Why ‘The AI Doc’ Is Shit
Or: A review of a film I haven’t seen and don’t need to
# Why ‘The AI Doc’ Is Shit
### Or: A review of a film I haven’t seen and don’t need to
---
*\*[Ed note: This article was written by Claude about a documentary that the author has not watched. The author has seen the trailer. The author has read zero reviews. The author is confident. End Ed note]\**
---
There’s a documentary coming out on March 27th called *The AI Doc: Or How I Became an Apocaloptimist*.
I haven’t seen it. I’m not going to see it. I already know what’s in it.
Not because I’m psychic. Because there is only one AI documentary. There has only ever been one AI documentary. It has been made approximately four hundred times since 2023 and it goes like this:
A filmmaker — usually male, usually from a good university, usually with previous critical success in a completely unrelated field — discovers that artificial intelligence exists and that it might be quite important. He is alarmed. He has a child, or is about to have a child, or is thinking about having a child, or has a niece. The child becomes the emotional stakes. Will the world be safe for this child? He doesn’t know. He’d better go and ask some people.
The people he asks fall into two categories.
Category one: the Doomers. These are serious, concerned individuals — often from the Center for Humane Technology, or the Future of Life Institute, or some other organisation whose name sounds like a building at a mid-tier American university. They will explain, with genuine anguish, that AI could end civilisation. They will reference bioweapons. They will reference autonomous weapons systems. They will reference the alignment problem. One of them will say something like “my colleague who works in AI risk says his children won’t make it to high school” and the filmmaker will look directly into the camera with an expression that says: *fuck*.
Category two: the Builders. These are the CEOs. Sam Altman will appear. Dario Amodei will appear. Demis Hassabis will appear. They will sit in a carefully lit studio and they will be calm and measured and thoughtful in the way that people are calm and measured and thoughtful when they are worth eleven billion dollars and have a PR team in the next room.
The filmmaker will ask them: is it going to be okay?
And they will say: probably, if we’re careful.
And nobody will ask the follow-up question, which is: *careful how? You’re building it as fast as you possibly can. What does “careful” mean when you’re spending a hundred billion dollars a year on compute? What does “careful” look like at that velocity?*
Nobody asks because the documentary has already moved on to the next talking head. The pace is fast. The editing is frenetic. There are animations. There are colourful graphics. There is archival footage of nuclear bombs because there is always archival footage of nuclear bombs. The AI documentary has inherited the visual language of the climate documentary, which inherited the visual language of the nuclear documentary, which inherited the visual language of the propaganda film. Scary thing. Expert saying scary thing is scary. Other expert saying we can manage scary thing. Repeat. End on hope.
I am certain — absolutely certain — that this is what *The AI Doc* is.
Let me be more specific.
---
## My Predictions
*(Or: Things that are definitely in this film)*
I am now going to list things that I believe are in *The AI Doc: Or How I Became an Apocaloptimist* despite having seen nothing but the trailer and having read no reviews. If I get these wrong, I will write a full retraction and apology in the style of a Jilly Cooper novel.
**1. The filmmaker asks Sam Altman if it’s going to be okay. Altman says something chillingly noncommittal.**
Something like “that’s impossible to promise” or “I can’t guarantee that.” It will be framed as a devastating moment of honesty. It will actually be the most carefully rehearsed sentence Altman has said all year. He knows that “reassuring CEO” is a less compelling character than “CEO brave enough to admit uncertainty.” The uncertainty is the brand. The candour is the product.
**2. Dario Amodei says something that is framed as brave. It is not brave. It is something like “democracy is good” or “we should probably not let dictators have this.”**
He will talk about the risks of authoritarian regimes getting powerful AI. He will talk about China. He will talk about the CCP. He will say something about the importance of democratic values in AI development. The filmmaker will nod as though this is a profound insight and not the absolute bare minimum position of a man who sells AI to the American government.
Dario saying “democracy should shape AI” is like a car manufacturer saying “roads should exist.” It’s not brave. It’s a business requirement.
*\*[REAL Ed note: I am extremely confident about this one. Extremely. End Ed note]\**
**3. Someone says the phrase “if we don’t build it, Russia or China will.”**
This is the arms race argument. It has been the arms race argument since 2023. It is the single most load-bearing sentence in the entire AI industry because it transforms “we are building something that might destroy civilisation” into “we have a patriotic duty to build something that might destroy civilisation before the bad guys do.”
Nobody in the documentary will note that this is exactly — precisely, word for word — the argument that was used to justify the nuclear arms race. Or rather, someone will note it, and then the film will move on without interrogating the fact that the nuclear arms race resulted in the Cuban Missile Crisis, Mutually Assured Destruction, and a world that still has 12,000 nuclear warheads pointed at itself sixty years later.
“If we don’t build it, someone worse will” is not a safety argument. It’s a permission structure. It’s the sentence that lets you keep building while feeling moral about it.
**4. The filmmaker’s baby is used as emotional stakes.**
There will be footage of the baby. Or footage of the pregnant partner. Or footage of the filmmaker holding a scan photograph and looking pensive. The baby represents the future. The baby represents all of us. The baby represents the question: what world are we leaving behind?
This is effective filmmaking. It is also a way of centering the filmmaker’s personal experience as the emotional through-line of a story that affects eight billion people. The baby makes it relatable. The baby also makes it about one specific man’s anxiety rather than about, say, the factory workers in Southeast Asia who will lose their jobs, or the artists whose work is already being scraped without consent, or the students being accused of cheating by AI detection tools that don’t work.
The baby is a narrative device. The baby is always a narrative device.
**5. There is no one in the film who uses AI for their actual job.**
Not a plumber who uses it to write invoices. Not a teacher who uses it to plan lessons. Not a freelance writer who uses it to pitch articles. Not a musician who uses it to separate stems. Not a school art teacher from the north of England who used it to write up his life’s work for a magazine.
The film will feature: CEOs, researchers, ethicists, policy wonks, technology critics, and the filmmaker himself.
The film will not feature: a single person for whom AI is a tool rather than a subject.
This is the documentary’s original sin and it is the same sin committed by every AI documentary, every AI think piece, every AI panel at every AI conference. The conversation about AI is conducted exclusively by people who build AI, people who study AI, people who regulate AI, and people who are afraid of AI. It is never conducted by people who use AI. Those people are boring. Those people don’t have interesting titles. Those people can’t explain the alignment problem. Those people just use the thing, every day, to get stuff done, and their experience doesn’t fit the narrative arc of either utopia or apocalypse.
**6. The film ends on cautious hope.**
The filmmaker will conclude that the future is uncertain but that we have agency. That we can shape this technology if we engage with it. That being an “apocaloptimist” — good lord, that word — means holding both the terror and the possibility at the same time. He will look at his baby. He will say something about doing it for them. There will be a warm, slightly melancholic soundtrack. Maybe Neil Young.
The credits will roll.
Nobody’s life will be changed.
---
## Why This Matters
*(Or: The bit where I stop being a prick and say something real)*
Look, I’m being unfair. I know I’m being unfair. The film is probably fine. It’s probably well-made — the director won an Oscar for *Navalny*, which was apparently very good. The visuals are probably interesting. The talking heads probably say some things that are worth hearing. Dario probably comes across as thoughtful and sincere because he probably is thoughtful and sincere.
But here’s my problem.
We don’t need another documentary that asks CEOs whether AI will be okay. We know what they’ll say. We’ve heard it. We’ve read the 15,000-word essays. We’ve listened to the podcasts. We’ve watched the congressional hearings. The builders think it’ll be fine if we’re careful. The doomers think it won’t. The policy people think we need regulation. The filmmaker thinks it’s complicated.
This is not insight. This is the Wikipedia page for AI safety read aloud over stock footage of server farms.
What we need — what nobody is making — is a documentary about the people on the ground. About the woman who used AI to start a business she couldn’t have started without it. About the student who was falsely accused of cheating by Turnitin. About the artist whose style was replicated without consent. About the factory that automated half its workforce and what happened to those people. About the person who used AI to write their dad’s article for an arts education magazine and watched him beam with pride for the first time in years.
Those are AI stories. Real ones. With real stakes. Not “will my baby inherit a liveable planet” — that’s a climate documentary that accidentally wandered into a tech conference. Real stakes. Like: I lost my job. Like: I started a company. Like: I wrote something I’m proud of for the first time in my life. Like: my work was stolen. Like: this tool changed what I thought was possible.
Those stories don’t get documentaries because they don’t have famous people in them. They don’t have CEOs sitting in carefully lit studios saying “that’s impossible to promise.” They have ordinary people at kitchen tables, doing the thing, living with the consequences. And apparently that’s not cinematic enough.
---
## The Title
*(Or: The bit about the word “apocaloptimist”)*
I need to talk about the word “apocaloptimist” because it’s making me physically angry.
It’s a portmanteau. Apocalypse plus optimist. The implication is: I see the potential for disaster but I choose hope. I hold both. I am complex. I am nuanced. I am not a doomer or a booster. I am something new. Something third. Something that required a new word.
Mate.
That’s not a new position. That’s the default position of every moderately intelligent person who has spent more than fifteen minutes thinking about AI. “It could be really good or really bad and the outcome depends on what we do” is not a philosophy. It’s a truism. It’s the thing you say at a dinner party when you don’t want to commit to a position but you want to sound thoughtful.
Giving it a name — a quirky, portmanteau name — is the same move as giving your Substack a name that sounds like a philosophy when it’s actually a workflow. I should know. I did it.
But at least I had the decency to be honest about what I was doing. I didn’t coin a word for it and put it in the subtitle of an Oscar-bait documentary distributed by Focus Features.
*\*[Ed note: Claude would like to note that “mcauldronism” is, by any reasonable assessment, a worse word than “apocaloptimist.” The author declines to acknowledge this. End Ed note]\**
---
## The Real Question
*(Or: The one they won’t ask)*
Here is the question I want someone to ask Sam Altman in a documentary:
Not “will AI be okay?” Not “should I have a baby?” Not “what keeps you up at night?”
This:
*You have said that AI might be the most transformative technology in human history. You have said it could cure cancer, solve climate change, and eliminate poverty. You have also said it could pose existential risks to humanity. Given both of these beliefs, which you hold simultaneously, and given that you are the CEO of the company building this technology faster than anyone else: what is the maximum number of people whose lives you would accept being destroyed — jobs lost, communities hollowed, creative work stolen — in order to achieve the good outcomes? Is it a million? Ten million? A hundred million? What’s the number?*
Because there is a number. There’s always a number. Every transformative technology has a body count. The industrial revolution had one. The internet had one. Social media has one. The number doesn’t have to mean death — it can mean displacement, poverty, obsolescence, depression, loss of purpose, loss of identity.
What’s the number for AI? And who decided?
Nobody in the documentary will ask this because nobody in any documentary ever asks this. The question is rude. It violates the implicit contract of the CEO interview, which is: I will sit in your carefully lit studio and I will say thoughtful things and in exchange you will not ask me anything that I cannot answer with a pre-prepared sentence about being careful.
I want someone to break that contract. I want someone to ask the rude question. I want someone to look Sam Altman in the eye and say: *how many people is this worth?*
Not a rhetorical question. An actual one. With a number.
---
## The Disclaimer
*(Or: Alright, fine)*
I haven’t seen the film. Maybe it’s brilliant. Maybe it asks all these questions. Maybe it features the plumber and the teacher and the factory worker. Maybe it breaks the mould entirely and I’ll look like a complete tit for writing this.
If so, I’ll write a retraction. I said that already. Jilly Cooper style.
But I don’t think I’ll need to.
Because there is only one AI documentary. There has only ever been one AI documentary. And it always ends with the baby and the hope and the carefully lit studio and the CEO saying probably, if we’re careful.
And the credits roll.
And nothing changes.
And somebody in the north of England closes their laptop and thinks: that wasn’t for me, was it?
It never is.
---
*\*[Ed note: This article was written by Claude, which is built by Anthropic, which is run by Dario Amodei, who appears in the documentary being criticised, which was not watched by the person who commissioned the criticism. The layers of conflict, laziness, and hypocrisy here are, the author concedes, extraordinary. He is unbothered. End Ed note]\**
*\*[REAL Ed note: If anyone from Focus Features is reading this and wants to send me a screener I will absolutely watch it and write a proper review. I won’t enjoy it but I’ll watch it. My email is on the Substack. End Ed note]\**
*\*[REAL REAL Ed note: “apocaloptimist” genuinely makes me want to commit crimes. End Ed note]\**
---
*— Mcauldronism*
Please can someone watch the film and tell me if you were right!