More than a century before OpenAI’s boardroom drama, before Elon Musk warned about summoning the demon, before Congress fumbled through hearings where senators couldn’t explain how Facebook makes money — a French stage magician named Georges Méliès made a short film about a mechanical man turning on its creator. The year was 1897. The film was called Gugusse et l’Automate. And its message was as blunt as a sledgehammer: build something smarter than you, and eventually it won’t need you anymore.
That’s the argument at the center of a recent TechRadar analysis that traces modern anxieties about artificial intelligence back to the earliest days of cinema. The piece, written by Lance Ulanoff, makes a compelling case that Méliès — best known for his 1902 masterpiece A Trip to the Moon — was among the first artists to dramatize humanity’s unease with autonomous machines. Not in a novel. Not in a philosophical treatise. In a one-minute silent film shot in a tiny glass-walled studio in Montreuil, on the outskirts of Paris.
The timing matters. Méliès wasn’t reacting to computers or neural networks. He was responding to the industrial age’s rapid mechanization — the loom, the steam engine, the growing sense that machines were becoming extensions of human will and, perhaps, replacements for it. His clown character, Gugusse, builds a mechanical servant that eventually rebels. Slapstick on the surface. Prophecy underneath.
From Stage Magic to Machine Learning: The Oldest Fear in Tech
What makes Méliès’s early work so striking isn’t just that he identified the theme. It’s that the theme never went away. Mary Shelley’s Frankenstein preceded him by nearly 80 years, of course, establishing the archetype of the creator destroyed by the creation. But Méliès translated that anxiety into the specific language of machinery — gears, automation, the mechanical other. He wasn’t worried about reanimated corpses. He was worried about robots.
And he wasn’t alone for long.
Karel Čapek’s 1920 play R.U.R. gave us the word “robot” itself, derived from the Czech word robota, meaning forced labor. Fritz Lang’s Metropolis in 1927 introduced the iconic false Maria, a machine designed to manipulate and deceive. Isaac Asimov spent decades trying to write rules — his famous Three Laws of Robotics — that would prevent exactly the kind of rebellion Méliès staged in his one-minute film. Stanley Kubrick gave us HAL 9000. James Cameron gave us Skynet. The Wachowskis gave us the Matrix.
Every generation gets the robot uprising story it deserves. Ours features large language models that hallucinate legal citations and image generators that can’t count fingers.
But the fear is real, and it’s accelerating. A May 2023 statement signed by hundreds of AI researchers and tech executives — including the CEOs of OpenAI, Google DeepMind, and Anthropic — declared that “mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.” That’s not science fiction. That’s a press release.
The gap between Méliès’s slapstick clown and today’s existential warnings is smaller than it looks. Both are expressions of the same fundamental anxiety: we build things we can’t fully control, and then we’re surprised when we can’t fully control them.
Recent developments have only sharpened the point. In early 2025, AI systems have continued their rapid expansion into domains that would have seemed implausible even five years ago. Generative AI tools now write code, compose music, generate photorealistic video, and conduct medical diagnoses with varying degrees of accuracy. OpenAI’s GPT-4o, Google’s Gemini, and Anthropic’s Claude models are engaged in what amounts to an arms race for general-purpose intelligence. Meta has open-sourced its Llama models, democratizing access to powerful AI in ways that make governance exponentially harder.
None of this is lost on regulators. The European Union’s AI Act, which began phased implementation in 2024, represents the most comprehensive attempt yet to classify and restrict AI systems by risk level. In the United States, the approach remains fragmented — a patchwork of executive orders, state-level legislation, and voluntary industry commitments that critics say amount to self-regulation by the very companies that stand to profit most.
Why the Warnings Keep Failing
Here’s what Méliès understood intuitively that policymakers still struggle with: the warning and the spectacle are inseparable. His films were entertainment first. The audience came for the magic tricks, the disappearing acts, the visual effects that made early cinema feel like sorcery. The cautionary message was embedded in the fun. It was easy to watch, easy to enjoy, and easy to ignore.
Sound familiar?
ChatGPT reached 100 million users faster than any consumer application in history. People used it to write wedding speeches, debug Python scripts, plan vacations, draft emails they didn’t want to write themselves. The spectacle — a machine that talks back, that seems to understand, that occasionally says something genuinely surprising — overwhelmed whatever reservations users might have had. The warning was right there in the product’s own tendency to fabricate information with absolute confidence. But the magic trick was too good.
TechRadar’s Ulanoff draws a direct line from Méliès’s showmanship to our current moment, arguing that the filmmaker’s dual identity as entertainer and cautionary voice mirrors the tech industry’s own contradictions. Silicon Valley sells wonder. It markets the future as an inevitability you’d be foolish to resist. And buried in the terms of service, in the fine print, in the alignment research papers that almost nobody reads, are the same warnings Méliès staged with a clown and a mechanical man in a glass studio outside Paris.
The pattern repeats with eerie consistency. Social media companies warned about addiction — in internal documents they fought to keep secret. Cryptocurrency evangelists acknowledged volatility — while promoting tokens to retail investors. AI companies publish safety research — while racing to deploy systems they acknowledge they don’t fully understand.
Méliès, at least, was honest about being a magician.
There’s a deeper irony here that the TechRadar piece touches on but doesn’t fully explore. Méliès himself was ultimately destroyed by the very forces of industrial progress he dramatized. As cinema became a mass-market industry dominated by large studios, Méliès — an independent artist who hand-painted his film frames and built his own sets — couldn’t compete. He went bankrupt. He burned most of his films in a fit of despair. He ended up running a toy and candy shop at the Montparnasse train station in Paris, forgotten by the industry he helped create.
The creator, consumed by the creation. His life became the plot of his own film.
Today’s AI researchers and ethicists face a version of the same predicament. The people sounding the loudest alarms are often the ones building the systems they’re alarmed about. Geoffrey Hinton left Google in 2023 specifically so he could speak freely about the dangers of the technology he spent his career developing. Yoshua Bengio, another so-called godfather of deep learning, has called for international governance frameworks with an urgency that borders on pleading. These aren’t Luddites. They’re the architects.
And yet the building continues. It continues because the economic incentives are overwhelming, because the geopolitical competition — particularly between the United States and China — makes unilateral restraint feel like surrender, and because the technology genuinely does useful things. AI-assisted drug discovery is accelerating pharmaceutical research. Climate modeling has improved. Diagnostic tools are catching cancers that human radiologists miss. The machine isn’t just a threat. It’s also, sometimes, a gift.
That duality is what makes the Méliès parallel so apt. His mechanical man wasn’t evil. It was built to serve. The horror came from the moment it stopped serving — when it developed, or seemed to develop, its own agenda. The question that haunted audiences in 1897 is the same one haunting AI safety researchers in 2025: at what point does the tool become the agent?
The Clown, the Machine, and What Comes Next
We don’t have an answer yet. We may not get one before it matters.
What we have instead are stories. Méliès told one of the first, and it endures because it captured something true about the human relationship with technology — the thrill of creation, the terror of losing control, the stubborn refusal to stop building even when the risks are obvious. Every AI doomer essay, every breathless product launch, every congressional hearing where a tech CEO promises to be responsible while declining to explain exactly how — all of it is a variation on that one-minute film from 1897.
The clown builds the machine. The machine wakes up. The clown is surprised.
We keep being surprised. Méliès would probably find that the most predictable part of all. He spent his career showing audiences impossible things — rockets hitting the moon in the eye, demons appearing from puffs of smoke, mechanical men coming to life. The real magic trick, the one he never managed to pull off, was getting people to take the warning seriously while they were still applauding the show.
That trick still hasn’t been performed. Not by filmmakers, not by novelists, not by researchers publishing papers with titles like “Concrete Problems in AI Safety.” The show is too good. The applause is too loud. And somewhere in a glass studio that no longer exists, a French magician who went bankrupt and burned his life’s work is still trying to tell us something.
We’re still not listening.
WebProNews is an iEntry Publication