There is something almost instinctive about the way we reach for old stories when something new makes us uncomfortable. Not because the old story fits, but because it has edges. A beginning, a shape, an ending we already know. The new thing doesn’t have any of that yet. So we borrow the container and hope the contents don’t spill over.
I was at a panel recently: AI and science fiction, or so it was presented. Within the first few minutes, someone mentioned the golem. Then Prometheus. Then HAL. The references came quickly, one after another, and then the conversation moved on, leaving all three of them sitting there like unopened boxes. Nobody went inside any of them. Nobody had to. The names alone were enough to create the impression that something meaningful had been said.
The golem is a figure from Jewish folklore, most fully formed in the 16th-century legend of Rabbi Loew of Prague. A being made from clay, animated by inscribing the Hebrew word for truth on its forehead, created to protect a community under threat. It cannot speak. It has no inner life. It follows instructions without understanding them, and the danger it poses is not rebellion but literalism: it does exactly what it is told, including when no one remembered to tell it to stop. Control is maintained through language. One letter erased, and it returns to clay.
You can see immediately why the comparison gets made. Language-animated, obedient, potentially dangerous through excess compliance rather than intent. There is a real structural similarity between the golem and certain anxieties about AI. Prometheus works differently: his story is about overreach, about stolen fire and punishment, about the creator suffering for what they’ve unleashed. Two myths, two completely different fears, dropped into the same conversation as though they were saying the same thing. They weren’t.
The container and what spills over
This is what bothers me about how these comparisons tend to work. Not that they’re made. The impulse is understandable, and myths exist precisely because humans have always needed tools for thinking about things that feel too large or too new to hold directly. Stories are how we process what we can’t yet name. That’s not a weakness. That’s culture doing its job.
The problem is when the comparison stops there. When the myth gets offered as a conclusion rather than a starting point. Because the golem was never designed to think about artificial intelligence. It was designed to think about persecution, community, and the very specific danger of a protector that cannot exercise judgment. The connection to AI is real but partial; an overlap, not a map. And the moment you treat it like a map, you stop looking at where the territory actually goes.
Interestingly, science fiction does this better. HAL, the Replicants in Blade Runner, Asimov’s robots – these stories were built specifically to think about the problem we are now living inside. The complexity is already in the text. You don’t need to build the argument yourself, because the writers already did it; they thought hard about exactly where the comparison holds and where it breaks, and they put that thinking into the story. Which makes it quietly strange that a panel convened to discuss AI and science fiction reached for ancient mythology instead.
What the reaching is really about
I don’t think the people in that room were being intellectually careless. I think they were doing something very human in a situation that called for it: finding the nearest familiar shape and holding onto it. Because AI, as it currently exists, does not have a stable shape. The people building it can’t fully agree on what it is. The people studying it are still working out the vocabulary. And sitting with that, with the genuine open-endedness of it, is uncomfortable in a way that an old story with a known ending is not.
Myths feel authoritative in a way that science fiction doesn’t, because they’re old enough to feel unchallengeable and communal enough to feel like shared property. Nobody has to have read anything specific to feel like they have access to the golem. It belongs to everyone, vaguely. And that vagueness is part of its appeal. It’’s hard to argue with something that ancient and that collectively held.
But a myth used this way isn’t doing the work of a myth. It’s doing the work of a security blanket. It gives the feeling of historical grounding without requiring anyone to go anywhere uncomfortable, without asking what specifically connects them and what doesn’t, without sitting in the gap between the old story and the new thing. And that gap is exactly where the interesting questions live.
We don’t have a good enough story for AI yet. Maybe that’s the more honest thing to say. The golem gets us part of the way there, Prometheus gets us somewhere else, HAL gets us closer but still not all the way. None of them quite fits. And borrowing the ones that almost fit, without acknowledging the almost, is a way of avoiding the work of building something new.
I’m not sure we’re ready to do that work. But I’m not sure the old containers can hold this much longer either.
If this got you thinking about the frameworks we use to make sense of AI, you might also find The line we needed to draw and The other we invented worth a read.