Only half a decade has passed since the early days of generative AI: the days of multimodal systems outputting six-finger hands, AI sound producing nothing but lo-fi gibberish, and not-so-large language models predicting a character, rather than a word or more, at a time. Though seemingly retro by now, it was these glitches and their aesthetics of misrepresenting the world, rather than accurately predicting it, that opened both technological and social cracks for artistic explorations. Now hidden behind neat interfaces and more “accurate” predictions of the world, AI has arguably been “tamed”... or so they want to make us believe.

It is a sunny afternoon in early November when artist-researcher Beth Coleman invites me for the inaugural exhibition, symposium, and book launch of her work Reality Was Whatever Happened: Octavia Butler AI and Other Possible Worlds (2023). Coming in from Montreal, where I am pursuing my doctorate in science and technology studies, it feels like a pilgrimage to arrive at the coach house that is home to the University of Toronto’s Centre for Culture and Technology. Founded in 1963, the centre is where famed philosopher Marshall McLuhan brought his groundbreaking thinking around media, communications, and technology not only to his students, but to the world. At the beginning of the centre’s written constitution, McLuhan writes that “the object of the Centre is to pursue by a wide variety of approaches an investigation into the psychic and social consequences of technologies.”¹ And psychic it is, what I encounter once entering the ground floor of the coach house on that sunny afternoon.

There, framed, on the far wall, a painting. A distorted grimace. Slenderman-like but with an Afro? A person, maybe. Familiar, yet alien. Features perceptible and, at the same time, distinctly exceeding the expected. What we want to see. What we are trained to see, coerced into looking for. In this moment, trying to decipher what I am looking at, it is not the depicted subjectivities that I relate to in this piece titled Alice Freeman Engineer (2023). It is rather the StyleGAN aesthetic that I am holding on to the most.

GANs, short for generative adversarial networks, are a particular class of AI systems used to generate a variety of different media ranging from text, to sound, to images, and beyond. The project This Person Does Not Exist famously demonstrates the power of GANs to create photorealistic images—created by software engineer Phillip Wang, the website generates a portrait of a non-existent person on each reload.² Seemingly retro by now, early versions of StyleGAN carry a particular aesthetic. Low in resolution with a sort of washed-out quality to them, outputs were veiled in something akin to Photoshop’s soft-focus filter. Depicted features were lacking clear lines, making the borders of subjects and objects appear to melt into each other and with the background.

Encountering the materialized StyleGAN aesthetic in a physical painting hanging on the wall of McLuhan’s coach house, I am reminded of the collective artistic research laboratories I cocurated in late 2019 and early 2020. Hosted together with the digital arts and electronic music festival MUTEK in Tokyo and Montreal, we brought together interdisciplinary groups of artists, researchers, and technologists to jointly explore the uses and misuses of early generative AI systems in and through artistic practice. Sparked by the lo-fi StyleGAN aesthetic, the memories of working with artists such as Isabella Salas and Lucas LaRochelle draw me in. And this drawing in is necessary. The flip side to the rather approachable soft-focus textures and lo-fi aesthetic is the alienness of StyleGAN’s depicted subjectivities. Unlike today’s Midjourney or DALLE, early StyleGAN systems were notoriously bad at making “accurate” predictions of the world. Rather, they presented distorted versions of reality, particularly uncanny when trying to generate humans and lifelike features. Hands with any number of fingers. Floating limbs appearing out of nowhere. Washed-out and distorted faces injecting an atmosphere of ghostly haunting into generated images. In the 19th century, German psychiatrist Karl Ludwig Kahlbaum coined the term “pareidolia” to describe the human tendency to perceive objects, patterns, and meanings from sensorial inputs where there might be none.³ The most common example is people seeing human-like features, especially faces, in the most random places. The somewhat familiar lifelike subjectivities depicted in Coleman’s images trick us into looking for recognizable human features. This initial impulse is followed by a quick disappointment that things are not what they appear to be. The mutation of the familiar into something else is really where the alienness and uncanniness of these images arise. Inspired by Octavia Butler’s Xenogenesis book series, Coleman creates a generative AI system that embraces the soft yet uncanny StyleGAN aesthetic in exploring the alien.⁴ In Xenogenesis, the main character, Lilith, a Nigerian American woman originally from California, is abducted by extraterrestrials after Earth’s environmental apocalypse. Only by eugenic cross-breeding with the Oankali, the story’s alien species, can Lilith ensure the survival and arguably the evolution of not only herself but humanity at large. It is this cross-breeding that Coleman explores in a series of visual experiments titled Alice. Alice in all her facets is a product of mixing (or shall I say algorithmic cross-breeding) of the Black feminine and “... something else,” as Coleman describes in the exhibition’s accompanying book publication: “animal, alien, what we might call decidedly non-human.”⁵

It is here, then, where I return to the opening paragraph. Reality Was Whatever Happened resists reduction and futile attempts to “tame” AI. In a way, “taming” these systems becomes the latest colonial rendition of sorting people and things into predefined boxes aimed to accurately reproduce a certain categorical version of the world. This is ultimately where binary utopian/dystopian fantasies propagated in popular discourses surrounding AI emerge from. If we manage algorithmically driven sorting properly, it will lead us to utopia. If not, we receive the dystopian flip side of the same coin. This is really the fundamental question that is currently at stake in binary discourses surrounding AI, art, and creative practice.

In contrast, by embracing the glitchiness and arguable inaccuracy of generative AI systems, Coleman opens pathways to “other possible worlds,” the wilder places they might lead us to. By doing so, Reality Was Whatever Happened embraces Coleman’s earlier call in the text “Technology of the Surround” to imagine an AI “that can be free—if not to imagine then to generate—speeding through possibilities, junctures that are idiotic until they are not.”⁶ Through releasing and following AIs into wilder territories, Coleman’s work troubles the narrative and technological borders of current modes of AI-driven content production. Rather than obsessing with binary futures—the obsession with generative AI either replacing or enhancing human creativity—an Octavia Butler AI explores the multiplicities of alternative worlds, spaces as places, which feral AIs might come to generate. What these territories come to be won’t be known until we reach them—until, as Indigenous technoscience scholar Michelle Murphy wonders, we “produce a world where undecided futures are brought into the present.”⁷ Reality will be whatever has happened.