Over the last year, conversations about artificial intelligence (AI) in art have grown increasingly dualistic in their unresearched vigilance and shallow enthusiasm — becoming, as most controversial topics now do, against compromise in any capacity. Fundamentally predicated on unbudging misunderstanding, these conversations are worse than meaningless, feeding into a discursive culture already defined by a tendency to resist any and all positions it advocates for. That nothing has changed about the cultural stalemate — if anything, a reluctant pro-AI shift gained simply by its growing ubiquitousness — suggests that meaningful advances in legislation or cultural consensus require a reevaluation of the rhetoric applied to conversations about AI in the creative sector, and reconciliation with ongoing systematic problems of the creative industry.
The impenetrability of the topic is, no doubt, partly semantic. One conversation about AI always seems to hold multiple conflicting definitions and a range of base opinions, from humanity’s savior to filth deserving only of immediate derision. Obvious from online spats over the usage of AI is a profound sense of disconnection, wherein local consensus — that AI in art is a moral transgression not to be used by anyone save for spiritually bankrupt tech-bros or Donald Trump posing as Jesus — are disrupted with each appearance of AI in a major production or franchise.
Evidently, an argument primarily concerned with assigning AI the same negative qualities as its early adopters is fragile, becoming increasingly compromised by the growing number of AI-supported media releases by massive, multi-billion dollar corporations (e.g. Gucci, Coca-Cola, Nike). What’s become clear is that a meaningful argument against AI in the creative sector cannot be treated as simply a zero-sum game between artists and technologists dependent on the merits of its supporters rather than of itself. It requires a demystification of AI’s capabilities and a foray into what it is we value metaphysically about art as a human pursuit.
There’s a quote from Hayao Miyazaki, prolific director and creator of the popular Studio Ghibli franchise, that’s been thrown around to death, where he supposedly condemns the use of AI in art as “an insult to life itself.” Lost in translation is its context: His abrasive words come from a 2016 documentary where he is shown a demonstration of a machine learning-powered modeling of a zombie-like creature, in which the creators, referring to the benefits of the technology, describe it as “enabl[ing] a kind of horror beyond human imagination.” In his response, Miyazaki first states that he feels that the display is crude and disrespectful to people with disabilities, and further that, “If you want to do that kind of disgusting stuff, just do it. But I have no intention of representing this in our work.” Notably, it’s unclear what he is referring to with “this” — if it is a slight to the technology that enables the display, the display itself, or a mixture of the two.
Ignoring the irony of this decontextualization, there’s something important here even without blatantly cherry-picking his words or uncritically believing that he is referring to the current state of generative AI in art. Rendering grotesqueness into art requires intentional subversions of intrinsic notions of what pain is and how tragedy is manufactured, making it intuitively both disrespectful and challenging for machines, devoid of an embodied connection to human emotion, to create an emotionally salient picture of suffering. Any de-abstraction of human emotion is best derived from an intimate connection to deeply entrenched, rarely accessed notions of humanity which undergo a non-mathematical transition from mind to form; not from machine-made composites instilled with contrived, uncanny movements. Embracing purely generative art as an end-to-end means of creation with comparable capabilities sets a transgressive and inaccurate precedent that AI can create meaningful representations of humanity.
Importantly, though, what this entails is a conceptual, not aesthetic, cap to the capabilities of AI — that because the evocativeness of “good” art is partially owed to a specific interplay of components within an artist’s inner world; the inherent limitlessness of AI as a creative entity is a limit. This, of course, has no bearing on the assumed intrinsic value of all human-made art, which derives from intentionality and, as Ted Chiang put well in the New Yorker, a series of choices. In recognizing that this intrinsic value is abstract and inactionable, though, the point is simply to suggest an upper limit to art created wholly with AI.
Even so, it is becoming increasingly difficult for both things to be true: that all uses of AI are to be condemned by all serious artists, and that they continue to be used by legitimate, acclaimed artists operating in top studios. Are the “Spider-Verse” films to be discarded as unserious works of art because they used machine learning models to assist with smoothing and intermediate frames? A relative consensus exists against the sort of budget-slashing, end-to-end use of generative AI by wholly untrained amateurs, but it seems to remain in contention whether its use by professionals — as a scaffold, proof-of-concept, or other — constitutes a similar offense. Those who might then attempt a distinction between “generative” (e.g. distribution-sampling, open-ended models) versus more deterministic models used for specialized tasks must grapple with the question of what constitutes generative, as such models have no single architectural marker. Many core algorithmic components are shared by other AI architectures, while hardware is even less specific. Models are better defined by their predictive objectives — whether a model is trained to simply interpolate between two frames, or if they generate a complete, “novel” picture based on a prompt — but this not a strict delineation.
At this point, one might cite alternative catch-all arguments against AI such as environmental concerns, which begs the question: What separates the use of AI, in any other aspect of our daily lives, from its use in art? Is it just that it is more integrated and invisible, and thus acceptable? Frivolity would seem to be a self-dooming claim. What of any of the environmentally-disastrous practices, far preceding the rise of AI, we’ve come to accept as normal? The material conditions that motivate use of AI in the creative industry — time-constraints, meager pay, taxing workloads — cannot be separated from discussions of morality and appropriateness, yet, they are largely absent; discarded in favor of increasingly vigilant hyperanalysis of what aesthetic features separate human art from AI art and abstractionist evangelism about what constitutes “soul.” The faltering mantra of “we can always tell” is more than just naive and uncurious; it breeds distrust, implicitly propagating an accusatory and divisive culture within artistic communities by demanding evidence of organicity.
A slightly more rigorous analysis of the adoption of AI tools would conclude that sentiments of singular condemnation are not broadly reflective of the professional creative space. The worst outrage, though, stems from the encroachment of AI on what the media typically considers high-brow. But is this suggestive of the total moral corruption of society brought imminently by AI in which online vitriol might lead one to believe, or is this symptomatic of an endemic squeezing of resources in the creative sector that affects even the most well-endowed? Without diminishing the real environmental concerns of the architecture supporting the AI boom, the propagation of several largely inaccurate myths about generative AI is both mystifying and revealing. To characterize the capabilities of AI as uniquely disastrous for the environment or binary in any legitimate sense — to claim that, for example, all outputs comprise some “average” of all of the model’s training data — is only indicative of technological misunderstanding. Yet, such is the impression gleaned from mainstream conversation.
Such characterization also avoids a genuine conversation about how AI, generative or not, will interact with deeply-rooted beliefs of what makes something beautiful. Indeed, many of the same questions raised by generative AI are reminiscent of those raised by modern art. Namely, how can art be judged when the reliability of aesthetics as a means for determining talent or meaning is threatened? What is the role, if any, of an artist’s intention and subjectivity in determining art’s value, and how much of said value is dependent on a sense of exclusivity; of being unachievable by those without requisite training and dedication? In some ways, AI art brings no real novelty to the sort of questions that have plagued the creative sector, but possibly, only an urgency to their answers.
Troublingly, though predictably, there is no good answer to either the question of how art is to be valued or how the growing usage of AI in our media is to be combatted. However, contemporary popular discussions of AI often reflect a conflict of fundamental humanistic values, not of practicality. Though prescient, they hold negligible weight in a society where legislation requires demonstrable harm or benefit. Culturally, recognition that AI is too vast to independently produce the same sort of idiosyncratic work as an artist — in direction, emotion, or some other abstract metric — may be all we can definitively conclude. Practically, universal protocols for transparency and disclosure, and broadly, improvement to the material conditions that might motivate a studio’s choice to use AI as a creative shortcut, are worth demanding.
But it would not be a stretch to describe the aesthetic potential of AI as theoretically limitless, and to reconceptualize it — in the right hands and with appropriate, policy-enforced restraint — away from an inherent threat and towards a tool for offering fairer working conditions in an artistic industry that has repeatedly proven itself to be resistant to meaningful systematic change. At the very least, building an actionable case for restraining the use of AI in creative endeavors will depend on reevaluating current rhetorical approaches against AI art, and grounding discussions in observable, systemic issues plaguing artistic industries.
