Do algorithms have epiphanies?

In the final chapter of A Portrait of the Artist as a Young Man, as he expounds his aesthetic theory in dialogue with his friend Cranly, Stephen Dedalus faces a basic problem: how can beauty be defined if it varies from one viewer to another? He has two possible responses. The first is that, as in the adaptations of human beings to varying climates and ecosystems, beauty constitutes a specific set of standards (in his example, those of the ideal woman) that supports the continuation of the species in each territory. Dedalus finds this idea rather limiting, but takes more hope from his second hypothesis, in which beauty’s constitutive qualities—wholeness, harmony, and radiance—transcend man’s animalistic and tribal nature; in which “the WHATNESS of a thing” passes from the mind of the creator into a relationship with the other.1

Intriguingly, James Joyce develops this passage differently in Stephen Hero, the literary precursor written a decade before the publication of Portrait. There, rather than raising the loaded question of embodied female beauty (particularly as a man in love), Stephen describes culture as a noisy thicket surrounding the pure concept of beauty:

No esthetic theory…is of any value which investigates with the aid of the lantern of tradition. What we symbolise in black the Chinaman may symbolise in yellow: each has his own tradition. Greek beauty laughs at Coptic beauty and the American Indian derides them both. It is almost impossible to reconcile all tradition whereas it is by no means impossible to find the justification of every form of beauty which has ever been adored on the earth by an examination into the mechanism of esthetic apprehension whether it be dressed in red, white, yellow or black…This is the moment which I call epiphany.

Stephen’s approach is relevant not only because it asks us to think about where beauty resides, but also because its reference to “every form of beauty which has ever been adored on the earth” expands the territory of beauty to a seamless and infinitely scalable field. His understanding of beauty need not be (and in fact cannot be) hierarchical, but instead becomes shorthand for the aesthetic encounter as a way of confronting reality in general, or at least its visual counterpart.2 Therefore, the main questions remain: if we take the “whatness” as the content or currency of the aesthetic encounter and the epiphany as its medium, then what is their relationship to reality? How is that reality altered by the encounter? And does the participation of new forms of intelligence, specifically the algorithmic, complicate not only the encounter but also reality itself?

It would be easy, of course, to say that reality exists only within the epiphany; that it is endlessly recreated in each encounter; and that the multiplication of realities emerging from aesthetic encounters is governed by the same economy as that which exists in any other realm of technologically-mediated human activity. But the aforementioned “whatness” presents a thorny contradiction: though we might acknowledge two simultaneous realities as diametrically opposed, the “whatness” suggests an underlying stratum of unqualified meaning, upon which our encounters play out in concert with the billions of encounters effected by previous generations. Simultaneously, the development by humans of complementary forms of artificial intelligence makes it impossible to conceive of reality as merely another invention, since the two conduct a large part of their relationship independently of our interference. Indeed, the strange encounter between the digital gaze and the “whatness” may force us to confront a very different kind of reality than the one we have thus far experienced.

If we speak exclusively of the pre-digital human encounter, then Stephen Dedalus’s two-pronged theory of the encounter works as a microcosm of an ongoing debate on aesthetics. On one side, the image is seen as inextricably bound up in our biological nature, our genetic material as a particular formula for decoding visual input into meaning. While Stephen imagines Darwinism as anathema to aesthetics (leading instead to eugenics), the root argument lives on in the biophilia theory of E.O. Wilson. He qualifies our encounter with the landscape as a deep-rooted form of “savanna gestalt”, in which we subconsciously desire, seek, and remake those biomes that recall the environment in which our ancestors first evolved bipedalism and other definitive traits for the modern human. At the same time, he explains our instinctive as well as cultural responses to threatening climates and organisms, such as hellfire or the serpent, as a deeply embedded biological response. The biophilia theory makes reality a savage battleground, using resonant images—both visual and metaphorical, artistic and literary—to guide our animalistic nature against our “better” judgment. The image becomes a weapon against the encroaching reality. Truth is not, in this case, the correlation between image and reality, but the survival of the fittest.

Stephen’s second theory, that of the transcendent quality of the image (at least in terms of aesthetics, if not semantics), differs in its flow of causality. Art, as conceptual armature, precedes our animalistic adaptivity and represents a purer form of human consciousness. In fact, the cognitive appraisal of our surroundings as an imageable landscape is understood primarily as a creative act. This explanation recalls a section of Leonardo da Vinci’s A Treatise on Painting, in which the artist describes “a way to bring out the genius by various inventions” by looking at stained walls, piles of ashes, clouds, and rocky escarpments in order to find the complex forms and landscapes necessary to paint scenes (whether battlegrounds, faces, or “monstrous things”) of sufficient intricacy and diversity.3 The counterpart to this active artistic technique is the more involuntary phenomenon of pareidolia or apophenia, in which meaning and pattern are construed from ostensibly random fields of data.4 In this sense, reality is an amorphous soup that can only be organised and imbued with intention by the human power of creation. Leonardo celebrates this default state of chaos, saying that “in confused things genius is aroused by new inventions”—as long as there is a precise understanding of the individual elements that inhabit it.

While the debate on the human aesthetic encounter, on a spectrum between hardwired response and aleatory creation, is still in play, the introduction of digital intelligence alters that polarity. Computer code grants us the privilege of directly witnessing the interaction of digital logic with mass quantities of numeric, lexical, or visual data. To that end, as the active “evolvers” of digital intelligence, we condition its encounter based on the things we design it to look for. Initiatives like Google’s DeepDream, which seeks and interprets dog faces (among other visual references) in any image fed to it, hint at the creative possibilities buried within that process, or at least the computer’s potential to create “art” for human eyes. Yet this understanding works mostly on our terms as creators and users of digital tools. To understand what the encounter means in terms of digital intelligence for its own sake, we require another framework—and we may find it, ironically, within art history itself.

In Real Spaces, the art historian David Summers constructs a lineage of art on two pivotal moments. The first moment is the invention of the image as a way of “[making] present in social spaces what for some reason is not present” physically: the hand-stencils and animal figures painted on cave walls in Sulawesi, El Castillo, and Chauvet in the Palaeolithic era are the earliest evidence of this cultural advance. Crucially, the invention of the image allowed humans to translate dynamic reality into static surfaces, making it possible to reorder the world through spatial hierarchies, frames, orders, ratios, symbols, and representational systems. Over the next thirty thousand years, the surface would develop an internal logic, absorbing technological innovations like the perspectival grid and propagating them through the circulation of artworks in the emerging global network.

Roughly one hundred years ago, however, as Joyce was developing Portrait, the ability of art to represent reality amid the overall context of technological and social progress met its reckoning. On one hand, the mechanical and chemical process of photography had become an accepted artistic form of systematically reproducing the visual evidence of reality; on the other, the nascent field of psychology questioned whether that evidence documented reality as such. At this point, the second moment arose, epitomised by René Magritte’s La trahison des images. Below a painting of a pipe, Magritte writes the words “Ceci n’est pas une pipe”—this is not a pipe. This work reveals a turning point: the image is no longer the conquest of reality; it is both a fantasy and a lie, a surface and a physical object.

In the intervening century, as the field of art underwent exponential diversification and expansion, it was far less concerned with the image as a tool for ordering and making sense of the world. At least, we humans were not. Computers, meanwhile, have an enormous affinity with the image. Their ability to swap meaningful interpretation with quantitative cataloguing allows for a massive increase in the amount of surfaces they can regard. At the same time, they can continuously apply their growing awareness and recognition of individual elements (as Leonardo suggested) to a sea of images without first developing a top-down explanation, something we humans can barely resist doing. While we continue to create and nourish ourselves with the human-generated images we deem valid as art, we are simultaneously devouring, submitting to, and reenacting the image encounters of computers. What would it mean, on the other hand, if we made art for the digital gaze? If the algorithm has an aesthetic encounter, can it also undergo an epiphany?

In May of 2015, Richard Wentworth created the artwork Agora on the roof of a car park in Peckham, South London. A thick stroke of aluminium paint curves and squiggles across the entire concrete surface, the curiously useless border of an idiosyncratic zone, a children’s maze without directionality. While a multitude of factors—human perspective, the fickle British weather, and the precarity of underused London property—may condition the ability of visitors to see Agora in person, the work’s natural viewing state may be through the form of the digital encounter, mapped in planimetric perfection through the aerial vision of Google Earth.5 From this perspective, the digital gaze not only takes in the pleasingly precise turning radius of Agora, but also compares it to the round clusters of trees to the north and contrasts it with the orthogonal ordering of row house and warehouse roofs to the east and south. Agora will be a work continually reproduced as every cycle of satellite imagery documents its reflective lines, as long as the substratum of undisturbed car park remains.

In other views, this digital visual system also reveals dimensions of reality that our own eyes may obscure. Switch to bird’s eye, view, and the virtual representation system of Google Earth stitches together the flat, distorted perspective of central Peckham with the three-dimensionally mapped agglomeration of the upper half of Southwark, continuing into the city. Along Commercial Way, Peckham Hill Street, and Goldsmith Road, the boundary between Peckham and the wards of Faraday, East Walsworth, and Livesey may not be immediately readable to the pedestrian, but the digital gaze both reveals that difference (of information? of technology? of power?) and gives it an aesthetic language. At the same time, it’s worth seeing how we, the humans, fare from this perspective. At the aerial scale, we are too small to notice, but the panoramas taken at roof level show the bodies of humans in beguilingly surreal configurations. People vanish from the waist down, overlap each other, and morph through different scales from the left hand to the right hand of the body (the panoramic software does a decent Francis Bacon). In these images, we see the hint of a reality translated through a radically different aesthetic encounter; this time, however, we are fixed inside the amorphous reality, and the epiphany is purely in the digital.

  1. “…that, though the same object may not seem beautiful to all people, all people who admire a beautiful object find in it certain relations which satisfy and coincide with the stages themselves of all esthetic apprehension. These relations of the sensible, visible to you through one form and to me through another, must be therefore the necessary qualities of beauty.”
  2. The change from the word “symmetry” in Stephen Hero to “harmony” in Portrait already suggests a loosening of the definition of beauty within Joyce’s decade of work.
  3. Somewhat snidely, Leonardo even claims that “as our Botticelli said, the study [of landscape painting] was in vain, because only by throwing a sponge dipped in diverse colours against a wall, it makes a stain in which one sees a beautiful landscape…but even if these blotches encourage inventions, they do not represent anything in particular. And that painter made very sad landscapes.”
  4. Apophenia refers to the interpretation of meaningful patterns in any kind of data, while pareidolia is strictly visual, often used to describe the detection of faces in landscapes or inanimate objects.
  5. Amusingly, some websites list the work as on view only until the end of September.