Back to selection
Emergence
New work in new media by Deniz Tortum
Speaking in Tongues: How Koyaanisqatsi Was Reborn as The Vivid Unknown
The Vivid Unknown (Photo by Zachary Schulman) When I step into the exhibition hall, three large screens display rapidly shifting, AI-generated images of natural and industrial landscapes. The overstimulating visuals resist easy interpretation, and they are also responsive to the viewers in the room. The video stops when the room is empty and speeds up as the room becomes more crowded. The longer one stays in the room, the more one’s “shadow” on the screen becomes visible—effects all made possible by a complex real-time vision-tracking system.
This is The Vivid Unknown, an immersive installation that recreates the groundbreaking 1982 film Koyaanisqatsi using AI. The piece is a continuum of cinema, technology and collective experience. Unfolding over hours, it defies complete comprehension. The ghostly traces of the original film fill the room, but its meanings are elusive—I sense things that I cannot yet articulate.
Koyaanisqatsi explored the tensions between nature, humanity and technology; The Vivid Unknown, a collaboration between artist John Fitzgerald and Godfrey Reggio, director of the Qatsi trilogy, adapts the film to our current technological moment, in which we step into an unknown territory, a techne incognita led by rapid developments in artificial intelligence.
I saw the first iteration of The Vivid Unknown when it was presented at IDFA in 2023. In January 2025, it arrived in New York for BAM’s TECHNE exhibition, co-produced by Onassis ONX. For each successive iteration, the technology is tuned, the staging is refined and the interactive component gains new layers of responsiveness. However, throughout its different iterations, the piece evokes similar central questions.
I first met Fitzgerald when he was a visiting artist at MIT along with Matthew Niederhauser, with whom he later co-founded the immersive media studio Sensorium. Together, they produced some of the most interesting early work in this space: Rachel Rossin’s The Sky is a Gap, Lincoln in the Bardo for NYT VR and Zikr: A Sufi Revival. Lately, they’ve been involved with ONX Studio, a global XR residency of which I am also a member, and they have both been working in AI, though creating separately.
Fitzgerald credits Reggio as his inspiration to work in film. As a student, he often projected Koyaanisqatsi onto his ceiling and imagined stepping inside Reggio’s “total cinema,” as his work was described in the title of a Museum of Modern Art series in 2023. Reggio, Fitzgerald notes, prefers the term “experiential” over “experimental.” The distinction resonates—experiential work aligns closely with the immersive art scene, where Fitzgerald has been a pioneer.
Fitzgerald and Reggio first met through a common friend a few years ago, when Reggio was, as he put it, “1,000 months old.” Their initial meeting at Reggio’s studio in Santa Fe, New Mexico, turned into a four-day conversation. Fitzgerald was working in VR; Reggio had no interest in wearing a headset. Despite his deep interest in technology, Reggio remains indifferent to its tools—he dismisses them as “gizmos” and owns neither a cellphone nor a computer.
At the time, Fitzgerald was researching early GANs (generative adversarial networks) and was drawn to the open-source ethos of the Stable Diffusion community. “Closed platforms like Sora and Veo are fun to experiment and ideate, but you don’t have control over things,” he explains. To generate never-before-seen visuals, he turned to animation pipelines like Deforum and AnimateDiff. This led to an AI-driven reinterpretation of Koyaanisqatsi. Fitzgerald began making tests and meeting with Reggio online every other week.
Collaborating with the artist/technologist Dan Moore, they built the “Koyaanisqatsi image generator,” training a custom machine-learning model using 5,000 frames from the original film alongside metadata describing each frame and pixel-based feature extraction. This model generated new visuals from which Fitzgerald and Reggio curated the final installation.
A similar approach shaped the sound design. Using Stable Audio 2.0, Zachary Novack trained an AI system that learned the patterns, instrumentation, timbre and moods of Philip Glass’s score. Then, using text prompts, Ben Shirken generated short sound clips and used them as the base materials for the final composition, which is a hybrid score—part machine, part human.
When I spoke with Fitzgerald, I had many questions about the ontology of the AI image. For me, The Vivid Unknown provides a generative ground to think about this topic. Koyaanisqatsi is rooted in physical reality—human faces, crowded streets, sprawling industrial sites. What does the AI translation of The Vivid Unknown do to the actuality and realness of the original film?
“The images in The Vivid Unknown do not capture reality; they reassemble it,” Fitzgerald responded. “They are not records, but impressions—visual echoes shaped by patterns, light and motion rather than memory. These images do not document; they translate—a machine’s attempt to recall a world it never truly saw.”
Throughout film history, theorists from André Bazin to Roland Barthes have suggested that the photographic image can transcend language. In Devotional Cinema, Nathaniel Dorsky introduces the concept of “self-symbol”—the idea that film can portray the inherent magic of things, presenting them as they are without imposed meaning. Barthes describes the “punctum”—the unexplainable detail of an image that pierces the viewer. In the recently translated Duras/Godard Dialogues (published by The Film Desk), Jean-Luc Godard stresses the importance of images existing independently of words and scripts.
Film and photography, then, offer a world unmediated by language. But large language model (LLM)-generated AI imagery begins with the word—the prompt precedes the image. Does this make AI images closer to language than to photography? “This is the great paradox of AI-generated imagery,” Fitzgerald says. “It emerges from text prompts, from perceivable descriptions. In that sense, it is closer to language than photography. But what it generates is not linguistic—it is a diffuse visualization that lacks the specificity of traditional images. The AI-generated visuals in The Vivid Unknown do not depict reality; they synthesize it. They are the visual equivalent of speaking in tongues—a neural network approximating meaning but never quite landing on it.” Fitzgerald describes these images as “like symbols, but symbols generated by a system that does not think in symbols.”
The paradoxical language of Fitzgerald is not accidental—both Koyaanisqatsi and The Vivid Unknown embody this same paradoxical duality. In 1977, several years before Koyaanisqatsi was made, NASA launched Voyager 1 and 2, both carrying a copy of the Golden Record—a curated selection of sounds and images representing life and culture on Earth. Among these were depictions of industrial civilization chosen to showcase humanity’s progress and achievements. Yet in Koyaanisqatsi, similar imagery takes on a darker meaning, conveying the descent of civilization and the violence inflicted upon nature. The same visuals can signify both ascent and descent.
I see a similar duality in the AI-generated imagery of The Vivid Unknown, finding myself caught between fascination and unease. As Fitzgerald observes, “The ascent and the descent are always intertwined.” The Vivid Unknown simultaneously reveals “a civilization rising, a civilization unraveling.” This refusal to offer simple explanations—this tension between progress and collapse—may be the work’s most powerful aspect.
As AI images become more and more prevalent in our digital lives, is our aversion toward them growing? I’ve heard filmmaker friends expressing a growing feeling of “nausea,” an actual physical response, toward AI imagery. This might be because not just the internet, but also the physical world, is becoming filled with AI images.
“The discomfort of AI imagery is more than the ‘uncanny valley’—it is the vertigo of recognition without understanding, a digital déjà vu, a dream just beyond coherence,” Fitzgerald says. “AI-generated images mimic the aesthetics of reality but lack the weight of experience. There is a deep fatigue in encountering these images at scale.” These images unsettle something deeper. That representation of reality is detached not only from the mediator but from the world itself. “The nausea we feel is not just visual—it is existential.” The AI images represent a world that is outside of our human consciousness. It is the nausea of losing the upper hand in the interpretation of the world.
Released just before the dawn of the internet, Koyaanisqatsi has an ability to anticipate what is to come. This is especially evident in its final scene, which shows an Atlas-Centaur rocket explosion from the early 1960s. When the Challenger disaster occurred in 1986, four years after the film’s release, the scene took on an eerie new significance.
A similar sense of prefiguration runs through The Vivid Unknown. We can already sense its implications, but their full weight will only be understood in retrospect. Just as Koyaanisqatsi explored the tension between humanity—technology’s creator—and the natural world, The Vivid Unknown arrives at a moment when AI has begun to shape reality in ways we have yet to fully grasp. Much has been written about AI’s potential (some of it, no doubt, written by AI itself), yet the true scope of its impact remains uncertain.
Fitzgerald wonders, “What if the violence we unleash on nature returns as technology’s vengeance upon us?” We designed machines to serve us, but now we serve them. Algorithms dictate what we see, automation defines labor, AI-generated imagery replaces human imagination. The violence is cyclical. We shaped technology in our image, and now we are being reshaped by it.”
As The Vivid Unknown draws to a close, Godfrey Reggio’s voice reverberates: “The very ground we stand on, the terra firma, is now terra techno firma. We are on fire. We are indeed the future now.”
