About an hour before the start of the AI Film Festival, a gathering meant to “offer a glimpse at a new creative era,” a line down Broadway had already formed.
Attendees were buzzing with the chance to see the latest shorts made with tools by the event’s backer, Runway AI, and other models, though an unknown percentage were themselves involved with the showcased films. The ticket buyers formed one of the first large-scale gatherings of humans to celebrate the creativity generated by machines. In honor of the occasion, Runway had rented out Lincoln Center’s Alice Tully Hall, a status-conferring site of countless concerts and the New York Film Festival.
One man lingered in an “I [Heart] AI” T-shirt modeled after the I Love NY conceit; others took selfies in front of Runway’s minimalist black-and-white signage plastered across Alice Tully’s windows. “I just want to see something interesting,” one attendee said.
That they certainly got. Over the two hours that followed the company showcased ten AI-produced films of varying quality and mood, though nearly all had a dream-like experimental aesthetic — shaped, in part, by limitations on sound and the movement of real people in AI films.
One of the company’s trio of co-founders, the 36-year-old Cristóbal Valenzuela, addressed the crowd. “Three years ago, this was such a crazy idea,” he said. “Today, millions of people are making billions of videos using tools we only dreamed of.”
He said last year’s festival — this is its third edition, though the first at Lincoln Center (a similar event will be held next Thursday at Los Angeles’ Broad Stage theater) — yielded only 300 submissions. But this year brought 6,000, a statement that drew oohs from the crowd.
Founded in 2018, Runway began gaining notice in Hollywood last year after Lionsgate made a deal to train a Runway model using its entire library. Other pacts have since followed, as the firm has sought to convince Hollywood it comes in peace, or at least with a serious amount of film cred. (Valenzuela claims he is a cinephile.) So far this year, the company has released “Gen-4” and “Gen-4 References,” tools that aim to give scenes a consistent look throughout an AI-created short, one of the medium’s biggest challenges.
The L.A.-raised film and music producer known as Flying Lotus, born Steven Ellison, took the stage with Valenzuela.
“A lot of people are terrified to talk about [AI],” Lotus said. “I empathize but at the same time I’m the kind of person who uses the tools.” He walked off the stage with an exhortation to the audience. “Don’t let anybody tell you there are rules to this shit.”
Then they showed the films. Several brought a distinct style. The slice-of-life More Tears than Harm had an animated painterly look as it reflected on a hardscrabble childhood in Madagascar. Another movie, 6,000 Lies, used a series of rapidfire sonograms to tell a moving semi-dystopian story about a fetus that may or may not have been aborted as part of a culling. In the most affecting and least experimental piece, Jailbird, a chicken was shown being rescued from a factory farm to serve as a companion to a prisoner, part of a real-life British program aimed at compassionate rehabilitation.
Many of the movies sought to turn liability into asset and integrate the unpredictable reality-defying qualities of AI-generation – which can unexpectedly do things like make people float or rain fall upward — into the story.
A film titled Fragments of Nowhere used wild images of bodies shapeshifting and cars hanging from buildings as it suggested a time-space continuum that allowed for this surrealism. It closed with the line “Reality is just another layer of perception.”
As the films played, the room gathered the energy of an art-school summit, even as the larger goal is a wholesale transformation of the very nuts-and-bolts billion-dollar business of Hollywood. The presentation made a case for originality — the films undeniably exhibited a new grammar — while omitting any training-data talk of how these images were ultimately enabled by all the human-crafted cinema that came before.
The subtext to the evening was that AI not only wouldn’t oppose art but would enable a new form of it for those with the right creative instincts; Lotus noted that when anyone asks him if he worries that others would get the tool to mimic professional artists, he counters, “But I got the tool too.” Of course Hollywood executives are not always interested in turning over production means to A-list artists, not when keeping it from them could be a hell of a lot cheaper.
After the movies played, Valenzuela took the stage again to hand out awards for the machine-generated moving images. Notably, the runner-up prize went to the most humanist of the group, Jailbird.
The Gran Prix went to Total Pixel Space, a cone-headed look from the filmmaker Jacob Adler at all the possible images and films of the world (bigger than a googolplex) and a movie that seemed, in its contemplation of the vast possibilities of cinema, to implicitly argue on behalf of AI filmmaking. If the possibilities are literally almost infinite, how could a machine be reducing cinema to genericism?
Valenzuela made his case for the power of what Runway was doing. “AI is beginning to alter the fabric of our culture and of course the art that comes from it,” he said.
As people streamed out into the elegant marble and glass lobby, the din grew as the crowd hashed about the various films in the manner of any festival, though this time with an added procedural layer. “Is it just me,” one young bearded man could be heard asking his friend, “or did you think about what prompts were being used the entire time?”