No account required
With most current research into AI generated music, the buck stops at the model’s ability to fool humans into thinking the piano piece it generated was actually composed by Bach. Modern music seems somewhat uncrackable, yet what we have seen with the recent DALL-E-2 explosion, indistinguishable imitations of pop and electronic music could be just around the corner.
This raises interesting questions about how the modern music landscape as we know it will change, and perhaps, already is changing. Does this mean doom for the common musician?
For the sake of this post, let's assume a new AI, MOLL-E, has been made available to use for everyone, on the same creative license as DALL-E. This can be used by anyone to create any music they can conceive of, including vocals and lyrics, in any style using just a text prompt.
Starting with huge, major artists, I think very little will actually change for this demographic. Infact, their process now will likely only become easier with MOLL-E.
Let’s imagine MOLL-E creates a song which perfectly synthesizes their voice. The artist has now made a song without paying for studio time, mixing engineers, mastering engineers or ghost writers. Most pop songs are ghost-produced by other artists anyway, so using MOLL-E probably fits right into their workflow and replaces the ghost producer overhead. All they need to do is turn up to live events and perform.
The success of top pop stars is not down to the unique creativity in their output. The I–V–vi–IV chord progression famously dominates pop music, and listeners are often not drawn to these artist for deviations from the musical status-quo.
These musicians rely on branding, image and relatability to their target audience. When a fan says they love a pop artist - they likely mean exactly that. Their image and persona are more attractive to them than the music itself. That’s not me saying they don’t like the artist’s music - of course they do, but the main draw to the artist is their image.
So these artists will be fine. Their work may even become easier, and maybe we even see a larger pool of these types of people, similar to the Instagram influencer boom creating a whole slew of paid supermodels.
Does this harm the industry and remove the creative element of producing music? Yes, probably. Will their fans care? I doubt they’d even realize.
But what about artists that don’t rely on image to succeed? For those who create music in smaller, niche genres - this is where I start to worry.
As a personal fan of less popular, ‘underground’ music scenes, what keeps them alive is innovation and a desire to push the boundaries of a genre forward. As we have seen with DALL-E, AI will likely imitate in order to create something new, rather than being truly creative.
Note that my definition of ‘truly creative’ is the process of combining two or more influences together to create something else, and there’s really nothing to say that a future AI couldn’t do that.
So in the same way it became easier for large artists, it also becomes easier for the small ones. And when everyone can create a song automatically in 5 seconds, suddenly everyone is a music producer, and the world becomes flooded with AI generated music.
This will give us an ever expanding flow of new music that suits our tastes. But perhaps this all becomes noise - I would be lying if I said I wasn’t already fairly desensitized to DALL-E’s work. While it’s impressive, the only aspect that has truly captured me (bearing in mind I’m not an art buff) is the ability to ask it deep and introspective questions and see how it responds. For me the art itself is lacking something human, perhaps by its nature of being instantly producible and infinitely repeatable.
It’s worth pointing out though that this flooding of the market phenomenon has already happened, albeit to a much lesser scale. Music production is more democratized than ever; dance music hits which would have taken thousands of pounds worth of studio equipment can now be made on a laptop. The numbers support this, with 35,516 albums being released in the year 2000 and 79,695 albums being released seven years later. With over 10 years having passed since then, and Spotify now adds 60,000 new tracks each and every day.
But rather than destroy the music industry, it instead just changed our method of consumption from physical releases to digital streaming. You can now listen to music online without even knowing the smaller artist that made it, which perhaps paves the way for AI generated music to abstract your knowledge of the artist even further.
For certain use cases, such as ambient background music, I actually think this can be beneficial. Throw on a playlist and go do whatever it is you want to do. Ambient music as a genre, which contains artists you know and like, can co-exist with this in an active vs passive listening sense.
You could even become your own small artist. Tell MOLL-E what you want to listen to, and it generates a playlist of fresh music just for your listening pleasure.
With practically infinite music being pumped out into smaller genres, I do however worry that this music will become lackluster, and it’ll become harder and harder to find diamonds in amongst the rough. Maybe this makes those gems all the more sweet, but I can’t be sure.
Also as a side note, if everyone is now a music producer, that means the money paid to artists dries up in its current format. Similar to Instagram influencers now taking private sponsors, new money avenues will need to be found, as the streaming platform royalties pot only goes so far.
As I just outlined, it goes without saying that MOLL-E will exponentially increase the number of people attempting to become artists, which poses the question of how this will be handled by streaming platforms in the short term.
Recently, we saw Getty Images block photos generated by DALL-E & friends from being uploaded to their site. They cited a worry around the grey area surrounding the copyright of images used in training data, and it’s not inconceivable that music streaming platforms won’t do the same.
However it does make you think - where would MOLL-E’s training data, and MOLL-E itself, come from in the first place? Perhaps from a company such as Apple who has access to pretty much all music, and are known for skirting around legal issues by throwing lawyers at lawsuits? But I digress.
Spotify for one already has a ton of playlists which contain music from artists with no given identity. A percentage of these could already be AI produced. The ambient music here on Flowful is all algorithmically generated, snippets of which I have posted on Spotify.
Personally, I think these platforms will embrace the AI artist. Other than the decreasing royalties pot, there’s almost no reason for them not to. They can pad out their playlists and give people more to listen to.
And even though I would still listen to the AI generated music, I for one would favour a seperate section or even streaming service that only contains ‘purist’, human-made music.
This last point here is something I find exciting. In my opinion, there's a difference in the underground music scene between producers, DJs and people who do both. It’s becoming increasingly rare to find pure DJ’s who don’t produce - those who spend all their time collecting music and crafting perfect sets. This is in contrast to producer/DJ hybrids who are often more comparable to a live set as they play a significant amount of their own music.
AI could bring back a new wave of pure DJs similar to those of old. The DJ will now have an infinite collection of music they deem fit, which puts them in a position to once again become artists in their own right.
Those who get into producing just so that they can make a name for themselves as a DJ can instead focus solely on DJing. The art form of true DJing may become loved once again.
And the output of a producer, what with there being so many of them, will be under scrutiny more than ever, and standing out will become more difficult. Live sets, even underground ones, may be that route to doing so. Some of the best small dance music shows I’ve seen have involved a producer creating music in front of the crowd, rather than a producer DJing their tracks.
TLDR; more music will be released and it’s up to you to decide if it bothers you that it was made by an AI.
If you’ve made it this far, you might be interested in the algorithmic music that I make here on Flowful. Each track is procedurally generated in real time which means it lasts forever and is always varied.
I have a strong input as to how these are made - it’s not quite at MOLL-E’s level yet. Infact, it's more like music I’ve written is being performed by algorithms. But, baby steps… our AI overlords will get there eventually.