Ridiculous Opinions #282
I experimented with AI to bring my art thief movie idea to life...
So, if you missed last week, I was having some adventures with AI. My adventures with Claude did some very cool things, but I learned some of the limits of what Claude was able to do.
But in my adventures, I wanted to be more ambitious. I started to have visions of what AI might be able to do for me. Thus, I decided to do some further experiments.
Each day in my journal, I will write down an “Idea of the Day”. Longtime readers have probably heard me speak of this. Basically, it’s a warm-up exercise for me where I try to think of something to write about from scratch. It could be an idea for a story or a character, or even an app. I have, literally, thousands of these ideas in my journals.
One day last week, I was writing crime-based ideas. Basically, I was trying to come up with concepts for heist movies and/or robbery movies that did not involve murder and killing. In the process, I came up with a great idea for a movie about an art thief.
The basic concept of the movie is that there was a master art thief that was put in jail twenty-plus years ago and has been there ever since. But when a series of thefts start to happen with his particular signature style, a very reluctant detective gets him out of prison to help the detective investigate the thefts. It was going to be a bit of a buddy-cop film with a very grumpy detective having to deal with an art thief who was just fascinated with the world around him. After all, the art thief had missed the thrust of the tech revolution, so he has never seen things like an iPhone or high definition televisions or streaming. The thief becomes a bit like a child when presented with all of this cool technology and the detective is super-pissed that he has to deal with him. It was a fun premise, one that I would probably never touch again (I’ve got thousands of these ideas…I’m not kidding…).
But I had a notion: What if I fed this idea to Claude to see if he could flesh out a story?
The idea that I wouldn’t have to do the hard mental labor of coming up with a story for this pretty throwaway premise was very appealing to me. Sometimes, I will labor over a concept for a very long time, because telling a story like this is hard. What if I was able to just feed Claude my basic story idea and see what it would come up with?
I decided to try it. I fed Claude the following prompt:
Write me the synopsis of a screenplay. The screenplay is about a series of robberies of important art around the world. Someone is going through art galleries and private collections, stealing extraordinarily valuable pieces of art that are worth hundreds of millions of dollars. An Interpol detective has been assigned the investigation to find the art thief. However, the only thing that he can find is a series of clues that point to a specific person as being the thief. The problem is that that thief has been in jail for the last twenty years for stealing many valuable works of art. So, the Interpol agent has an idea. When he checks on Pierre, the French thief who has been languishing in jail for so long, Pierre is still in jail. His idea is that he will get Pierre out of jail in order to figure out how to catch the "new" thief. Thus begins a globe-trotting, buddy comedy about the uptight Interpol detective (a British man named Nigel) and Pierre, who is happy to be out of jail and absolutely astounded by the pace of technology and the wonders of the modern world. There should be three big heist set pieces in the synopsis that involve elaborate trickery in order to steal the art. Each of those set pieces should take place in different, exotic European countries and involve intricate details on how to avoid security systems that would keep the audience on the edges of their seats. The final, large scene should be the theft of one of the most famous paintings in the world, as Nigel and Pierre try to stop the art thief (who has an identity related to Pierre). In the end, they will stop the thief and Pierre will escape his captivity, as does the mystery thief. We should be surprised at who the mystery thief is. This should be a well fleshed-out synopsis with important details that enable me to write a well-formed screenplay from it.
Good enough, right? It’s got a basic concept for a movie and a particular (if vague) structure. Claude went to work and within 30 seconds, it spit this out:
"MASTERPIECE"
Synopsis
Act 1
In the elegant halls of the Louvre, a priceless Vermeer vanishes without a trace. No alarms triggered, no guards alerted, and the security footage reveals nothing. This is the seventh high-profile art theft in three months, with paintings worth over $600 million missing from galleries across Europe. The thief leaves behind only a single white chess piece—a knight—at each crime scene.
NIGEL HARRINGTON (45), a methodical, by-the-book Interpol detective with a reputation for solving impossible cases, is assigned to the investigation. Divorced and married to his job, Nigel's apartment walls are covered with case files and timelines. When analyzing the theft patterns, he discovers something remarkable—they mirror the legendary heists of PIERRE DURAND (65), a French master thief who has been imprisoned in a maximum-security facility outside Paris for the past twenty years.
Nigel visits Pierre, a charismatic, silver-haired Frenchman who maintains an air of sophistication despite decades behind bars. Pierre is intrigued by the copycat but insists he knows nothing. However, when Nigel shows him photos of the crime scenes, Pierre notices details only a true professional would recognize—his protégé is leaving him messages.
Convinced Pierre is the key to solving the case, Nigel proposes an unorthodox plan to his superior, DIRECTOR COLLINS (60): release Pierre into Nigel's custody to help catch the new thief. Collins reluctantly agrees, but with conditions—Pierre must wear an ankle monitor, and if he attempts to escape, Nigel's career is finished.
ACT II
Pierre steps out of prison into a world he barely recognizes. He marvels at smartphones, self-driving cars, and drone deliveries. His childlike wonder contrasts sharply with Nigel's rigid personality, creating immediate tension. When Pierre spots a bakery and insists on stopping for a proper croissant—his first in twenty years—Nigel reluctantly agrees, establishing their dynamic.
FIRST HEIST - VENICE, ITALY
Intelligence suggests the thief will target the Peggy Guggenheim Collection. Nigel and Pierre arrive to assess security and discover an elaborate system of laser grids, pressure-sensitive floors, and biometric scanners. Pierre identifies vulnerabilities immediately, explaining how he would bypass each measure.
That night, they stake out the museum when suddenly the power goes out across the district. Pierre realizes it's a diversion and leads Nigel to the canal entrance—the thief's actual entry point. They catch a glimpse of a figure in black slipping away on a speedboat with a Picasso under their arm. Pierre smiles, impressed.
Later, examining the scene, Pierre finds a hidden message only he would understand—coordinates to Budapest and a reference to "the first lesson."
BUDAPEST, HUNGARY
In Budapest, Pierre takes Nigel to an old apartment building where he once trained young thieves. He reveals that twenty years ago, he mentored several promising students before his arrest. The most gifted was ISABELLE RENARD, a young woman who Pierre thought of as a daughter.
SECOND HEIST - BUDAPEST, HUNGARY
The coordinates lead to the Hungarian National Gallery, where a Rembrandt exhibition is taking place. Nigel and Pierre attempt to warn security, but the staff dismisses them. Pierre educates Nigel on the art of misdirection, explaining that great thieves never take what seems most valuable.
That night, they watch as an elaborate water main break forces the evacuation of the museum. Amidst the chaos, Pierre spots the pattern—the thief is inside, disguised as emergency personnel. They give chase through the labyrinthine halls of the gallery. The thief, face concealed, demonstrates acrobatic skills identical to Pierre's techniques.
When they corner the thief, Pierre hesitates, allowing their escape. The Rembrandt remains safe, but a less valuable (though still worth millions) Dutch masterpiece has vanished. Nigel is furious until Pierre explains—this was never about the art. The thief wanted them to follow.
In the stolen painting's frame, they find another message—coordinates to Switzerland and the words "final exam."
Nigel confronts Pierre about his hesitation. Pierre confesses that he believes the thief is Isabelle, his former student who disappeared after his arrest. He trained her since she was a teenager, taking her in when she had nowhere else to go. Their relationship was complex; she wanted more than mentorship, but Pierre kept boundaries. Her disappearance always haunted him.
ACT III
THIRD HEIST - ZURICH, SWITZERLAND
The coordinates lead to a private collector's heavily fortified mansion outside Zurich. The collector, HEINRICH MUELLER (70), possesses one of the few authenticated Da Vinci paintings in private hands.
Upon arrival, they discover Mueller knows of the threat but refuses police protection, relying instead on his state-of-the-art security system. The mansion is a fortress with armed guards, a moat with sensors, thermal imaging cameras, and a vault that would challenge even Pierre in his prime.
Pierre realizes the heist is impossible through conventional means. He explains to Nigel that the only way in would be if Mueller himself opened the vault. That night, they stake out the mansion, but nothing happens.
At dawn, Mueller's security chief reports the painting safe. Relieved, Nigel prepares to leave when Pierre notices something—the painting's frame is slightly askew. Examining it closely, they discover it's a flawless forgery. The theft already happened, possibly days ago, without anyone noticing.
Mueller is outraged. The investigation reveals that someone posing as an art authenticator examined the painting weeks earlier—gaining access to the security protocols and creating the forgery.
Pierre finds one final message hidden in the forgery's canvas: "The student becomes the master. For my final masterpiece, I'll take what you couldn't. Mona awaits."
FINAL HEIST - PARIS, FRANCE
They realize the thief's ultimate target is the Mona Lisa at the Louvre—the one painting Pierre attempted to steal before his capture twenty years ago. It's not just a heist; it's personal.
Back in Paris, they alert Louvre security and develop an elaborate trap. The museum remains open to avoid tipping off the thief, but security is doubled. Pierre guides them in setting up countermeasures based on his own failed attempt decades ago.
On the night of the predicted heist, Nigel and Pierre wait in the darkened gallery. Hours pass with no sign of the thief. Pierre suddenly realizes they've been outmaneuvered—the real heist is happening elsewhere in the museum as a diversion.
They race to the Mona Lisa, arriving just as a figure in black is replacing the painting with a forgery. The thief turns—it's ELISABETH DURAND (35), Pierre's daughter, not his student as he believed. She reveals that Pierre never knew of her existence; her mother was Pierre's lover who was pregnant when he was arrested.
Elisabeth grew up studying her father's methods, determined to surpass him. She orchestrated the entire series of thefts to draw him out of prison, wanting to meet the legend before executing the heist he never could.
In an emotional confrontation, Pierre must choose between his daughter and justice. As alarms blare and security closes in, Elisabeth offers him a choice—escape with her or return to prison.
EPILOGUE
A month later, Nigel sits in his office, reviewing the case. The Mona Lisa was recovered, but both Pierre and Elisabeth vanished that night. Other stolen masterpieces begin reappearing in museums worldwide, shipped anonymously.
Nigel receives a postcard from Monaco with no message—just a white knight chess piece taped to it. He smiles despite himself, placing it on his desk next to a newspaper headline about a new security consultant helping museums protect against art theft—a consultant whose methods seem remarkably familiar.
At first glance, one would sit back and say, “Well, this is actually pretty good!”
But my reply would be: Is it?
It’s amazing that an artificial intelligence can spit out something like this so quickly. It’s pretty much a fully-formed story. It’s got a beginning, middle, and end, and it does exactly what I wanted it to do, based upon the premise that I came up with and asked it to execute.
But the problem here is that it’s generic. Now, that might not be a problem with most people nowadays, because if we’re really looking at things here, the plot of this movie could easily be a Netflix original. Of course, that doesn’t speak well to the tripe coming out of Hollywood as of late, but it’s true. The concepts lurking behind his particular film that Claude spit out are concepts that have been around for decades. And that speaks to the overall problem with AI.
Large Language Models like Claude are not thinkers. They are logic machines. The reason that they are able to do what they do is because they use logical reasoning (at a lightning pace) to figure out what is the next most probable outcome. So if I plant the seed of a movie about an art thief, it does not try to come up with an interesting twist on art heist movies. It simply produces the most logical outcome based upon the meager premise that I fed to it. The results are not necessarily good, but lean more towards the bad.
But it did what was asked of it. It cranked out a story. And I, like you might be, was suitably impressed with what it did. It was good enough for what I was requesting. As cliche-ridden as it was, it was a start. I didn’t like it, but it did the job.
So what did I do next? Well, I asked it to do a simple thing:
Can you turn this into a feature-length screenplay?
Almost immediately, Claude started spitting out a screenplay. It moved so quickly on this. I could see the words coming, one after another, faster than I could ever type them. And my knee-jerk reaction was, This is a miracle.
It took Claude about three-minutes to write the screenplay for Masterpiece. I was sufficiently shocked and surprised. I went from idea to finished screenplay in about fifteen minutes.
But you know there’s a catch, right? Of course there is.
I copied the text from Claude and pasted it into a screenwriting program so that it could be correctly formatted. Your average screenplay is between 90-120 pages. When pasted into the program, Claude’s screenplay was 53 pages. But that wasn’t the worst of it.
The worst of it was that Claude seemed to be swimming in cliches with this writing. The screenplay it produced was just…bad. This is the way it began:
INT. LOUVRE MUSEUM - SECURITY ROOM - SAME TIME
Three SECURITY OFFICERS monitor a wall of screens showing feeds from throughout the museum. One sips coffee, another scrolls on his phone.
SECURITY OFFICER #1
(in French, subtitled)
Another quiet night.SECURITY OFFICER #2
(in French, subtitled)
Eight more hours to go.Suddenly, the screens flicker. For a split second, they go black, then return to normal.
SECURITY OFFICER #1
(frowning)
Did you see that?SECURITY OFFICER #2
System hiccup. Happens all the time.
How many times have you seen this in a movie? A BILLION. There was more:
INT. CLAIRVAUX PRISON - VISITOR AREA - DAY
Nigel sits at a metal table, waiting. A heavy door clangs open. Two GUARDS escort in PIERRE DURAND (65).
Despite two decades of imprisonment, Pierre maintains an air of dignity and charm. His prison uniform is impeccably neat. His eyes sparkle with intelligence and mischief.
PIERRE
(with French accent)
Interpol. How novel. I haven’t had such distinguished company in years.The guards remove his handcuffs and step back. Pierre rubs his wrists, sits opposite Nigel.
NIGEL
Mr. Durand. I’m Agent Nigel Harrington.PIERRE
And to what do I owe this pleasure, Agent Harrington? Have you come to admire my watercolors? Prison has given me time to develop my artistic talents.NIGEL
I’m here about a series of art thefts.PIERRE
(amused)
I’ve been rather occupied, as you can see.Nigel opens his briefcase, spreads photos across the table.
NIGEL
Seven paintings. All stolen using methods remarkably similar to yours.Pierre studies the photos with professional interest.
PIERRE
Impressive work. Clean. Elegant.His expression changes slightly when he sees the chess pieces.
NIGEL
The thief leaves these at each scene.
Or how about this?
INT. PEGGY GUGGENHEIM COLLECTION - EXHIBITION HALL - DAY
A stunning gallery with Picasso works displayed prominently. "Woman with a Book" occupies a central position.
Marco explains the security measures.
MARCO
Motion sensors. Laser grid activated after hours. Pressure-sensitive flooring. Biometric access for staff. Bulletproof glass with vibration detectors.Pierre walks the space, studying everything with professional interest.
PIERRE
Impressive. May I?He gestures toward a control panel. Marco nods reluctantly.
Pierre examines the panel, noting model numbers, wiring.
PIERRE
Modern. Much better than twenty years ago.(pointing to a vent)
That ventilation system. Where does it lead?
MARCO
(defensively)
It's too small for a person.PIERRE
For an adult, yes.Marco and Nigel exchange concerned glances.
Some of you may sit back and say, “Well, I couldn’t write that.” And you would be correct. Most people can’t even churn out a page of awfulness like that at all. There’s nothing wrong with that. Everyone has skills in different areas.
But from a writing perspective, this is as generic as it comes. It’s an echo of a movie you’ve already seen. It’s a feedback loop of the cliches of a heist film. There’s nothing remotely interesting or thoughtful about this writing. It is simply what the computer thought it should write.
Now, you could consider Claude to be a tool to help something like this get started. Sure. I’ll give you that. If I wanted to, I could go in and start to rework this script. I could massage the dialogue. I could flesh out the scenes. I could come up with new concepts.
But why not do that in the first place?
Being a creative writer is tough work. It’s hard to come up with ideas. It’s hard to stare at a blank page and have to think of something to put there. And when you do these things, they’re not always good!
But that process, whether what you produce is good or bad, is what gives us as human beings our soul. Sitting in front of a computer or a piece of paper and having to spit something out is what makes us awesome as human beings. Is it easier to have a computer do it? Absolutely! But there is nothing of you in that writing. It is something that simply exists.
Cory Doctorow recently wrote about this in an essay on his blog, Pluralistic. It was an essay titled, AI Will Never Take Your Job. In it, he wrote something that I think is quite profound:
AIs want the future to be like the past, and AIs make the future like the past. If the training data is full of human bias, then the predictions will also be full of human bias, and then the outcomes will be full of human bias, and when those outcomes are copraphagically fed back into the training data, you get new, highly concentrated human/machine bias…
Basically, this means that AI is a giant feedback loop that gets smaller and smaller. If the AI is producing the information that the AI is absorbing, the AI is going to get dumber. You can already see this happening with algorithms at the moment. You buy a pair of shoes on Amazon, then Amazon knows you like those shoes, so it feeds you more ads about those types of shoes. Then, you buy more shoes like the ones advertised to you and you end up with shoes that are the same as the shoes you previously bought.
If the current AI models have absorbed all of the information on the internet, then that’s it. There’s no more. And if the AI starts to produce more information for the internet, then those models will just start producing dumber versions of the information it has already produced. It’s an ever-diminishing cycle.
From a creative perspective when this happens, you end up with an extraordinarily generic screenplay that goes through the motions of writing, but does not say anything new or create anything new. As I said, it’s merely an echo of what a screenplay should be.
I’m sure Hollywood executives are salivating at the notion of using this method. But you can see the results in the Marvel Cinematic Universe. Each movie is just a reverberation of one that came before it, until it has lost all meaning, nuance, and creative spark that created it in the first place. It was why the new Captain America film was so bad.
So the next time you watch a movie on Netflix, think to yourself, Is this original? Or is this merely an echo of what might have been original at one time?
You might be surprised at the conclusions you come to.
Do I think that AI is a problem? Sure. It’s definitely not going away. Do I think AI is going to rob me of my livelihood? Definitely not.