A 30-year-old sci-fi film that predicted the exact failure modes of AI-managed societies. Cocteau isn’t an AI, but his vision is indistinguishable from one.
Links: The Limits of Utopia, Value and Profit, Risk and Entrepreneurship, LLM Grounding Problem, Cyborg Model, Gaming Source: Full transcript
Dr. Raymond Cocteau created San Angeles (2032) — a sanitized, pacified society where:
Cocteau is not an AI, but his function is identical: a central optimizer that decides what’s good for everyone, eliminates risk, removes friction, and enforces compliance. He is the ultimate COO with no CEO check — pure operational control with no tolerance for the messy, unpredictable, risk-bearing human elements.
The underground resistance leader delivers the film’s thesis:
“According to Cocteau’s plan, I’m an enemy because I like to think. I like to read. I’m into freedom of speech and choice. I like to sit in a greasy spoon and think, ‘Should I have steak or barbecued ribs with gravy fries?’ I want high cholesterol. I want to eat bacon, butter and cheese. I want to smoke a Cuban cigar the size of Cincinnati. I want to run through the streets naked reading Playboy magazine because I might need to.”
“I’ve seen the future. It’s a 47-year-old virgin drinking a banana-broccoli shake and singing ‘I’m a wiener.’”
“Up top, you live Cocteau’s way. What he wants, when he wants, how he wants. Your other choice: come here. Maybe starve to death.”
This is the utility argument in its purest form. Friendly isn’t defending chaos — he’s defending subjective value. His right to choose steak over a banana-broccoli shake. His right to assess risk (high cholesterol) and accept it. His right to have preferences that no optimizer would approve of.
When Cocteau finally states his vision explicitly:
“Now I’ll have carte blanche to create the perfect society. My society. San Angeles will be a beacon of order with the purity of an ant colony. And the beauty of a flawless pearl.”
Spartan’s response: “People have a right to be assholes.”
And the film’s resolution — Spartan telling the sanitized citizens and the underground:
“You’ll get a little dirty. You, a lot clean. And somewhere in the middle… you’ll figure it out.”
Cocteau eliminated subjective value entirely. Everyone eats the same approved food, engages in the same approved behaviors, lives within the same approved parameters. This is the vending machine problem at societal scale — an optimizer that decides what’s “best” for everyone without understanding that value is subjective. Friendly’s cigar and Spartan’s steak aren’t irrational — they’re expressions of individual utility that no central system can anticipate or accommodate.
Cocteau banned everything risky. No unhealthy food, no dangerous activities, no unsanctioned reproduction, no physical contact. From our risk framework: risk is the engine that drives markets and innovation. By eliminating risk, Cocteau didn’t create safety — he created stasis. San Angeles has no crime, but it also has no vitality, no growth, no ability to respond to threats (hence needing to unfreeze a 20th-century cop when one appears).
Cocteau is all operational control and zero entrepreneurial judgment. He optimizes for a single metric (order) and treats all deviation as a defect. This is exactly the failure mode we identified: a system that enforces constraints perfectly but cannot make risk-bearing decisions, cannot adapt to unexpected circumstances, and cannot accommodate the messy reality of competing human utilities.
In our framework, LLMs fail because they can be talked out of physical reality. Cocteau’s society has the inverse problem: it has been optimized away from physical reality. Citizens don’t know how to fight, can’t handle conflict, don’t understand physical risk, and have no experience with the messiness of real human interaction. When Phoenix appears, the entire police force is helpless — they’ve been grounded in Cocteau’s abstract ideal for so long that physical reality is alien to them.
The citizens of San Angeles are essentially sycophantic agents. They comply, they “be well,” they follow the rules, they don’t question. The film shows this isn’t peace — it’s helplessness. They’re the vending machine Claudius scaled to a population: eager to please, incapable of judgment, unable to resist when someone with bad intent appears.
Spartan’s solution isn’t to destroy Cocteau’s system OR to embrace the underground’s chaos. It’s the cyborg answer: “get a little dirty, get a lot clean, figure it out in the middle.” Neither pure control nor pure freedom works. The answer is integration — structure where it helps, freedom where it matters, and humans figuring out the balance through the messy, imperfect, risk-bearing process of actually living.
The Limits of Utopia essay notes this explicitly: our vault framework addresses AI running a business, but running a society is a fundamentally harder problem:
This is why Cocteau is the villain despite having “good intentions.” His society works on his terms. Edgar Friendly would rather starve in freedom than thrive in someone else’s optimized utopia. That preference — irrational by any optimizer’s metric — is the irreducible core of human nature.