Notes on the Relational Aesthetics of Ambient A.I.

By Jason Hoelscher

Technologies become socially interesting only when they become technologically boring, to paraphrase NYU new media professor Clay Shirky. That is, people marvel at a new gadget when it debuts, because its rarity causes it to stand out. At that point it is technologically interesting but socially dull. It is only when a technology becomes integrated into everyday life, so prevalent as to be largely invisible, that its social effects become interesting.

Artificial intelligence has already reached this point even before it has fully arrived, going straight to the technologically dull but socially compelling stage. While science fiction has long presented A.I. as some sort of large, well-spoken mainframe computer, in reality A.I. is largely invisible, less an object than a set of immaterial tools and processes that underlie systems of transport, security, finance, healthcare, law enforcement, and so on. Put simply, A.I. has become deeply integrated into the fabric of everyday relations without our noticing, and the A.I. systems most of us encounter daily do not even register as A.I. in the first place.

Art, on the other hand, operates quite differently. If technology must become materially dull before it becomes socially compelling, artworks can sustain interest in both senses, sometimes for centuries after their creation. Recall Erwin Panofsky’s description, in Perspective as Symbolic Form, of how Renaissance linear perspective established and reinforced a spatial and organizational logic for western visuality. This visual logic created an objectivity of human experience that anticipated the scientific revolution and the Enlightenment. Consider further how, in the 20th century, cubist fragmentation not only ruptured this artistic plane of objective certainty, but opened the door to such fragmented and scattershot modes of contemporary experience as quick-cut film edits, memes, and the fleetingly fast samples of contemporary music. Unlike the increasing mundanity of the smartphone or A.I., however, the artworks that have triggered these social effects remain interesting as technological or technical artifacts in and of themselves, as the crowded galleries of the Uffizi or MoMA attest. Art is not either/or, but both-and.

The inversely-proportional relations between technology and the social on the one hand, and the mutually reinforcing relations between art and the social on the other, suggest ways to consider relational aesthetics within the context of invisible, ambient technologies. Relational aesthetics, according to Nicolas Bourriaud in his 1998 book of the same name, describes a mode of artmaking wherein the work of art is not an individual object, but the relations that might unfold around or between individuals. A relational artwork is less a distinct thing than a linking element or encounter that uses the whole of human relational and social context as its medium.

Strangely enough, the above description is applicable also to ambient artificial intelligence-a linking technology so pervasive and easy to use we often do not recognize the extent to which it mediates our relational and social encounters. Whereas it took centuries for western art to become untethered from the object and to dematerialize into conceptual art, systems aesthetics, and relational aesthetics, A.I. has been post-object all along, or at least since its seamless emergence into widespread use. This prompts a question: what happens when social relations and aesthetic experience are permeated by atmospheric fields of invisible, technological intelligence? To update Guy Debord’s notion of the spectacle as a social relation mediated by images, what happens when social relations are mediated by ambient technologies?

In order to explore some of these questions and ideas, this article is structured akin to the theory equivalent of a poetry slam, privileging the fast and rapid-fire over the methodically reasoned through. I offer ten prompts, ranging from the general to the specific, and from the plausible to the wildly speculative, regarding ways the accelerating emergence of artificial intelligence might influence art and creative culture in coming years, and vice versa. My goal here is less to provide an authoritative take on the issue-an impossibility, as we are too enmeshed within the moment to see it with any great clarity-than to suggest avenues for further consideration and exploration.

One: Art’s indeterminacy will become a crucial counterpoint to A.I. quantification

Although art has been a problematic philosophical subject since at least the era of Plato, aesthetics did not emerge as a distinct discipline until the mid-18th century. This emergence was in large part a response to the rise of Enlightenment thinking and the scientific method, both of which focused on quantification and objectivity at the expense of the subjective and non-quantifiable. Following on this, what roles might art and aesthetics play today, when every aspect of society, culture and experience is undergoing quantification at levels of high-granularity detail unimaginable even a decade ago?

I would propose that art’s ability to generate and sustain indeterminacy and difference might serve as a counter to just such quantification by information technology. Consider that information in a technical sense is a measure of difference, but most differences settle. A new fact constitutes a momentary difference relative to its context-which is what we call information-but then quickly settles into the equilibrium of knowledge. Art’s difference, however, remains different over time, less a difference than a process of differencing. While we all know the factual information and purpose of a snow shovel, we still debate Duchamp’s readymade snow shovel a century later, because its difference never settles, but continues differencing. This capability of sustained differencing reveals art’s potentials for resisting high-tech standardization and quantification, as a complex mode of open information. An A.I. could tell us everything about the snow shovel or the Mona Lisa down to the atomic level, while never quite reaching whatever aspects make these things art.

Two: Art will become increasingly material-specific relative to A.I.

Resurrecting and updating Clement Greenberg’s notion of medium specificity-whereby each artistic discipline must focus only on elements specific to that discipline-in an era of A.I., I predict artists will increasingly emphasize those aspects of art specific to material existence, and thus resistant to representation via digitization. We see some of this already, as painters like Franklin Evans and others push painting past its reliance on planes or screens, and explore materiality as it exists and operates in lived space and time. Where artists of the 1960s explored dematerialization to resist art’s drive toward objecthood, future artists will explore materiality qua materiality to resist, or at least interestingly complicate, sociocultural drives toward immaterialization and digitization.

Will Penny, Hypershade III, 2018, (still), projection mapped 4:00 minute single channel digital video on RENshape 5014 panel, 46” x 94” x 3 ½.” 3D scanned forms converted into a sculptural relief with projection mapped animation overlay.

Three: Art will become conceptual in new ways, as A.I. textualizes the world beyond anything Roland Barthes or Jacques Derrida ever conceived

A.I. systems operate by way of algorithms, which are sets of instructions. In this sense, algorithms are a performative mode of operational textuality-not just text that describes things, but text that does things. I would argue that A.I. marks a new, pervasive, and accelerated textualization of the everyday. In a world saturated with Wi-Fi and 4G signals zipping back and forth, constant data flows from smart-watches and facial recognition systems, and GPS-based location trackers in every pocket or purse-not to mention the trillion other data points that undergo A.I. correlation every second in vast databases-contemporary life offers a new and algorithmic take on Derrida’s famous quote that “there is no outside-the-text.”

As we continue to move beyond what Lev Manovich describes as older narrative modes of culture toward a database mode of culture, aesthetic approaches to such algorithmic textuality will open new ways to re-organize and re-integrate culture and creativity along relational rather than deterministic lines.

Four: Art’s relation to information will change as A.I. transforms human modes of communication and relation

In his recent book The Second Digital Turn, architectural historian Mario Carpo describes human history as a series of increasingly thorough ways to compress information. For instance, we’ve compressed the vast range of human sounds into representation via spoken words, then compressed those still further into phonetic letters and so on. More recently, the scientific method allows us to sum up vast cosmological ideas into simple equations, using the smallest possible sets of rules or constraints.

Today these compression approaches are no longer necessary, when A.I. systems can by brute force aggregate and recombine options billions or trillions of times a second. The eight months of representational and spatial exploration that went into Picasso’s Les Demoiselles d’Avignon would take less than a nanosecond of combinatorial pattern exploration for an A.I. like IBM’s Watson system.

This opens up potentials for reorienting human linguistic and aesthetic communication away from issues of compression, and toward issues of expansive potential and possibility. To give one, perhaps shallow, example: as information processing and storage capabilities have increased, and as bandwidth needs have loosened, we have seen a correlative rise of less-particular communication modes as emoji, less reliant on specificity of meaning and more open to interpretation. This is arguably the first time in history that a widely-adopted written language has been designed as deliberately less specific than its predecessors. Further, the fact that emoji and related modes of textuality are entirely visual suggests interesting potentials in terms of relational art and aesthetics.

Five: A.I. art is not fine art-not yet, anyway

A web search of art + artificial intelligence yields a range of interesting results, mostly focused around A.I. systems that generate images both beautiful and strange. Of these, the results of the Google Deep Dream project are the most widely known, being hauntingly bizarre and often downright surreal.

That said, unless they are created in tandem with a human creator, I do not believe the results of these works qualify as fine art-at least, not in the sense of “fine art” as promulgated by MFA programs, MoMA, the global biennial scene, and art history as constituted in the west over the last two centuries. Figuring out such things is why art students get MFA degrees in the first place, whether they know that fact beforehand or not. One thing that separates the fine arts from everyday vernacular art is art’s relationship to questions, to other art, and to the larger field of discursive relations in which art operates. Vernacular art succeeds on its own terms as long as it looks and feels interesting, beautiful, or engaging. Although a Picasso, Pollock or Piper artwork might also succeed along these lines, these works also test, interrogate and complicate their world. Beautiful and engaging or not, these works also challenge those ideas of vision, experience, or performance staked out by other art of the past or present.

Jason Hoelscher. Arbitrary Signage. 2019. Google Deep Dream Generator text-to-image transform, inception depth level 3. Courtesy of the artist.

By these standards, then, an A.I. does not create a work of art any more than a bank surveillance camera creates art by occasionally capturing a dynamic or well-composed scene. This is not to say A.I. might not do so in the future. For now, however, A.I. is essentially limited to pattern matching and reconfiguration-based approaches, and thus creates little more than interestingly art-like artifacts. The first A.I. to become aware of fine art’s entanglement with artistic discourses, transtemporal relations, and ideological expectations will create works that pose serious challenges to how we have come to define art in the western world since the emergence of modernism.

Six: A.I. + Art = Ambiguous/Aesthetic/Indeterminate Artificial Intelligence, or A/A/I.A.I.

Like all contemporary computational architecture, current A.I. systems are based on stark either/or 0/1 binary contrasts. What would a fuzzy, aestheticized A.I. look like, an A/A/I.A.I. (Ambiguous/Aesthetic/Indeterminate Artificial Intelligence) with input/output options that range from the literal and specific to the increasingly conceptual, metaphoric and interpretive?

Consider that, if A.I.s can already simulate fairly convincing analogs of Beatles songs or Modigliani paintings through nothing more than brute force data-mining and analysis of pitch and tempo, or value and line quality, might something like the Singular Computing company’s fuzzy processing chip, which is hardwired to resist precision, simulate the slippery ambiguities of aesthetic experience? In other words, A.I. is already doing things thought impossible a decade ago using nothing more than brute force iteration and generative processing. What would happen if that brute force approach were applied not to specific data but to fuzzy, ambiguous, and approximate calculation? Harvard mathematics professor Leslie Valiant describes this as probably approximately correct computation. Such computational ambiguity seems rife with potentials for art world adoption as the field of aesthetic software studies pursued by Matthew Fuller and Anna Munster et al. finally begins turning its attention toward A.I.

Among the best examples of quasi-aesthetic, computational ambiguity is the Google Deep Dream project, briefly mentioned in note five above. We have all likely seen these images before, which result when a neural network, trained on image recognition with tens of millions of images of faces and animals, is suddenly given little or no data to work with. The result is the generation of vast complexity and strangeness as the system undergoes a kind of A.I. equivalent to Kant’s free play of the faculties, its processing power operating with no definable goal to work toward. In his book Beyond Zero and One: Machines, Psychedelics and Consciousness, engineer Andrew Smart describes such processes as the A.I. analog of an acid trip, suggesting ways A.I. can be trained to approach and handle ambiguous and indeterminate situations-and thus, perhaps someday, aesthetic experience in its full richness.

Will Penny, Plein Air XI – Savannah, 2018, digital rendering, digital still life utilizing 3D photogrammetry scans taken around Savannah, GA.

Seven: Art will take on a role of reality stabilizer as social relations undergo flux within so-called post-truth contexts

The Internet was hailed in the 1990s as a peer-to-peer blow against centralized one-to-many authoritarian media control. What we now have, alas, is something quite different, a reverberative feedback ecology of interweaving bot-swarms, profit-driven distortions, and misinformation. While literary and aesthetic theorist Mikhail Bakhtin described the emergence of ideological consensus and belief systems as a polyphony of shared speech that weaves together the social sphere in which it operates, A.I.-enhanced rogue bots, news algorithms, and deep-fakes allow for a simultaneous construction of disinformational discursivity paired with an unraveling of socioculturally shared consensus. Here, Barthes’ easily-pierced tissue of quotations reveals its tissue-thin fragility, its susceptibility to disruption and manipulation.

Along similar lines, Louis Althusser described how a discourse or ideology works best when it is not recognized as an ideology, but is assumed to be simply the way things are. In an information context where the notion of “that’s just the way things are” is always up for grabs, if not under active attack, art’s historical role of showing alternative options to the real will come also to include revealing the real among the flood of options and alternatives.

Will Penny, Plein Air X – Savannah, 2018, digital rendering, digital still life utilizing 3D photogrammetry scans taken around Savannah, GA.

Eight: If the blank cube sculpture of Minimalism activated the two-way relational and theatrical space of the viewer, the inscrutable black cube of invisible A.I. processes will change the direction of our individual and social relations

How does the gaze operate when the feedback loop of subject-to-subject relations is broken by one-way surveillance? Jacques Lacan, by way of Kojève and Hegel, describes the gaze as a kind of feedback process between individuals: You and I become who we are in part because our interactions bring each other forward into ourselves. I see you seeing me even as I know you see me seeing you, and vice versa. Humans are social animals, and we are ourselves the most whenever we are around other people.

This relationship-possibly the core of relational aesthetics-changes with the advent of the one-way, algorithmic gaze. The all-seeing eyes of technological, biomedical, and location-based surveillance systems, with their far-beyond-human A.I. data-aggregation capabilities, changes our existence as individuals when the relational feedback between the observed and the observer is replaced by the one-way flow of observation and aggregation. If relational aesthetics is activated by the flow of relations around and between, this process becomes unbalanced when the flow goes only one direction.

Nine: Portraiture will experience a renaissance

What are the possibilities for portraiture in a cultural context of real-time facial recognition systems? I predict that A.I. will generate new modes of portraiture as different from current modes as a Matisse portrait is from Vermeer’s Girl with a Pearl Earring-not only in obvious visual difference, but also in underlying material logic and pictorial goals.

Consider how modernist approaches to paint changed due to the shift of focus from representation circa 1500-1800, to the material and semiotic tools of representation as explored circa 1800-2000. A.I. will prompt a similar shift from our present-day focus of isolated visual appearance to data aggregations and predictive algorithmic processes. Just as the emphasis of the modernist portrait shifted from a realistic likeness to a fusion of likeness + the materials used to construct that likeness, portraiture in an age of omnipresent A.I. will explore social, morphological, behavioral and economic flows, and hitherto unseen patterns of gait and circumstance, rather than the outer particularities of single sitters.

Jason Hoelscher. Symmetry Breakage. 2019. Google Deep Dream Generator image, inception depth level 3. Courtesy of the artist.

Ten: Artistic approaches to space will move beyond the perspectival, optical, and networked in weird new ways as relational space becomes cognified

A.I. is so far not monolithic and general, but distributed, specialized and ambient. This suggests interesting relations to Genevieve Bell’s observations that a technology is only truly transformative when it changes our relations to space, time, and each other. As Bell, formerly the Director of Interaction and Experience Research at Intel and now an anthropology professor in Australia, notes, the more of those three components a new technology changes, the more transformative it is. It is now practically impossible to remember what life was like prior to the smartphone or Internet, or to imagine a world without cars, trains or planes, because these technologies transformed not only our relations to each other and to time, but also our experience of space.

Along these lines, consider the fact that A.I.-not the large technical object we imagined, but rather a field of invisible computation capabilities-changes our relation to space by becoming spatialized, by cognifying space itself. If, as per Sol LeWitt, the idea becomes the machine that makes the art, what happens when idea and space become indistinguishable-when the entirety of our surroundings, visible and invisible, is even more saturated with data flows and information swarms than it already is?

Jason Hoelscher is gallery director and assistant professor at Georgia Southern University. Hoelscher has exhibited his artwork in Atlanta, New York, Berlin, Hong Kong, Paris, Stockholm and elsewhere, and has presented papers at such venues as CAA, SLSA, SECAC, Harvard University, and the University of Copenhagen. Hoelscher has also written for BURNAWAY, ArtCore Journal, ARTnews, and Evental Aesthetics. He received an MFA in painting from the Pratt Institute, and a PhD in aesthetics and art theory from IDSVA.