After AI Art

[T]hese days pretty much everybody is in the loop with AI.

Helena Sarin¹

In the AI Loop

This article approaches the problem of digital entanglement by examining “the AI loop”: an ontological condition of our human co-emergence with the discourses, practices, and technologies of artificial intelligence at a time of its current “summer.”² To posit the existence of such a loop is not to fall prey to any kind of technological determinism or uncritical adoption of the industry narrative. It is only to recognize the intensity and breadth of articulations, understood as forms of both data linkage and linguistic–affective expression that machine learning technology (which is a more accurate term for what gets referred to as “AI”) has generated over the last decade. We can mention here machine translation, algorithmic programming of culture through media platforms such as Netflix or TikTok, as well as various applications of large-scale data processing in medical imaging, prediction, and surveillance. However, these articulations also include stories about our supposed liberation from drudgery, coupled with anxieties about the loss of income, the end of all purposefulness and meaning, or even the very survival of the human species.

While the starting point for the examination conducted here will be philosophical, the specific cultural context for the analysis will be provided by the transformation of artistic practice, and of the associated concepts of creativity and invention, by machine-learning models that go under the moniker of “AI.” The postulation of an “after AI art” moment in this article’s title may therefore seem premature, given the current buzz around various domains of AI-enabled artistic creativity, from generative neural networks producing images in the form of style transfer to text-to-image generators conjuring images seemingly from scratch: DALL-E, Stable Diffusion, Midjourney, Luma, Runway. Yet, seeing past the summery haze of the current AI wave, especially with regard to its generative variety, UK technology critic John Naughton has recently used the hype cycle designed by American technology research company Gartner to postulate that we have just climbed “the peak of inflated expectations.” Disillusionment is bound to follow this peak,³ subsequently leading to a new rise in interest and then, eventually, mainstream adoption of AI technology.⁴ This article aims to cut through the timeline of AI’s hype cycle by outlining a double proposition conveyed in its title, “After AI Art,” and enabled by the temporal ambiguity of that “after.”

On the one hand, “after AI art” suggests that we need to pull away from the set of visual ideas around AI art produced over the last decade—and from the rather narrow concept of AI art that emerged as part of those early experiments—in order to fully consider what art co-created with, against, and around AI can become. In this sense, the “after AI art” proposition should actually read “after AI Art,” serving as it does as a follow-up to the argument of my polemical book, AI Art: Machine Visions and Warped Dreams, published in 2020.⁵ The book argues that the then nascent form of AI art was effectively a version of the mindless app experience provided by Candy Crush. It offered a garish aesthetic aimed at numbing the public, while anaesthetizing viewers to the damaging effects of algorithmic technology being implemented in so many areas of our lives. Yet I want to suggest that the time has come to move on from that first set of artworks, from the reactions generated by them—and from their critical assessments (including those produced by myself). This is where we get to the second meaning of the “after AI art” proposition of the article’s title. Being “after AI art” involves searching for and exploring art’s yet undetermined potentiality in the aftermath of AI’s embeddedness in artistic and wider cultural production and circulation cycles, from automated image, text, and sound generation to algorithmic curation and blockchain.

After AI Art

The term “AI art” itself serves as shorthand for a set of artistic practices that engage with technologies of machine learning (ML). This engagement can take place on the level of programming (designing and training ML models or adjusting the already existing ones), usage (making use of existing models), or—perhaps a little controversially for some—discourse (creating works about AI). Equivalent terms adopted by the art community include “ML art” or simply “media art,” with cognitive scientists preferring “computational creativity”⁶ and the technical community opting for “creative AI.”⁷ Therefore, “AI art” functions as a heuristic, positing a new field with somewhat frayed edges while articulating a relationship between computer science and art that is itself a form of artifice. It is in this broad and enactive sense that the term is used in my article.

AI art came to public awareness around 2017 through either works based directly on style transfer (i.e., machine-generated copies of van Gogh or pictures that looked like they could have been painted by Rembrandt) or their uncanny combinatorial variations. Works by computer science-trained creators, such as Mike Tyka, Memo Akten, and Mario Klingemann, all associated with the first wave of AI art, were premised on understanding creativity in engineering terms. At the same time, a rather conservative idea of art as the averaging of possibilities underpinned these works. Rendered by GAN models trained on photographs of human faces or historical artworks, the works were approximations of the source material, with their artistic quality residing in the divergence from the input data. Even though they were rhetorically presented by their creators as machinic “hallucinations,” the images in fact resulted from statistical analysis gone (somewhat) wrong, with the calculation error deemed a satisfactory visual effect by the engineer-artists. The posited loss of control on the part of AI artists enveloped ML-driven creativity in the language of mystery. We must remember, however, that in the previous centuries, artists who produced works in the stage of mania, through supposed divine inspiration, or under the influence of drugs embraced a similar creative approach involving suspended or distributed agency.⁸ Moreover, artists such as Michael Noll or Herbert W. Franke have been sharing creative and technical agency with computers, scanners, and plotters to produce their works since the 1960s. Yet, even though programmability may have been part of artistic practice for a long time, the first generation of AI artists working in the early decades of the twenty-first century adopted the loss of control over their output as their creative modus operandi. This loss stemmed from the ultimate unknowability on the artists’ (or anyone else’s) part of the ML model’s logic, aka from its black-boxing. At the same time, choosing the original training dataset, adjusting the model, and then selecting the most desired outputs involved a significant degree of human artistic agency. Visually, the works obtained looked strangely psychedelic, merging the Francis Bacon “distorted face” aesthetic with a surreal compositional style in the vein of Salvador Dalí.⁹

AI Art: Machine Visions and Warped Dreams offers a critique not only of this garish and somewhat kitsch aesthetic but also, more importantly, of the sociopolitical assumptions about both creativity in its human and machine guises and AI’s supposed magic. In the book, I propose to reread art history through its nonhuman influences, attempting to cut through the smoke and mirrors effect of the AI rhetoric. I also suggest that the question as to whether machines could be creative is not the best one to ask about AI, because we should not see creativity as a uniquely human trait but rather as a property of systems—some of which include humans.¹⁰ As New Scientist has pointed out, “If it is so easy to break down the style of some of the world’s most original composers into computer code, that means some of the best human artists are more machine-like than we would like to think.”¹¹ To say this is not to devalue human creative outputs—or humans tout court: it is, in fact, to investigate conditions for doing precisely the opposite. Nevertheless, this investigation will need to take into account the latest research in natural and human sciences which indicates that the key signal points of human uniqueness, from sentience to communication and tool use, have all been found across the species barrier.¹² It will also need to recognize, after philosophers such as Henri Bergson, Gilles Deleuze, and Vilém Flusser,¹³ that “the program of life” far exceeds the parameters and current capabilities of human imagination. In other words, human creativity is always entangled; it is always partly nonhuman.

The more important question to ask about creativity today, and one that has become even more pertinent in the light of the deluge of artefacts produced by generative AI (of visual, textual, and sonic variety), is how we can discretize and valorize creativity. What kinds of creative outputs do we want to value? And how can we create propitious social, educational, and technical conditions for the enactment of human-entangled creativity? As part of this inquiry, we need to take seriously the legitimate concerns of many creators about their livelihoods and professional opportunities, especially in the light of the incursion of automatized tools and platforms into creative professions—from script writing and stage design to music making and acting. Therefore, we must interrogate the current sociopolitical conditions in which what we humans understand as creativity—with its associated institutions and practices such as art schools, public art galleries, and creative education—can prosper. While art has never been free from capital and it would be naïve to posit its existence in the sphere of pure spirit, its monetization, coupled with the overproduction of variations of the same thing, has accelerated and intensified in the age of late capitalism.¹⁴ Thus, the optimization narrative of Big Tech, which presents the likes of Stable Diffusion or Midjourney as being able to create “more” and “faster,” forms part of the broader agenda of devaluing the modern liberal understanding of art as a human pastime or lifestyle that goes against the logic of productivity. This narrative repositions art as a career choice with only a few winners. Therefore, perhaps we can see the first wave of “AI art” as the industry’s somewhat ham-fisted attempt to “disrupt the art world” by seemingly opening it up to all and sundry, while safeguarding the investments of the rich of this world in “real artefacts” held in freeport art storage spaces, or what Hito Steyerl has called “Duty Free Art.”¹⁵

Yet, just as the program called life escapes the constraints of human imagination, the programmable parameters of the art world always manage to activate the glitch within its system. Many artists, other creatives, institutions, and counter-institutions have successfully worked their way through the AI terrain to come up with interventions and artefacts that exceed the technical and discursive framework of the early phase of AI-driven creativity. It is to these forms of “exceeding AI” that we turn in the next section.

After AI Art

Now that the first wave of hype around the novelty of AI technology has subsided somewhat, at least in the art world, the current cultural moment can be described as one driven by a search for what art, all art, may become in its engagement—technical, cultural, critical—with AI. Being “after AI art” involves searching for and exploring art’s yet undetermined potentiality in the aftermath of AI’s embeddedness in art’s (and wider culture’s) production and circulation cycles, from automated image, text, and sound generation to blockchain and algorithmic curation (already familiar to us thanks to YouTube or TikTok). This search for what art after AI will look like needs to entail coming to terms with the reconfigured potentiality of the human artist, one whose humanist self-image has undergone a necessary correction in recent years. Yet the human artist’s technical co-dependence has a history almost as long as that of humanity itself, as evidenced in stones and shells used for carving in prehistoric times to the camera obscura becoming a seeing and drawing device or the photographic camera capturing the world while constructing its picture for us.

The recent scholarly and curatorial publications on the topic have reflected the growing conceptual maturity of the field of AI art. They have defined, overviewed, and finally, yet importantly, given a history to the concept of “AI art” (with its cognate iterations listed at the beginning of this article). Scholars in the fields of media, communications, art history, and computer science have considered the growing technical expertise of artists, critics, and the public,¹⁶ as well as AI art’s link with other technologies, such as blockchain.¹⁷ They have simultaneously offered broader reflections on the shifting understanding of creativity that AI technology demands of its audiences.¹⁸ Last but not least, the developments around ML and AI in art have been placed on a longer trajectory of human–machinic collaborations in creative fields.¹⁹

I want to focus here on two collectively produced accounts that have attempted to capture the state of the field of AI art via modes of delivery that go beyond conventional scholarly or artistic publications. The first, AI and the Arts: How Machine Learning is Changing Creative Work, a report published by the Oxford Internet Institute in 2022, is based on a series of interviews with artists and scholars working in the field of AI. The second, Creative—Critical—Constructive—Collaborative—Computational: Towards a C5 model in Creative AI, is a position paper authored in 2023 by the Creative AI Lab (of which I am a member), a collaboration between the Department of Digital Humanities at King’s College London and Serpentine Galleries. While the dominant technology and discourse of AI originates in the US, with the Silicon Valley voices working hard to overpower all others, it is not insignificant that those two accounts of AI-enabled creativity should hail from the UK. A country on the literal margins of Europe, enfeebled by Brexit, the COVID pandemic, and the increasingly unstable global geopolitics, the UK is currently attempting to reclaim its position and authority by promising investment in STEM subjects more broadly and AI specifically. In November 2023, the then UK Prime Minister Rishi Sunak hosted the AI Safety Summit at Bletchley Park, a country estate famous for hosting a government code unit which cracked the code of the German Enigma machine during the Second World War. The event aimed to position the UK as a key player in the field of AI. Attended by, among others, US Vice President Kamala Harris, European Commission President Ursula von der Leyen, as well as key tech figures such as Sam Altman, Demis Hassabis, and Elon Musk, the summit articulated the UK government’s conviction that “there could be nowhere more fitting for the world to come together,” outlining its vision “to seize the opportunities of the greatest breakthrough of our own time … while giving people the peace of mind that we will keep them safe.”²⁰ When the UK summit was unfolding, the US announced, on its home turf, a whole roster of radical ways of policing AI, resulting in multiple headlines along the lines of the Financial Times’s verdict: “US upstages Rishi Sunak with AI regulation plan.”²¹ In distinct contrast to its imperial history and overarching ambitions, nowadays the UK art world harbors (or at least outwardly manifests) less grandiose beliefs about its geopolitical positioning—and its role in capturing and owning a field. Many artists, curators, and critics share a sense of weakness, or the experience of an inherent vulnerability of our bodies, minds, and the human-machine networks we are part of, coupled with a recognition of the need to work together with consideration and care. The two collaborative texts on the state of AI art in the early 2020s encompass this spirit of vulnerable curiosity, but they also introduce an important element that is often missing from the government’s and industry’s more bombastic pronouncements about AI, be they hype or horror stories: play.

The AI and the Arts: How Machine Learning is Changing Creative Work report reflects the growing confidence of many artists working with, or around, AI technology. Positing the idea of “human/AI creative complementarity,”²² it proclaims an opening toward a less antagonistic, more playful adoption of the technologies and models of machine learning in artistic practice. Some artists are working “against the grain” of AI technology, celebrating machine failure and the glitch, while others are focusing on exploring the creative potential of AI models.²³ Many consider technical competence valuable, but it seems crucial to emphasize that artists use ML in a variety of ways that manifest various levels of technical skill and wider computational literacy, from undertaking technical research, to choosing, constructing, and training models, building datasets, and curating outputs.²⁴ Importantly, it is not just machine failure that gets revealed in the process of working with AI but also human weakness, with artists reporting their bodies cramping up when having to take a huge number of images for the dataset, or suffering from visual fatigue after sorting out through endless outputs. “Respondents sometimes thought about these physical demands within the context of the contrast between the finite resources of the human and the infinite resources of the machine.”²⁵

Notably, the Oxford report contributes to ways of theorizing AI art by offering a new set of concepts and frames for the discourse around it. The first concept relates to thinking about the artefacts produced through the generative process not in terms of individual images but rather as classes of images. As artist and theorist Frieder Nake explains, “The entire set is the work of art … the individual piece now is only an instance of the class that it belongs to.”²⁶ This allows artists to adopt the concept of the “latent space” from the field of ML. Originally conceptualized as “the statistical topography of the aggregated features of the training images, mapped onto a high-dimensional mathematical space,”²⁷ in AI-driven art practice the latent space becomes a figurative way of mentally capturing all the possibilities of an output, or its virtuality, to use a now somewhat obsolete term from new media studies. The artist’s role consists in carving out a smaller set of individual outputs from this virtual, or latent, space. This idea is coupled with the proposed shift—also articulated by Nake—in AI art from the static to the dynamic image, with the latter containing the possibility of always being something else, over and over again.²⁸ These articulations lead the report’s authors to propose the term “statistical creativity” as a more apt description of the relatively constrained algorithmic method of creation. With this method, the context is confined to the dataset, the data itself—truncated and biased, while connections and similarities—established on the basis of “statistical, not qualitative, proximity.”²⁹

While it would be difficult to argue with the report’s sober conclusion that “[h]umans are still fundamentally needed to generate outputs meaningful to other humans,”³⁰ it seems curious that its authors foreclose any uncertainty concerning human–machine cocreation to come out firmly on the side of humanity, as if the technical opening toward complementarity threatened not just artists’ livelihoods but also our lives. Thus, they opine: “artists ultimately work to address human—rather than technical—questions whether human/algorithm synergies may help unlock human creative potential.”³¹ By keeping the ontological categories of humanity and technology separate, what could be termed “the latent space of the critical analysis of art” as enacted in the Oxford report ultimately carves only a few (rather still) snapshots from the artistic landscape after AI. What is missing from this account is not only the exploration of the broader issues of human–machine co-constitution but also the expansion of the field of AI art to its parergonal function, whereby the distinction between the main text and its commentary, the picture and its frame, disappears in the perception of what counts as “the work.”³² Yet, even if the report itself perhaps forecloses the recognition of humans’ originary technicity, many of the artists discussed—for instance Robbie Barrat, Sarah Meyohas, or Lauren McCarthy—do embrace and enact the concept of art as machinic co-creation. However, for some, such as Jake Elwes or Anna Ridler, rooting their AI-driven work in the wider artistic and art historical context is of equal value. One should acknowledge that the report and the artworks together form part of the field of AI art: they all participate in a debate of what AI art is, what it can be, and what we can become with it.

The Creative AI Lab more explicitly adopted this parergonal approach, where the discursive articulations and framings become an inherent part of that field rather than just a commentary on it. Defining “Creative AI” as “a research field”³³ and not just a method for executing artworks, its position paper, Creative—Critical—Constructive—Collaborative—Computational: Towards a C5 Model in Creative AI, describes the Lab’s practice as premised on a shift “from artefacts to their wider contexts, processes and infrastructures.”³⁴ Borrowing from its founders Mercedes Bunz and Eva Jäger’s proposal to investigate the “backend” of artistic processes involving AI, the Lab positions itself as a facilitator of both the debate on, and practical knowledge about, the use of ML in creative practice. This has entailed a curation of a database of resources—texts and tools—aimed at making the field accessible to those without prior technical training. Gene Kogan’s Machine Learning for Artists website, originally launched on GitHub and then transferred to the ml4a.net website, has had similar practical ambitions and has delivered educational material on various AI models, while also setting up a virtual community. However, ml4a has mainly focused on tools. In turn, the Lab has intended to situate such tools in the context of the critical debate on AI creativity. The Lab has also worked with artists on the R&D side of their projects, from 0rphan Drift exploring how an octopus can transform our understanding of technological development in an encounter with AI to Danielle Brathwaite-Shirley playtesting her gaming attempts to tell Black trans stories through different media, including retro consoles and AI models. Therefore, the “backend” here involves the infrastructures of knowledge and critical debate, not only those of code. Locating AI art in the historical context of developments in computational creativity such as cybernetic art or artificial life art, as well as data visualization, the paper identifies two dominant strands in the current practice: a visual and a “situated” one. In the visual strand, artists explore new stylistic possibilities of generative AI while engaging in AI-driven forms of algorithmic visualization. The “situated” strand, in turn, goes beyond aesthetics to incorporate ML tools into artistic “complex systems.”

Yet, the paper’s role is not just diagnostic. Its authors have also offered a position on the possible directions of AI art’s development. Recognizing, in the McLuhanian vein, that art making “can help us get closer to the algorithmic logic we are all increasingly living under,” the Lab members have stated that “an engaged critical reflection on the AI/ML technologies and their sociocultural underpinnings” is necessary as part of Creative AI’s agenda.³⁵ Thus, the paper calls on the art community to work toward developing “fairer and more democratic technical and social systems by way of an informed and hands-on public,”³⁶ an approach that includes shifting curatorial and public attention to the conditions of art’s production in the age of AI, while recognizing collaborative effort and the need to extend the “creative attribution to technical roles.”³⁷ In its title, the paper repurposes the “C4 communication model (‘Command—Control—Communication—Computer’), with its orientation toward mission accomplishment based on the cybernetic logic and its original military associations” to propose instead “a C5 model for Creative AI, a model which is, in its very premises, “Creative—Critical—Constructive—Collaborative—Computational.”³⁸ Highlighting “an ethics of cooperation that involves building horizontal, noncompetitive and research-driven alliances of institutions and stakeholders interested in Creative AI: museums and galleries, universities and art schools, technology and media companies, NGOs,” it ends up positioning AI art as a space for designing “alternative metrics and values—beyond optimisation, efficiency or profit.”³⁹ At a time when “AI ethics” is often presented as a stop-gap solution to deeper structural inequalities and blind spots that shape the technical and economic infrastructures of ML systems, raising ethical questions about ways of living in a society whose planetary resources, including those of a human kind, are being plundered at scale, has become extremely timely. But what is needed here is more than a post-factum intervention, often issued with corporate approval, into Big Tech’s operating practices. To be truly capable of addressing the multiple problems of AI, ethics must be embraced as a lived practice of taking responsibility for life, with its multiple demands and scales. This practice needs to be coupled with a second-level reflection on what it means to live a good life, on whose idea of good is being espoused here, and, last but not least, on how competing demands for articulating goodness can be negotiated and resolved. Therefore, what we need is not a form of procedurally driven yet ultimately inefficient “AI ethics” enacted as a form of corporate responsibility. Instead, we need ethics, full stop.

Who Cares?

As we can see from the above, the field of AI art has entered a new phase, one that allows for technical and conceptual experimentation, ethical reflection, and sociopolitical critique. I recognize that the analysis offered here has focused more on recent documents about AI art and less on specific artistic practices as framing devices and articulators of these new tendencies and trends. Yet, the very existence of such papers testifies to the richness and vibrancy of the art scene around AI—and, equally importantly, to the engaged and responsible curatorial and institutional debate about what AI art is and what it could be. This debate involves not only the technical capabilities but also the broader question of machinic, planetary, and human resources that seems essential to drive the development of AI. The problem of exhaustion—be it of our bodies, minds, or energy resources used to fuel the working of ML engines—manifests itself as an inextricable part of the debate on creative AI. This problem is also connected to the issue of labor, both hidden human technical and regulatory labor involved in running the (supposed) machine intelligence while safeguarding it and disappearing human labor currently being replaced by generative AI in a variety of creative sectors.

The exhibition and public events program, AI: Who’s Looking After Me?, curated by FutureEverything (with Irini Mirena Papadimitriou at its helm) at the Science Gallery London and held between June 21, 2023 and January 20, 2024, rightly put the problem of care at the center of the debate. Alongside Luba Elliott and Hannah Redler-Hawes, Papadimitriou has established herself as a key curatorial voice in AI art working in the UK and wider European context. Her groundbreaking exhibition, Plásmata: Bodies, Dreams, and Data, curated for Onassis Stegi and held in a public park in Athens in the summer of 2022, explored human and nonhuman bodies and their energy sources. The exhibition featured works from twenty-five international artists, including Morehshin Allahyari, Refik Anadol, LaTurbo Avedon, SUPERFLEX, and Dries Verhoeven, to explore questions such as: “Why do sunsets in video games never end?” and “Would you ever heed a prophecy spouted by a disembodied artificial intelligence?”⁴⁰ In AI: Who’s Looking After Me? the problem of energy and use returned to the fore, but it was framed more explicitly by the concern about humans’ fragile, broken, or diseased bodies and lives, as well as their habitats. In a series of digital cartoons, Mimi Onuoha’s The Future is Here! revealed the domestic spaces of the anonymous crowdsourced workers involved in the invisible tagging work for AI databases. Wesley Goatley’s Newly Forgotten Technologies, smart speakers dumped on piles of soil, told a futuristic tale of the technological obsolescence and the possible survival of humans and machines. Does AI Care?, a sound and visual installation developed by Sofie Layton with young adults who experienced cancer and with Wellcome/EPSRC Centre for Medical Engineering researchers, engaged patient and medical communities in conversation about the use of AI in medical diagnosis and treatment. With these and other projects, the curatorial team asked some crucial questions that connected AI art to broader issues of responsibility, ethics, and repair: “Who holds the power, who distributes the benefits, and who bears the burden of existing AI systems? Can we rely on these technologies for our wellbeing and happiness?”⁴¹

Less beholden to the power of Big Tech and increasingly less important on the global business arena contrary to its Brexit fantasies of Global Britain, the UK can afford to ask these questions because there is less at stake here. Thus, the UK’s sociopolitical fragility, coupled with fewer dependencies on Big Capital coming from Big Tech, also provides an opportunity for the creative exploration of the wound and the glitch. The story of AI art that emerges from the current landscape is by no means parochial: it draws in artists, curators, and issues from various geopolitical locations across the globe, but it attempts to do so with concern and care, as evidenced in the projects and narratives presented in this article. Challenging the sovereignty of the human body and the body politic, the UK’s AI art scene does embrace human–machinic coexistence and cocreation, but it also asks probing questions about the material conditions of creativity and its institutions: art education; public universities, museums, and galleries; youth clubs. In so doing, it posits the problem of AI art as one of infrastructure, context, politics, and policy—and not just technology. In this sense, the problem of AI art does not differ much from the broader (and longer-standing) problem of art.

Entangled with Art

Indeed, now that the first wave of hype around the novelty of AI technology has somewhat subsided, at least in the art world, the current cultural moment can be described as one driven by a search for what art, all art, may become in its technical, cultural, and critical engagement with AI—and for what we may become with it. As part of this current phase, time also seems ripe to put to bed once and for all the oft-unquestioned humanist pronouncements about the supposed true affect conveyed by human artists in their work, in a way that “an AI” would never be able to do. To believe the latter is to misrecognize that much work produced by human artists is often generated with a calculating mindset, be it of the artist themselves or of the gallery circuit and wider artworks in which they operate, resulting in a mode of operation that is itself algorithmic, even if it does not involve any use of AI. To say this is not to ignore or dismiss the ongoing endeavors to develop alternative trajectories of art on the part of individual artists and artist-activists, art collectives, as well as institutions, from Nan Goldin’s attempts to take on the Sackler family to the Hyperallergic online platform’s efforts to hold the powers of the art world to account through their critical and often irreverent publications. Nevertheless, in our human assessment of what AI art is and, more importantly, what it can become, we need to move beyond unquestioned humanist assumptions about our supposed human uniqueness to be revealed in the artefacts themselves and shift the scope of analysis to broader contextual issues. In other words, we need to stop both projecting our human sentiments onto artworks and anthropomorphizing AI as our weakened, albeit threatening, “good Other.” Feelings, impressions, and even brush strokes can be simulated, while experts can be easily taken in when it comes to the species provenience of particular artefacts.⁴² The question of “art after AI” needs to shift to that of value: not just the financial value of art and the monetization of human authenticity but also broader cultural and societal values established by and for humans with regard to the activities that we hold dear and that matter to us.

Indeed, the question of why art matters remains important, as does the justification of art in terms other than evolutionary, pragmatic, or economic. Art matters to us because it is a form of excess, involving the “thwarting and perverting patterns” and the generation of “more complex syntheses.”⁴³ The process is often as important as the work produced, not just in terms of the effort to arrive at the artefact but also as a form of technical, conceptual, and affective endurance, a practice of “getting there.” Just as it does not make much sense to recognize “AI ethics” as a discrete form of enquiry, in the longer term it will not make much sense to continue seeing “AI art” as a discrete form of creative engagement. Indeed, in time AI will be folded into both the human as a technically constituted being and the art world as a field of techne, another one in a long series of our co-constituting technologies, albeit possibly having more far-reaching consequences for own self-understanding and self-image.

Drawing on research in paleontology and on our human history and prehistory which have always involved making things and their images, in his book The Entanglement: How Art and Philosophy Make Us What We Are, philosopher Alva Noë goes so far as to suggest that art should not be seen as the outcome of our history but rather as one of its conditions.⁴⁴ In his phenomenological understanding of the human as a processual being, there is room for art to be recognized as a formative part of this process rather than a mere add-on. Noë writes:

[W]e are ourselves a happening, a becoming. Wherever we first show up, we show up not only as creatures of habit, but as creatures of habit whose very habits incorporate our own acts of resistance. This is entanglement. The things we know best, that make us what we are—our mental powers and personalities—are made up by art, or by art and philosophy. We ourselves, then, are the very stuff of art. We are living in the entanglement.⁴⁵

His definition of “entanglement” names the process through which our human way of being human is altered by the work of reflection as enacted in both art and philosophy.⁴⁶ Based on this proposal, we can suggest that not only are we all currently positioned in the AI loop, as was stated at the beginning of this article, but also that AI becomes for us a form of entanglement, a formative process of becoming something (new), of becoming different from ourselves and creating possibilities for living with and as that form of difference. To say this is not to replicate the optimization rhetoric of Big Tech which sells enhancement and multiplication of volume as desirable outcomes of their products. It is only to recognize the inherently unfinished nature of Project Human, as well as the creative evolution that drives it. According to Noë, “Art loops down and changes the life of which it is the artistic representation.”⁴⁷ Therefore, AI can become a creative force of becoming for us,⁴⁸ but there are no guarantees that its outcomes will themselves prove creative rather than destructive, or that they will be life-enhancing rather than entropic. Our ethical task, the task of self-care and care for life in its multiple demands and responsibilities, is to rise to AI, the same way in which we previously rose to painting, photography, or digitality—and emerged creatively and socially altered with them.⁴⁹ However, it also means acknowledging that this form of rising to AI may (need to) become an uprising.

After AI Art, Art

To conclude this article, I would like to present my short video (Figure 1). Made in 2023, it served as a prompt for the argument subsequently outlined in this article. This prompting action, qualitatively different from the mechanical process involved in getting an AI model such as ChatGPT or Midjourney to generate a statistical output on the basis of a sequence of words provided, was used by a human agent known as a philosopher-artist that goes by the name of “Joanna Zylinska”, with all her bio-technical entanglements, against the very same agent. In line with the etymology of the term, prompting here involved a revealing, a process of bringing something to light from a latent space of ideas and images.

The After AI Art, Art video explores our anxieties around human and machine creativity at a time of intense investment in machine learning technology. To make the video, I used three off-the-shelf avatars that were AI-generated. The avatars were then programmed to each deliver a script responding to the following query: “What will art after AI look like?” Blinking, smiling, and occasionally glitching, the avatars merge a friendly professional disposition with an “Uncanny Valley” look. Their narratives—written by GPT-4, ChatGPT, and myself—invite a reflection on an increasingly thin line that separates original thinking from algorithmic banality, in humans and machines. Readers may recognize that some of the sentences from the video script have made their way into the article. For reasons of academic probity, I want to confirm that no text included in the written part of this article comes from the automated models, although I embrace artist Mark Amerika’s confession made in his book, My Life as an Artificial Creative Intelligence: “I too, as an improviser of spontaneous poetic riffs and self-reflexive artist theories focused on the creative process, continually train myself to transform my embodied praxis into a stream of consciousness writing style that doubles as a kind of onto-operational presence programmed to automatically scent new modes of thoughts.”⁵⁰

The narrative of entanglement of not just humans with AI but also of humans with art, with creation, lends itself to a closing line that has served as a framing device for this article: after AI art, art. But what this art will look like, what forms it will take, what media it will use—and what effects it will generate—remains to be seen. Importantly, we can (still) play a role in designing the scene.