirely new business around piracy’s successor technology – streaming. P2P infringed copyright, it took control out of the hands of the traditional business, and it created previously impossible use cases. AI is doing the same. What is different now is that the very ecosystem that streaming created (along with social platforms) puts AI into the hands (and ears) of billions of people, whereas P2P reached just tens of millions. Consumers will experience AI at scale before the industry can shape it. And in the digital world, consumers tend to get what they want.

Guitar or tape machine?: These two old technologies both reshaped music. One was about creating, and one was about copying. AI is a mix of both, which is what makes the response so difficult. Assistive and generative AI is already a mainstay of music creation, such as iZotope’s Neutron 4 and Splice’s CoSo. AI music is a continuum, from tweaking mixes through to composition, with virtually everything else in between. There is not one single, simple answer for ‘what to do with AI?’

Enforcement will be difficult: With the best will in the world, copyright law was not designed for AI. Music rightsholders will do their best to apply existing law, but they will face challenges in doing so. Meanwhile, there will simply be too much output to effectively pursue plagiarism cases, which take time and ultimately depend on the personal interpretation of non-expert judges and juries. If you think 100,000 tracks being uploaded per day to streaming now is a problem, when generative AI goes mainstream among consumers (which it most likely will), the number of new ‘songs’ created daily could easily be a hundred times that – perhaps even a thousand. 

Focus on the input not the output: So, the most scalable solution for music rightsholders will be to fix the problem at the top, by ensuring that generative AI tools only learn from what they have permission to learn from. ‘heart on my sleeve’ can only sound like Drake and The Weeknd because the tech learned from theirmusic. A number of generative-AI companies already only learn from selective, pre-authorised datasets. If this becomes the norm then an entire new licensing opportunity emerges for music rightsholders. Artists and songwriters will likely need to consent first, similar to how sync works. The alternative (trying to license and / or collect royalties on the millions, billions or trillions of songs that will be created) would be a fool’s errand.

The reason why AI feels so frightening to much of the music business is not just because of what it is, but also because it is a catalyst for pre-existing market shifts. The last half decade was characterised by the rise of non-traditional music, in the shape of ‘fake artists’, mood music, and independent artists. All of which have eaten into the market share of traditional music companies and creators. 

Streaming’s finite royalty pot makes revenue a zero sum game. Whatever may be done to try to ‘formalise’ AI music, it is almost certainly going to accelerate the fragmentation paradigm shift, by putting music creation in the hands of consumers. Radiohead once sang that “anyone can play guitar”. In practice, most people cannot, and do not. But literally anyone can ‘play’ AI.

There is growing concern among investors that this will mean market share erosion for the majors (and it probably will), but there is still a play for traditional labels and publishers, by licensing AI at the top. In doing so, they can benefit from the shift, just in the same way that major labels benefit from the rise of independent labels and artists through owning distribution platforms. That opportunity, though, requires the right approach and for it to be taken fast. The time is now.

I will leave the final words to President Biden, whose comments on AI as a whole apply just as neatly to AI in music:

“Look what’s happening with artificial intelligence right now. It poses enormous promise and enormous concern. Our world stands at an inflection point. The choices we make today are literally going to determine the future of this world.”