Why I love and hate AI film

For some time now, I’ve been hiring myself out for AI films and hybrid productions. At the same time, I once co-founded an association called the Society for Digital Ethics. So anyone who accuses me of being full of shit is actually understating it.

The criticism of AI productions rests on two pillars. First, there’s the climate: both the CO₂ emissions of data centres and their water consumption. Second, there’s the fundamental ethical question: the models were trained, every last one of them, on the real work of real people — ultimately in order to make exactly those artists obsolete.

Christopher Nolan vs. the Earth

AI bros like to claim that an AI production is far kinder to the climate than a traditionally made film, because the biggest ecological cost of a classic production is transporting crew and equipment. That isn’t entirely wrong — but put that bluntly, it’s a lie. What is true is that, ecologically speaking, a flight isn’t a guilty pleasure. It’s an absolute bomb. That rarely gets said this drastically, because we all want to see the world. Because air travel weighs so heavily on the balance sheet, you have to split traditional filmmaking into film with flights and film without. With flights, a traditional production causes considerably more CO₂ than an AI production. Full stop. A single flight is enough. Usually, though, people want to fly back again — and anyone who insists on shooting at real locations come hell or high water should, in my opinion, be taken to court.

And yet Christopher Nolan is one of the most revered directors of our time. We love him because he wants to do everything analogue — and on real locations. He’s regularly celebrated as the saviour of handmade cinema while everyone else churns out dead digital images. This isn’t about whataboutism (it is whataboutism!), but it’s remarkable that Nolan flew a gigantic crew to six (!) countries for The Odyssey — and gets praised for it. Tenet was even shot in seven different countries. Nothing could be further from my mind than condemning Christopher Nolan (I condemn him!). My point is about perception and felt truths.

Tenet was even shot in seven different countries.

Commercials, too, get off lightly on the ecological front, even though a crew gets flown to Cape Town and back just so a company can sell more toothpaste. Which brings us to data centres — the place where AI films physically come into being. Unfortunately, we hit a black box here, because the big AI players don’t disclose their power consumption. Qualified estimates at least allow us to say that a minute of AI video is considerably less harmful to the climate than a minute of travel-heavy film. In fact, AI video is roughly comparable to a local TV series production — which is still too much. What really matters is quantity: because AI video is so cheap to make, far more of it gets produced, and in sum that genuinely adds up to dramatically high emissions. After all, a data centre isn’t built for a single AI filmmaker, but for all of us.

A lesson in capitalism — ethics and AI

With AI, we’ve created a technology whose main purpose is saving money. Whenever people get the chance to save, they’ll gravitate towards it. One of the most important metrics in the film industry is cost per view: how expensive was each view? Naturally, the client side of the industry is rubbing its hands right now, because for the first time the cost component can be slashed. On the other side are the artists, and unless they happen to be called Christopher Nolan, they have to fight for every cent of budget. Film is expensive; very few filmmakers pay for their films themselves. All year round, we wrestle with people who couldn’t care less about the narrative vision of our film and cling to the cost-per-view principle instead. The average filmmaker always gets the smallest possible budget for their work — that’s part of the job. By Seedance 2.0 at the latest, AI video reached a level that can no longer be told apart from traditional film. Which means: AI is already priced into the budgets.

We’re long past the point where we get to ask the ethics question, and I think artists are the last people anyone should complain to about using AI. We don’t set the prices. If a budget allows for nothing but AI, then we work with AI. The alternative would be to turn the project down, which isn’t a realistic alternative. It would be the morally right thing, but that’s not how the business works. You can’t just say no to a good client. You can’t even say no to a bad one. Follow-up jobs depend on it, reel material, relationships. Filmmakers have to film. Ninety-nine percent of everyone working in film scrape by from project to project; if you don’t shoot for six months, you have no money for six months.

That’s why I advise every traditional filmmaker to retrain in AI as fast as possible. Not because I think AI is so great, but because I want as many artists as possible to be able to live off their work. It’s a dance with the devil and a textbook lesson in capitalism that Marx wrote down 150 years ago: workers produce the means of their own superfluousness. The only new thing is that now they take online courses to do it. For each individual filmmaker, retraining is rational. For all of us together, it’s ruinous, because the collective effect devalues the very work everyone wanted to save. It’s a classic coordination failure: nobody can get out without losing alone.

Workers produce the means of their own superfluousness. The only new thing is that now they take online courses to do it.

We filmmakers are highly visible right now, because we use AI to make images that are easy to take a stand on in the comments. But the fundamental shift — capitalism coming to a head through AI — affects all of us. In my view, anyone who has an LLM check their work, or simply do it for them, is no more guilty than the workers who were driven into the factories by industrialisation.

At the same time, we’re not assembling machines. We tell stories and work on something beautiful. Rest assured, I don’t flog myself after every minute of AI video. I enjoy being able to tell stories completely free of budget restrictions. And I’m happy for everyone who can realise their vision without first begging for a micro-budget on bloodied knees, at the price of dramatic compromises.

By the way: most films aren’t made by the most talented filmmakers, but by the ones who sell themselves best. If you were too modest to call yourself a fucking genius, you didn’t stand much of a chance in the old film world. That can and will change with AI. I’m also thinking of the many women in the industry who are tired of having to act like men just to get funding for their film.

And then there are regions where film has so far mainly been something you watch, not something you make. There’s never been a shortage of talent or stories there. What was missing was the money for images that can compete with Hollywood. A science-fiction epic in Lagos, a historical drama about the Kingdom of Benin, a fantasy saga drawn from Yoruba mythology: films like these didn’t exist because nobody wanted to pay for them. And the ones who would have paid were the same studios that decide which stories are globally relevant.

I’d love to leave that thought standing as a happy ending. But capitalism only has happy endings for the very few. The tools belong to corporations in California and China, cost subscriptions in dollars and were trained on images that come overwhelmingly from the West. Anyone who has ever tried to generate an African market with AI that doesn’t look like a charity appeal knows what I’m talking about. And then there’s the water: the servers these clips are rendered on are often located in precisely those regions of the Global South where water is already scarce.

It’s complicated.