I went to an AI film festival at Hot Docs. It wasn’t great

The Cinema Shift Festival, a “curated cinematic evening on how AI is reshaping film,” was held at the Hot Docs Ted Rogers Cinema on June 5, capping off an auspicious (or ominous) week for the rise of artificial intelligence. Mark Carney’s federal government unveiled its “AI for All” strategy, going all-in on the new tech while “building trust, creating opportunities, and reinforcing Canadian sovereignty.” Meanwhile, German AI company Black Forest Labs sent shockwaves through the film world by announcing that Martin Scorsese—one of our most celebrated living filmmakers—had joined them as an adviser. In a statement, Scorsese said that he hopes AI will “push the bounds of creativity to create deeper and richer experiences for audiences,” citing its potential to streamline the storyboarding process.

Related: A conversation with Prime Minister Mark Carney

This last point was perplexing. As a teenager, I saw Scorsese’s hand-drawn storyboards for the climactic shoot-out in Taxi Driver, which were included in the film’s DVD extras. They were perfect: a blueprint for the final cut that was still loose enough to invite input from his team. These sketches helped define my understanding of both the role of the director and the collaborative nature of filmmaking. So why, at age 83, does Scorsese now want to outsource this creative work to a computer?

Full disclosure: few things make me want to disengage from art, culture and the world itself more than the spectre of AI. I hate going on YouTube and seeing ugly computer-generated thumbnails for every video. I hate seeing social-media prognosticators claiming that “Hollywood is COOKED!” and then posting awful videos of fake Harry Potter scenes. I hate that my artist friends who used to make good money designing posters are now losing work to machines who do the job worse but for free. But I also hope to be open-minded enough to acknowledge the potential benefits of AI in the arts, should they appear to me. Offhand, I can think of at least two artists who have succeeded in using AI to evoke feelings other than revulsion: Romanian filmmaker Radu Jude in his outrageous post-modern adaptation of Dracula (2025) and comedian Conner O’Malley in his videos and stand-up shows. That said, both of these artists used AI to make a point about how bad its output looks.

I attended Cinema Shift hoping to be convinced that AI could actually create something beautiful. The event kicked off with a discussion on “artistic disruption” between Canadian filmmakers Ann Marie Fleming and Ryan Patterson. Fleming is a veteran director whose most recent film, Can I Get a Witness?, warns against the dangers of technology. But she said she wants to stay open and curious: “In the end, I may decide to just do what my film said and never use this technology at all, but right now, I’m still in exploration mode.” Fleming was introduced as a “filmmaker,” Patterson as an “AI filmmaker.” He talked about being an ’80s kid, growing up with Star Wars and Indiana Jones. Like Spielberg before him, he was that little boy who ran around his backyard with a camera. His love of storytelling was matched only by his love of tech, and in AI he claimed to have found not just a tool but a collaborator. Sometimes an unexpected look from one of his characters will propel him in a new storytelling direction. He expressed misgivings only when it came to scaling up his practice, which would involve working with more fellow humans. “I find the collaborative process very new, and to be honest, I don’t know how to do it in a lot of cases,” he said. Fleming pointed out that filmmaking, by definition, is collaboration. “I’ve heard, but I don’t know much about it,” Patterson replied dryly.

Related: I spilled my secrets to an AI therapist. It didn’t go well

Other panels addressed the legal issues surrounding AI in the arts, from the use of famous actors’ likenesses to run-of-the-mill plagiarism, as well as AI’s increasing integration into the visual-effects industry. More relevant to my interests was a discussion on “The Future of Human Authorship” between Katerina Cizek, co-founder of the Co-Creation Studio at the MIT Open Documentary Lab, and Laura Mingail, the evening’s host. “Authorship is actually a really new concept,” said Cizek. “The first mention of any kind of producer or author in theatre was only made about 150 years ago.” (For the record, the First Folio of William Shakespeare’s collected plays was published in 1623, and it certainly mentioned the name of the author.) Cizek added that we need to understand these new tools while considering equity and justice. She acknowledged labour issues as well as the environmental impacts of the technology: “The people who live near these data centres that are running at such high heats that literally the trees beside the data centres just burn.”

I hoped we would get a more fulsome discussion of how to address these issues, but of course we didn’t. AI has been enthusiastically embraced by segments of the online right, and this seems to be a difference between the discourse about it in conservative versus liberal cultural spaces. A conservative-coded  AI event would be all-in on the destruction, and stuff your feelings. In a liberal space, we acknowledge the bad things, then move right along.

Finally, we saw 16 AI-generated short films selected by a panel of film-industry professionals. In a pre-screening discussion, juror Samantha Wan told us, “You’ll see when there’s a real idea—like, when there’s a kernel of an idea—and maybe it’s not perfectly executed. And you’ll see when it feels generated.” Moderator Fayeque Townsend-Rahman added, “I invite you to watch with curiosity, not only for how these films were made but for where you see intention and authorship and story on screen.” These remarks did not fill me with confidence. At most film festivals, the audience doesn’t need to be instructed to look closely for rare moments of intention, authorship and actual ideas.

Patterson’s Legend is a Ridley Scott–esque fantasy adventure that crams a whole TV season’s worth of plot into about 20 minutes that still somehow feel interminable. Martin Haerlin’s Grandmère is a garish tribute to his grandmother, beginning with fake 8mm home-movie footage before turning to magic realism (revolting computer-generated beasts and an Old Dark House fit for a Sega Dreamcast game) and finally culminating in a tender encounter between the filmmaker and his grandmother’s spirit. Compare it with the recent Blue Heron—another memory film about family and loss, but one in which every aspect of the production was carefully considered by human minds—and it comes across as trite and tasteless.

A few of the shorts make pathetic attempts to engage with the zeitgeist. Afro Futcha’s The Stars is a cloying monstrosity about a little boy in an unnamed war-torn area (obviously Gaza) who collects “the stars” that fall from the sky. (He holds a little jar of glowing bullets.) Mark Wachholz’s unbelievably awful Céremony, also about a child from a war-torn area, is a maximalist effort that evokes virtually every contemporary societal ill and flirts noncommittally with maybe half a dozen overarching themes before finally settling on “The kids will save us.”

Buried in all this incompetence were two shorts that seemed good enough to have comfortably blended in at a real film festival. Max De Donato’s Hard Dayzzz takes the POV of a fly as it buzzes through a series of travails before its quick, gruesome death. The concept is well-executed, and the punchline made me laugh. And Catherine Hang-Hicks’s Dial-A-Mate is a funny send-up of 1980s video dating featuring animals dressed like middle-class office drones. Hang-Hicks is a long-time animator at Pixar, and the short’s style and humour have that studio’s polish. Both projects set modest goals and managed to achieve them.

One other short that at least gave me pause was Sinuca de Bico (Cornered) by Odair Faléco, a presumably autobiographical story of a Brazilian wage slave who self-actualizes through AI. As with a lot of the shorts, it’s overloaded with story and visual information, but at one point its protagonist says, “Making films in Brazil has always been for the rich…. The tool they use to exploit me is the same tool that gives me a voice.” Well, point taken. It’s nice to think that a less exploitative world is possible, but in the meantime, I won’t begrudge someone trying to game a broken system.

In the jury talk, Townsend-Rahman said, “Filmmakers live inside change and have always been working within limits: limited money, limited access, limited time, limited permission. That’s part of the history of independent cinema, and filmmakers have always found a way to make the impossible feel possible.” Judging by their blockbuster-scaled AI shorts, many of these filmmakers would do well to put their computers away and embrace limitations. Too many seem to believe that AI has liberated aspiring artists from the hard work of learning their craft and developing a personal style.

During the ordeal that was Legend, I flashed back to Edgar G. Ulmer’s film noir classic Detour (1945), a film shot on a tiny budget in under two weeks. Lacking the resources to shoot a crucial scene in New York, Ulmer used a street sign and a fog machine to create a dreamy, minimalist Manhattan. A computer doesn’t have the ingenuity and taste to make an artistic decision like that, but how many of the evening’s AI filmmakers do? “Even if you have AI, which is a powerful tool, it doesn’t inherently mean you know how to story-tell,” said Samantha Wan. On this point, I agree.