Grok can launch a thousand ships – but this AI remake of the Odyssey can’t make us feel

Why can AI create a passable landscape for a few seconds, but fail to capture a fully fledged world, make characters look lifelike or tell a story that moves us?

Ads

GrokImagine a father pretending to be mad because he wants to stay with his son and his family and not go to war. When his son’s life is in danger, he is forced to disclose the truth and is obliged to leave his son to go to war.

This is a heart-breaking story, but when artificial intelligence created it, critics found it unmoving, motiveless and meaningless.

The seven-minute film introducing the story of the Odyssey was made by Elon Musk’s Grok Imagine system, and is the first episode in what is planned to be a full-length feature.

Guardian film critic Peter Bradshaw said it had an “uncanny-valley stock-footage quality”. The Telegraph called it “sloppy gruesomeness”, while Rolling Stone noted viewers found the acting “soulless” and “wooden”.

The occasional reviewer has been slightly more muted in their criticism, finding brief visual moments that worked. Tech Crunch said “a few individual shots might provide a reasonable simulacrum of reality”, and in Digital Trends, Paulo Vargas wrote:

Some of the best shots […] don’t contain actors. Ancient landscapes stretch into the distance. Close-ups of grimy hands suggest people who have touched something rougher than a keyboard. For a few seconds, the illusion holds.

But, he continues, it falls apart when “the people begin to move, and worse, they start speaking”.

Why can AI create a passable landscape for a few seconds, but fail to capture a fully fledged world, make characters look lifelike or tell a story that moves us?

Guessing the next emotion

AI systems can generate sequences of actors, ships or battlefields by calculating from learnt patterns what image is likely to follow the last.

These images can be detailed. In Grok’s Odyssey, we see the sunlight reflecting off polished bronze breastplates; dry dirt clods flying behind a straining ox; the dark, windblown folds of Penelope’s tunic.

But there is a difference between generating detailed images, and making them look natural and believable.

A static landscape or a few seconds of a close-up might look fine. But once people start moving, AI video tools struggle to keep things consistent from one shot to the next, leaving viewers with jarring cuts and plastic-looking characters.

For a story to work, a scene must be connected to previous events and the characters must have reasons for their actions. The audience needs to understand what characters want, why they want it, what stands in their way and what could happen if they fail.

Showing a fleet of ships sailing to Troy in the distance doesn’t make a film look lifelike, and it doesn’t give us a reason to care about the people who are aboard the ships.

AI-generated films can’t capture the imperfections which make us human.
Grok

There is another challenge for AI filmmakers: making AI-generated characters express emotions naturally.

We can ask AI to make a character look sad or angry and it can generate tears, facial expressions and gestures. But human emotions are more complicated than that.

Someone might smile while feeling nervous, remain silent when they are deeply upset, or try to hide their fear. When digital characters get these facial expressions slightly wrong, our brains quickly notice and it feels eerie and unnatural.

Being human means being imperfect. AI-generated films can’t capture this imperfection in the same way a human actor can.

And many memorable moments in films were unplanned. In Good Will Hunting, Robin Williams improvised a story about his late wife farting in her sleep. Matt Damon’s laughter was genuine, and the filmmakers kept the scene.

AI can produce striking visual surprises. Because generative models combine patterns in unpredictable ways, they can throw up happy accidents like an eerie shaft of light cutting across a character’s face, an unusually dramatic low camera angle, or a startling colour contrast that the prompter never asked for.

These can look like bold artistic choices on the surface. But to an algorithm, a poetic shadow and a visual glitch are both just statistical variations. It can’t know if an unexpected shot enhances the heartbreak of a scene or completely ruins its mood, because it has no concept of human grief, love or dramatic tension.

Genuine, unscripted human moments make a story feel real because they come from lived experience, not mathematical probability.

Eye of the beholder

Grok’s Odyessy also suffers because we know it is AI.

We are biased against artworks when we know they are made by AI versus humans, but we sometimes prefer AI-generated images to human-made ones when we aren’t told their origin.

How we feel about art depends not only on what we see, but also on who we believe created it.

The debate about Grok’s The Odyssey is about the essence of being human and what we value in art as the technology becomes easier and cheaper to use.

The story of Odysseus is not simply memorable because it contains battles, monsters and a wooden horse. It has survived millennia because it is a very human story we can all relate to: a story of a man struggling to return home to the people he loves.

AI can generate a shot of a thousand ships; the harder task is making us feel the cost of one goodbye.
The authors do not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.

Tags: Academic Journalism, The Conversation, The Conversation Academic News