by Simone Martinotta
Since the advent of AI-based image production tools, an increasing number of media artists have sought to reintegrate the unique traits of human sensitivity into their work, embracing the deviations and disturbances that define (at least until now) its distinctiveness and inimitability. Recently, this range of positions has broadened into a veritable school of thought. Those who interpret the imperfections and gaps in AI (“AI hallucinations”) as valuable creative resources stand out within this school of thought. They see these hallucinations as a glimmer of hope that heralds a new era of imagination. But let’s clarify what we mean by “AI hallucinations” right away 〈1〉.
“AI hallucinations” are technical anomalies caused by a generative AI’s misalignment when dealing with false or misleading data. In essence, the AI becomes confused. Even a simple lack of available data or overly literal adherence to primary sources by information providers can give rise to malfunctions in the translation of the initial input. Thus, deviating from pure computational logic, AI generates outputs containing erroneous elements incongruous with the initial assumptions. Instructions from humans result in outcomes that differ from initial expectations (the “hallucinations”). Needless to say, there has also been a proliferation of AI programs that do not make mistakes at all but, if anything, simulate errors and slips of the tongue of various kinds, based on instructions that give rise to delusional, “drugged” effects.
According to the aforementioned theoretical guidelines, when AI is unable to respond clearly and unambiguously to requests, it produces so-called “hallucinations,” which give rise to unconscious activity comparable to human creativity. Generating unexpected, distorted outputs would prove that the machine can escape its destiny as a docile, passive servant precisely because of the error. The horizons of certainty and predictability would disappear, giving way to a phenomenology characterized by uncertainty, enigma, and the unexpected instead. Although the error is due to the machine malfunctioning, the way it manifests could be related to the concept of “lateral thinking,” coined by psychologist Edward de Bono 〈2〉, which today forms the basis of many reflections on the fundamental mechanisms of human creativity. In short, “AI hallucinations” appear to be a case of creativity in machines, similar to how it manifests in humans.
Based on this line of thinking, which is considered salvific with respect to the anthropological, aesthetic, and ethical unknowns that AI poses to humanity, multimedia artists like Refik Anadol 〈3〉 express keen interest in instances when machines produce deceptive, ambiguous images. In their view, it’s as if error does not fall within the normal, ever-perfectible horizon characteristic of all man-made technological instruments but rather within a metaphysical design aimed at redeeming the machine from its original sin: the absence of emotional involvement and compromise in its actions, rigorously programmed to carry out assigned tasks. An example of this perspective is the following statement by Anadol, from one of his conversations available online: “[…] I felt that there is not art here – when the system works perfectly for example an algorithm imitating the reality – but when the system is broken […]” 〈4〉.
Essentially, Anadol adopts the romantic argument, which was later taken up by various avant-garde movements and continues to influence contemporary psychedelic, alternative, and cyberpunk cultures. In a society dominated by technology and characterized by the constant refinement of production methods in every field, the primary task of the artist-outsider is to break the mold, reject convention, and propose alternative paths that diverge from the cold rationality that stifles utopia and breeds dystopia. From this perspective, advocating for “another” art and interrupting the purposeless cycle of productivity is equivalent to seeking authenticity detached from calculation and restoring a profound connection with the world and its languages.
This attitude is similar to that seen in the 1960s with the advent of Pop Art. During a time of unprecedented economic growth and the rise of a consumer-driven society, Pop artists responded by engaging in a dialogue with capitalist mass production on equal terms. They introduced images and objects into the art world that were indistinguishable from mass-produced items to challenge the concepts of originality, authorship, and the autonomy of art in relation to the economic and social systems from which they originated.
In today’s age of pervasive, purposeless, and nihilistic technology 〈5〉, how should artists respond to this encirclement? Should they hope for a mistake by the machine on which they depend, yet from which they wish to dissociate themselves? Should they imagine themselves both “inside” and “outside” the logic of turbo-capitalism? Or by always claiming the need to find a flaw in the system’s functioning? Co-optation and liberation are the two extremes between which Anadol’s conception of the artist’s condition struggles. However, both extremes are two sides of the same coin: powerful technical means, which are the beginning and end of everything. Furthermore, the multinationals that rule the web are often the first clients and sponsors of artists like Anadol, who require substantial resources for their impressive installations.
What if this desire to crash the system, this messianic obsession with spying out its weak points, were itself a mistake, a bit like when a roulette player deludes himself into thinking he can beat the bank? What could be more futile than seeking (or simulating) imperfection in a machine that is by its very nature always perfectible, and precisely for this purpose – to continuously improve its performance – has at its service a host of engineers, scientists, and technicians, as well as an army of underpaid workers connected online from developing countries 〈6〉? Thanks to their monitoring efforts, the glimmer of error recorded today will be reduced or even eliminated tomorrow. The imbalance in which all hope was placed, will have revealed itself for what it was: a hesitation quickly resolved, the illusion of a moment.
It takes much more than a few minor glitches that are quickly resolved to escape the nightmare of Black Mirror, the dystopian world depicted in the british television series of the same name 〈7〉. If anything, a significant change in approach to the tool is needed. That is, we must no longer focus on finding errors that are quickly corrected by rapidly advancing efficiency, but on establishing a new paradigm of dialogue with the machine. A machine that is too often improperly seen as an end, a reason sufficient to justify itself, when in fact it is still only a means, albeit one with formidable power of suggestion. Using this medium with a view to revealing, to taking a look at what is truly unknown, truly salvific, is the task that a new generation of media artists must take on as soon as possible.
〈1〉 See M.T. Hicks, J. Humphries, J. Slater, ChatGPT is bullshit, “Ethics and Information Technology”, volume 26, article number 28, june 2024. 〈2〉 E. De Bono, Lateral Thinking, Ward Lock Education, London 1970. 〈3〉 Turkish-american Refik Anadol (1985) is probably the most famous media artist today and has contributed greatly to raising awareness of digital art through immersive installations exhibited around the world. The websites of the artist and Refik Anadol Studio: https://refikanadol.com/ , https://refikanadolstudio.com/ 〈4〉 The link to the video from which the quote is taken: https://www.instagram.com/p/DQ9vRsojwE6/ 〈5〉 On this topic: M. Heidegger, The Question Concerning Technology, Garland Publishing, New York-London 1977; U. Galimberti, Psiche e techne. L'uomo nell'età della tecnica, Feltrinelli, Milano 1999. 〈6〉 Cultural, environmental, and social unknowns of AI: K. Crawford, Atlas of AI. Power, Politics and the Planetary Costs of Artificial Intelligence, Yale University Press, New Haven 2021. 〈7〉 Two reflections on the loss of aesthetic experience in the digital age: Byung-chul Han, Saving Beauty, John Wiley and Sons, Hoboken 2017; G. Perretta, Il sensore che non vede. Sulla perdita dell'immediatezza percettiva, PaginaUno, Milan 2023. Homepage: Refik Anadol, Machine Hallucinations - Sphere, 2023, 3D mapping, Las Vegas, The Sphere (courtesy Sphere Entertainment). Below: Sound waves as an image, AI-Art image generated with a text prompt by Deep Dream, a program developed by Google and distributed since 2015 (Wikimedia).
