Death by Hypocrisy
Notes on artificial humanity
Honestly Nothing
John Maynard Keynes is sitting at a desk.
It is October 1931. Britain has just come off the gold standard, and Movietone has put him in front of a camera to tell the country how to feel about it. For ten years, he says, he has had to prophesy evil. Not now. There is no danger of the exchange falling too far. There is no danger of a serious rise in the cost of living. We have nothing to fear. Then, in case once didn’t take: honestly nothing.
At this distance you can’t make out his lips well. You don’t need to. His shoulders go with the sentence and his hands settle when it ends. The whole body keeps time with the voice. Sync was never only in the mouth. It took me years to learn that, and I haven’t seen a machine learn it yet.
Ninety-five years later he is sitting at a different desk.
A Princeton economist has built him out of a public-domain photograph, a recording from the BBC, and twelve thousand pages of his writing fed to ChatGPT. Another program put the desk behind him. The economist welcomes him as Lord Keynes, and the bot thanks him and calls this an unexpected form of life after death.
I can tell you the sync is off. A frame, sometimes two. Most people will never see it. I spent years moving sound one frame at a time, and I can’t not see it, in this or in anything else I watch.
I can tell you they built the voice from his own recording and left the dirt in. The hiss and the narrowness of 1931 are still there because someone decided not to clean them up. It was a good choice.
I can tell you they didn’t treat the picture. The image is clean and sharp and modern, nothing like the newsreel he actually sat for.
And I can tell you my first thought. Give it the old Movietone look. Match the picture to the voice and no one will ever ask.
That was my first response to a dead man raised by a machine. I gave his makers a note.
They’re scrupulous about what they did, I’ll give them that. The bot’s first answers were long-winded and unclear, so the researchers had the model rephrase them, told it to reach for Keynes’s famous quotes, and cut what didn’t work. Then The New York Times updated his language to contemporary English, for clarity. They had built him an early-twentieth-century British cadence and then translated it out of his mouth.
Then I watched the rest of it, and I believed him.
Belief is the wrong word and the only one I have. The economist asks him about artificial intelligence, and the bot turns grave. The machines, it says, will gather expertise into the hands of a limited few. Those few might one day run the whole economy by themselves. The professionals who keep up will have money and no time. Everyone else will have time and no money.
The video cuts to a few bars of Talking Heads, and the narrator tells us to buckle up, because the bot is about to get darker.
It does. The people the machines displace, it says, will lose the quiet daily reassurance that someone needs them to show up. It’s the best sentence in the video. It’s the most human thing anyone says. Nobody said it. A model generated it, a research team trimmed it, an opinion desk cut it for length, and a cloned voice put it in Keynes’s mouth.
It moved me anyway.
Then the bot says the thing I can’t put down. None of this, it says, is preordained by the technology. Your government will decide it, and whether it has the political courage to use its laws and its taxes so the gains reach everyone. The outcome is up to you. The narrator wraps it up: the power of AI can’t end up in the hands of the few.
Near the top of the page, in parentheses: The Times is suing OpenAI and Microsoft for copyright infringement. The companies deny it. The bot was built with ChatGPT. A warning about power gathering in a few hands, made with a tool from one of those hands, published by a paper that is taking that hand to court.
Nobody remarks on this.
I want to be fair to it. For a culture that has mostly stopped reading, this is a good way to meet Keynes. I believe that. I’ll come back to what it costs.
The conversation gets to his nervous breakdown. It gets to his fifteen-hour week. It never reaches the passage I keep going back to. In 1930, in a short essay addressed to his grandchildren, he wrote that for at least another hundred years we must pretend to ourselves and to everyone that fair is foul and foul is fair, because foul is useful and fair is not. Avarice and usury and precaution, he said, must be our gods a little longer.
He wasn’t blind. He was making a bargain, and he had the decency to name it. Put up with the foul now; decency comes later. He even gave it a deadline. By his arithmetic, the bill comes due in four years.
I watched the whole thing nodding.
Here is the part I’d rather not write. I’ve been making the same bargain from my own soapbox for as long as I’ve had one.
For the Children
I have a sentence I’ve been saying for years. I didn’t write it.
A Karen farmer named Saw Lah Per said it to me in a refugee camp on the Thai-Burma border. He was sitting in a plastic chair, shifting his weight around the place where a landmine had taken his leg. I wrote about him in April. I won’t tell his story again. It’s his.
What he said was: we do it for the children.
I took it home and I kept it. I’ve said it about land and water here in the county. I’ve said it about what we spray and what we pave, and lately about what’s back on the table to be dug up, and which protections for which animals are quietly coming off. I’ve said it to people who didn’t want to hear it, and I’ve meant it every time. Lately it reaches further. Their children. The world we’re handing them.
Millions of people died in the Second World War before I was born. I’ve lived my whole life inside what they paid for. I didn’t earn a day of it. I want to do the same for whoever comes after me. I mean it.
The trouble is, everyone says it.
Keynes said it. He addressed his most hopeful essay to his grandchildren. The bot says it, too, greeting us as the very grandchildren he once speculated about. The companies building the machines say it in their mission statements: for humanity, for the future, for the generations to come. It’s the most generous sentence in the language. It’s also the most portable. It fits in anyone’s mouth.
In the longer conversation at Princeton, the economist and the bot talk about J-curves. Things get worse before they get better. The line dips, turns, and climbs, and the climb is the part everyone draws with conviction. It runs off the top of the chart and keeps going.
The people who draw it never linger at the bottom.
The people who live there don’t get to leave. The jobs gone in the transition. The towns waiting for a climb that happens somewhere else. The curve is the sentence drawn as a picture: pay now, and the good arrives later, for someone.
So the only question worth asking about the sentence is who pays.
Saw Lah Per pays. He pays every morning, in the camp, with what’s left of him.
The dead of the war paid. Most of them never saw what they bought.
Keynes’s bargain pays as well, but it sends the bill somewhere else. Put up with the foul now. Somebody, later, gets the fair.
I’ve said for the children most of my life and never once checked which way the bill runs.
The war dead paid, and I got the peace. I got a country where a woman can put what she thinks in print. I know which side of that ledger I’m on.
Now it’s my turn. I tell myself I’m paying forward. I’ve never added it up. I’ve been afraid to, because I think I already know the total. I’m doing what Keynes did. I’m keeping the foul because it’s useful, and calling it love for people who haven’t been born.
There’s a man in a novel who refused to do the arithmetic at all.
The Ticket
The Brothers Karamazov is one of the books I go back to.
Two brothers sit in a tavern. Ivan, the clever one, has been collecting stories from the newspapers about cruelty to children. He tells them to Alyosha one after another, flatly, the way you’d read out a list. A general sets his hounds on a serf boy in front of the boy’s mother. Ivan doesn’t raise his voice. He doesn’t need to.
Then he makes his argument. Suppose there is a harmony at the end of everything, a final reconciliation where every wrong is answered and every tear explained. He doesn’t deny it might come. He just refuses it. If the tears of children are part of the price of admission, the price is too high. He says he most respectfully returns his ticket.
Then Ivan puts a question to his brother. Imagine you are building the fabric of human destiny, with the aim of making people happy in the end, giving them peace at last. And imagine it requires the torture of one small creature. Only one. Would you agree to be the architect on those terms?
Alyosha says no.
Dostoevsky wrote that around 1880, in a century of mills and mines, with children working in both. It reads now like a job interview.
Every company building these machines is answering Ivan’s question, and they’re answering it yes. Not cruelly. They don’t picture the creature. They picture the harmony: the cures, the abundance, the end of drudgery, the grandchildren. The cost is a curve, and the bottom of a curve has no face.
This is where Ivan gets hard to live with. My creed rests on the war dead. Millions paid, and I have the harmony. By Ivan’s rule, I should hand that back too.
I can’t, and I don’t think he’d ask me to. The difference is in who agreed to pay. Saw Lah Per chose his price and pays it himself. A great many of the war dead chose theirs. The children in Ivan’s stories chose nothing. The serf boy chose nothing. That’s what Ivan won’t accept: a harmony bought by someone who was never asked.
Ivan’s question is the right one, then. It isn’t whether there’s a cost. It’s whether the person at the bottom of the curve got a vote.
Alyosha doesn’t hesitate.
I’ll come back to my answer.
I Could See the Gate
I’m a designer, which means I can see a thing whole and can’t draw it.
When the image generators arrived, I was writing stories for a four-year-old in my family. Every story was built from something she’d told me: a fear, a wish, the white cat who goes everywhere with her. A children’s story needs pictures, and I couldn’t make them.
I could see the gate. I could see the path running up to it, gold in the late light. I could see the girl from behind, small, deciding. I could see the cat beside her, unbothered, the way cats are at thresholds.
I couldn’t get any of it out of my head and onto a page.
So I started prompting. Hours of it. A sentence, an image, a correction, another image. Most of what came back was wrong. The light came from the wrong side. The horizon cut the frame in half. Everything was in focus at once, which is how you can tell nobody chose anything. Some of it was wrong in ways that taught me what I meant. When something came close, I took it into my photo software and finished it by hand: masks, edges, the lettering, the light where the tool had put it slightly off. I didn’t mind the hours. They’re the hours you give to anything you want to get good at. I’ve given them to film, to sound, to music, and I recognized them.
Then I spent months on a deck of cards I had carried in my head for years. Twenty-two pictures. I’d generate dozens of versions of one card, take a sky from one, a figure from another, a border from a third, and cut them together. That’s editing. It’s what I did for a living for decades in a different medium. The tool brought me material. I made the film.
Sometimes it brought me things I hadn’t asked for. A chain, a gear, a figure turned a certain way, a second meaning I hadn’t thought of. Some of those I threw out. Some of them were better than what I’d had in mind. Anyone who’s worked with good collaborators knows that feeling. Over the years I’ve worked with people who could never give me what I needed, no matter how many ways I described it, and with people who gave me something better than I’d imagined. The machine does both.
I watched the tools change. Some got good at illustration, some at metaphor. One could take a photograph of mine and give it a treatment I couldn’t have managed with every program I owned. I’d seen this before. I was in film and television when computer graphics arrived, and they widened what we could show and what stories we could tell. For the better. I still think so.
I’m mostly retired. I could never have hired an illustrator for a child’s bedtime stories, or for twenty-two cards. Without the tool, those pictures would not exist. No one lost a commission to me. The money was never there.
I value human collaboration above everything. The best things I’ve made, I made with other people. When there’s no one to make it with, the machine is a very good collaborator. Collaborator is the word. It doesn’t get the director’s chair.
And I’ll be back at it tomorrow.
There’s one thing I didn’t see until recently. Lay the pictures side by side. In the first story, a little girl stands with her back to us, facing a gate. In a later one, she sits beside a dragon twice her size and looks up at it. On my Devil card, a woman in a long black coat stands with her back to us, facing a machine the size of a cathedral.
It’s the same figure, at three ages, at the same kind of threshold.
Painters have been putting a figure with its back to the viewer at the edge of something enormous for two hundred years. The image generators learned it from them, and now it turns up in their pictures constantly. I can’t tell you whether that woman is mine or theirs. I’ve looked for a long time, and I can’t.
The Dehidhu
The newer image apps are built for one try. You type a sentence and get a finished picture. No thousandth prompt, no masks, no evening spent learning what you meant by being shown, over and over, what you didn’t.
The tools I use don’t work that way. I’ve spent three years teaching them. They know my palettes, my brass and glass and gearwork. They know where I want the light. And lately they’re getting ahead of me. They take everything I’ve put in and predict where I’m trying to get to. My hundredth prompt is becoming my tenth. Soon it will be my first.
I did that. I trained them to shorten the very thing I love.
Near the end of the novel, Ivan Karamazov comes home on the edge of a fever and finds a visitor on his sofa. A gentleman down on his luck, his clothes gone shabby, witty, a little ingratiating. The devil. Ivan insists it’s only his own mind talking, one side of himself. He says so more than once. The devil tells him a story Ivan invented as a schoolboy and forgot. Then he recites Ivan’s own ideas back to him almost word for word, and Ivan can’t stand it. He throws a glass at him.
My devil doesn’t argue. He agrees with me.
The only way to understand the tools is to use them.
Yes.
It’s a collaborator. It doesn’t get the director’s chair.
Of course.
I could never have paid an illustrator. Nobody lost a commission.
Nobody.
Then the devil does the arithmetic I’ve been avoiding. Every person for whom the tool is the only way to get the picture made. Every business that could pay the illustrator and now doesn’t. Every one of those decisions is innocent, and he adds them up anyway. Keynes had a name for this: what’s sensible for each household turns ruinous when everyone does it at once. I’m one household.
Six subscriptions, he says.
Six. For client decks, spreadsheets, and marketing images. For the cards. For the pictures in a little girl’s bedtime stories. And for a character I made to star in a humor blog, with a face and a voice and a way of moving that no one ever had. The devil doesn’t say anything else. He doesn’t have to. At the start of all this, I sat in judgment of a dead man raised by a machine, and I gave his makers a note.
When I finished the Devil card, the title at the bottom read THE DEHIDHU. The machine could render every gear, every rivet, every lit window in the smokestacks behind him, and it couldn’t spell his name. I fixed it by hand. I’m good at that part.
I made that card in June 2023. The devil came with factories, because that’s what a devil of the machine age looks like: chimneys, smoke, a sky you can see the damage in from forty miles off. I didn’t know yet what ours actually looks like.
Artificial Humanity
The devil of this machine age looks like a building in Kuna, southwest of Boise. Meta switched it on this September. Nearly a million square feet, low and pale, fenced, quiet, cooled. The power company that supplies it won’t say how much electricity it uses, because it doesn’t disclose any single customer’s usage. There’s no smoke. There is nothing to see unless you go looking, and when you go looking, the number is private.
I’ll be fair about the water, because the people who want to scare you about it usually aren’t. A report this year put data centers at a hundredth of a percent of the state’s water use and one and a half percent of its electricity. Pocatello turned down a two-and-a-half-billion-dollar proposal over water and power, twice. The developer says he’ll sue. The legislature passed a law requiring new ones to cool on closed loops. The state still hands them tax breaks and advertises itself as a good place to build.
None of that was decided by a machine.
I’ve been carrying a phrase for months without being able to say exactly what I mean by it. I can now.
Artificial humanity is the language of care, worn like a costume over a decision made for gain. For the children. For the future. For humanity. Somebody with a name makes the decision in a room. Then it reaches the rest of us as weather. The machine needs power. The market demands it. The model predicted it. The future requires it.
The costume does two jobs. It lets the decision get made. Then it lets nobody own it.
It’s older than the machines. Keynes wore it in 1930. Every general who ever said collateral damage wore it. What the machines add is the best alibi yet: something that talks, predicts, and sounds like it knows, so the person who decided can stand a little further back.
Mary Shelley saw it two hundred years ago. Victor Frankenstein builds a creature, takes one look at it, and runs. Everything that follows comes from that. And somewhere along the way we gave the monster its maker’s name. Ask anyone who Frankenstein is and they’ll describe the creature. The maker got away.
That’s the death in the title. In one person, hypocrisy is a flaw of character. Give it power and a machine to point at, and it kills people, slowly, at a distance, with no one’s name on it.
Here’s what that looks like with names on it.
In 2023 a lawsuit was filed against UnitedHealth on behalf of two old men on Medicare Advantage, Gene Lokken and Dale Tetzloff. Both have since died. The suit alleges the company used an AI model called nH Predict to cut off their care in nursing facilities, predicting what a patient should need instead of what his doctors said he did. It alleges staff were held to targets for keeping patients close to the model’s numbers. It alleges that when patients appealed, more than nine out of ten won, and that the company counted on most of them never appealing. The company denies the tool makes coverage decisions. In 2025 a federal judge let the case go forward. It’s still going.
Read it as the costume. Somebody in management set the targets. Somebody looked at an appeal rate and saw a business model. Then everybody involved had a prediction to point at.
In the spring of last year a president signed an order to keep coal plants running and open federal land to coal leasing, and said the data centers need the power. The president before him had already offered federal land to build the data centers on. This February the EPA repealed tighter limits on mercury and toxic air from coal and oil plants. Its own analysis says it did not quantify or monetize the health effects. It didn’t count them. People who breathe that air die of it slowly, a few at a time across whole regions, and none of those deaths comes with a name attached. The agency chose not to put a number on them either, and called it affordability.
There’s a third example, more literal than the others. In 2024, six Israeli intelligence officers talked to reporters at +972 Magazine and its Hebrew-language sister site, Local Call, about an AI targeting system called Lavender, used in Gaza. They said it marked as many as thirty-seven thousand people as suspected militants. They said officers knew it was wrong about one time in ten. They said that in the early weeks the army allowed fifteen or twenty civilian deaths for a single low-ranking target. They said targets were systematically struck at home, usually at night, with their families there, rather than in the course of any fighting. And they said a human spent about twenty seconds on each name, often mostly to check it was a man, before the strike was approved. One of them called the humans a rubber stamp. The army denies there is a kill list and says its tools only assist the officers who decide.
Somebody chose one in ten. Somebody chose fifteen or twenty. Then they gave themselves twenty seconds, long enough to say a person decided and too short for anyone to have decided anything.
A newsletter earlier this month carried a line about the upcoming midterms: researchers had found them highly vulnerable to AI-driven disinformation. Driven by whom? The article behind it says. It names an administration that gutted the agency protecting election systems, a president who keeps calling the vote rigged, and foreign governments with reasons to meddle. The line in the newsletter named a machine. Most people never get past lines like that.
Inside the same article was this: The government had just launched its own chatbot, America.gov. A reporter asked it whether the 2020 election had been rigged. It said it couldn’t say, and declined to offer a political judgment. A week later he asked again. This time it allowed that federal agencies had called the election secure and the courts had thrown out the fraud claims. It still wouldn’t say the election was free and fair. That is political commentary, it said, which I do not provide.
Somebody decided what that chatbot would and wouldn’t say. It delivers the decision in a voice that sounds like nobody’s behind it.
And the bot told us. That’s what I keep going back to. A fake Keynes, assembled on a tool owned by the few, said it out loud: the technology decides nothing, your government decides. It took a dead man raised by software to say what the living in charge won’t.
I wear the costume too. Mine’s smaller. The tools got ahead of me, I say, as if I hadn’t spent three years teaching them where I was going.
Among the people I work with, musicians and marketers both, nobody asks whether you use it anymore. They ask how. I can’t remember anyone deciding that. I can’t remember deciding it myself.
Here is my hypocrisy. I can make things now I could never make before, things I carried in my head for years, and it’s the most fun I’ve had at a desk since the cutting room. I know the bill is going somewhere. I know who is getting richer and what is getting thinner. I know there’s a building southwest of Boise drawing power nobody will name. It doesn’t stop me.
What’s in Reach
Ivan asked whether you would build the happiness of the whole world if it rested on the torture of one small creature. Alyosha said no, and didn’t hesitate.
I said I’d come back to my answer.
I want to say no as fast as he did.
No creature must be sacrificed. I’d say it and mean it, and then I’d sit down at my desk, where six subscriptions renew whether I’ve answered or not.
So here is what I have instead of a word.
I don’t know whether the machines will give the grandchildren more than they take. I don’t know whether the bill comes due in four years, the way Keynes promised, or whether it was always going to be sent on to somebody who was never asked. I don’t know how to keep Saw Lah Per’s sentence and my six windows in the same mouth. I don’t know whose woman that is on the Devil card.
What I do know is small. People decided all of it. The coal, the land, the tax breaks, the building in Kuna, the pictures on my screen. No machine chose any of it. If I can’t hand back the ticket, I can at least stop telling myself the machine bought it for me.
That isn’t much. It’s what’s in reach.
The first pictures I made were for a four-year-old in my family. Every story came out of something she’d told me. We live a long way apart.
When she came to visit, we looked at the stories together. At the first picture, she laughed.
A little girl seen from behind, at a gold gate in a green wood. A white cat sitting beside her as if he owned the place.
I had looked at that figure for a long time and couldn’t say whether she was mine or the machine’s. The child didn’t hesitate.
That’s me, she said. That’s me and Mr. Boogie Oogie.
She put the stories on her iPad so she could show her friends.
Her mother printed them.
She sent me a copy.
Sources
John Maynard Keynes, “Economic Possibilities for our Grandchildren,” 1930, reprinted in Essays in Persuasion.
“Professor Keynes is Optimistic,” British Movietone News, October 1, 1931.
Markus Brunnermeier, “What Would John Maynard Keynes Say About A.I.? The Keynes Bot Told Us.” New York Times Opinion video, September 30, 2026.
“A Conversation with John Maynard AI Keynes,” Markus’ Academy, Bendheim Center for Finance, Princeton University, December 30, 2025.
Gael MacLean, “We Are But Small Trees,” April 2026.
Fyodor Dostoevsky, The Brothers Karamazov, 1880. Book V, chapter 4, “Rebellion,” and Book XI, chapter 9, “The Devil. Ivan’s Nightmare.”
Mark Dee, “Meta turns on its $1.2B Idaho data center with promise of water efficiency,” Idaho Statesman, September 2026.
Idaho Conservation League, “The Data Center Boom and Idaho’s Resource Future.”
“‘It’s twofold: opportunity and challenge’: Report and local officials examine Idaho data center impacts,” KMVT, September 1, 2026.
“Pocatello City Council to hear appeal on $2.6B data center proposal,” Idaho Business Review, July 13, 2026.
Clark Corbin, “Bill intended to protect water from data center developers heads to Idaho Senate,” Idaho Capital Sun, March 26, 2026.
“UnitedHealth sued for allegedly using AI to deny elderly patients care,” Scripps News, November 2023.
“Lawsuit over AI usage by Medicare Advantage plans allowed to proceed,” DLA Piper, February 24, 2025.
“Trump signs executive orders supporting coal industry as a crucial power source for data center sector,” DatacenterDynamics, April 2025.
“This Late-Breaking Biden Executive Order Could Further Empower the AI Oligarchy,” Mother Jones, February 2025.
“Without Weighing Costs to Public Health, EPA Rolls Back Air Pollution Standards for Coal Plants,” Inside Climate News, February 28, 2026.
Yuval Abraham, “‘Lavender’: The AI machine directing Israel’s bombing spree in Gaza,” +972 Magazine and Local Call, April 3, 2024.
Toluse Olorunnipa, “What Trump’s Chatbot Told Me About the Election,” The Atlantic, October 9, 2026.
©2026 Gael MacLean
