In AI Art: Machine Visions and Warped Dreams, Joanna Zylinska pulls no punches. Most of what passes today for AI art is bad. Really bad—a dismal portfolio of robot-made Rembrandts, Van Goghs, and vapid AbEx takeoffs ejaculated by the R&D wings of the world’s favorite tech corporations. “Kindly put,” Zylinska writes, “much of generative AI art celebrates the technological novelty of computer vision, fast processing power and connection-making algorithms by regaling us with a dazzling spectacle of colours and contrasts as well as the sheer volume of data” (76). But
unkindly put, it becomes a glorified version of Candy Crush that seductively maims our bodies and brains into submission and acquiescence. Art that draws on deep learning and big data sets to get computers to do something supposedly interesting with images often ends up offering a mere psychedelic sea of squiggles, giggles and not very much in-between. (76)
Such art is lamentable not just for its aggressive vacuity. It also frames AI research as a benign pursuit, eliding the ways in which social and political life become ever more subject to AI’s algorithmic machinations. But artists can do better; indeed, as Zylinska repeatedly shows across her twelve short chapters, some already are. Fusing the electricity of a polemic with the measure of the best art criticism, AI Art takes stock of the range of narratives today’s art tells about artificial intelligence, from the insipid to the incisive. Narrative is the keyword here: Ultimately, Zylinska’s aim is to urge artists to counter the AI industry’s “warped dreams” by telling “better stories” about AI (27).
Of the utmost urgency for Zylinska is the need to connect artificial intelligence to another AI: the “Anthropocene Imperative.” Artificial intelligence is inextricably bound up with environmental crisis, she argues, but it has so far disavowed that entanglement. Employing the language of Freudian dream analysis, she remarks that most AI research “look[s] like a displacement activity with regard to the planetary issues concerning our climate and, ultimately, our survival” (43). Such displacements buttress AI’s most grandiose fantasy: “to reverse-engineer extinction…to Make Man Great Again” (40). Underwriting those “warped dreams” is the persistent fiction, much maligned for decades by ecocritics, posthumanists, and new materialists, that humans are ontologically distinct from the rest of Earth’s inhabitants. Zylinska has offered nuanced philosophical ripostes to this myth in several of her books.1 Likewise, here she takes aim at a “misguided” pop-philosophical question frequently triggered by AI art: “Can computers be creative?” (49). Why “misguided”? Because it makes man the measure of ontological value: What it’s really asking is whether computers can be creative in the same way humans are. This humanist understanding of creativity obscures the fact that human exploits “hav[e] always been technical, and thus also, to some extent, artificially intelligent” (13). To construct less myopic AI narratives, Zylinska argues, we need to waylay the anthropocentric idea that art evinces humans’ ontological distinction from technology. Rather than asking whether computers can be creative (just like humans are), our narratives should really be exploring the ways in which human creativity is always already “partly computational” (54).
Zylinska also takes issue with AI art’s affective habitus. Leonel Moura’s Robot Art (2017), a work Zylinska handles with care, shows how AI art can be fun and joyful, at best. In the exhibition space, Moura’s squad of little robots (engineered by the artist) traverses a big sheet of paper, the aggregation of their periodic, multicolored marks forming a complex abstract drawing. Recalling the exhibition, Zylinska notes that the robots’ “clunky movement, wobbly gait and unwieldy architecture often evoked a smile from members of the audience,” fostering an atmosphere of mesmeric amusement (61). “It could be argued that the ‘joyfulness’ of robotic and AI art is where its force lies,” Zylinska ventures, “opening up the possibility of not only reaching to audiences outside the traditional exhibition spaces but also of redefining the very idea of what it means to make, display and distribute art” (61–2).
But too much AI art—the “kitsch and derivative,” corporate-driven stuff—capitalizes on the fun while foreclosing chances for critical speculation (62). In such cases, playful provocation shades into “an odd combination of the fuzzy, the mindless and the bland” (72), resulting in nothing more than “mild bemusement” (81). Such is Zylinska’s assessment—harsh, but measured—of a batch of art projects by Memo Akten, Mario Klingemann, Gene Kogan, and Mike Tyka, most of which were funded by tech giants. For example, Tyka’s Portraits of Imaginary People (2019) uses Google’s DeepDream algorithm to suture together bits of different people’s faces (sourced from a Flickr repository) into portraits that amalgamate ages, sexes, and races. “Slightly uncanny, already boring” is Zylinska’s verdict (81). And worse, while such work may seem progressive in its celebration of social difference, it ultimately “creates an illusion of diversification without being able to account for differences that matter—for why and where they matter, when and to whom” (83).
All this creates the need for artists to “tell better stories” about AI. Thankfully, compelling examples do exist, and Zylinska offers us several. Trevor Paglen’s Sight Machine (2017), made in collaboration with Kronos Quartet and Obscura Digital, investigates machine vision, a major field of AI research. Machine vision refers to systems in which images facilitate data transmission between machines, with humans left out of the loop. (For example, in facial recognition technologies, a computer processes an image of a human face into machine-readable visual data to communicate with other computers.) Sight Machine aimed to make visible machine vision itself by projecting AI-derived images onto a screen above the string quartet. For Zylinska, what the work revealed was the unbridgeable “translation gap” between machine and human vision (93). In so doing, it “shows us that a new network of visibility, much of which remains permanently obscured from human vision, has now emerged which has the potential to redefine radically what counts as visible and what doesn’t—or what counts, full stop” (94). Given as an example of a critical AI artist, Paglen shows us how human life is bound up with machine vision in ways that remain inaccessible to us—a blind spot both literal and political.
Notably, several of Zylinska’s examples of critical AI art do not involve the use of AI technologies. In Lauren! (2017), artist Lauren McCarthy equipped the home of a participant with a network of cameras, microphones, switches, and other hardware, creating a DIY version of an AI-driven smart home akin to Amazon’s Alexa. But McCarthy substituted herself for the AI system, manually controlling the apparatus to respond to the inhabitant’s whims. Zylinska forwards a posthumanist take on McCarthy’s work: It “allows us to interrogate [the] ongoing process of human becoming (with) AI…what kind of technology we enter into a relationship with, in whose interest, and to what effect and purpose” (134). In this way, Zylinska makes the work speak to the Anthropocene Imperative, but she says remarkably little about another imperative that Lauren! clearly baits: the history and politics of “women’s work.” McCarthy’s piece warrants a more thorough analysis than Zylinska gives us here; it’s a moment when the short-chapter form she chose for AI Art works against her.
Toward the end of the book, Zylinska sketches a thumbnail genealogy of computer-based art. She observes that net art of the 1990s, post-internet art circa 2010, and today’s AI art share several characteristics: a preoccupation with the estranging effects of computer technologies in daily life; a tension between critical engagement and spectacular displays of programming skill; a market-friendly rematerialization of contemporary art (post-internet art appealed to galleries because it gave them edgy, tech-savvy art in the form of physical, sellable objects ). Like the cursory history of artificial intelligence set out earlier in the book, this genealogy occupies just a few pages. Needless to say, historical depth is not Zylinska’s main aim here. Still, extending Zylinska’s genealogy back a few decades might help address some of her other concerns about AI art, particularly as regards corporate patronage. Like today’s AI artists, television artists of the 1960s and ’70 s often relied on the financial support of large corporations, in their case to gain access to state-of-the-art televisual technologies. But while, as Zylinska opines, corporate-funded AI art tends toward technophilic fluff, TV artists like Nam June Paik channelled a rebellious, countercultural animus into their experiments with TV. This was an attitude not just tolerated but actively encouraged by his backers at the Rockefeller Foundation and elsewhere. Taking a cue from Paik could help AI artists leverage corporate support to more provocative ends.
In earlier work, Zylinska has used her own art to bolster her verbal argument, and AI Art continues this rhetorical strategy. Here she presents View from the Window (2018), a project in which she paid some of Amazon’s Mechanical Turk employees to take a photo from the window nearest their workspace. The project refers to Joseph Nicéphore Niépce’s View from the Window at Le Gras (c. 1827), thought to be the first photograph ever made. In Nonhuman Photography, Zylinska offers a posthumanist meditation on Niépce’s image: due to its eight-hour exposure time, the photograph registered a catalog of highlights and shadows impossible for the human eye by itself to entertain.2View from the Window summons Niépce again to address the curiously liminal status of the Mechanical Turk employees, who are hired (under exploitative terms) to perform computer-based tasks that, though highly repetitive, cannot yet be entrusted to AI. (Amazon describes its Mechanical Turk service—ironically?—as “artificial artificial intelligence.”) In conjoining art criticism with art practice, Zylinska exhibits a modus operandi that remains rare in humanities scholarship, despite openings made in emerging disciplines like artistic research and the digital humanities. Here it works well, and I for one hope more researchers will follow her lead.
AI Art offers artists, critics, and scholars an indispensable field guide to the stories art tells about artificial intelligence. Its brief, bulletin-like chapters inevitably leave threads loose, but where the writing foregoes comprehensiveness, it gains in energy, tone, and urgency. I take the book as a valuable exemplar of contemporary critical writing: It’s acerbic when needed yet affirmative in intent, profound in ambition yet rendered in snappy, witty prose. With this book, Zylinska offers convincing proof—if yet more were needed—that critique has not, in fact, run out of steam.
See for example Joanna Zylinska, Minimal Ethics for the Anthropocene (Ann Arbor, MI: Open Humanities Press, 2014).
Joanna Zylinska, Nonhuman Photography (Cambridge, MA: The MIT Press, 2017).