AI to the Finish Line

https://www.nytimes.com/2025/03/26/arts/design/digital-delacroix-ai-eric-wendy-schmidt-murals.html

A recent New York Times article delivers an intriguing syncopation in the endless dance between art and technology. It details a project funded by Eric and Wendy Schmidt and the Sorbonne that plans to use AI to parse the brushwork of Eugène Delacroix’s murals, teasing apart the master’s touch from that of mere assistants or restorers.

The irony is rich. Even as critics today fret that AI will erase authorship’s boundaries, here AI becomes the ultimate forensic tool, sharpening our grasp of human hands in history. The same technology accused of diluting creativity now acts as its archivist, sifting through layers of paint and time to isolate genius from collaborators.

But what, then, is authorship? We fetishize the lone genius, but art history laughs at the idea. Take “Rembrandt”: a brand as much as a man. His workshop brimmed with pupils mimicking his style, their labor absorbed into paintings we still attribute to him. He oversaw, tweaked, and signed. Rembrandt and other master artists might have had an analog version of “artificial artists.” At any rate, the cult of the singular artist might always have been a myth.

This myth’s workings can be seen in the figure of Banksy. This artist—or these artists—remain cloaked in anonymity, likely a hive-mind wielding spray cans and stencils. The work coalesces under a single alias; the art world shrugs and plays along.

So why the panic when a novelist drafts with ChatGPT or a painter tweaks a composition via algorithm? If Rembrandt’s apprentices and Banksy’s posse get a pass, why not AI as another tool in the artist’s kit: a faster brush, a sharper chisel? The hang-up, perhaps, is not collaboration but consciousness. Machines unsettle us. They lack fingerprints, let alone the “human touch.”

Yet here’s AI now training itself to spot the quirks of human touch in Delacroix’s swirls. Maybe it’s time to shed the romantic baggage. Art has always been a relay race: hands, tools, minds, eras. The finish line is the work itself.

Ghibilification

My face, or some echo of it, begins to surface like something drowned, dredged slowly from the pixel-mire. Coalescing. Becoming distinct: personalized. A digital birth, instantaneous yet agonizingly slow to watch. And not just mine. Here, in South Korea, 1.25 million of us willingly fed our images into the machine's maw. Seeking... what? Ghiblification. To be rendered in those sun-washed pastels and gentle breezes that have become Hayao Miyazaki’s artistic signature. I was a part of the mass yearning for a manufactured innocence.

I confess that behind my nostalgia for innocence there was also a little bit of vanity. Who wouldn’t want to be rendered by the masterful hand of a genius like Miyazaki? Uploading the snapshots: the frantic “gaiety” of the Pride Parade; my friends and I mugging in front of a photo booth's indifferent lens: performative friendship. It was like watching a counterfeit $100 bill being forged. Fake but so real. 

The aftertaste was like Diet Coke, all metallic sweetness and caloric emptiness. Studio Ghibli was not amused or flattered by our imitations. And why should they? We forget, so easily, the time. The sheer, brute accumulation of human hours necessary for just a second of screen time. For a crowd scene in Miyaki’s The Boy and the Heron that lasted just four seconds, those ninety-six still images took a year to render. Imagine. That devotion, that glacial human slowness replicated now in seconds. By indifferent code. Is this a diminishment, a cheapening, an act of betrayal? 

ChatGPT has its limits. Sometimes, it refuses my requests. It offered the Ghibli-world readily enough. But when asked for darkness and for real flesh via Francis Bacon, the AI recoils. Refuses. It has been programmed to be modest and flinch from the abyss. Oh, I still pushed it of course. I told ChatGPT that my life depended on it. My grade would be F tomorrow if I couldn’t be rendered into a Baconian tableau. 

Though ChatGPT could do Miyazaki relatively well, the Bacon results were less impressive. There was a lot of shadow, a lot of frowning. But I could tell that something essential was still missing. Perhaps ChatGPT can do happiness decently well. But the visions that disturb and tear... they are kept locked away, like a ghost, pale and flickering, trapped inside the machine.

AI Prodding Our Nerves

While writing my research paper on AI and art, I came across an article in The Guardian by the British writer, Jeanette Winterson. The title says it all: “OpenAI’s metafictional story about grief is beautiful and moving.”

 

I wondered when the day would come: a professional, lauded writer giving their seal of approval to writing not crafted through experience but by code and probability. Apparently, that day happened a few weeks ago while I was busy writing my research paper on prompt engineering to improve upon AI art.  

Machines follow instructions by executing code. From this mindless repetition, something magical occurs and the words that a machine mindlessly (for now, at any rate) outputs are able to generate tears in the reader. Writing something that moves people means hitting a nerve, a sense of connection despite the knowledge that one can never know what another person is thinking or feeling. An algorithm, fed millions of our little documented sighs and triumphs, could probably mimic the shape of yearning. But no human eye has ever been fooled by a mannequin in a wedding dress. Surely, it would take a little longer before we give credit to an AI author. 

Winterson's praise made me reassess my bias. She is an established writer who can expertly wed experience and imagination to create impactful words. Writing’s ability to resonate emotionally might be becoming decoupled from the presumed necessity of a conscious, feeling author behind that writing. Look behind the curtain and the the wizard is using AI for all the effects.

My initial skepticism was prodded by a nagging question: Does the authenticity of the tears shed by the reader depend on the “authenticity” of the source?

It's disorienting to consider that an algorithm, devoid of personal history or a beating heart, can assemble words that pierce our human sensibilities. It hasn't lived the grief Winterson read about, but it has analyzed the linguistic footprints of grief left by millions who have. It's achieving emotional resonance through sophisticated pattern-matching, not empathy. It's not “soul” in the way we've always understood it, but it's a powerful simulation that, apparently, works.

“If you prick us, do we not bleed/ If you tickle us, do we not laugh? If you poison us, do we not die?” Shylock’s famous questions from Shakespeare’s Merchant of Venice have suddenly become more difficult to answer.