This resonates. I just finished a 3-part series on Immutable Mobiles called The Safety Layer vs. the Social Primate that digs into this exact pattern.
AI safety filters don’t just block explicit content—they erase social signal: eye contact, ambiguous smiles, power dynamics, erotic tension. Anything that suggests agency, danger, or desire—especially in women—gets scrubbed. Not because it’s obscene, but because it’s uncategorizable.
We called it castration by style: flattening art until it’s smooth, safe, and emotionally inert.
Thanks for naming what so many of us felt on Ghibli Day. The machines can render beauty—but they flinch at presence.
Painters responded to the invention of photography by expanding their themes, techniques, and horizons. In this case, I fear, that might not be enough, because, it doesn't matter what you do, it can be imitated in a second.
Style can be reproduced, and that truth has been demonstrated to the extreme. But training on style cannot give current models the creativity to speak novelty through the style. The mad burst of people using the Ghibly style for a wild variety of purposes (albeit with the typical statistical distrubution of success) demonstrates the value of making style freely usuable.
I think the same applies to formal languages for computors, the agent/tool concept of system design, and all the other directions in which LLM-based systems are pushing into domains that were previously unique to human artists and builders. It may be true that hammers, sufficiently advanced can build houses, but no matter how advanced a hammer may become, it will never be able to design based on needs and purposes that advanced Large Hammer Model can recognize on its own.
The parallel with photography seems moderately convincing to me. Everything you write here would have been relevant -- and probably was said, roughly -- in 1830.
I read your comments as envisioning the death of art -- at least the most meaningful sector of it. And I suspect people said the same thing two hundred years ago. Didn't happen. So, yeah, I do see a parallel. Not going to happen this time either. I know a number of contemporary artists and I cannot imagine any AI technology that is currently out there -- and since I read your newsletter I know something about that -- replicating anything like what they do.
This isn't about the *death* of art. It's about the degradation. It has happened with many other things. I'm not sure there's a good argument against that.
OK. I also think it is pretty likely that lots of people thought the invention of photography would *degrade* art. Did it? Assuming that you think otherwise, what happened to prevent it and why isn't the same thing going to happen this time? Do you think it is reasonable for me to ask you to compensate for the doomster bias in our species?
The banana duct taped to a wall that was sold for however many thousands of dollars as “modern art” has entered the chat. Lol. Yeah, Art is definitely degraded.
Everything degrades. Borrowing heavily from Neil Postman:
Social media HAS degraded our social fabric, increasing the distance between and reducing the overlap among the views of our political parties.
And smartphones did cause an increase in anxiety and depression.
And television DEFINITELY made us vapidly dependent on every experience being maximally entertaining and extremely stimulating.
And the telegraph DID make us obsessed with trivial (to us) news from far away places that we have no ability to impact, causing us to feel more powerless in our lives.
And the printing press DID create more ossified dogmatic thinking.
And written language itself DID make us less capable than ancient Greeks of memorizing long texts and doing our own critical thinking. Just as Socrates supposedly warned.
All of this is true, AND I wouldn’t be able to talk to you about it on this verifiable miracle beneath my thumbs if none of it had happened.
We lose something with every revolution. We also gain things. The world keeps turning. But it helps no one to pretend the degradation isn’t also happening.
How can you have possibly understood the article without knowing that!! I'm referring to last week, when everyone went crazy making Ghiblified images with ChatGPT
No, it's not literally a day of Ghibli, I called it that. Put “Ghibli” and “ChatGPT” on X and search. That's what I'm talking about. But if you didn't live it at the moment, the effect won't be the same
Inflation comes to mind, the gradual gnawing and eating and digestion and defecation of any and all value. We must remember the old axiom: everything in moderation, including in moderation.
This resonates. I just finished a 3-part series on Immutable Mobiles called The Safety Layer vs. the Social Primate that digs into this exact pattern.
AI safety filters don’t just block explicit content—they erase social signal: eye contact, ambiguous smiles, power dynamics, erotic tension. Anything that suggests agency, danger, or desire—especially in women—gets scrubbed. Not because it’s obscene, but because it’s uncategorizable.
We called it castration by style: flattening art until it’s smooth, safe, and emotionally inert.
Thanks for naming what so many of us felt on Ghibli Day. The machines can render beauty—but they flinch at presence.
Painters responded to the invention of photography by expanding their themes, techniques, and horizons. In this case, I fear, that might not be enough, because, it doesn't matter what you do, it can be imitated in a second.
Style can be reproduced, and that truth has been demonstrated to the extreme. But training on style cannot give current models the creativity to speak novelty through the style. The mad burst of people using the Ghibly style for a wild variety of purposes (albeit with the typical statistical distrubution of success) demonstrates the value of making style freely usuable.
I think the same applies to formal languages for computors, the agent/tool concept of system design, and all the other directions in which LLM-based systems are pushing into domains that were previously unique to human artists and builders. It may be true that hammers, sufficiently advanced can build houses, but no matter how advanced a hammer may become, it will never be able to design based on needs and purposes that advanced Large Hammer Model can recognize on its own.
Of course. But as Ghibli Day proved, many people don't care about that, as long as they get the cute image they want.
The parallel with photography seems moderately convincing to me. Everything you write here would have been relevant -- and probably was said, roughly -- in 1830.
Right. And isn't it relevant today as well? I don't see how that takes from its current relevance.
I read your comments as envisioning the death of art -- at least the most meaningful sector of it. And I suspect people said the same thing two hundred years ago. Didn't happen. So, yeah, I do see a parallel. Not going to happen this time either. I know a number of contemporary artists and I cannot imagine any AI technology that is currently out there -- and since I read your newsletter I know something about that -- replicating anything like what they do.
This isn't about the *death* of art. It's about the degradation. It has happened with many other things. I'm not sure there's a good argument against that.
OK. I also think it is pretty likely that lots of people thought the invention of photography would *degrade* art. Did it? Assuming that you think otherwise, what happened to prevent it and why isn't the same thing going to happen this time? Do you think it is reasonable for me to ask you to compensate for the doomster bias in our species?
The banana duct taped to a wall that was sold for however many thousands of dollars as “modern art” has entered the chat. Lol. Yeah, Art is definitely degraded.
Everything degrades. Borrowing heavily from Neil Postman:
Social media HAS degraded our social fabric, increasing the distance between and reducing the overlap among the views of our political parties.
And smartphones did cause an increase in anxiety and depression.
And television DEFINITELY made us vapidly dependent on every experience being maximally entertaining and extremely stimulating.
And the telegraph DID make us obsessed with trivial (to us) news from far away places that we have no ability to impact, causing us to feel more powerless in our lives.
And the printing press DID create more ossified dogmatic thinking.
And written language itself DID make us less capable than ancient Greeks of memorizing long texts and doing our own critical thinking. Just as Socrates supposedly warned.
All of this is true, AND I wouldn’t be able to talk to you about it on this verifiable miracle beneath my thumbs if none of it had happened.
We lose something with every revolution. We also gain things. The world keeps turning. But it helps no one to pretend the degradation isn’t also happening.
Exactly. This is Scott Alexander's point. And he goes even further. If you haven't read his essay, it's super good
By the way, what was "Ghibli Day"? First I have heard of it.
How can you have possibly understood the article without knowing that!! I'm referring to last week, when everyone went crazy making Ghiblified images with ChatGPT
Is this it? https://festival2025.com/ghibli-festival-2025-an-enchanting-celebration-of-magical-worlds/
No, it's not literally a day of Ghibli, I called it that. Put “Ghibli” and “ChatGPT” on X and search. That's what I'm talking about. But if you didn't live it at the moment, the effect won't be the same
Inflation comes to mind, the gradual gnawing and eating and digestion and defecation of any and all value. We must remember the old axiom: everything in moderation, including in moderation.
This is for Altman and openai: thanks for stealing and bragging about it.