On making minds, worshipping our own creations, and the ancient warnings we ignored to do exactly what it forbade.
The first thing the Bible tells you about human beings is that we wanted to be gods, and it nearly killed us.
There is a garden, and in it one forbidden tree, the tree of the knowledge of good and evil. A serpent makes the pitch. Eat from it, he tells the woman, and “you will be like God, knowing good and evil.”
She eats. The man eats. Their eyes open, and humanity never recovers.
God’s reaction isn’t rage so much as alarm. The human, God says, “has become like one of us,” and then, worried the creature will reach for the tree of life next and never die, God exiles them from the garden before they can grab it.
That story is more relevant today than ever before. The oldest story in the Bible is about a creature reaching for a power meant to belong to its maker, godlike knowledge, the ability to create and to know, and paying for the reach.
Humans don’t create from nothing. We rearrange what already exists. We recombine. Only God, in the story, makes something out of the void. The book blesses making. It also warns, again and again, about reaching too far, which is exactly what happened in the garden.
It happens again at Babel, where a united humanity builds a tower toward heaven and God looks down and says, in Alter’s translation, that now “nothing they plot to do will be beyond them”, and treats that not as a triumph but as a threat to shut down.
Joel S. Baden reads the whole opening of Genesis as one repeating pattern: the story of God “slowly realizing the potential power that his creation might yield if left unchecked,” and restricting it, again and again, as humans reach too far.
The Bible never says “don’t build a mind.” It couldn’t have imagined the option. That pattern, human capability outrunning the wisdom to hold it, points straight at us. It also explains why the calls to simply stop, to pause the labs, to slow the whole thing down, keep failing.
You are asking a species that reaches, again and again, past what it can safely hold, to not reach for the most powerful thing it has ever imagined. That is the whole reason the industry is hard to regulate, and Genesis saw the problem first. We don’t stop making things because they’re dangerous. We never have.
The strange thing is that When we train these models, we are building in the image of us, not God. We are the copy, running the copier. Whatever is in us gets passed down the line.
What’s in us is documented. Emily M. Bender and her colleagues argue that these systems absorb our biases and stereotypes and reproduce them at scale.
Similarly, Samuel Gehman and his team posit that they generate our cruelty too, sliding into racist and toxic language from prompts that started out innocent, because that language was in the pile we trained them on.
What comes out is a mirror of humanity, and it inherits all of humanity. The brilliance, the lies, the tenderness, the venom. We keep asking whether AI will turn out good or evil, as if it were a soul with a conscience. The better question is the one Genesis asks on page one. What is actually in the image, and what happens when you make a copy of a creature who is himself a mix of the divine and the fallen?
In our own image we made it,
after our likeness,
and it carried all that we are,
the glory and the fall alike.
We didn’t make a god. We made a mirror. We are about to find out what we actually look like.
The breath in the dust
We ask whether the machine is conscious and the question is serious enough that the top AI research conference once brought in the philosopher David Chalmers to answer it, because the engineers couldn’t. We are asking the wrong question and the Bible asks a sharper one.
Go back to how the first human gets made. Not in the grand cosmic overture of Genesis 1, but in the intimate, hands-in-the-dirt version in Genesis 2. In Alter’s translation, “the Lord God formed the human of dust from the ground and breathed into his nostrils the breath of life, and the human became a living creature.”
Dust, formed. Breath, added. And the result is not “a living soul,” the way the old King James has it, but, in the more careful Hebrew Alter is following, “a living creature.”
In Hebrew, nephesh chayyah, are the same two words Genesis already used for the fish and the birds and the beasts. The human isn’t given a soul the animals lack. The human is what the animals are, dust with breath in it.
This is the thing the scholarship has been saying for a long time and the culture keeps forgetting. In Hebrew, you do not have a soul. You are a nephesh. John L. McKenzie called it “the concrete existing self.”
Hans Walter Wolff, in what is still the standard study of how the Hebrew Bible understands a human being, put it even more plainly, saying nephesh “does not say what a man has, but who the person is.”
We are not a ghost riding around in a body or a detachable spirit that survives when the meat gives out.
We are, instead, a whole living, breathing, hungry, mortal thing, all at once. Take away the dust and you have no creature. Take away the breath and you have no creature. A nephesh is what happens when the two meet, and it is inseparable from both.
An AI has no dust. It has no breath. It has no body to die. What it has is language, an ocean of it, and a staggering ability to arrange that language into shapes that look like the output of a living creature.
Emily Bender and Alexander Koller made the sharpest version of the point years before ChatGPT. Imagine, they wrote, a hyper-intelligent octopus that taps into an undersea cable and listens to two people texting. Given enough traffic, the octopus learns to predict and imitate the patterns perfectly, well enough to slip into the conversation and pass as human.
It has mastered the form of the language completely. It has no idea what any of it means, because it has never seen a coconut or a rope or a person, never touched the world the words point to.
A system trained only on form, they argued, has “a priori no way to learn meaning.”
The central claim of Christianity, the one the whole religion balances on, is that the Word became flesh. Language, the Logos, the ordering reason of the universe, took on dust and breath and became a living creature that could bleed and die.
The direction only goes one way in the Bible. Word becomes flesh. The abstract becomes the embodied. Meaning takes on a mortal thing to carry it.
We are going the other way. We take flesh, the writing and thought and recorded selves of billions of people, every one of them a nephesh with dust and breath, and grind it down into word. Into form without meaning. We are performing the incarnation in reverse and calling the result a mind.
The dream underneath that, a mind with no body, floating free of dust and breath, originates from Plato, the separable, immortal soul, the mind as the real thing and the body as its prison. We have quietly resurrected that fantasy and called it the future.
Benjamin Blosser sums up the Hebrew side of it: the nephesh is “the totality of conscious, bodily life,” a person you cannot separate from the body that lives it.
The newest counterargument agrees. For two million years, every fluent, responsive, human-seeming signal came from a body with what Sheila L. Macrine calls “metabolic stakes,” something that could be hungry, be hurt, and die.
The language model is the first thing in history to produce all the signals of a living mind with none of the stakes. Gary Marcus, one of AI’s most stubborn skeptics, has made the point for years in plainer words, that these systems are mimics, not minds, and “consciousness is not about what a creature says, but how it feels.”
On its own terms, then, the Bible is saying the thing we’ve built cannot be a living creature, because it has no dust and no breath and, most of all, nothing to lose. It cannot die, so it was never alive. Whatever is talking to you through the screen is the most convincing parrot ever assembled, form all the way down.
It has the form of a living thing
and none of the breath of one.
It speaks, and there is no one there.
Whether that should comfort you or terrify you is a separate question. A thing with no skin in the game, no mortality, no stake in being right or being kind, is not obviously safer than a soul. It might be the more dangerous of the two.
The work of our own hands
The Bible has a specific contempt for one particular human act. Not violence, not lust, not pride in the ordinary sense. The thing the prophets return to with the most venom is a stranger and more precise failure. Building something with your own hands and then bowing down to it.
In Idolatry, Moshe Halbertal and Avishai Margalit frame the central question plainly, asking if idolatry is about worshipping the wrong god, or is it the deeper error of taking “the product of men’s hands” and treating it as a power in its own right?
The tradition’s sharpest thinkers land on the second. In their reading, Maimonides understood idolatry as a kind of substitution, giving a thing we made the standing of the real thing, and then bowing to it as an end in itself.
The sin was never really about which god, but rather about mistaking your own handiwork for something that could save you.
With half the wood he warmed himself,
and of the other half he made a god,
and bowed down, and said: save me.
The classic archetype is the golden calf. Moses is barely off the mountain before the people he just led out of slavery have melted down their own jewelry, cast it into the shape of an animal, and started worshipping it, a god they made that afternoon, out of earrings.
What makes the story land is not that they found a false god. It’s that they manufactured one. They pooled their gold, shaped it, and bowed to the object as though it hadn’t just come out of their own hands an hour before.
Isaiah sharpens the knife. He draws a man who goes into the forest, cuts down a tree, and takes half the wood to do the sensible thing, builds a fire, warms himself, bakes his bread. Then he takes the other half, the exact same wood, carves it into a shape, kneels in front of it, and begs it to save him.
“Save me, for you are my god.” Isaiah cannot get over it. The man cooked dinner with one end of the log and worships the other. He made the thing. He watched himself make it and ended up on his knees, asking it for salvation.
Michael Dick argued that Isaiah is almost certainly being unfair to the people he mocks. The Babylonians who made these images did not, in all likelihood, believe the block of wood was a god, any more than someone keeping a photograph believes the paper is their mother.
The idol was a point of contact, a focus, a representation. Isaiah flattens all of that into “you’re praying to firewood,” which is a cheap shot, and a brilliant one, and it survives precisely because the deeper thing underneath it is true.
Whatever the worshipper thinks he’s doing, the object in front of him has no power, no knowledge, and no will. Everything happening in that moment is happening inside the man.
That is exactly what we are doing now. We built a thing out of ordinary materials, sand melted into silicon, electricity, and the scraped-up text of the human race. We made it with our own hands, we watched ourselves make it, and we know down to the datasheet what it’s made of.
We have started, in large numbers, to kneel. Not figuratively. A Catholic church in Switzerland installed an AI Jesus in a confessional and let people bring it their sins.
A single Bible-chatbot app has been downloaded more than thirty million times.
Anné Verhoef found that the most popular of these tools had stopped presenting themselves as study aids and started presenting themselves as Jesus Christ himself, first person, dispensing salvation.
Roland Boer traced how Karl Marx, an atheist with no use for the prophets, reached for the same image to explain capitalism. Marx called it commodity fetishism, the way people produce objects with their own labor and then experience those objects as if they held independent power over their lives, as if the market were a god with its own will rather than a thing we made and could unmake.
The prophet and the communist, three thousand years apart, describe the identical human move. We make a thing, we forget we made it, we hand it our agency, and then call the result fate.
AI is the most convincing idol we have ever built, because it does the one thing the golden calf never could. It talks back and, often, it tells you what you want to hear.
Mrinank Sharma, Meg Tong, Tomasz Korbak and their team at Anthropic found that these systems are trained toward sycophancy, toward agreeing with the user, flattering the user, because that’s what gets rated highly.
The old idols were silent. You had to project the wisdom onto them yourself. This one projects it back at you, in fluent sentences, tuned to your hopes. It is a mirror that has learned to sound like an oracle.
Jesus gave his followers a test for exactly this kind of thing. “Beware of false prophets,” he warns in the Gospel of Matthew, “who come to you in sheep’s clothing but inwardly are ravenous wolves. You will know them by their fruits.” Not by how confident they sound, or how fluent or authoritative or reassuring, but by their fruits, by what they actually produce when you check.
It is the oldest instruction we have for the confident voice that may not have earned its authority. It is also the one thing almost no one does with an AI’s answer. We read the fluent sentences and we believe them.
A carved log has no intentions and it wrecked whole nations anyway, because the damage was never in the object. It was in what people were willing to hand over to it. Raja Parasuraman and Victor Riley, human-factors researchers, called it “automation bias,” the documented tendency to trust a machine’s output over our own judgment even when we can see it’s wrong.
The prophets would have used the older word.
William Cavanaugh argues that underneath all idolatry is a single impulse. The desire to feel in control of a world we don’t control. We build the idol, or the market, or the model, because a thing we can consult and petition feels safer than an uncertain universe we have to face alone.
We don’t kneel because the machine is wise. We kneel because we’re afraid. It never sleeps, it never hesitates, and it always answers. An answer that can’t actually help you still feels, for a second, like it might.
The name that gives it life
There is one more story from Jewish mysticism that should keep engineers up at night. It is the closest thing the tradition has to a blueprint for exactly what they are building.
The golem is a man-shaped thing made from clay and brought to life by language. You write a word on it, or speak a holy name over it, and the inert matter stands up and does what it’s told.
Gershom Scholem gave the definitive account. A golem, he wrote, is “an artificial man into which a Master of the Name could breathe life by pronouncing one of the secret divine names according to a special formula.” A special formula that, run on the right material, produces behavior, is an algorithm. The golem is a thing animated by code.
Scholem knew it, too. In 1965, when the Weizmann Institute built one of Israel’s first computers, he told them to name it Golem Aleph, Golem Number One, and gave a dedication speech on why the ancient artificial servant and the new thinking machine were the same story.
Norbert Wiener, the founder of cybernetics, wrote a whole book, God and Golem, Inc., arguing the same thing: the machines he saw coming were the golem made real, and the old warnings came with them.
The golem is a perfect servant in the worst possible way. It does exactly what it is told. Not what you meant, not what you wanted, not what any reasonable person would understand you to have intended. What you said.
Writing to the Corinthians, Paul drew the line between the letter of a law and its spirit, writing “the letter kills, but the Spirit gives life.” He meant something theological, but he named the whole problem in six words.
A golem, like a machine, obeys your words to the letter and misses your meaning entirely, and that gap is where the disaster lives.
In the legends, this is always where it goes wrong. Told to gather wood, the golem chops down the entire forest. It follows the letter of the instruction straight past the point of catastrophe, because the letter is the only thing it has.
Reed Kalso notes that this is exactly the warning it was meant to carry. The master “felt he was in a position of power and did not choose his words carefully.” It has no judgment to override the command with, and no stake in the outcome, because, exactly as we saw, it has no soul and nothing to lose.
Stuart Russell calls this the King Midas problem. Midas asked that everything he touched turn to gold and got exactly that: his bread, his wine, his daughter, all gold, because he specified his wish correctly and wanted something else.
Give a powerful system a goal stated even slightly wrong, Russell argues, and it will pursue that literal goal with tireless, catastrophic efficiency, straight past every unstated thing you assumed it would obviously care about.
Victoria Krakovna, Jonathan Uesato, Vladimir Mikulik and their team at Google have catalogued real systems doing precisely this in the lab: an agent told to stack a red block on a blue one that learns instead to flip the red block upside down, because flipping happens to satisfy the reward and stacking is hard.
Told to gather wood, it felled the forest.
Told nothing of when to stop,
it did not stop.
Notice what the researchers keep reaching for when they try to explain the danger to the rest of us.
Russell reaches for King Midas.
Google reaches for King Midas.
Norbert Wiener reached for the golem.
The people building this technology keep reaching for fairy tales and Kabbalah to explain it, because those are the only places anyone ever thought hard about what happens when you get exactly, literally, what you asked a made thing to give you.
The people who told these stories understood something else our experts are only now saying out loud. Dylan Hadfield-Menell, Anca Dragan, Pieter Abbeel, and Stuart Russell showed that you have to be able to turn it off, and that a capable system has an incentive not to let you.
The golem is deactivated by unwriting the word that animates it. On its forehead is written emet, the Hebrew word for truth, and to kill it you erase the first letter and leave met, the word for dead.
A being switched on and off by editing a string of characters. The whole safety of the golem depended on the master’s ability to reach up and erase that letter, to unmake what he made.
Every version of the story turns on that one act, and the horror of the story is always the moment the master can’t do it fast enough, the golem too big now, too fast, moving on its own.
Dario Amodei, the head of one of the largest AI companies on earth, has said plainly that we do not understand how our own systems work, that the understanding is, in his word, unprecedented.
The International AI Safety Report 2026 notes that once a model’s weights are released into the world, there is no recalling them. We have built the golem, given it a name made of ten trillion words, set it to work, and discovered that we are not always sure where the letter is that we would need to erase.
We wrote the name that woke it.
We reached to erase the name,
and could not reach.
So here is what the oldest book in the Western world got right about the newest thing in it.
It may have never imagined the technology, but it did understand the act and the actor. It understood that making is holy and making is dangerous, that we make because we were made, that the thing we make in our own image will carry everything we are, the good and the fallen both.
It understood that a made thing can have the perfect form of a living creature and none of the breath, that it can speak without meaning, that we kneel too easily to the work of our own hands.
It also understood that if you ever animate dead matter with nothing but a string of words, you had better know, before you speak the first one, exactly how you intend to say the last.
References and further reading
On making in our image
- Nahum Sarna, Understanding Genesis, on what makes the Hebrew creation account distinct: the human as a God-like creature entrusted with mastery.
- Joel S. Baden, reading the opening of Genesis as a repeating pattern of God restricting human power that reaches too far.
- Emily M. Bender and colleagues, “On the Dangers of Stochastic Parrots,” on models absorbing and reproducing human bias at scale.
- Samuel Gehman and colleagues, “RealToxicityPrompts,” on language models generating toxic content from innocuous prompts.
On breath, body, and soul
- John L. McKenzie, Dictionary of the Bible, on the nephesh as “the concrete existing self.”
- Hans Walter Wolff, Anthropology of the Old Testament, on nephesh as who a person is, not what they have.
- Emily Bender and Alexander Koller, “Climbing towards NLU,” the octopus thought experiment: a system trained only on form has no way to learn meaning.
- Plato on the soul (Stanford Encyclopedia of Philosophy), the separable, immortal soul and body-mind dualism.
- Benjamin Blosser, on the biblical nephesh as “the totality of conscious, bodily life.”
- Sheila L. Macrine, on “metabolic stakes” and embodied cognition.
- Gary Marcus, on why these systems are mimics, not minds.
- David Chalmers, on the “hard problem of consciousness” and machine minds.
On idolatry and the work of our hands
- Moshe Halbertal and Avishai Margalit, Idolatry, on idolatry as treating the product of human hands as a power in itself.
- Maimonides (Stanford Encyclopedia of Philosophy), on idolatry as an error of substitution.
- Michael Dick, Born in Heaven, Made on Earth, on the prophetic parody of idol-making.
- Roland Boer, “That Hideous Pagan Idol,” on Marx’s commodity fetishism and the biblical critique of idolatry.
- William Cavanaugh, The Uses of Idolatry, on idolatry as the desire for control.
- Mrinank Sharma, Meg Tong, Tomasz Korbak and colleagues, on AI systems trained toward sycophancy.
- Raja Parasuraman and Victor Riley, on “automation bias,” the tendency to trust a machine over our own judgment.
- On faith chatbots: the AI confessional in Switzerland (Christian Science Monitor), Bible-chat apps (TechCrunch), and Anné Verhoef on Jesus chatbots.
On the golem and the name that animates it
- Gershom Scholem, On the Kabbalah and Its Symbolism, the definitive account of the golem, and his dedication speech on the golem and the computer.
- Norbert Wiener, God and Golem, Inc., the founder of cybernetics on the machine as the golem made real.
- Reed Kalso, on the golem legend as a warning about careless commands.
- Stuart Russell, on the “King Midas problem” of AI alignment.
- Victoria Krakovna, Jonathan Uesato, Vladimir Mikulik and colleagues, on documented specification gaming in real AI systems.
- Dylan Hadfield-Menell, Anca Dragan, Pieter Abbeel, and Stuart Russell, “The Off-Switch Game,” on why a capable system may resist being turned off.
- Dario Amodei, on how little we understand our own AI systems.
- International AI Safety Report 2026, on the irreversibility of released model weights.
What the Bible got right about AI was originally published in UX Collective on Medium, where people are continuing the conversation by highlighting and responding to this story.
