Luke Girgis has written one of the sharpest analyses of what AI is doing to the music industry’s economics that I’ve read. The scarcity argument holds up. The displacement spiral is real.
The part about the pipeline collapsing, destroying the training ground for the next generation of music professionals, is the most important thing in the piece, and he’s right to name it. But the entire scenario he maps out rests on an assumption so deep he never questions it: that creatives are passive.
That artists sit inside whatever economic structure the industry builds around them and wait to see if it survives. History doesn’t support that. When sheet music publishing collapsed, composers found radio.
When piracy gutted the album economy, musicians found touring, merch, Bandcamp, Patreon. Every supposed extinction event in music turned out to be a forced evolution. The ones who called it death were the ones too attached to the previous form to see the next one taking shape.
This isn’t an AI problem. It’s an imagination problem. The creatives genuinely at risk in any version of 2029 aren’t the ones technology replaced. They’re the ones who spent three years waiting for AI to go away, still trying to optimise a model that had already moved.
The assumption that the future will look like the past, that success still means streaming numbers and sync deals and label investment, is the real crisis underneath the one Luke is describing. In a world of infinite synthetic content, the rarest thing in the room isn’t a great-sounding track. It’s a documented human story. A mythology.
A community built around something real. AI can generate music. It cannot generate thirty years of cultural authority, a specific city, a specific moment, a specific journey that people witnessed and remember.
The door Luke says is closing is not the only door in the building. The middle-class musician he mourns was always one platform decision away from losing everything, because the income was never really theirs to own.
What his 2029 scenario actually reveals, if you read it as a signal rather than a sentence, is where human value relocates when synthetic content becomes abundant. The artists building direct relationships with their audience, converting listeners into members, turning cultural authority into something a community wants to fund and be part of, those artists aren’t in the royalty pool AI is diluting.
They’re running a different economy entirely, one that gets stronger as synthetic noise floods the market, because belonging and authenticity become the scarcest things available. The building isn’t burning.
The map is wrong. The creatives who’ve always found the next form will find it again.
This is definitely a really sobering article. I personally think the evolution will need to be around strengthening and enforcing copyright laws on the music these AI models are being training on FOR FREE that should be the crime. Want to train an AI model on someones catalogue? u gotta pay for it, simple.
LMMs will never be good enough to make entire songs indistinguishable from human artists "in 60 seconds", much less entire playlists' worth of those in that time. To my knowledge now, they require repeated prompts + multiple iterations, in the dozens, to churn out the Lowest Common Denominator kludge-gruel they serve today. And model-collapse (or more likely, intentional model-poisoning by competitors) is slowly appearing over the horizon; bc everyone owning "AI" will stop at nothing to be the only one left standing after the bubble pops. They will corrupt each other's training-data streams, on purpose or inadvertently, + at best their further development will be largely hampered.
Monopolies over live music venues and ticket sales, the buying and trading of copyright by people other than artists, and big music biz investment in streaming that favors artists working for big music biz all preceded AI.
These are the current problems for the independent artist.
The home studio that sounds as good as the hired one is the laptop or even the phone when the artist knows how to use one. And the use of AI tools to make music doesn't necessarily make that AI music. Indie artists using AI for DIY and punk reasons born out of poverty are going to be more common than type 2 luddites if they're not already.
Still, a lot of great music is still made in that middle zone you predict the collapse of. There are places that got there first - medical and scientific publishing, a field I've worked in, was screwed by Maxwell sr long ago and now seems like the model for Spotify.
Luke Girgis has written one of the sharpest analyses of what AI is doing to the music industry’s economics that I’ve read. The scarcity argument holds up. The displacement spiral is real.
The part about the pipeline collapsing, destroying the training ground for the next generation of music professionals, is the most important thing in the piece, and he’s right to name it. But the entire scenario he maps out rests on an assumption so deep he never questions it: that creatives are passive.
That artists sit inside whatever economic structure the industry builds around them and wait to see if it survives. History doesn’t support that. When sheet music publishing collapsed, composers found radio.
When piracy gutted the album economy, musicians found touring, merch, Bandcamp, Patreon. Every supposed extinction event in music turned out to be a forced evolution. The ones who called it death were the ones too attached to the previous form to see the next one taking shape.
This isn’t an AI problem. It’s an imagination problem. The creatives genuinely at risk in any version of 2029 aren’t the ones technology replaced. They’re the ones who spent three years waiting for AI to go away, still trying to optimise a model that had already moved.
The assumption that the future will look like the past, that success still means streaming numbers and sync deals and label investment, is the real crisis underneath the one Luke is describing. In a world of infinite synthetic content, the rarest thing in the room isn’t a great-sounding track. It’s a documented human story. A mythology.
A community built around something real. AI can generate music. It cannot generate thirty years of cultural authority, a specific city, a specific moment, a specific journey that people witnessed and remember.
The door Luke says is closing is not the only door in the building. The middle-class musician he mourns was always one platform decision away from losing everything, because the income was never really theirs to own.
What his 2029 scenario actually reveals, if you read it as a signal rather than a sentence, is where human value relocates when synthetic content becomes abundant. The artists building direct relationships with their audience, converting listeners into members, turning cultural authority into something a community wants to fund and be part of, those artists aren’t in the royalty pool AI is diluting.
They’re running a different economy entirely, one that gets stronger as synthetic noise floods the market, because belonging and authenticity become the scarcest things available. The building isn’t burning.
The map is wrong. The creatives who’ve always found the next form will find it again.
Excellent stuff LG.
Thanks dude!!
Frightening stuff Luke. I fear you are too close for comfort on this one.
This is definitely a really sobering article. I personally think the evolution will need to be around strengthening and enforcing copyright laws on the music these AI models are being training on FOR FREE that should be the crime. Want to train an AI model on someones catalogue? u gotta pay for it, simple.
LMMs will never be good enough to make entire songs indistinguishable from human artists "in 60 seconds", much less entire playlists' worth of those in that time. To my knowledge now, they require repeated prompts + multiple iterations, in the dozens, to churn out the Lowest Common Denominator kludge-gruel they serve today. And model-collapse (or more likely, intentional model-poisoning by competitors) is slowly appearing over the horizon; bc everyone owning "AI" will stop at nothing to be the only one left standing after the bubble pops. They will corrupt each other's training-data streams, on purpose or inadvertently, + at best their further development will be largely hampered.
Monopolies over live music venues and ticket sales, the buying and trading of copyright by people other than artists, and big music biz investment in streaming that favors artists working for big music biz all preceded AI.
These are the current problems for the independent artist.
The home studio that sounds as good as the hired one is the laptop or even the phone when the artist knows how to use one. And the use of AI tools to make music doesn't necessarily make that AI music. Indie artists using AI for DIY and punk reasons born out of poverty are going to be more common than type 2 luddites if they're not already.
Still, a lot of great music is still made in that middle zone you predict the collapse of. There are places that got there first - medical and scientific publishing, a field I've worked in, was screwed by Maxwell sr long ago and now seems like the model for Spotify.