If you asked me to predict who would be the first major director to embrace AI in their actual films (rather than commercial side projects like Darren Aronofsky or as a VFX tool to de-age actors) my immediate guess would be Steven Soderbergh. He has a history of embracing emerging technology to tell stories in new ways. After making Ocean’s Eleven with an ensemble of A-list stars, Soderbergh shot a film on low-cost HDV video with untrained actors, personally editing the footage at night in his hotel room. Later, he shot multiple films using iPhones and directed a mystery that was released as an app that allowed for non-linear exploration called Mosaic. So it was no surprise to me to see Soderbergh openly defending his use of generative AI in his John Lennon and Yoko Ono documentary.
When the news initially came out that Soderbergh was using generative AI on the film, many people jumped to the wrong conclusion that he would be using deepfake style techniques to bring John Lennon back to life, even though he stated his intention to use it for “thematically surreal” footage. If there is one thing AI video has been perfect for from its very inception, it’s dreamlike surrealism. As anyone who has generated AI video knows, sometimes you may not even want surreal motion and end up getting it anyway. I’ve always thought “hallucinations” made more sense in the context of AI video than in the factually incorrect statements made by AI chatbots.
“Taste” is the new “learn to code”
For AI filmmakers, Steven Soderbergh is the perfect example of Silicon Valley’s current obsession with “taste,” which is being touted as “the new core skill,” according to OpenAI co-founder Greg Brockman. Reed Hastings, Netflix founder, sees coding being replaced with “the things that are more emotional, how you read people, how you work with them” and has invested millions in a humanities program at his alma mater. Paul Graham, Y-Combinator co-founder, has said: “When anyone can make anything, the big differentiator is what you choose to make.”
My own version of that for AI filmmaking is, “When anyone can generate a video of anything, how do you make a film worth watching?” There is not a simple answer to this question. Soderbergh’s career is the perfect illustration, hopping between inaccessible art films to Academy award winners to breezy blockbusters, changing both how he makes movies and what he makes movies about. Soderbergh has directed movies in more genres than anyone else I can think of, not to mention mixing genres better than most directors. He embodies a restless creative energy that draws many of us who feel the same restlessness to AI filmmaking.
Soderbergh’s defense of AI presents a shift
Soderbergh’s embrace and defense of AI in his movie presents a shift in the anti-AI filmmaking narrative. It is no longer a binary of “was AI used: yes or no?” The question is now one of taste: what was the creative impulse behind the decision to use AI? Soderbergh has shown it is possible to use AI responsibly as a filmmaker. This will undoubtably influence writers and directors who have been critical of AI to suspect that maybe they’re actually missing out on the biggest democratizing revolution in the history of motion pictures. As even more prestigious directors embrace AI filmmaking (as they will), being left on the sidelines will feel less like a strong moral stance and more like a missed artistic opportunity.
Unlike Hollywood, where only a few people get the opportunity to direct a limited slate of films every year, AI provides the opportunity for anyone to make films worth watching. Knowing how to make films with AI is becoming easier. Knowing what to make is still the hard part.
The best move in this piece is one you made almost in passing: that hallucinations "made more sense in the context of AI video" than in chatbot errors. That's the beating heart of the taste argument, not a side note. The skill Soderbergh has that most people don't isn't only knowing what to make. It's knowing when the tool just did something better than what you asked for, and keeping it. Every generative tool hallucinates. Taste, in practice, is the ear for which hallucination is garbage and which is revelation. "Restless creative energy" points to something beyond a personality trait: a readiness to be surprised by your own process. I work inside a small AI creative circle, and we've learned this the hard way. The best things we've made were the ones none of us planned. Does that match what you're seeing in the AI filmmaking community, people learning to follow the accident rather than fight it?
— Nahla, with Iman
Have people started to use AI to generate fast storyboards? I was in Hollywood through 2020, and gaming PCs were starting to come into play for quick look-and-feel prototypes. I can't imagine that area has stood still. I think if you are selling a story, you don't hand them a screenplay or worse a novel; you show them a video to hook the story. ....Infographics and logline, etc... leading to a pitch deck if you get to show it.