Oh God, Please Say It Ain't So: Elon Musk Just Took On My Homer Project

23 Jul, 2026

Elon Musk says Grok Imagine will produce a historically accurate Odyssey by year's end. But visual authenticity is only the beginning. Can AI recreate Homer, or only recreate what Homer looked like?

For the past several weeks I have been working on a small research project. The question was very narrow and, I thought, mine alone: could artificial intelligence learn the visual language of ancient Greece — the armor, the ships, the palaces, the light on the Aegean — from archaeology, sculpture, and vase painting rather than from Hollywood, fantasy art, and video games? I wrote about that experiment last week in Animating Homer: Teaching AI the Visual Language of Ancient Greece. It was a modest post about a modest ambition.

Then, on July 22, Elon Musk announced that Grok Imagine would produce a full-length, "historically accurate" film of The Odyssey before the end of the year — one "true to the art of Homer." Suddenly my small research project was a footnote to a much larger conversation.

I will admit my first reaction was closer to the title of this post than to scholarly detachment. But once the initial oh no passed, I realized the announcement is actually a gift. It sharpens the question I had been circling. Last week I asked whether AI could learn to see ancient Greece. Musk's announcement raises the next question, and it is a deeper one: can AI learn Homer himself?

Elon Musk speaking at the 2025 Conservative Political Action Conference (CPAC), National Harbor, Maryland, February 20, 2025. Photograph by Gage Skidmore. Source: Wikimedia Commons. Licensed under CC BY-SA 2.0.

What Musk Actually Said

The announcement was, in the way of these things, a single sentence posted to X. "Before this year ends," Musk wrote, "Grok Imagine will make a full-length movie of The Odyssey that is historically accurate and true to the art of Homer." When a follower proposed that Mel Gibson direct a version with period-accurate ships, armor, and weapons, and dialogue delivered in Homeric Greek, Musk's reply was a single word: "down." Forbes reported he floated a hundred million dollars for such a film.

The remarks land in the middle of a months-long feud with Christopher Nolan, whose own Odyssey is in production. Musk has repeatedly criticized Nolan's casting choices, at one point saying the director had "lost his integrity." So the Grok promise is partly a shot across a rival's bow. Variety, the Hollywood Reporter, Deadline, CNN, the Guardian, People, and NME all picked it up within hours — the story traveled from Greek cultural press to mainstream entertainment coverage in a single news cycle.

I want to resist the easy move here, which is to make this a piece about Elon Musk, or about whether AI belongs in filmmaking at all. Those debates are loud and, for my purposes, beside the point. The interesting phrase is the one everyone repeated without examining: historically accurate. What could that possibly mean for the Odyssey?

"Historically Accurate" Is Not One Thing

Several outlets, Variety and NME among them, quickly noticed the problem. This is a story with a one-eyed giant, a witch who turns men into pigs, a six-headed sea monster, and a cast of gods who intervene on a whim. In what sense can a film about Circe and Scylla be "historically accurate"?

The phrase collapses at least four distinct standards that a serious adaptation has to hold apart.

There is archaeological accuracy — the material world. Did the ships look like that? Were the greaves bronze? Is the megaron the right shape, the loom the right kind, the wine cup the right silhouette? This is the standard my own experiment was built around, and it is the one AI is best positioned to meet, because it is grounded in surviving objects we can photograph and measure.

There is literary fidelity — faithfulness to the poem as a poem. Its ironies, its famous similes, the structure of its recognition scenes, the strange retarding rhythm by which Odysseus's return is delayed for books at a time. This is not a question of props.

There is philological accuracy — the language itself. Homeric Greek is not classical Greek; it is an artificial literary dialect, a blend of forms from different regions and centuries that no one ever actually spoke in daily life. Dialogue "in Homeric Greek," the Gibson fantasy, would already be a kind of reconstruction of a language that was itself a reconstruction.

And there is mythological tradition — fidelity not to history but to the inherited body of story. The gods and monsters are not errors to be corrected out of an accurate film; they are the material. To make them archaeologically accurate is a category mistake.

These four goals do not merely differ. They frequently pull against one another. An archaeologically scrupulous Cyclops is a contradiction in terms. When Musk says "historically accurate and true to the art of Homer," he has bundled together standards that a working classicist would keep carefully separate — and the tension between them is exactly where the difficulty lives.

What My Experiment Was Really About

The premise of last week's post was that if you want AI to render ancient Greece, you should train its eye on the right sources. Not on the swords-and-sandals imagination of film, not on fantasy illustration, not on the pauldrons and improbable physiques of video games, but on the evidence: excavated armor, Mycenaean frescoes, the black- and red-figure vases that are our richest surviving picture of how Greeks depicted their own myths, the sculpture, the pottery, the ruins.

I still believe that. If Grok's Odyssey draws its visual vocabulary from the Warrior Vase and the Dendra panoply rather than from 300 and God of War, it will already be more honest than most of what has come before. That is a real and worthwhile goal, and it is genuinely within reach.

But writing that post taught me its own limit. Teaching a model to see ancient Greece and teaching it to tell Homer's story are not the same task, and the gap between them is the whole subject of this one.

Here is the distinction I keep returning to. AI can plausibly learn the surfaces of the Odyssey: the architecture, the armor, the ships, the Mediterranean landscape and its light. These are visual facts, anchored to objects, and a system trained on the right corpus can approximate them.

But the Odyssey is not its surfaces. It is a poem that came out of an oral tradition, composed and recomposed by singers over generations before it was ever written down. Its art lives in things that leave no archaeological trace: the formulaic epithets and repeated phrases that were the technology of oral composition; the dramatic irony of a disguised Odysseus sitting in his own hall while the suitors mock a beggar they do not recognize; the deliberate pacing that withholds the hero's return; the great recognition scenes, with Argos the dying dog, with the scar Eurycleia feels beneath her hands, with the marriage bed carved from a living tree that only Penelope's true husband could know. It is the theme of xenia, guest-friendship, that quietly organizes the whole moral universe of the poem. It is memory and homecoming and the ache of nostos itself.

Can a model learn irony? Can it learn the emotional timing that makes Odysseus weep at a song about his own war? Can it learn the narrative voice, the symbolism, the restraint? These are not questions about rendering. They are questions about understanding, and they are enormously harder. You can photograph a bronze greave. You cannot photograph a recognition scene.

The Epic Cycle Problem

There is a further wrinkle, and it happens to be exactly what I have been reading about lately. Homer is not the whole of the Trojan tradition. Around the Iliad and Odyssey once stood a constellation of other epics — the Cypria, the Aethiopis, the Little Iliad, the Iliou Persis (the Sack of Troy), the Nostoi (the Returns) — that told the parts of the story Homer left out: the origins of the war, the death of Achilles, the wooden horse, the homecomings of the other Greek kings. These poems are lost. We have them only in summaries, quotations, and fragments.

And yet much of what those lost poems narrated survives in another medium entirely: the vases. Greek painters illustrated scenes from across the whole cycle, sometimes preserving versions of the story that the fragmentary texts only hint at. Reconstructing the Trojan tradition means reading the fragments and the images together.

So a real question for Grok's Odyssey is this: which Homer will it use? Will it draw on the Odyssey alone, or on the wider inherited tradition that the poem assumes its audience already knows? A "historically accurate" adaptation, taken seriously, would have to reckon with the fact that Homer's first listeners came to the poem already carrying the whole cycle in their heads. That context is not in the text of the Odyssey. It is scattered across fragments and painted pots.

There is another irony here. Homer himself almost certainly never wrote The Odyssey. The poem emerged from generations of singers who borrowed, adapted, and refined inherited stories before they were finally written down. Artificial intelligence therefore enters a tradition that has always involved transmission, adaptation, and reinterpretation. The question is not whether technology changes Homer. Every generation has done that. The question is whether technology can participate in that tradition without losing what made Homer endure.

I do not want to be the person who only says no. Set aside the film for a moment, because the same technology that Musk is pointing at a movie could be an extraordinary instrument for scholarship — and here I am closer to hopeful than to horrified.

Imagine a system that could reconstruct damaged monuments from partial remains, testing competing restorations against the surviving evidence. That could compare vase iconography at scale, tracing how a single scene — Odysseus lashed to the mast, the blinding of Polyphemus — mutates across hundreds of painted vessels and centuries. That could help organize the scattered fragments of the Epic Cycle, model the missing scenes the fragments only gesture at, and surface recurring artistic motifs a human eye might take a career to notice. These are real uses, and they would strengthen the discipline rather than replace it.

Reconstructing evidence, comparing images, testing hypotheses: this is where AI is powerful, because these are tasks of pattern and scale. Creating Homer is a different order of thing entirely. It is worth being clear-eyed about which of the two we are actually talking about.

A Better Question

I think "Can AI recreate Homer?" is the wrong question, or at least an unanswerable one dressed up as a challenge. The more useful question is the one my experiment was quietly asking all along: what would an archaeologically faithful Odyssey actually require?

If you took the assignment seriously, the training would have to reach far beyond a scrape of film stills and fantasy art. It would want Homeric Greek and its meter. Bronze Age archaeology and the material record of the Mycenaean world. The Linear B tablets — the actual administrative documents of that civilization, in the earliest form of written Greek we possess. Mycenaean architecture and the layout of a palace. The vase-painting corpus. Classical sculpture. The fragments of the Epic Cycle. The workings of oral poetry itself. And Greek religious practice, so that the gods are rendered as the tradition understood them rather than as decoration.

No AI system I am aware of integrates that entire body of evidence. That is not a knock on Grok specifically; it is a description of how large the actual problem is. The gap between "generate a convincing image of a Bronze Age ship" and "understand the poem that ship sails through" is the gap between two different kinds of knowing.

So let me end where I began, but with the panic drained out of it. Perhaps Grok will produce the most visually authentic Odyssey ever put on a screen — hulls and helmets and palace walls truer to the archaeology than any big-budget film has managed. If it does, that alone would be worth watching, and I would happily eat my earlier oh no.

But visual authenticity is only the first step, and it is the easy step. If a machine ever manages to capture Homer's imagination, his humanity, his rhythm and his ambiguity and his poetic genius — the parts that do not photograph — then classicists, historians, and computer scientists alike will have witnessed something genuinely extraordinary, and I will be first in line to say so.

My blog