✘ Will AI correct the anomaly of the 20th century music business?
The disruptive impact of AI will be seismic. Our questions should be of similar scale.
Musicians and other artists worry about their futures in the age of AI. Rightfully so. We may have assumed that we had fully entered the Information Age when PCs got hooked up to the internet in the 90s, or when smartphones found their way into every pocket, and practically all information was available at our fingertips via community-built datasets on Wikipedia, Google Maps and modern social media. It certainly looked like the Information Age, but I’d argue it was just a warmup.
The Information Age has only just commenced, now that advanced algorithms can be trained on information to produce novel information. Essentially anything that can be digitised, or described digitally, can become information. That includes art. So what was true in the earlier days of networked media, the heydays of internet piracy, is true now: anything that can be digitised will be subject to the principles of the Information Age.
One of these principles has always been that that which can be reproduced freely, will be. This principle is sometimes also referred to as Information wants to be free. For music, this meant that the value proposition of placing information on a carrier (like a CD) as the primary way to access that information suddenly lost most meaning to most (former) buyers, because they could now exchange files. The value propositions shifted: how streaming services managed to compete with free (piracy) was by offering a more convenient alternative, which is the primary reason for many people to pay a monthly subscription.
The principles of networked media have been supercharged. Now, through AI, networked media can replicate itself. Not just replicate, but come up with all kinds of novelty. Information creating information; no longer depending on humans.
When the internet came along, innovative minds in the industry imagined a future of people buying CDs via the internet. What actually happened through piracy, remix culture on SoundCloud, and eventually 120k tracks being added to streaming services daily, would far surpass that initial imagination of the future. Lately, services like Okio and Suno have started offering ways to create music simply by describing it in words, similar to how you’d create images in DALL-E or Midjourney.
The hot question on everyone’s lips now is: what happens when anyone can just make a song with a text prompt? The question, and any intermittent answer to it, will be obsolete before we ever figure out a satisfying answer. The disruptive impact of AI will be seismic. Our questions should be of similar scale.
The tradition of official and unofficial music
The 20th century created a music industry that has been able to maintain a stranglehold on culture. Instead of singing or playing folk music at home as our primary way to hear songs, we started buying information carriers (vinyls, cassettes, CDs) from corporations and playing that back over and over. In the centuries before, traditions of informal and formal music had been established in many places around the world. In the twentieth century a new tradition of ‘official’ and ‘unofficial’ music has established itself, with the official music being the economic and cultural giant.
This has allowed for the establishment of a vast industry around this official music and introduced an era with the largest amount of professional musicians since the history of mankind.
Being born into this era, we assume this as ‘the normal’, but a 100 years from now people may look at it as a historical anomaly. What if we look at it as an anomaly right now? Instead of forcefully projecting ‘the normal’ into the future, what if we consider the implications of advances in technology whilst seeing what we’ve had until now as an anomaly?
I see a world in which many more people are creatively active again. For a long time, in part due to the professionalisation of creativity, it has been hard to create any unofficial music that was on par with official music. That barrier is gone now. Soon, if not already at the time you’re reading this, most people will be able to create basic, professional-sounding music. This will blur the lines between official and unofficial music, at the expense of the ‘official music’ tradition, but with the benefit that far more people are able to actively participate in cultural expression through music again, without their expression somehow being ‘lesser’ than those of professional creatives.
There is an important ethical discussion to be had here. After all, the reason why those barriers are gone is because AIs have been trained using information (‘official music’) created by professionals. However, like with piracy, I don’t believe a black & white ethical discussion will accomplish anything. The genie is out of the bottle. When ethically complicated aspects are abstracted away from the end users enough, they will not behave as if they care. I don’t think I have a single millennial musician friend that never pirated music. All the ‘piracy bad’ discussions didn’t reverse the changed status quo, though they did inform future debates for how we could build a new landscape.
This is why initiatives like Spawning.ai are important, co-led by artists Holly Herndon & Mat Dryhurst, which among other things lets artists opt their information out of training algorithms. Anyway, as you can see, I’m digressing and the importance of these complexities don’t fit in the scope of this article. So for now, I’ll leave them for a future piece (or conference talk - book me here).
In summary, I think we had an era in which a concept of official music was introduced. This era even saw a large semi-professional amateur class, but the average person’s relation to music was changed. Whilst musicians were grappling with the question of how to stand out among 120k tracks being added to streaming services daily, the last barrier that distinguished official and unofficial music was dropped. Now everyone can sound pro and express themselves quickly. What will this mean for music?
Small music
We are leaving the age of Mass Media behind us and all media, including music, is responding to this. The Age of Mass meant mass audiences, superstars, mass shared experience, and one-directional distribution. Meanwhile, we have entered the Age of Networks, which means distributed collaboration (e.g. Wikipedia), social media, networked forms of distribution (viral posts, memes, piracy), long-tail platform economies, but also decentralization.
New innovation means an acceleration of trends set in motion decades ago. Whilst Networked Media will not completely supplant the realities introduced by Mass Media, it will augment these realities and cause foundational cornerstones of entire legacies to crumble. One of the effects of this is that media is becoming less mass; music is becoming smaller.
First, let me create some context to make my point easier to understand. Let’s talk about internet memes.
Early internet memes, like the rage face comics, were somewhat universal. The internet was a smaller place and the amount of people creating (or digitising) information even more so. In 2008, when the first rage face comics were shared, there were around 1.5 billion people connected to the internet - many of whom would just sign on occasionally. In 2023, we have over 5 billion daily internet users. Furthermore, the ability to create and share art or information has changed. We share much more. On social media, through messaging apps, by leaving reviews, or simply by unintentionally creating information footprints around the internet through visits, likes, and clicks.
Memes have become more specific. Highly relevant to certain contexts and completely irrelevant to people outside of those contexts. It is easier than ever to alter a meme, so more people participate, more people express, and memes find ways to become relevant to increasingly specific contexts. Memes have become smaller.
The same is happening to music. First through internet forums, aided by music hosting sites like Mp3.com and Soundclick, but also targets of counter-piracy enforcement like Rapidshare and MegaUpload, and later through SoundCloud. This was relatively early remix culture. One person would post a novel musical idea in the form of a track online, the next person might download it, load it into their DAW and work on a remix for a few days, or create something new altogether, and then upload it. It’s what fuelled genres like moombahton and edm trap, but also many modern forms of rap. Music started behaving more like memes.
However, that is nothing compared to the velocity we see today. People create quick edits on their phones, sometimes without even leaving TikTok. They make specific music, like memes, for specific niches, that TikTok’s algorithm knows exactly how to reach due to the vastness of its network. It seems like the era that produced megastars like Michael Jackson, David Bowie or Madonna is over and in place of that, we’ve gotten smaller, more specific music and popstars that are slightly less ubiquitous. It would be much easier to find someone now who is not aware of ever hearing a Taylor Swift song, especially in Europe, than it was in the 90s with a megastar age equivalent.
There is probably a better way to phrase this phenomenon. Maybe it’s the atomisation or segmentation of music. However, I like to think of this in the context of the transition from the age of Mass Media to the age of Networked Media. We’re now able to move beyond music for the masses and instead music becomes small, on a massive scale.
So what is small music? Small music can be:
Music that exists only between friends on a messaging app, like an inside joke, for example a song you listen to to hype yourselves up before going out that contains references to things specifically relevant to you as a group.
The soundtrack to a meme that is only relevant to a specific scene of people, like a meme making fun of a specific situation in the techno scene paired with a track that sounds like it was produced in a circus 🤡
Hyperlocal music that is bound to specific places.
Music that can only be heard a limited number of times, for example someone makes a song that can be heard 50 times. It can only be played once and from one device at a time. Think of it a bit in the vein of a Snapchat pic that can only be viewed once or an IG Story that can only be viewed for 24 hours.
Music that changes through interaction, for example some Skrillexcore: the more often you play it, the heavier the bass gets. (This doesn’t exist as far as I’m aware, but give me a ping if you ever make something like this)
Jam sessions and derivations, through apps like Endlesss which let people create loops of music that people can then iterate on, similar to how you might fork code on Github.
And it can be so much more.
The challenge in describing ‘small music’ is that it can be everything that ‘official music’ isn’t. Official music was rather specific. It’s songs that traditionally had been on carriers. They’re often released by labels or self-released by artists. Distributors make sure they’re available on music streaming services where they sit on artist profiles and can be added to playlists. Small music doesn’t have to be any of that, yet the specificity of what music has become in the 20th century almost demands a clearer answer.
That’s where I feel things get most exciting: not only can everyday, non-professional people express their ideas into song, but they’ll be able to express their ideas as variations of music as a medium.
If you wanted to make music in the past 60 years, you basically had to make songs; ‘official music’. Now that is not necessary anymore, unless you want to participate in the music streaming rat race.
What does this imply? An acceleration of an existing trend: more music, more derivatives, and more people involved in the creation and distribution of this music.
We are moving from a world where music creation, especially the creation of good music, was something exclusive. Now that that is open to all, artists are rightfully worried about their position. The age of mass media has created a familiar frame for what an artist is supposed to be and how they’re supposed to make a living. What does being a professional artist mean in an age with more music, infinite music, and more musically engaged people?
From dead to adaptive music
The above describes a challenge. One that is existential for many artists. Yet that is the hopeful part, a very human story. For all the human creativity that AI is about to enable, we will also see something that to many would feel more dystopian. I’ve alluded to it in the previous paragraph: infinite music. Music no longer needs people.
Music separating itself from people has become true multiple times in the past. With the advent of broadcasting and recorded music, some people rebelled against speakers: music is supposed to be heard from musicians and their instruments, not from mechanical devices. In the decades following, avant-garde artists would sometimes refer to recorded music as ‘dead music’ in contrast to live music. But you don’t have to go to the 20th century to find composers who take the human being out of the performance: 18th century composer C. P. E. Bach, son of J.S. Bach, has composed 30 works for mechanical instruments.
Since transforming the humble MP3 player into powerful pocket computers called smartphones, the writing has been on the wall: sooner or later these computers, like the record before them, will play a decisive role in redefining the role that people play in music. Suddenly we could think of all the sensors that can be used in a phone, or the fact that our primary devices for music listening are communication tools, or how this all may interact with the ‘internet of things’ - connected devices in our environments like smart lamps and speakers.
Early machine learning algorithms started turning music into infinite versions, such as The Infinite Jukebox by early pioneer Paul Lamere, a researcher for The Echo Nest, a music data startup acquired by Spotify in order to power its recommendation algorithms. It was around this time in the early 2010s that I would talk to anyone and everyone about this topic of ‘non-static music’, during which I had the pleasure to meet Oleg Stavitsky, creator of numerous successful playful, game-like apps for children such as Bubl Draw which lets kids make music drawings or Bubl Gelato which turns ice cream into music, well, virtually.
Oleg later went on to found Endel, a generative music app that creates soundscapes for specific purposes like sleep, focus, relaxation and recently also errr… nsfw fun times. This journey from games to new types of music experiences is more obvious than it may seem. As a matter of fact, I think many creatives who have spent time in games are better prepared for music in the age of AI than many professional musicians. There are two major reasons why.
Firstly, games contain worlds that players can interact with. These are worlds of data and any signal the player gives can be used. For example, when a person crosses a certain point in an environment a monster appears, or if they make two objects collide there’s a specific sound those things make. More importantly, it has allowed for the rise of dynamic music compositions, which means music that adapts to the player. You might not know how long the player spends in one environment, so you need a composition to be able to stretch in time. You might want that composition to transform based on whether the player is in a fight or is moving from one environment into another. The reason why this is possible is data. Now that we have all kinds of sensors around us, the worlds around us have become as rich in data as the worlds of video games were decades ago, if not richer than that.
Secondly, games are created with user experience in mind. Designers spend a lot of time testing and trying to get environments and interactions exactly right. It’s not enough to design for the expected: when let loose, players will perform unexpected actions, so you have to design with that in mind too. For the age of AI, this means that designers can create algorithms that will react in certain ways to people’s behaviours and then adjust their compositions based on how the people react to them. Media theorist and philosopher Marshall McLuhan said “We shape our tools and thereafter they shape us.” In the case of AI, it’s: “We shape our tools and thereafter they shape us, and thereafter we shape them, and thereafter they shape us, etc. etc.”
Music will be more scaleable than it has ever been. More individualised than it has ever been. AI will be able to work with our everyday environments to create soundtracks for us that hit our reward centres harder, with more consistency, than any single professional artist might be able to do.
But it’s not just music that will become scaleable. So will the artists, as AI will allow us to create avatars of ourselves, so we can exist in many places at once.
The scaleable artist
Infinity does not limit itself to music. Fans will be able to have direct conversations and other types of interactions with their favourite artists. Individually. All over the world. Simultaneously with millions of others.
The artist as avatar presents a new type of interface for music. One that is personal yet depersonalised. It allows people to request, jam together, dance, improvise, and whatever their imagination desires and the avatar’s boundaries will allow.
Avatars will not just consist of existing artists, the way you currently have artists creating chatbots of themselves. This is a trend which actually predates GPT-like AI and goes back to around 2016 when Facebook Messenger opened its API for automations. Yet another trend saw the creation of artists who do not exist as real people. Miquela comes to mind, but also Warner’s attempt at creating an NFT supergroup out of Bored Apes called Kingship.
This personalised delivery method for music is also highly monetisable: you can let people unlock features in the AI, like bringing their favourite artist to a karaoke party, or unlocking a more flirtatious songwriter. Some of what some fans may desire will be creepy. Different artists will set different types of boundaries for what they will and won’t allow their avatars to entertain. Yet, not all avatars will be based on actual people and sooner or later it will be hard to know whether an artist is an actual person or not in the first place.
This interface for music, in my mind, will be one of the most important shifts for music culture. It picks up on networked media trends like influencer culture, metaverse aspects, but also hyper individualism and personalisation. It is straightforward in terms of monetisation in a way that will be much harder to do with other types of music that don’t have a strong user experience aspect designed in (because there will be an infinity of that music, ie. it’s not scarce anymore). This is the format that music has been waiting for since we reached the music record’s end game.
This will take a new place in culture. At first, we’ll see a race of artists (‘real’ and virtual) trying to offer the most exciting and vast ranges of experiences as possible. Yet, most likely, creators of completely virtual avatars will feel much more comfortable to be extremely loose with the boundaries of what fans can do with their avatars. This will present interesting cultural questions as it won’t be the behaviour of the artist as a person that will create their reputation, but rather the way fanbases interact with the avatars that will establish artists as pious or promiscuous (or both), family-friendly or provocative, etc. Boundaries set by creators can guide the way this reputation builds, but ultimately the way the mass interacts with avatars and shares these interactions that will shape artists.
Artists will be as much a creation of their fanbases as they are of the creative force directly behind the artists, including the artist themselves.
To be clear: I am not advocating for this. Rather, I am fairly certain that this will come to pass and in some ways the above can already be applied to stars of today.
We are already there
We might comfort ourselves with the thought that we’re not there yet, but if the last 2 years have shown us anything it’s that things can move fast. Lightning fast.
In the introduction I mentioned the question “what happens when anyone can just make a song with a text prompt?” but I believe this question to be irrelevant, because before we can formulate a meaningful answer, the above will have already come to fruition.
Even with the above just around the corner, we should act like we are already there. What we believe to be true in the future, is true now. If we hold it to be evident, inevitable, the best thing we can do is to behave like it is already the case, at least part of the time.
Afterword
Hey everyone, Bas here! While Maarten’s holding down the newsletter with his excellent weekly commentary and curated links on Thursdays, I will follow a slower cycle focusing on other formats like longer-form pieces such as the above instead.
I hope to see many of you at conferences or events during the year. I have logged out of Twitter for the last time (don’t DM me there - I won’t read it), so if you’d like to know what I’m up to you can follow me on LinkedIn instead.
Love,
Bas
Is music art? If so, what is art?
Is AI only capable of prior derivations?