This got me mulling over what would happen if AI became better at writing pop songs than humans.
So let's game it out...
Most of the pop hits of the last 30 years come from a small handful of songsmiths. Most pop stars already don't really write their own songs, they are more the 'Figurehead' of the whole experience. Their overall look and stage presence are the major selling point (some of them can't really sing without autotune, looking at you Britney).
So a handful of songwriters will be out of work, and nothing much will change. Mayyybe a few artists will sign away their likeness and voice to churn out album after album, hoping the public is none-the-wiser. Pop music may become still more homogenous, since part of the huge pop hits is the unexpected twist worked in that makes it "pop".
It'll be kind of a reverse Hatsune Miku, but things won't change that much. I think the appetite for a completely synthetic pop star is not there in the West - they'll have to be tricked into it. Pop stars are already arguably artificial in some form.
I think what you're describing is basically something I've pointed out in different contexts: for many areas concerned with AI, content is already effectively written by "AI", it's just humans following the algorithms rather than fully automated.
- content marketting is already AI writing, it's just humans churning through the SEO optimized algorithms.
- Mass digital stock photograph is AI generated "art". People that do that for a living have an exact formula for reproducing images that will sell en masse.
- As you point out, a good chunk of generic pop music is churned out by musicians following algorithms behind the scenes and then manufactured to look like the work of a public performer.
The parts of our content work threatened by AI are already cold, sterile, and machine generated.
> So a handful of songwriters will be out of work, and nothing much will change.
It would be more than that, though, if AI truly becomes as good at writing pop songs as that handful of songsmiths.
Currently, those songsmiths are the limiting factor for pop stars. As you said, there aren't many, and they can only write so many songs per year. Because of this, choosing which pop stars get to sing the songs they write is a big factor in determining who becomes successful. If suddenly AI can pump out as many pop hits as we like, that limiting factor goes away and the whole market changes.
>Currently, those songsmiths are the limiting factor for pop stars.
Anything to support that conjecture?
Pop songs take off because they're marketed properly AFAICT. Writers choose from many songs, all good, producers largely decide which songs and groups will make it. Songs can be around for years before being given to a star to make into a hit.
Of course marketing now can be 'going viral', and that can be various degrees of organic and paid promotion.
I'd take a step back, and ask what makes art (music, visual, etc) valuable.
To that, I'd respond "scarcity."
It's hard to imagine highly valued art (either culturally or monetarily) without scarcity.
Consequently, I expect the next few decades will see similar progressions to the past, when technology encroached on scarcity (e.g. painting after the photograph, live performance after consumer video/music playback).
Scarcity will reestablish itself in whatever guise remains technically feasible, and again become valued.
In the case of LLMs and diffusion, I expect it will be creating things that are so novel they could not have come from AI.
Hopefully skillful, deep parody and the absurd will reassert themselves, as post-JS-Daily-Show I think that's been missing in culture.
But it sure as shit isn't generic pop music, which will be the first thing to be churned out of humanless hit factories and flood the market.
>It's hard to imagine highly valued (either culturally or monetarily) art without scarcity.
I disagree entirely. The Mona Lisa's value is greatly expanded by its free availability -- the one-off image on my wall is scarce, but it will never be a cultural icon, that is antithetical to scarcity.
A value of art is reflection (the mental process), but this is magnified enormously if society can reflect on the work, reference it, abstract from it, view it and develop freely from it.
Great pop songs are great because everyone has heard them and we have shared experiences around them. There is no such thing as a scarce pop song.
(This is one of the great crimes that arise from copyright terms being great than a couple of decades, society doesn't get to riff on the important icons of its recent past.)
Is the Mona Lisa print on your wall distinct from the original? I.e. Was her face shape changed in your print to a version you prefer?
Arguments against against scarcity that point to copies of a singular work seem circuitous. There are innumerable copies because the single, canonical version is so valued.
Similarly, great pop songs are valued because everyone knows them, which is why individual song value peaked during the dawn of mass distribution (CD, then early digital) but before the market was flooded with volume.
The scarcity in the pop sense is the limited number of songs that everyone knows. (Largely through sophisticated media campaigns and forced radio/channel placement, but I digress)
If the Mona Lisa didn't exist we'd elevate some other work more. If not Bo'Rap then Stairway to Heaven, etc.
>Arguments against against scarcity that point to copies of a singular work seem circuitous.
It's a good point, but I think it misses something. It is experiences of the effect of an artistic work that matter, not the work per se. Experiences of the work are 10-a-penny -- tens of millions have probably seen one of the 'originals', billions have probably experienced the effect of the artwork (which I will deftly avoid defining ;o)) through photos, copies, videos, imitations, and emulations.
> I'd take a step back, and ask what makes art (music, visual, etc) valuable.
> To that, I'd respond "scarcity."
What is your definition of valuable here?
If you're referring to value to culture/society, I think you're very far off-base. The most culturally valuable artistic works are ubiquitous, the opposite of scarce. Art isn't really able to have any culture influence if it only impacts a small number of people.
If you're referring to monetary value, you're also dead wrong lmao. Look at the top 100 most-paid artists of the last decade, and tell me how hard it is to find and appreciate their entire artistic catalogue for yourself.
The argument that scarcity = artistic value doesn't have any basis in fact, and is the sort of thing that would only be shilled by someone trying to con you into buying an NFT.
> It's hard to imagine highly valued art (either culturally or monetarily) without scarcity.
hahahaha what? Compare the monetary and cultural impact of that one "ultra-scarce" Wu-Tang album (monetary: $2m, cultural: none) to the impact of Taylor Swift's last album, which is available on every streaming service (monetary: $200m+, cultural: very high)
Valuable as in culturally significant and monetarily expensive.
> The most culturally valuable artistic works are ubiquitous, the opposite of scarce.
Not so. Copies of those works are ubiquitous, but there is a singular, definitive work.
Name me a handful of world-famous works for which there are multiple, almost-indistinguishable but distinct copies.
The Mona Lisa has a few original alternates, and yet they pale in value to the famous one. Which itself, ironically, became popular famous mostly through being stolen (scarcity).
> Look at the top 100 most-paid artists of the last decade, and tell me how hard it is to find and appreciate their entire artistic catalogue for yourself.
> [Once Upon a Time in Shaolin] vs [Speak Now (Taylor's Version)]
Total artistic renumeration, especially in the modern period, is dominated by distribution volume.
But if we're talking about single work valuation, the Wu Tang album costs $2M.
> Copies of those works are ubiquitous, but there is a singular, definitive work.
When referring to recorded music, this isn't a distinction that has ever actually mattered in the real world, just a fiction made up to shill NFTs.
Are you going to pretend that anyone actually cares about a "singular, definitive FLAC file" that all of the streaming services' FLAC and MP3 playbacks are based on? This is pure fantasy, the copies are the same thing as the original piece.
The idea that Mona Lisa's (or any other artwork's) cultural influence comes from its scarcity is hilarious. Literally anyone can visit the Louvre and appreciate it for themself. Do you think it would have anywhere near as much influence if it was hidden behind closed doors and only 1 person was able to see it?
> But if we're talking about single work valuation, the Wu Tang album costs $2M.
Taylor's album costs $15.
Last time I checked, the sum of revenue from their discography is how artists and labels get paid, not based on the maximum amount that 1 person is willing to pay.
Speak Now is a single work, and it generated like 100x as much monetary value as Shaolin (with like 10,000x as much cultural impact). And those estimates are extremely conservative, when you consider that you can tour and sell merch off an album that people can actually listen to lol.
Literally every world-famous work has replicas and recreations, what's your point? Those copies are also part of the work's cultural influence, and in many cases (if the replicas are sold by the original artist) part of the monetary value as well.
This doesn't provide any more credence to the falsity that art's scarcity is the source of its value (when overwhelming evidence proves that the exact opposite is true)
> The one officially-blessed Taylor Swift album. The one version of Beethoven's Fifth.
Millions of people have bought vinyls/CDs/MP3 replicas of these albums. These are not scarce works. The demand is not for the scarce original (who even knows what that means when it comes to music), it's for a faithful recreation of the artist's work and creative output.
On top of that, nobody who listens to these albums really just wants one album... they want to listen to music that moves them. Taylor Swift fans will listen to other albums that she puts out, and music from other artists that they enjoy. When Beethoven released his 6th symphony, it didn't make his 5th any less valuable to those who enjoyed it.
Because for most people, the value of art comes from its intrinsic beauty, not its "scarcity". Unless if you're someone whose only attachment to art is as a vehicle for financial speculation, which is sad but unfortunately common in NFT circles, etc.
> But if that's not true, it should be possible to point to, say, a series of similar paintings or musical compositions that are all famous.
Have you heard the term genre before? It's literally a word for a group of similar musical compositions. Most genres contain plenty of similar compositions that are all famous.
If people enjoy a work of art, there is almost always other popular art made that's similar to it. Turns out, you get bored if you just listen to one album over and over again.
When it comes to art, most people enjoy variety, it doesn't "confuse them" lol. This is clearly reflected in the market, as there are hundreds of unique genres of music, each with thousands of unique artists.
I think something can be valuable and abundant, what you really mean is what makes the value capturable. If truly good songs were cheap to create then people making good songs won't be able to charge as much as they do, but consumers will arguable experience more value through an abundance of good songs.
>consumers will arguable experience more value through an abundance of good songs.//
It is people's shared experiences of a song that makes a song most valuable (it's not the only value, of course). Having more songs means less shared experiences, more is less.
Imagine going to a club where the DJ can play any of a million songs but only one person knows that song; compared to them playing a setlist of floor-filling bangers ...
"It is people's shared experiences of a song that makes a song most valuable"
You are definitely right that abundance leads to a decline in shared experiences. I’m the only person I know who listens to the music that I do, but of course I treasure that music. But beyond my anecdotal experience, from sites like Last.fm it appears that listeners really began to fragment by the 2010s; with so much on offer now, people don’t necessarily listen to the same music as even their closest peers.
DJs, too, have spoken about the decline of the well-known banger when they are being flooded with hundreds of new tracks every week. Moreover, computer mixing today means those tracks might get so cut up by DJs (e.g. taking a bass line from one track and a synth line from another) that they become well-nigh unrecognizable to even the savviest trainspotters.
The herd vs merit distinction is interesting, and I think they both have value. Maybe differently to different people, but non-zero.
Imagine a handful of objectively "good" songs.
Play them for a crowd where nobody knows them.
Or, in a more humorous example, imagine the Beatles played tracks only from their second album, while touring for their first album. Would audiences be disappointed?
I struggle to imagine an overly negative reaction. People love hearing songs they love, but they also fell in love with those songs for a reason. Well executed music is well-executed music.
Yes, agreed, music has a particular value to a person (I was going too say 'intrinsic' but it is personal) as well.
But shared love of music, or shared experiences of music transcends this value IMO; and is largely orthogonal.
"That song we danced to", "what we sang at the campfire", crab dance for LTT watchers, 'easy for ENZ' for CSGO players, a national anthem to a nationalist, ... the music matters a little but it's only really a rallying point for shared experiences.
You can buy a copy of Vincent van Goghs Starry Starry Night for a few dollars, totally legally. There is no scarcity here.
The only meaningful difference between it and the one in the Museum of Modern art is that it was painted by the Van Gogh directly. And sure it has a hefty price tag attached, but everybody who buys a copy buys it because they like it.
Growing up music was rare, but with access to Spotify I can listen to so much more music. That has not meant I appreciate music less. Sabatons "Primo Victoria" is still going to pump me every time I hear it and Bethovens ninth is as uplifting as it ever was, no matter how many times I listen to it.
Why do you think there will still be pop-stars? I'll just be able to generate music I like. Quite possibly I'll be able to provide AI generated avatars to help my meat brain associate menu options (with images) with a particular music. Possibly there will be curated such avatar/music generators, and more than likely, many of them will be naked. Hatsune Miku is a red herring I think. I expect we'll see more Milli Vanillis too.
Completely agree, and definitely what I was saying.
> Possibly there will be curated such avatar/music generators, and more than likely, many of them will be naked.
I would make the case that pop musicians would be less popular if they freely got naked. The teasing of maybe being able to see a bit of rarified (and real) flesh is definitely a big part of the equation, which you're not likely to be able to reproduce with what people think is an AI.
>I would make the case that pop musicians would be less popular if they freely got naked.
If the system knows you are a prude, then they won't get naked for you. They'll get very good at learning "how much is too much" for any given person, for any given "what" (skin, private messages, flirting, etc).
Human pop stars can't have relationships with every fan. AI pop stars can.
> Human pop stars can't have relationships with every fan. AI pop stars can.
I don't know if it will go that way, to be honest. But certainly, the film Her does show us what that might end up looking like. (Even though she was just a disembodied voice)
> It'll be kind of a reverse Hatsune Miku, but things won't change that much.
I don’t really agree with this for two reasons:
1. It is understood that vocal synths (e.g. Hatsune Miku) are not humans and that humans are not vocal synths. There are also humans covering songs originally sang by synths and vice versa, but everybody still understands what’s going on. (Heck, there are even synths based on the voice of real human singers, e.g. KAFU; it’s still understood that the singer and synth aren’t the same thing.)
2. Vocal synths are not fully AI and require a significant amount of ‘tuning’ to sound good (with various producers generally having their own ‘sound’ due to differences in tuning style). They are also just a voice, they don’t generate any instrumentals. Therefore, a human-produced song sang by a vocal synth is still a long way off from a fully AI-generated song.
Yeah, interesting to see how it plays out but my gut feeling is that it’ll get from bad to worse. Hope I couldn’t be more wrong and this will shake up the industry but with generated music and cloned voices I’m almost inclined to not even compare them. It’s a new thing altogether.
So let's game it out...
Most of the pop hits of the last 30 years come from a small handful of songsmiths. Most pop stars already don't really write their own songs, they are more the 'Figurehead' of the whole experience. Their overall look and stage presence are the major selling point (some of them can't really sing without autotune, looking at you Britney).
So a handful of songwriters will be out of work, and nothing much will change. Mayyybe a few artists will sign away their likeness and voice to churn out album after album, hoping the public is none-the-wiser. Pop music may become still more homogenous, since part of the huge pop hits is the unexpected twist worked in that makes it "pop".
It'll be kind of a reverse Hatsune Miku, but things won't change that much. I think the appetite for a completely synthetic pop star is not there in the West - they'll have to be tricked into it. Pop stars are already arguably artificial in some form.