The Gods of Rome Walk Into a Reality Show
How This Started
It began, as many things do, on Reddit.
Someone was generating images of the planets as people, which was visually interesting, stylistically consistent, and about as deep as a decorative plate. I found myself thinking about what it would actually mean to take the Roman gods seriously as a contemporary prompt. Not the planets named after them, not the aesthetic shorthand, but the actual deities, their domains, their relationships, their place in a specific culture's imagination.
So I asked an image model to render the Roman pantheon as a contemporary family, placed in the economic and social world of their cultural descendants. Modern Italians, or Italian Americans, or whoever the AI understood the inheritors of Rome to be.
What came back was not what I expected. It was funnier than I intended, and considerably more uncomfortable than the joke.
A Note on the Romans Before We Get to the Image
Roman religion is easy to condescend to, which is worth resisting. We tend to treat it as mythology-lite, Greek philosophy with the poetry stripped out and replaced by bureaucratic competence. And there's some truth in that. The Romans were not, as a culture, in the business of sitting with mystery. They were in the business of results.
The relationship between a Roman citizen and the gods was essentially transactional. You made the right offering at the right time in the right way, and the gods held up their end. Jupiter didn't love you. He administered you. Mars wasn't a figure of tragic fury like Ares. He was, more or less, a very senior defense contractor. Venus handled soft power and alliance-building. The whole operation was hierarchical, organized, and not particularly warm, which, if you think about it, is an almost perfect description of Rome itself.
What the Romans gave their gods was jurisdiction, not depth. And that transactional quality, that sense of divine power as something to be managed rather than experienced, turns out to matter quite a bit when you see what the AI did with them.
What Came Back: Jersey Shore Meets the Sopranos
The image is something.
A large group, staged like a prestige drama cast photo, draped across expensive furniture in what appears to be a Manhattan apartment or a very confident New Jersey living room. There's champagne. There's cash on the table. There's a small white dog being held by a woman in a champagne-colored slip dress who is the obvious center of the composition and is either Venus or the matriarch of a crime family, and the distinction feels unclear. An older man with silver hair and a well-cut suit radiates the specific energy of someone who has made serious decisions in rooms you were not invited to. A man in gold chains and a white blazer holds a drink and grins like he sold you something last week and would like to again.
The younger figures are a catalog of contemporary male affect: the sullen one in the fur coat, the one with his arms crossed in a track jacket, the one in sunglasses who has clearly decided that looking bored is a personality. There's a cigar. There are large watches. There is the posture of people who have power and want you to know it but also don't particularly want to explain it.
It reads, unmistakably, as a specific kind of Italian American wealth, the kind that has been codified by thirty years of television and film into something almost iconographic. It's funny. It's also something else.
What the AI Was Actually Doing
Here's the thing to understand about how image generation models work. The AI did not decide to dress Jupiter in gold chains. It did not make a joke. It has no relationship to irony and no opinion about Italian Americans. What it did was something more interesting and more troubling: it reached into its training data and synthesized the dominant visual vocabulary associated with its interpretation of the prompt.
"Roman gods as a contemporary family" required the model to make a series of inferences. Roman gods, it knows something about. Contemporary family in the economic and cultural world of Rome's descendants, that required a different kind of reach. And what it reached for was the dominant representation of Italian American wealth and power in American media for the last three decades. The Sopranos. Mob movies. Reality television. A specific aesthetic of excess that has been reproduced so many times it has become its own genre.
The AI is not making a judgment. That's the point. It is reflecting one back at us, assembled from the aggregate of what we have produced. The image is a kind of readout, not of what Italian Americans are, but of what the cultural record says they look like when they have money and power. Those are very different things, and the gap between them is where the discomfort lives.
The Weight of the Calcification
Italian Americans have an unusual position in the American cultural imagination. They arrived as laborers and faced genuine, documented discrimination, the kind that included lynchings and legal exclusion, not just the soft condescension reserved for groups considered merely foreign. They built communities, entered the professions, contributed to every sector of American life. And then popular culture handed them an identity almost entirely organized around organized crime and working-class caricature.
The Godfather and The Sopranos are genuinely great art. I'm not making an argument against them. But great art calcifies when it becomes the only story. When the AI assembles its image of Italian American power and reaches, automatically, for gold chains and champagne and a certain kind of studied menace, it is drawing on the sediment of that calcification. The image doesn't include Enrico Fermi, who split the atom. It doesn't include Frank Capra, who made some of the most hopeful American films of the 20th century. It doesn't include Don DeLillo, who has spent fifty years writing some of the most searching fiction about what American culture does to the people inside it. It doesn't include any of the painters, the architects, the scientists, the writers, the judges.
It includes champagne and cash and a small white dog.
What does it mean when a technology trained on the whole of human cultural output distills a community down to its most repeatedly broadcast image? It means the representation gap is real, and the AI just showed us exactly how wide it is.
The Mirror
This is what I keep coming back to when I think about what these image models are actually doing. They have no agenda. No malice. No awareness. They don't know they're making a statement about Italian American representation, and they certainly don't care. They just synthesize patterns from the record humanity has produced.
Which makes them a strange kind of critic. Not an intentional one, not a fair one, but a remarkably honest one. The image reflects the aggregate, and the aggregate is what we have collectively decided to make and distribute and reward and repeat. The AI didn't invent that aesthetic. We did. The AI just handed it back without the usual social filters that would make us hesitate to notice what we were doing.
That's what this series is going to keep exploring: what happens when you ask the machine to show you how it sees the world's sacred traditions in contemporary life, and then you actually look at what it shows you.
Ten Pantheons, Ten Worlds
This post is the first of ten, coming over the next several weeks and mixed in with other material, as is my habit.
The structure of each one is the same: take a pantheon, ask the AI to render it in the economic and cultural world of its modern descendants, and then examine what the choices reveal about how that culture is seen and how it actually lives. The gap between those two things is the subject.
The series will move through the Norse, Celtic, Egyptian, Yoruba, Hindu, Shinto, Aztec, Mesopotamian, and Chinese traditions. Each image tells a different story. Some of them will be generous and surprising. Some of them will be uncomfortable in the same way this one is, but from different angles.
There will also be a post on the Abrahamic traditions, specifically Judaism and Christianity, where the AI produced two of the most unexpected images in the entire project. And a post on Buddhism, built around a single image that says more about compassion in contemporary America than I was prepared for when I first saw it.
Islam is the notable absence in this series, and I want to say that plainly rather than let it pass as a footnote. The decision not to generate an image was deliberate, made out of respect for a tradition whose strong prohibition against figurative representation of the divine is not a technicality but a theology. That prohibition has meaning. It deserves to be met with something other than a workaround.
But the absence is also part of the conversation. A series about how AI sees the world's sacred traditions has to reckon honestly with the tradition that most forcefully resists being seen that way. What does it mean to have a mirror that cannot, by the terms of the tradition itself, reflect what it most wants to see? I don't have a clean answer. I think that's a good place to start.
The next post in this series looks at the Norse pantheon, which produced an image I did not expect and have been thinking about ever since.