“I myself, alone, have more memories than all mankind since the world began,” he said to me. And also: “My dreams are like other people’s waking hours.” And again, towards dawn: “My memory, sir, is like a garbage heap.”
― Jorge Luis Borges, Funes the Memorious

Memory is a creative act. When we conjure up images from our past, we imagine, embellish, reframe, distort, suppress, as we pull them into the present moment. This is especially true for childhood memories, often fragmented and dreamlike, marked by gaps and omissions—some vivid images stand out, but the logic connecting them is shaky.

Because it would be impractical to remember everything all the time, memories are usually aroused by some real world trigger, like the uncanny feeling of arriving somewhere you didn’t realize you had already been before. I get this sometimes walking in an unfamiliar neighborhood, turning a corner, and suddenly being hit with the view of a street I visited with my parents on some trip to New York 20 years ago—I recognize that storefront, that tree, that stoop. My mental map is filled in with pieces I didn’t know where missing. But inevitably, in the intervening time, the facade has been updated, the tree’s leaves faded, new people sit on the stoop. The shock of recognition is tempered by change. The world moves on.

The rules change when a memory is tied not to a physical location, but a media object. For most of recent history, media was fixed within a physical container or time-bounded experience—a performance, a book, a movie theater, a record, a VHS, a CD, a DVD. Over the last few decades, more and more media objects have been reduced to fungible data, accessible on a whim. At this point, just about anything is widely available in digital form with the right search.

One result of this transition is that we now have instant unlimited access to memory triggers—reopening portals to a past that remains as it was. We sort of can step into the same river twice, even if our feet have grown. This tendency has been exploited most obviously by the Hollywood dream factory’s IP recycling scheme, but can be seen everywhere—the social media nostalgia content industry, Disney adults, the Mandela effect. Increasingly, remembering and remixing what once was feels like the only thing left on the horizon.

I say this not as an outside observer, but a direct participant. Plenty of my early memories are associated with media objects, for better and for worse, and I’m just as prone to nostalgia as anyone. When a childhood memory flashes into consciousness, I’ll try to follow it back as far as I can. Why this memory? Why now?

Recently, I’ve been thinking about Garfield.

The Time Traveler

As a very young kid I loved visiting my aunt’s house. It felt heavy in a comforting way, like a weighted blanket. The beige-carpeted living room was perpetually dark, with smoky wood paneled walls, the shades pulled, soaked in an unfamiliar, musty smell (I’d later realize it was cigarettes). She had a big goofy Rottweiler, the first dog I ever befriended, and a twenty-something adult son who lived in her basement—my cousin.

In retrospect, my cousin was ahead of his time, a canary in the coal mine for so many trends that would play out over the next couple decades. He was a NEET and a weeb—studied Japanese in college, and would return from trips abroad bearing exotic souvenirs, like Japanese-language Pokemon cards or video game consoles that hadn’t yet been released stateside. I remember seeing a book cover with a nearly-nude woman on the floor of his room once (I would later piece together that it was a Ghost in the Shell manga) and feeling a little queasy, like I’d glimpsed a world I wasn’t supposed to know about yet. Because of both his age (halfway between me and my parents) and his access to this mysterious foreign land that seemed to exist in the future, he felt like a time traveller. Given the way that things have played out since, maybe he was.

Just as I idolized my cousin, I have always identified with Garfield, but my relationship to him has shifted over time. As a kid, I loved newspaper comics, and would tear through books of Garfield strips. Later on, mostly encountering him through GIFs and memes, I shared his love for coffee. For whatever reason, in recent years Garfield has become a magnet for nostalgia-mining content creators (there’s an unofficial subsection of YouTube called GarfTube, and people who make their living as Garfield historians).

I wonder if Garfield lends himself to this kind of content because he himself is a sort of algorithm.

He contains a few well-defined parameters—liking coffee and lasagna, hating Mondays—and each comic strip applies these basic variables to different scenarios that play out over a few panels. He’s predictable. When you see Odie approaching, you know conflict is imminent. When Jon is opening a can of tuna, he’s right there. When it’s Monday, he’ll hate it. This formulaic quality is the basis of “Garfield Minus Garfield,” a popular webcomic experiment that subtracts Garfield from his own strip, to humorous results.

But back to my aunt’s house. It was in her basement where I watched a VHS of A Garfield Christmas. Truthfully, I don’t really remember doing this—I just have vague flashes of the wooden walls and a fuzzy CRT TV. But I know that it happened, because of one very vivid image that stuck with me…

The Everything Machine

The special begins with a brief cold open. Jon, dressed as an elf, wakes up an uncharacteristically cheerful Garfield, who declares, “Christmas means presents. Lots and lots of presents!” Luring Garfield to the tree with “breakfast lasagnas,” Jon forklifts in a huge giftwrapped box, which falls away to reveal a gaudy, electronic Santa-shaped throne with a metal hat (resembling an electric chair skull cap) and a sack labelled Toys. “It’s the gift that keeps on giving,” Jon exclaims, before sitting on the throne to demonstrate: “This gift giving machine can read your mind. Whatever you think of it will give you!” The cap lowers to Jon’s head, some squeaky electronic sounds play, then the sack shudders and spits out a green bycocket hat. When Garfield takes a turn, the sack produces a handful of jewels, and he quips “Now this is what Christmas is all about,” before generating a deluge of consumer goods—guitars, cameras, TVs, radios, trains, etc. He dives into his sea of products like Scrooge McDuck: “Things, stuffs, gadgets, toys, greed, avarice, I love it.” After a credits song (“Gimme, gimme, gimme”) the screen ripples, revealing that the whole thing was Garfield’s dream.

I had no memory of the rest of the special—just the throne, which I dubbed The Everything Machine. It was my first encounter with an archetypal concept, going back to the mythical overflowing Cornucopia that fed a young Zeus with endlessly refilling food (see also, the Fruit of the Loom Mandela effect). Or, the Brothers Grimm’s Wishing-Table, the wish-fulfilling Kalpavriksha tree of Hinduism and Buddhism, Star Trek’s replicators and holodecks, Harry Potter’s Room of Requirement. The mechanics aren’t always exactly the same, but they share a core idea—generating something from nothing.

Because I’d only seen the show once and had no way to access it after the fact, I never fully knew if it was real, or something I had dreamed up. I was young enough, and the boundaries of my understanding of the world porous enough, that the concept could take on a life of its own. Naturally, it resurfaced around Christmas. I always loved the anticipation of the holiday, cutting out pictures from magazines to make elaborate wishlists of Legos and trampolines and scooters and video games. I didn’t care so much about actually getting presents as I did the act of creative fantasizing. Once I learned of The Everything Machine, my wishlist scrapbooks became redundant. I only needed one thing, from which all else would emerge. Ask, and you will receive.

Even before The Everything Machine appears, Garfield’s dream follows a logic of wish fulfillment, as Jon leads him to the tree with “breakfast lasagnas.” The dream appeals to a fantasy of mastery that runs counter to the creative act of memory—the closure of the gap between desire and outcome. Simply think “lasagna,” and you will get lasagna. It’s a perfectly closed loop. Daydreaming, imagining, chasing—no longer necessary. If Garfield is an algorithm, The Everything Machine is its terminal form.

Garfield Revisited

So thanks to the internet, where it’s only a matter of time before all half-remembered childhood media will be resurrected, I recently rewatched A Garfield Christmas, twentysomething years since I first saw it (almost 40 since it aired). This time I was able to absorb the episode as a whole, beyond The Everything Machine, which actually has very little relevance beyond the intro.

The plot arc is pretty standard for a Christmas special, seeing Garfield shift from Grinch-like materialism to the joy of the giving spirit, and the broad strokes were apparently based on Jim Davis’s own childhood memories. After the cold open, Jon drags a reluctant Garfield from his comfortable bed to his family’s farm in rural Indiana. On the drive over, Jon waxes sentimental about his love for this time of year: “I think the best part is watching the faces of children, and sharing memories of being a child at Christmas.” He launches into a song about his own fond memories, and Garfield cynically undercuts him at every turn (“dad was chopping a tree,” / “chores”).

At the farm, we’re introduced to Jon’s family, and together they cook, decorate the tree, and sing songs. Although Garfield is reluctant, his heart softens when Grandma Arbuckle pets him in her lap and reminisces about her deceased husband in the moonlight. Later, Garfield stumbles across a stack of old letters in the barn, and on Christmas morning, gifts the stack to Grandma Arbuckle. They turn out to be love letters from her husband from when they first met, and she’s moved. At last, Garfield has fully embraced the Christmas spirit, and delivers a heartfelt speech—“It’s not the giving. It’s not the getting. It’s the loving.”

Until now, The Everything Machine was my entire takeaway from A Garfield Christmas, a single image plucked from its context that returns from time to time. It never occurred to me that it was intended to be a plot device, a cautionary tale, a satirical fantasy of infinite consumerism that reveals through contrast the true meaning of Christmas—love, symbolized in the form of handwritten letters. There is something saccharine about the episode, transparently sentimental and manipulative. But from today’s vantage point, its simple appeal to emotion feels refreshing, even radical.

Despite my original intentions, in this writing process I’ve found myself much less interested in thinking about the linear inputs and outputs of The Everything Machine than about my cousin’s reappearances at different stages of my life, leaving books and DVDs behind in his wake, a cross-section of Gen X techgnostica including Philip K Dick, The Matrix, Fight Club, Cosmicomics, Aphex Twin, or Gurdjieff—portals to seeing and thinking in new ways, whose influence on the course of my own life I could only appreciate in retrospect. Or how he got me interested in drawing, or was the first person to employ me as a writer. All these moments were like seeds that would later grow in unpredictable ways, my aunt’s basement a ground zero for cultural currents that still ripple in me, and the world at large. The Everything Machine, with its promise of unlimited abundance, turned out to be a footnote—my way in to thinking about mind and matter, but not my way out. Which may explain why so many of the stories with magically abundant objects contain some kind of ironic catch—the satisfaction is not in the instantaneous fulfillment of a wish, but the gap.

A machine of infinite abundance is a seductive idea. More things, brighter things, shinier things. There is an undeniable excitement to fantasizing about all the new things this machine can do, the things it can create, the things. I was once the dreaming Garfield, swimming through a sea of toys. But time has passed. Seeing what changes and what stays the same, what lasts and what fades, the appeal of The Everything Machine has worn off. Now the most relatable, and aspirational, part of A Garfield Christmas was Grandma Arbuckle’s reaction to her handwritten love letters—the people and the traces they left, and how they continue to live inside us and change us long after they’re gone. I used to read Garfield, watch Garfield, idolize Garfield. Now, I just want coffee.

The World Moves On

I haven’t talked to my cousin in a while. Always ahead of his time, he was an early digital nomad and has spent the last couple decades hopping around Asia. Last I saw on LinkedIn, he’s doing B2B SaaS content marketing with AI.

At this point, everyone is familiar with the “Uncanny Valley,” the concept of the eerie liminal zone between human and non-human that provokes a primal feeling of unease. Humanoid robots, bad CGI (i.e. The Polar Express), and prosthetic limbs are all common examples of the Uncanny Valley, and the term has also experienced a bit of a renaissance lately with AI-generated images, which are often recognizably life-like but with enough details feeling slightly off to trigger the Uncanny sensation.

While the term itself is ubiquitous, the actual origins of the Uncanny Valley, and its connection to the Freudian concept of the Uncanny, is underexplored territory. Tracing back the roots of Uncanniness over the last century can help illuminate the mechanisms of Uncanniness in the current moment of AI, and I believe provide a key to the broader existential feelings that AI provokes.

Mori

Let’s work our way back from The Uncanny Valley. Written by roboticist Masahiro Mori in 1970 for the obscure Japanese journal Energy in 1970, his essay coined the term and outlined the basic principles in two charts, based on the axes of human likeness and affinity (as in, do people feel positively about it).

Mori begins by describing how, typically, industrial robots don’t aim to look human, and are designed for functionality:

“Nowadays, industrial robots are increasingly recognized as the driving force behind reductions in factory personnel. However, as is well known, these robots just extend, contract, and rotate their arms; without faces or legs, they do not look very human. Their design policy is clearly based on functionality.”

Approaching the first peak, he describes toy robots, which are designed with loosely human-like aesthetics in mind, and kids love them. So they’re medium in both likeness and affinity. Mori places Bunraku puppets even higher on the scale of likeness and affinity, just below the very top—a healthy person. The Uncanny Valley exists between these peaks, where likeness is high but affinity is negative—Mori uses the example of a prosthetic hand:

“One might say that the prosthetic hand has achieved a degree of resemblance to the human form, perhaps on a par with false teeth. However, when we realize the hand, which at first sight looked real, is in fact artificial, we experience an eerie sensation. For example, we could be startled during a handshake by its limp boneless grip together with its texture and coldness. When this happens, we lose our sense of affinity, and the hand becomes uncanny.”

Mori then reveals a new chart, with a second, dotted line. The dotted line factors in movement, and has an even steeper slope. Mori gives the example of the added eeriness of a moving disembodied prosthetic hand, and provides a somewhat fanciful anecdote:

“Imagine a craftsman being awakened suddenly in the dead of night. He searches downstairs for something among a crowd of mannequins in his workshop. If the mannequins started to move, it would be like a horror story.”

Also notable on the second chart is the line that travels backwards, from human (highest affinity) to corpse (Uncanny Valley). Mori explains:

“As healthy persons, we are represented at the crest of the second peak in Figure 2 (moving). Then when we die, we are, of course, unable to move; the body goes cold, and the face becomes pale. Therefore, our death can be regarded as a movement from the second peak (moving) to the bottom of the uncanny valley (still), as indicated by the arrow's path in Figure 2. We might be glad this arrow leads down into the still valley of the corpse and not the valley animated by the living dead!

I think this descent explains the secret lying deep beneath the uncanny valley.”

So, maybe, the secret is death? Let’s keep that in mind as we move back to the roots of the Uncanny in Freud, and even earlier, in Ernest Jentsch.

Jentsch

Sigmund Freud’s 1919 essay The Uncanny is a classic piece of psychoanalytic writing, a rare Freudian text that is mostly based around literary analysis, and a popular reference point in the humanities. But Freud’s essay is itself a response to an earlier paper by Ernest Jenstch, On the Psychology of The Uncanny (1906), the first psychological reckoning with the concept. Jentsch’s essay primarily focuses on the Uncanny’s association with “uncertainty,” and it lays an important foundation for Freud’s expansion of the Uncanny. Jentsch draws attention to the linguistic slippage between canny/uncanny (heimlich/unheimlich), situates the Uncanny in a psychological framing, emphasizing the effect rather than the cause; he focuses on the Uncanny in art and namedrops E.T.A. Hoffmann, who Freud will devote a significant portion of his essay to; he speaks of wax figures and panoramas, and even has a bit about movement that sounds very much like Mori’s definition of Uncanny:

“A doll which closes and opens its eyes by itself, or a small automatic toy, will cause no notable sensation of this kind, while on the other hand, for example, the life-size machines that perform complicated tasks, blow trumpets, dance and so forth, very easily give one a feeling of unease. The finer the mechanism and the truer to nature the formal reproduction, the more strongly will the special effect also make its appearance.”

Finally, Jentsch ends on a note similar to Mori’s, again evoking the fear of death that seems to be central to the Uncanny:

“The horror which a dead body (especially a human one), a death’s head, skeletons and similar things cause can also be explained to a great extent by the fact that thoughts of a latent animate state always lie so close to these things.”

Again, death appears to be at the root of the Uncanny, though it’s not yet entirely clear how.

Freud

Freud builds on Jentsch’s foundation of the Uncanny, honing it from the overarching idea of “uncertainty” into a more focused paradigm of interlocking anxieties. His essay is only around 20 pages, but it is highly allusive, sprawling, and sometimes contradictory, so rather than go through it point-by-point, I’m going to try my best to pull out the key points and interpretations.

Like Jentsch, Freud starts with an etymology of heimleich/unheimlich:

“The German word unheimlich is obviously the opposite of heimlich, heimisch, meaning ‘familiar,’ ‘native,’ ‘belonging to the home’; and we are tempted to conclude that what is “uncanny” is frightening precisely because it is not known and familiar. Naturally not everything which is new and unfamiliar is frightening, however; the relation cannot be inverted. We can only say that what is novel can easily become frightening and uncanny; some new things are frightening but not by any means all. Something has to be added to what is novel and unfamiliar to make it uncanny.”

Freud accuses Jentsch of not moving beyond this “uncertain,” then goes into a lengthy section of definitions for heimlich/unheimlich in various languages, and essentially comes the conclusion that the words have a dialectical relationship:

“Thus heimlich is a word the meaning of which develops towards an ambivalence, until it finally coincides with its opposite, unheimlich. Unheimlich is in some way or other a sub-species of heimlich.”

This linguistic slippage between familiar and unfamiliar is precisely the area in which Uncanniness operates, just on the border of normalcy but with something not quite right. That’s essentially Jentsch’s definition of the Uncanny—uncertainty—but Freud wants to take it further, and he develops his argument with a close reading of E.T.A. Hoffmann’s 1817 short story The Sandman.

The Sandman

Freud picks up on Jentsch’s brief mention of E.T.A. Hoffmann’s stories as a fount of uncanniness to do a deep dive into The Sandman. While Jentsch only briefly alludes to the story, Freud provides an extensive reading of The Sandman and the specific ways it evokes the Uncanny.

The Sandman is itself a long and fairly complex story that I’ll try to describe in brief. Through an epistolary structure, it follows a young man named Nathanael, who experiences a series of childhood traumas that haunt him throughout his life and eventually lead him to madness. As a young adult up he meets and falls in love with a girl named Olimpia, who turns out to be a human-like automaton. Throughout, Nathanael is haunted by the figure of the Sandman, a boogeyman who steals the eyes of children who won’t sleep, to feed them to his own children on the moon. The Sandman appears across the story in various doppelgänger guises. As a boy, Nathanael comes to associate the Sandman with a colleague of his father, the mysterious lawyer Coppelius. One night Nathanael hides in his father’s room, hoping to see the Sandman, and sees Coppelius performing some kind of alchemy in a fireplace. The climactic section follows:

“He looked like Coppelius, whom I saw brandishing red-hot tongs, which he used to take glowing masses out of the thick smoke; which objects he afterwards hammered. I seemed to catch a glimpse of human faces lying around without any eyes - but with deep holes instead.

'Eyes here' eyes!' roared Coppelius tonelessly. Overcome by the wildest terror, I shrieked out and fell from my hiding place upon the floor. Coppelius seized me and, baring his teeth, bleated out, 'Ah - little wretch - little wretch!' Then he dragged me up and flung me on the hearth, where the fire began to singe my hair. 'Now we have eyes enough - a pretty pair of child's eyes,' he whispered, and, taking some red-hot grains out of the flames with his bare hands, he was about to sprinkle them in my eyes.”

Nathanael’s father manages to save his eyes, but Nathanael is still punished (which I will expand on soon), and shortly after, Nathanael’s father dies in a mysterious accident, seemingly caused by Coppelius. This primal scene in the story forms the foundation of Freud’s first argument about the Uncanny—that the Uncanniness stems from Nathanael’s anxiety around losing his eyes, eyes which can be interpreted as a phallic symbol, and thus connected to castration anxiety and the Oedipus complex. Freud draws this into the idea that the Uncanny can sometimes proceed from repressed infantile complexes, or basically, “the return of the repressed.” This overtly psychoanalytic definition of the Uncanny may be useful for clinical purposes, but I’ll be focusing on Freud’s second definition of the Uncanny, the return of surmounted primitive beliefs:

“Let us take the uncanny in connection with the omnipotence of thoughts, instantaneous wish-fulfillments, secret power to do harm and the return of the dead. The condition under which the feeling of uncanniness arises here is unmistakable. We—or our primitive forefathers—once believed in the possibility of these things and were convinced that they really happened. Nowadays we no longer believe in them, we have surmounted such ways of thought; but we do not feel quite sure of our new set of beliefs, and the old ones still exist within us ready to seize upon any confirmation. As soon as something actually happens in our lives which seems to support the old, discarded beliefs, we get a feeling of the uncanny; and it is as though we were making a judgment something like this: ‘So, after all, it is true that one can kill a person by merely desiring his death!’ or, ‘Then the dead do continue to live and appear before our eyes on the scene of their former activities!’, and so on.

And conversely, he who has completely and finally dispelled animistic beliefs in himself, will be insensible to this type of the uncanny. The most remarkable coincidences of desire and fulfillment, the most mysterious recurrence of similar experiences in a particular place or on a particular date, the most deceptive sights and suspicious noises—none of these things will take him in or raise that kind of fear which can be described as ‘a fear of something uncanny.’ For the whole matter is one of ‘testing reality,’ pure and simple, a question of the material reality of the phenomena.”

The idea works on both a historical and personal level—older societies with belief systems where the boundary between the real and divine was more porous, as well as the development of an individual from the magical thinking of childhood to the rational thinking of adulthood. Unlike his first description of the Uncanny as the return of a specific childhood complex, this surmounted beliefs definition of the Uncanny is about the return of an earlier structure of thought where the basis of reality, and crucially, the self, is called into question.

Freud goes on to develop this theory of the Uncanny with a number of examples, including ghosts, which appear to confirm “the doctrine of spiritual beings”; the double/doppelgänger, which raises questions about the immaterial soul (more on this later); lifelike inanimate objects like dolls, automata, and waxworks; coincidences which appear to validate the idea of fate or destiny; magical powers like the evil eye; and psychical phenomena like telepathy, precognition, and wish fulfillments, which suggest the “omnipotence of thoughts.” But in a strange omission, or perhaps repression, he glosses over one of the most striking images in the Sandman story.

In the aforementioned primal scene, where Nathanael encounters Coppelius by the fire and is caught, Nathanael’s father convinces Coppelius not to pull out his eyes, but he instead inflicts a far stranger punishment:

“Whereupon Coppelius answered with a shrill laugh: 'Well, let the lad have his eyes and do his share of the world's crying, but we will examine the mechanism of his hands and feet.' And then he seized me so roughly that my joints cracked, and screwed off my hands and feet, afterwards putting them back again, one after the other.”

Freud references this moment briefly in a footnote, where he reduces it to another castration equivalent and Nathanael’s infantile attitude towards his father. It’s understandable that he takes this approach, given his desire to shift away the emphasis of Jentsch’s analysis of “uncertainty” towards infantile complexes, but in doing so he misses perhaps the most significant Uncanny moment in the story. This singular image of Nathanael’s hands and feet being “screwed off” like a doll suggests that the source of his madness, and the events that follow in the plot, lie not in his “uncertainty” of Olimpia’s ontological status as a human or automaton, but his own. It is an emblematic example of Freud’s own argument about the return of surmounted primitive beliefs, as well as the anxiety at the heart of the enlightenment—the lingering doubt that earlier beliefs and understandings are completely dead and gone.

This ontological uncertainty is fundamental to Freud’s Uncanny, especially in its connection to death, on which Freud writes:

“There is scarcely any other matter, however, upon which our thoughts and feelings have changed so little since the very earliest times, and in which discarded forms have been so completely preserved under a thin disguise, as that of our relation to death.”

As noted above, nearly all iterations of the Uncanny seem to circle back to the same point: death. To understand the significance of this movement, we will need to take a detour to another of Freud’s central concepts, the Death Drive.

The Death Drive

The Uncanny was written in the midst of the composition of one of Freud’s other essential texts, Beyond the Pleasure Principle (1920), and there’s considerable overlap between the concept of the Uncanny and the development of Freud’s thought on the Death Drive. The main idea of Beyond the Pleasure Principle is that humans are driven by two principles: Eros, aka the Pleasure Principle, the affirmative drive towards survival, pleasure, sex, creativity, etc. and its opposition, the Death Drive, the destructive drive towards self-annihilation. The two are in constant battle because of their underlying similarity—the Pleasure Principle is about the relief of tension, resulting in the pleasure of relief. The Death Drive is also about the relief of tension, but in the opposite direction—the relief of being alive—resulting in compulsive behavior, and is frequently associated with repetition, aggression, and self-destruction. The Death Drive is the desire to return to the state of non-existence which preceded biological life, as Freud describes, it is “an urge in organic life to restore an earlier state of things.”

Because the two pieces were written at the same time, the Uncanny is conceptually proximate to the Death Drive, made explicit by Freud in a footnote where he cites Beyond the Pleasure Principle and recommends it to readers interested in learning more about drives. Immediately after, he outlines the connection between the Uncanny and the Death Drive:

“It must be explained that we are able to postulate the principle of a repetition-compulsion in the unconscious mind, based upon instinctual activity and probably inherent in the very nature of the instincts—a principle powerful enough to overrule the pleasure-principle, lending to certain aspects of the mind their daemonic character, and still very clearly expressed in the tendencies of small children; a principle, too, which is responsible for a part of the course taken by the analyses of neurotic patients. Taken in all, the foregoing prepares us for the discovery that whatever reminds us of this inner repetition-compulsion is perceived as uncanny.”

So it could be said that one source of the Uncanny is things that remind us of our inner repetition-compulsion, from the Death Drive. Freud notes the uncanniness of repetitions in various forms:

“For instance, we of course attach no importance to the event when we give up a coat and get a cloakroom ticket with the number, say, 62; or when we find that our cabin on board ship is numbered 62. But the impression is altered if two such events, each in itself indifferent, happen close together, if we come across the number 62 several times in a single day, or if we begin to notice that everything which has a number—addresses, hotel-rooms, compartments in railway-trains— always has the same one, or one which at least contains the same figures.”

The Double

The repetition in the Uncanny takes on a human form in the concept of the Double, or doppelgänger. Freud draws from Otto Rank’s literary analysis of the Double, although looking ahead to Lacan can also deepen our understanding of the relevance of the Double. In his excellent analysis of the Uncanny and Lacan, I Will be With You on Your Wedding Night, Mladen Dolar writes:

“For all of them the shadow and the mirror image are the obvious analogues of the body, its immaterial doubles, and thus the best means to represent the soul. The shadow and the mirror image survive the body due to their immateriality—so it is that reflections constitute our essential selves.' The image is more fundamental than its owner: it institutes his substance, his essential being, his ‘soul’; it is his most valuable part; it makes him a human being. It is his immortal part, his protection against death.”

The Double, an eerie duplicate of oneself, appears as one does in a mirror, which Dolar connects to the mirror stage of development. Lacan describes the mirror stage, the recognition of the self as an imaginary figure, as the moment where one is split from their immediate childhood state of wholeness, of jouissance, and becomes aware of themselves as an autonomous being. In other words, the moment of castration, the constitution of the subject and their entry into language, from the Real into the Imaginary.

'“It is only by virtue of one's mirror reflection that one can become endowed with an ego, establish oneself as an ‘I.’ My ‘ego- identity’ comes from my double.”

The difference between looking in the mirror in real life, and the literary/Uncanny concept of the Double is that the Double exists in the world, can go off and commit acts that the original would never dare to commit, enacting the jouissance that they lost when first going through the mirror stage.

“We can now see the trouble with the double: the double is that mirror image in which the object a is included. So the imaginary starts to coincide with the real, provoking a shattering anxiety. The double is the same as me plus the object a, that invisible part of being added to my image.”

The Double, filtered through Lacan, presents an interesting picture of Uncanniness not as a lack, an almost-human lacking some human characteristic, but as something overflowing with too much human, more than human: human+object a, with object a being the Real, the original lack that was constituted when the whole infant recognizes itself in the mirror and is thus split and brought into the world of language as a subject. The Double therefore represents an impossible synthesis of the Real and the Imaginary.

Here we can begin to bring together some disparate points that have been floating around. The Double is Uncanny because it ultimately reminds one of the Real—of the reality that remains out of reach once one has been born into language—which subsequently reminds one of death, our ego, the awareness of our limited subjecthood existing within a symbolic order.

“One could say that the double inaugurates the dimension of the real precisely as the protection against ‘real’ death. It introduces the death drive, that is, the drive in its fundamental sense, as a defense against biological death. The double is the initial repetition, the first repetition of the same, but also that which keeps repeating itself…”

The Double embodies the core characteristics of the Uncanny—ontological uncertainty, the return of primitive beliefs, the repetition of the Death Drive—flashes of the Real erupting into the Imaginary. And when we think of certain long-gestating concepts like the singularity, human-AI immortality, simulation via Matrix-like computer/brain interfaces—the Double as repetitive defense against biological death comes into focus as a foundation of the modern Uncanny.

The Uncanny Valley and AI Images

Having developed a theory of the Uncanny rooted in psychoanalysis, we can now take a look at how these concepts apply specifically to The Uncanny Valley. On a basic level, lifeless objects—a prosthetic hand, a CGI character, an AI-generated image—are imbued with the illusion of life, undercut only by some difficult-to-quantify strangeness.

But not all Uncanniness is created equal. I would suggest that the first order of the Uncanny of AI generated images came from their content—like the above early viral GAN image, where everything looks almost familiar but you can’t quite identify anything—it seems to short circuit the pattern-recognizing parts of our brain. Acting on a similar level is the long period when AI image generators had difficulty rendering hands (also notable for the centrality of prosthetic hands in Mori’s essay). The tell-tale mangled hands (perhaps AI parapraxes) become the Uncanny signifier. In both of these examples, something in the pictures themselves tips us off to their Uncanniness.

But the second order is an Uncanny form—where the AI images are actually indistinguishable from “real” images. The medium becomes the message—within the first order, the images are the medium that delivers Uncanniness through their strange content. In the second order, the medium is the internet, social networks, media channels—and the Uncanniness is delivered through the form of perfectly convincing “fake” images.

I witnessed this recently when someone Tweeted some images of monks playing synths in the mountains. They look pretty good, but I could immediately tell they were AI-generated, having seen a lot of these sorts of images. The person who posted it was called out, but insisted they wouldn’t post it if it was AI, and that as far as they knew it wasn’t, showing an online tool that claimed it was likely not AI-generated. They changed their original reply to the above, acknowledging that they were fooled.

This scenario was more Uncanny to me than an AI image of someone with weird hands—the reminder that what it means to create, to relate, to validate, is undergoing a colossal shift, provoking anxiety about that future, and even moreso, a retroactive anxiety that what was thought to be most fundamental to the human soul could be recreated with lines of code (or the turning of gears, the strings of puppets, the shutter of a camera).

This may be why some artists have become the most vocal opponents of AI—because their specific background positions them to viscerally understand the uncanny form of AI. Artists spend massive amounts of time honing their unique abilities, and it inevitably becomes a part of their ego and self-identity. When an image model recreates and regurgitates artworks in their style in seconds, it is not only upsetting for the logical, practical, job-related reasons, but also because of the existential uncertainty it provokes. An AI forgery appears like a Double in their own image, the real bubbling up from the digital black box. Beyond “will this thing take my job,” the more disturbing question it provokes is, “am I any different from it?”

Put another way, think about the immortal quote from Syndrome: “When everyone’s super, no one will be,” but replace ‘super’ with ‘human.’

The Future of the Uncanny

The Uncanny moves through the dialectical opposition of heimleich/unheimleich, and its form is appropriately one of constant oscillation, whose waves continue to equally enthrall and unsettle us. As demonstrated by the figure of Olimpia in The Sandman, the questions raised by lifelike human automata have been confounding us for centuries. But the real Uncanny encounter is not taking place between the reader and Olimpia, as Jentsch would suggest, nor Nathanael and his father, as Freud would suggest; but between Hoffmann and the reader. In the primal scene of Nathanael’s hands and feet being rearranged, Hoffman’s puppet-show of a novel confronts us with the suggestion that it is not Olimpia who is the real automaton but Nathanael, and by extension, ourselves.

Similarly, the Uncanny Valley’s source of anxiety is not simply robots that look lifelike, with some deadness behind the eyes—because we see them as our Double, as an overdetermined Uncanny nexus point of repetition, automation, omnipotence of thought, wish fulfillment, and jouissance—and thus realize that we ourselves are lifelike with some deadness behind the eyes. As Lacan says, the ego is constructed like an onion with no center, just a succession of peeling identifications. Like the dotted line on Mori’s second chart that travels backwards from human to corpse, we are standing on the cliff’s edge of the Uncanny Valley, confronted with our own lifeless reflection.

As AI and LLMs perform more-human-like tasks with ever more efficiency, we can see the double movement held in the original concept of the Uncanny. As technological advancements continue to shift the emphasis between heimleich/unhemlich, the Uncanny Valley deepens. We are faced not with the fear that the robots are real, but rather, that we are not.

Today is the launch of a new project I’ve been collaboratively working on for the better part of a year, One World Moments. The video you see above is a short preview of what’s to come, consisting of 12 short clips of AI-animated videos, which hopefully don’t look like what one would traditionally expect from “AI video.” In the coming days, we will be releasing more short clips several times a week. This first wave of videos represent a brief introduction to One World Moments, a project that has been gestating for a long time and will continue to develop as technology evolves.

As a longtime writer and practitioner of film, I’ve been closely following the progression of AI images over the last several years with a mixture of skepticism and curiosity. One World Moments is a way to explore the possibilities of the emerging world of AI-generated video, intended to lean into the specific strengths of the medium–not as a replacement for traditional narrative filmmaking, but as an offshoot, something that hasn’t quite existed yet because it hasn’t been technically possible until now. I’d mostly like the video to speak for itself, but I’ll go into a little bit more detail about the background and future of One World Moments below.

Background

To go along with the video, we’ve created a website that showcases some influences on One World Moments in a scrolling moodboard style. Exploring the website is probably the best way to become experientially acquainted with our approach, but I’ll try to give some words here to fill in the background.

One World Moments is an audiovisual artwork in the ‘ambient’ media tradition, a somewhat marginal, unofficial canon with roots in the experimental art of the 20th century, with points that could be windingly traced between the Lumiere brothers, Andy Warhol, John Cage, Brian Eno, Chantal Akerman, HD TV video loops, computer screensavers, ASMR videos and Lo-Fi Hip Hop Beats to Study To. Put simply, ambient media places an emphasis on atmosphere and feeling above action or narrative, doesn’t demand or expect full attention, and often involves sampling or remixing. They’re sensory white noise, an emotional backdrop to subtly augment your world, whether for the purposes of relaxing, working, or socializing.

One World Moments is a new experiment in ambient media, which seeks to use the new possibilities enabled by AI image generation to create more specific, evocative, and artistic ambient visuals than have been previously possible on a mass scale.

Future

Given the limitations of the current (though fast-developing) technology, the teasers represent a proof-of-concept for the aesthetic, the phase 1 of the project. The next phase of One World Moments is to go long, to develop proprietary fine-tuned models that can sustainably generate high-quality videos ad infinitum, leading to a 24/7 stream of video and audio. In a way what we are trying to build is an imaginary camera that produces a specific type of image, one that draws from various traditions of film history–actualities, home movies, landscape films, slow cinema, early social media images, internet ephemera–and narrows infinite possibilities into a singular aesthetic.

Like the video, this post is just a small glimpse of all the thought, work, and inspiration that has gone into creating this project, and there will be a lot more to come soon. We hope that you enjoy this early introduction to One World Moments and continue to follow the journey ahead.

On the occasion of the release of my new AI-assisted short film, Oasis, and the emergence of Runway’s Gen-2 (not to mention Adobe Firefly), I’ve been once again thinking about AI video and its implications on the future of film. I originally started this essay two months ago, when Runway’s Gen-1 had just been announced, but after playing around with Gen-1 a bit and my own experience using AI for filmmaking with Oasis, it felt like the right moment to finally finish it. I’ll begin by getting up to speed with the current state of AI video tools, then detail how AI fit into my process for Oasis, and provide some speculation on how AI may fit into the future of filmmaking.

Make Way for Video

AI generated text-to-video has been around in some form since at least the days of CLIP+VQGAN, along with various Stable Diffusion options, but they’ve always required specialty knowledge and a lot of fine-tuning, making it effectively inaccessible to the layperson. In the same way that it took the user-friendly interface of ChatGPT to get a critical mass of everyday people using AI chatbots, text-to-video still has a ways to go in both quality of output and UX in order to reach a mass audience. Some promising contenders are the big players’ text-to-video models announced last fall, which remain in the development stage—Google’s Imagen Video and Meta’s Make-a-video both promise to generate a few seconds of HD video based on prompts. There are some other ones as well using Stable Diffusion, in some cases to generate even longer and more complex videos, but as far as I’m aware, none of these are available to the general public yet, and certainly aren’t in the user-friendly state of something like Dall-E or ChatGPT.

The most recent development in consumer-friendly text-to video is Gen-2 by Runway. When I first started writing this post a couple months ago, Gen-1 had just been announced. Now, we’re on the cusp of Gen-2, currently in closed beta, and some people with early access have been posting tests that certainly look more cohesive and stylish than Gen-1. The key difference is that Gen-1 was purely a “video-to-video” model—modifying existing input video with text prompts (or images), but Gen-2 offers text-to-video: generating video clips entirely from scratch with just text inputs, in the same way that Midjourney or Dall-e generate images. See the explainer video below for more details.

I messed around with Gen-1 a bit but was mostly unimpressed. The limitations of clip sizes (initially 3 seconds, it’s since been expanded) and the relatively primitive quality felt like a step down after being used to the high-fidelity text-to-image models. It basically just operates like a highly limited video filter, a weak version of rotoscoping, a tech demo with little to no practical artistic applications. At best, some of the Gen-1 video experiments could be called “cool,” but it never felt like a tool someone could genuinely use in a video workflow.

Gen-2’s text-to-video option is more interesting, at least on paper. I, again, haven’t used it yet, but the promise of a video generator on par with Midjourney or Dall-E could have the potential to shake-up traditional camera-based workflows. The oft-cited fantasy spoken breathlessly by AI speculators of being able to create customized Hollywood-level feature films at the push of a button, is a vision of endless entertainment that feels like the natural apex of the current algorithm-driven streaming landscape. But I think this vision of AI Hollywood, utopian to some, dystopian to some, is misguided, and misses an important point by fixating on the content at the expense of the medium: the difference between images and video.

Image ≠ Video

When thinking about the text-to-image generators that have become popular over the last year or so, and the text-to-video generators in their nascent state, it’s tempting to compare them 1-to-1 because of their superficial similarity. But this approach obscures the significant difference between the two mediums. Still images have a constrained number of variables to engage a viewer: they have to make their entire impression in the bounds of a rectangle, in one moment in time, silently. There is still infinite room for innovation within these boundaries, but the somewhat limited nature of the medium makes them ideal for AI—it’s easier to learn the rules of what makes a still image good, because “looking good” is one of the main thing an image needs to do to be successful. Certain iconic images, captured or created by skilled artists, can impart meaning that goes beyond “looking good,” but for many purposes, looking good is enough. Because AI text-to-image generators have one narrow variable, “looking good,” to focus on, there’s been a somewhat straightforward path to training and improving the image models, a clear goal with noticeable results.

Video isn’t so simple. Moving images have a much higher number of variables (editing, camera movement, camera/lens choice, performance, casting, sound, genre, and the infinite web of signifiers in-the-world that can be captured by a camera, to name a few) and “looking good” isn’t very high on the list; film is a language with a much larger vocabulary than static images, and therefore a more complex criteria for quality. Think about how many popular Tik Tok creators with millions of followers are just filming themselves in their bedrooms with their iPhones. How a great film can move you emotionally whether you watch it on a crisp 35mm print in a theater, or a washed-out 240p VHS rip you found dubiously online. Or how some films (i.e. Jarman’s Blue, or a bunch of experimental works) don’t even use images at all. Good video, good film, is operating on exponentially more levels than just visuals—visuals are just one layer among many, and Gen-1 merely added (low-fi) filters to that layer. Gen-2 goes one step further by generating those visuals whole cloth, and slightly better visuals at that, but it still barely scratches the surface of what actually attracts people to video as a medium.

I see this as a fundamental limitation that won’t change anytime soon. Even as more and more aspects of video workflow become automated (more photoreal and consistent video, AI-generated sound, editing, etc), a good video still relies on the confluence of so many factors that are incredibly difficult to automate. I don’t see a magic Hollywood movie-generating button replacing the human element of filmmaking, but rather, a gradual incorporation of AI tools into the still largely human process of moving image storytelling.

Making Oasis

I’ve always been drawn to “mysterious island” narratives. The Invention of Morel, Treasure Island, The Wild Boys, The Strange Tale of Panorama Island, The Witness, and of course, Myst, all captured my imagination at pivotal times, and the balance between spatial constraints and imaginative expansiveness felt like a good sandbox to play around with. So the idea of a Myst-inspired short film had been floating around in my head for a long time—I initially started writing notes and storyboarding in 2019—but I never had the means to practically bring the vision to life. I always intended for it to be animated, to recreate the feeling of a lost 90s video game, but it became clear that it would be difficult to achieve by any means other than 3D software. I tried teaching myself Blender, but it would have taken me years to learn it well enough to be able to create the visuals I wanted, and probably more years to actually make it. So I shelved the idea, thinking I might return to it someday, but with skepticism that day would ever come.

When I started getting deep into Midjourney, it occurred to me that I might be able to finally take a stab at the island film. Software that can create whatever image you imagine, in any style? It naturally seemed like the perfect fit. But my initial experiments with Midjourney v3 were disappointing; it wasn’t consistent, the images weren’t quite right stylistically, I couldn’t cultivate a persistence of vision strong enough to make a cohesive film. When v4 came out last November, I tested the idea again, and it quickly became clear that it might actually be feasible.

The images looked convincingly like old 3D graphics of a first-person POV game, the prompts worked consistently: an illusion of continuity could be created. And as soon as it became clear it was possible, I immediately got to work. The process was pretty smooth and intuitive—I began with a loose idea for the script, that I would take to Midjourney to create a flurry of images, then tweak the script based on those images, then write some more, then make some more images, etc. It was a nice feedback loop that felt sort of like spinning a film directly out of my brain, unfurling from the conversation between me and the AI’s latent image space. It’s either a happy accident, or cosmic destiny, that the film happened to also incorporate this interplay between reality and fantasy, dream and technology, into the core of its thematic fabric. Either the process of using AI unconsciously shaped my storytelling instincts in that direction, or the original seed of the idea was always looking towards a future that could only be realized when this magical technology happened to emerge. In any case, the film itself is probably a better reckoning with my thoughts on the past and future of artificial images than I could ever articulate in words.

I made the first “episode” of the initial version of Oasis (around 1.5 minutes) in one marathon 12-hour session, give or take some polishing. The gap between the length of time it would take for me to create the film manually vs with AI is pretty astonishing, collapsed from years to hours.

But on the other hand, the images are just one part of the film. Midjourney enabled me to practically implement the images of the film, but I still needed to have the overarching structure of the vision to create something good (and years of experience making, watching, and studying films). Aside from the images, I basically approached the other technical aspects in the same way I would any other film. I wrote a script, I edited in Premiere, I downloaded sound effects from the internet and made the music in Garageband. Holistically, it’s a human-created film that happens to consist of AI generated images as the baseline material, not all that different from a found-footage project. So if it is a good film (and I like to think it is), the AI only makes up a small percentage of the reason that it’s successful. I don’t say this to be ungrateful—I’m happy that Midjourney allowed me the means to finally make the film, but it’s ultimately more of a means to an end than an end in itself. And I don’t see AI videos being more than a means to an end for a long time, no matter how photorealistic they become.

Movies Are So Over… Not

It is an important caveat that Oasis is a Le Jetée style film edited together from still images, so it doesn’t exactly translate 1-to-1 with the type of full-motion videos generated by Gen-2. In some ways its visual language is closer to a graphic novel, or animatic, or a video game stream. But these distinctions are ultimately superficial—it’s still a film that tells a story through a succession of images and sounds. And all of film is just creating the illusion of motion through still images, mine just happens to be <1 frame per second rather than 24. The technical specs of any video are always just a vehicle for something else, a means to evoke a reaction between a hypothetical viewer and the screen. Although Oasis doesn’t necessarily resemble a typical film on the surface, it uses AI as a means to evoke a reaction in the audience in the same way any other film would.

Perhaps this is the sort of video best suited to the strengths of AI—rather than hoping text-to-video AI will be able to generate Hollywood-style feature films at the push of a button, AI can be used in targeted ways within a workflow to speed up or allow for the creation of new visual and narrative forms. Runway’s video tools beyond Gen-2, like their AI-assisted green screen rotoscoping, transcription, etc. might ultimately prove more liberating to filmmakers than text-to-video. I see it as analogous to special effects—in the recent documentary on Industrial Light and Magic, I was struck by how the transition between practical and digital effects mirrors the discourse around AI: the same fear of job loss, of the loss of soul, an era of authenticity falling to a new regime of cold calculating machines. But the reality didn’t turn out so dire. Films are still a mixture of practical and digital effects, jobs were changed more than lost, and digital tools have made filmmaking more accessible and been creatively liberating for new generations. The history of film, not to mention images in general, is the history of technology, and technological sea changes have always undergirded and guided the shifting tides of moving image culture. Yet, the real reason that people continue to be drawn to movies (story, emotion, humanity, communication), remain the same throughout every seismic shift.

Of course, all of this should be taken with a grain of salt. Maybe it’s more than a coincidence that I happen to see my film as the model for future AI-generated filmmaking. Maybe later iterations of Runway’s Gen-X will completely rewrite the book on AI filmmaking, opening up new possibilities that we could never imagine. Or maybe film will remain a medium that is hard to get right, takes a lot of thought and planning and skill, and relies on distinctly human proclivities to resonate with a human audience. Time, as always, will tell.

Disclaimer to subscribers: I’m working on another essay on AI art, but the release of the new Sight & Sound poll inspired me to publish this piece on film, as explained below. If you only want to read about AI, there will be more to come soon, but if you’re interested in film, read on.

The Greatest Films of All Time 2022

The new Sight & Sound “Greatest Films of All Time” poll came out yesterday, the once-a-decade list that amalgamates votes from global film critics (and separately, directors) in an attempt to capture the pulse of the film world. There’s a lot to be said about the lists, and I’m sure they will inspire plenty of dialogue, especially as the individual contributors’ lists roll out. But I’m not looking to delve into the gritty details of the picks; instead, I’d like to use the list as a jumping-off point into a larger discussion around canonization. Shortly after the Sight & Sound Poll was released, I saw tweets from Criterion and Janus celebrating that more than 60 of the Sight & Sound films are in the Criterion Collection, and 48 are distributed by Janus (the theatrical distributor symbiotically connected to Criterion). This rather striking fact prompted me to return to an essay I wrote for a film magazine in college, back in 2017, about the role of film distribution as the bottom line shaping film canons, and decided to revisit it.

Five years ago, the film landscape was in transition from the primacy of physical media to the ubiquity of streaming, and I was writing very much in the midst of this shift, focusing mainly on Criterion Collection DVDs/blu-rays, with only a passing mention of the growing spectre of streaming. But many of the trends I was picking up on have proven more relevant as time has gone on. I also happened to spend much of the essay on the films of Jean-Luc Godard, which takes on a new significance with his recent passing, and the way his body of work is being memorialized retrospectively. Considering these events, I’ve decided to release a revised version of the essay here, mostly intact but with some additions and omissions in light of recent history.

Cinephilia™

Most “cinephiles,”1 people who consider themselves film lovers to the point where they make it a part of their identity, have likely entertained, or even cherished, the idea that they have “good taste.” Movie watching is inherently a passive activity, so there has to be some countervalent active process to distinguish oneself among the crowd, and that action happens to be the internal cultivation of taste. The concept of taste, multifaceted and nuanced in its own way, ultimately comes down to selectivity. Broadly speaking, one watches and enjoys certain movies, which they sort into the category of “good movies,” and ignores or dislikes others, which they categorize as “bad movies.” Over a long period of time, this selection pattern forms the basis of any individual cinephile’s taste, which they may then wear as a badge of honor, a testament of their own individuality and uniqueness as a shrewd and intelligent film-watcher. “Good taste” becomes an essential tool for engaging in conversation with other cinephiles about which films they choose to watch, which ones they value or reject, or which directors they like and dislike; the kinds of conversations that film people live for and everyone within earshot finds insufferable.

There are endless social dimensions to taste to be explored—like the ways in which taste is constructed along gender, class, and race lines, and how it changes over time. For my purposes, I’d like to hone in on a particular kind of internet cinephilia, that you might find on Letterboxd, Twitter, Reddit, or other online communities, frequently (though not always) driven by young men. There’s familiar landmarks that can be recognized in a certain type, and certain stage, of cinephilia: the fetishization and brandification of auteur directors, the currency of obscurity, the limiting of discourse into name-dropping or one-upping. In real life, social conventions might stifle these impulses a bit, but on the unaffected public square of the internet, they often run wild. This phenomenon is somewhat understandable considering that the activity cinephilia is centered around is so inherently consumptive. Within a passive hobby where nothing is directly produced, taste becomes the primary measure of any individual participant’s authority—the cinephile’s action is selection, which becomes the bottom line of identity formation. And because streaming means that watching movies is no longer bound to specific times and locations in the real world, there’s boundless opportunity for online film watching and discussion.

My attempt here has been to illuminate a particular narrative around how taste is understood and enacted within cinephile circles. But this framing of individual taste as the driving factor of cinephilia obscures a more materialist question—the question of where the selection comes from in the first place. The personal locus of “good taste” perpetuates an individualist idea of agency, which belies the external, institutional factors that ultimately shape taste. Though there are numerous complex reasons people may gravitate towards certain films more than others, the underlying factor for every film-watcher’s viewing habits is availability. To like and dislike films, one must be aware of them, and be able to see them. And this problem leads to the questions that this essay will actually be dealing with, namely: how are notions of “good taste” formed, and what underlying factors influence the process?

The Canonical Loop

Film canons are one of the strongest influences on viewing habits in classrooms, theaters, and at home, forming both the base and superstructure of availability. Film canons exist in many forms, but the most prominent examples are the lists of “The Greatest Films of All Time,” most famously by Sight & Sound and AFI. These expansive lists can, and often do, provide a starting point for any aspiring cinephile. Usually topped by Citizen Kane, Vertigo, Tokyo Story, and other “unimpeachable masterpieces,”2 these lists can open any young filmgoer’s eyes to a dazzling world of movie magic, and set the precedent for their own personal lists of favorites, providing a compass for future explorations. But the general similarity of the “Greatest of All Time” lists speaks to the self-perpetuating quality of the film canon. The “Greatest Films of All Time” will be re-watched, remain in circulation, and provoke critical conversations over the course of generations, even as contemporary developments in film shift under our feet. It’s a feedback loop, a closed circuit that has largely sustained itself for decades with only minor changes. And the gatekeepers (critics, artists, academics) who shape these lists with their “good taste” in turn shape the “good taste” of the viewing public, some of whom go on to be gatekeepers themselves and reinforce the cycle.

The consequences of this model—critical gatekeepers choosing which films to enshrine in canons to be consumed and internalized by the next generation of gatekeepers—results in a limiting of the films that people watch, and a shift of the agency of “good taste,” from viewers to gatekeepers. But in fact, the agency lies beyond the gatekeepers. Because in order for a film to be slotted into the category of “good movie” in the first place, it has to be seen. The true locus of power, then, is in the higher echelons of distribution, where decisions are often made with predominantly economic motivations.

Depending on its distribution, any film may be seen by a few or a lot of people; if it fails to reach an interested audience, it may toil in obscurity indefinitely, waiting for the day when it might be rescued and re-entered into circulation in the public consciousness. There are many such cases of films that failed to reach an audience on initial release, but have since become canonized classics, and metropolitan repertory film scenes notoriously have the power to shift the critical perception of films over time. Especially in the pre-digital era, the materiality of film reels ensured that a specific film had to be shown at a specific place at a specific time; if the right gatekeeper never saw a certain film, it may never be seen again. Same goes if gatekeepers saw a particular film and, for whatever reason, didn’t like it enough to champion it. No matter how intuitive one’s “good taste” may seem, it is inseparably contingent on whatever films one has seen throughout their life, which (until recently, at least) was contingent on film distributors and circumstance.

The Criterion Effect

Of course, modes of distribution have changed drastically over time, and viewing habits along with them. The most significant shift in the business of film distribution, until the advent of online streaming, was the rise of home video, giving viewers a new level of agency over what films they watch, and when and how they watch them. And no company better epitomizes the canonizing power of home video than the Criterion Collection, which specializes in “important classic and contemporary films” marketed to savvy cinephiles who are willing to pay premium prices for the company’s extensive releases.

Over the years, the quality of Criterion’s library and enduring popularity among cinephiles has made their brand a de facto canon in and of itself. Once a film is consecrated in the prestigious company of Criterion’s numbered spines (counting among their ranks classics from the silent era until now) it has entered into a canon that guarantees an audience of critics and casual fans alike. Criterion, often described as “film school in a box,” is positioned as a beacon of “good taste,” which hungry audiences are happy to bask in and absorb the benefits. Like Penguin Books, A24, or the Sight & Sound list, Criterion is above all a brand, that brings with it an expectation of quality, in both the films and the surrounding packaging—elegantly designed products that make perfect coffee table accessories.

It’s telling that Criterion “haul” videos have become an entire Youtube subgenre, a cottage industry, usually consisting of midwestern teens sifting through the shelves of a Barnes & Noble, gathering stacks of Blu-rays while they narrate their opinions on the films, like a funhouse mirror version of the Criterion closet videos. There’s something a little charming about it, but it’s also off-putting—the ultimate reduction of a transcendent, emotional relationship with an ephemeral artwork, into pure commodity fetishism.

On a base level, Criterion’s success is understandable and well-deserved. They’re good gatekeepers, doing laudable work restoring classic films and bringing lesser-known films to a wider audience. Their library is strong, despite some inevitable biases and blind spots. And their brand now extends into the digital age—after a few fledgling forays into streaming (Hulu, Filmstruck), their service The Criterion Channel, launched in 2019, has finally become a worthy home to their library, matching or even exceeding the quality of their physical releases with excellent programming and a wealth of supplements.

But, for all the good they’ve done, Criterion isn’t immune to the material forces of the world. They are, after all, a company—and a company must make a profit at the end of the day. Profit demands an audience; Criterion has to release movies that, theoretically, somebody will want to see—and hopefully enough people for Criterion to make their money back. They generally navigate this push and pull between crowd-pleasers and more obscure fare successfully. For every Breakfast Club or Wes Anderson release, there’s a Shoah or Stan Brakhage box set. But the correlation between accessibility (in terms of distribution) and accessibility (in terms of a film’s perceived challenge or difficulty to a broad audience) is not always so straightforward. Take, for example, Jean-Luc Godard.

Godard’s Long New Wave

The popular image of Godard is as the 1960s critic-turned-filmmaker enfant terrible poster child of the French New Wave, and in the wake of his recent death, it appears to be the image that will forever be immortalized in the public consciousness. As the familiar narrative goes, Godard’s New Wave streak blew open the formal possibilities of film and left the landscape forever changed. His early run from 1960-67, Breathless to Weekend, scrambled the tropes of American filmmaking in a postmodern blender, creating a fun and free-wheeling body of work that would go on to influence everyone from the New Hollywood filmmakers to Quentin Tarantino and Wong Kar-wai. The reverberations of the New Wave à la Godard persist even today—every year or two there’s sure to be a new indie film that references the New Wave era implicitly or explicitly. Criterion has nearly all of Godard’s New Wave films in their library, ensuring the canonization and enduring popularity of his work and status as a Cinema Icon™.

The problem is, Godard’s career continued for over 50 years after the New Wave subsided, and his later work is arguably even more formally radical and politically vital than the celebrated 10-year period that gets all the attention. You’ll only find two post-New Wave films in the Criterion Collection—Tout va bien (currently out-of-print) and Every Man for Himself. The Criterion Channel’s selection is a bit better, adding Film Socialisme, Goodbye to Language, and The Image Book. But these are still barely scratching the surface of his dozens of later-period films.

For a brief—but by no means comprehensive—sampling of the effects that Criterion has had on Godard, I’ll mention some figures drawn from the popular film database Letterboxd. Of the top 20 most-seen Godard movies, 14 are available in the Criterion Collection on DVD or blu-ray, and with an additional 3 on the Criterion Channel. So that’s 17/20 on Criterion. That leaves the remaining three. A Married Woman, one of the best of his New Wave films which wasn’t widely available until a recent blu-ray release by Kino (itself probably the second most prominent boutique home video distributor), sits relatively low on the list at #18—surely if it had been released by Criterion with the others it would sit among his other new wave hits. Histoire(s) du cinéma, his masterful, lengthy exploration of the history of 20th century images (included on the recent S&S poll), was completed in 1998 and not released on DVD until 2011, via Olive Films. First Name Carmen (1983) sits at #20 on the list, released by Kino in 2019. The further down the list you go the less accessible the films are—again, both in terms of availability and aesthetically. Some, like King Lear3, don’t even have DVD releases. Of course, part of the problem is the increasingly demanding difficulty of some of Godard’s later work. But even something as watchable, and essential, as the Jean-Pierre Léaud and Juliet Berto-starring Le Gai Savoir (1969) sits low on the list, at #33, as it was largely unavailable until Kino’s 2017 blu-ray release (and now on the Criterion Channel). So it can’t be said that his later films are underseen because they’re too “difficult”—the patterns all leads back to how easy the films have been to see.

All of this is to say that availability has played a substantial role in the public’s perception of Godard and his films. And it’s an issue that bleeds from reception into production. How would the film landscape of today be different if more filmmakers were influenced by Wind from the East or Six Foix Deux instead of Breathless and Band of Outsiders? And what future films are we missing out on by the continued obscurity of Godard’s most vital films? The work that late Godard does towards deconstructing the grammar of film images could push the discourse of cinema ahead ten years if impressionable filmmakers got their hands on some of his lesser-known work.4 But for the time being, we may have to settle for the pithy echoes of the French New Wave’s detached “cool.”

It goes without saying that Jean-Luc Godard remains one of the most famous filmmakers of all time, and was a white European man from a wealthy family with a career that anyone would envy. I say this not to diminish his accomplishments, but to acknowledge that, as far as choices of subjects go, I’ve focused on one who occupies a particularly privileged position. I could have just as easily devoted an article to any number of truly overlooked filmmakers, but that would be missing the point. The myopic canonization of a big name like JLG suggests the problem is inescapable: if a filmmaker as prominent as Godard can slip through the cracks, what else we are missing? And why? Godard is just one example of the influence Criterion wields, itself a microcosm of the way that distribution and economics shapes film culture as a whole.

Streaming the Future

The streaming revolution is a complicated issue, on the one hand devaluing the theatrical experience that most cinephiles treasure, but on the other, bringing infinitely more films to a wider audience than previously possible. Compared to the average filmgoer 50 years ago, anyone outside of a major metropolitan area now has access to films that they could have otherwise never been able to see. The internet is the new theater, and it levels out many of the material restrictions that previously limited gatekeeping roles of good taste to denizens of a handful of cities and institutions. This year’s Sight & Sound poll, somewhat controversially, included a fresh batch of internet-based critics, which some feared would disrupt the canon and lead to a diluted mess. These fears were largely unfounded; the list, despite some omissions and inclusions that will certainly be argued over for years to come (just ask Portrait of a Lady on Fire), it mostly resembles previous years. The internet hasn’t caused a tectonic shift in film taste, yet. The old canons still hold sway, perhaps a testament to Criterion’s enduring influence more than anything.

The next Sight & Sound list will be a better indicator of whether the internet and streaming really have the power to meaningfully shift canons in new directions, but in the meantime, we’re still feeling the downstream effects of a cumulative 100+ years of distribution trends. This article is far from exhaustive, and will hopefully be the beginning of a conversation rather than any kind of conclusion. If there’s anything to take away, it’s to continue to question what “taste” really means. Because taste is never purely an act of spontaneous generation, of individual will. It comes from somewhere—somewhere where commerce plays an inevitable role in what is seen and unseen.

Rabbit Holes

I’ve had a handful of experiences where a chance encounter with a book has radically changed the course of my life, the act of flipping through the pages rerouting my future like the flap of butterfly wings on the other side of the world. Stumbling across William Gaddis’s The Recognitions, a massive postmodern novel of art forgery and fraud in a Portland library, was one of these moments. On a whim I decided to check it out, mostly for the challenge of reading something long and difficult over an uneventful summer. It turned out to be one of the most formative reading experiences I’d ever had, reframing my conception of what a novel could be and do, inspiring me to return to college to finish my degree, and shaping the next decade of the way I made and thought about art. Tracing back the series of branching paths that unravelled from my encounter with The Recognitions, it’s clear that my life would look very different if it hadn’t happened to be on that library shelf that day.

Shortly after finishing my essay on Vilém Flusser and AI, I fell down another rabbit hole of AI-adjacent literature from an unexpected source. While reading through the recent New Yorker article on Ashgar Farhadi’s plagiarism, the word “agape,” defined as “a Greek term for love that persists without the expectation of reciprocation,” caught my eye. Seeing this unusual word in the wild triggered a memory of Gaddis’s final work Agapē Agape, a novella loosely about the invention of the player piano as the origin of art’s move away from authenticity towards mechanized uniformity. It had been sitting on my to-read shelf for years, but its central theme suddenly felt like it might be relevant to the research I’ve been doing, so I finally took the plunge and read it in one sitting.

My intuition turned out to be correct—Agapē Agape is relevant to the AI art conversation, albeit as the antithesis of Flusser’s writing. While Flusser expressed a detached uncertainty about the potential of AI and the internet tinged with an ultimate note of hope, Gaddis’s diagnosis couldn’t be more pessimistic. Their methods are also opposite: Flusser’s theoretical work is ostensibly non-fiction, rooted in history, but veers into the speculative, while Gaddis’s presents as a work of fiction, but is actually the result of decades of research originally intended for a non-fiction text on the history of the player piano.

The History of Agapē Agape

The origins of Agapē Agape go back to 1945, when Gaddis was working as a fact-checker at the New Yorker. He was assigned to edit an article on the player piano, which sent him down his own research rabbit hole, developing into an obsession that would last the rest of his life. Gaddis’s first writing on the player piano was a short non-fiction article published in 1951, the seed of the idea that would crystalize in Agapē Agape nearly 50 years later:

There was a place for everyone in this brave new world, where the player offered an answer to some of America's most persistent wants: the opportunity to participate in something which asked little understanding; the pleasure of creating without work, practice, or the taking of time; and the manifestation of talent where there was none.

The player piano perfectly symbolizes one of Gaddis’s core ideas: the decline of art as a sacred conversation linking the soul and mind of a Great Artist and their audience, and its slide into frivolous entertainment, mere pleasure, reflecting a broader trend in American society towards mechanization and homogeneity, with the reduction of the human element.

It’s the same shift that Flusser describes in less scathing terms—the fall of the era of Great Artists replaced by a mechanically democratized future, elevating communication into a global artistic conversation in which anyone can participate, replacing the rareified pact between special individuals. While Flusser’s book reckons with this shift matter-of-factly, without much of a moral judgment, Gaddis makes no secret of the fact that he finds this mechanization a force of apocalyptic hopelessness, as a powerless observer looking on in despair at the necrotizing foundation of the art he’s devoted his life to. Accordingly, Agapē Agape is his final desperate scream into the void, taking the form of a long unbroken monologue from a terminally ill narrator trying to synthesize decades of scattered research material on the player piano and related subjects into something before he dies. The unnamed narrator is a barely fictionalized avatar for Gaddis—himself suffering from cancer when he assembled the book, similarly brain-fogged from prednisone and surrounded by mountains of disorganized material that he’d gathered over the years.

Rather than the non-fiction form he originally struggled with, the thinly-veiled fictional conceit allows Gaddis to transfigure his research into a work of art itself, making it a terminal example of the distinctive human-centric artworks that he opposes to the mechanizing force embodied by the player piano. He finished the book shortly before his death in 1998, and it was published posthumously in 2002, intended as his final statement.

The All-or-None Machine

Below is the first sentence of the Agapē Agape, in full:

No but you see I’ve got to explain all this because I don’t, we don’t know how much time there is left and I have to work on the, to finish this work of mine while I, why I’ve brought in this whole pile of books notes pages clippings and God knows what, get it all sorted and organized when I get this property divided up and the business and worries that go with it while they keep me here to be cut up and scraped and stapled and cut up again my damn leg look at it, layered with staples like that old suit of Japanese armour in the dining hall feel like I’m being dismantled piece by piece, houses, cottages, stables orchards and all the damn decisions and distractions I’ve got the papers land surveys deeds and all of it right in this heap somewhere, get it cleared up and settled before everything collapses and it’s all swallowed up by lawyers and taxes like everything else because that’s what it’s about, that’s what my work is about, the collapse of everything, of meaning, of values, of art, disorder and dislocation wherever you look, entropy drowning everything in sight, entertainment and technology and every four year old with a computer, everybody his own artist where the whole thing came from, the binary system and the computer where technology came from in the first place, you see?

Gaddis announces up front that the central subject of his work is “the collapse of everything, of meaning, of values, of art.” In other words, “entropy.”

He continues, “entertainment and technology and every four year old with a computer, everybody his own artist where the whole thing came from, the binary system and the computer where technology came from in the first place, you see?” Alluding to technology’s infantilization of everyone, Gaddis goes on to situate the origin of entertainment and technology as one and the same, emblematized by the player piano.

Gaddis venerates the Great Artists of history, who struggle through suffering to deliver something of enduring value and meaning to humanity. Technology, on the other hand, has its origins in parlor tricks, toys, the tinkering of scientists and inventors that results, at its most innocuous, in amusement, and at worst, in the dehumanization of the entire population. The key distinction can be located in the binary system, “the all-or-none machine,” which allows for only two outcomes, on or off, with none of the shades of grey in human decision-making. He associatively traces its origins from Charles Babbage’s Difference Engine, a calculator credited as the first mechanical computer in the 1820s, back to Joseph Marie Jacquard’s loom, a mechanized textile machine patented in 1804, itself a refined version of the first automated loom invented by Jacques de Vaucanson in 1745. Vaucanson’s loom was powered by a system of punch cards he had first used in his 1734 invention “The Flute Player,” a life-sized automata that played the flute, via a rotating drum pierced with holes diverting the airflow—the predecessor to the binary system.

“It’s the same principle Vaucanson used for his flutist, this drum pierced with holes and levers controlling its fingers and lip and tongue movements the air supply driven through the lips against the edge of the holes in the flute it was actually playing the notes selected by the holes in the drum, the notes selected by the holes in that roll of paper because the piano was the epidemic, it was the plague spreading across America a hundred years ago with its punched paper roll at the heart of the whole thing, of the frenzy of invention and mechanization and democracy and how to have art without the artist and automation, cybernetics you can see where the, damn!”

When Vaucanson presented The Flute Player at the Academie d’Sciences, it drew outrage from Johann Joachim Quantz, an accomplished flutist, for its “shrill tone.” So in some sense, the origins of the digital world we live in now, the underlying structure of our incredibly sophisticated big data AI models, emerges from a toy, a lifeless robot poorly simulating the act of artistic performance. Maybe the AI then, rather than bringing us into a brave new world, has actually brought us full-circle, closing the binary loop that opened some 300 years ago. That’s certainly how Gaddis would see it if he were alive today. This historical context is established in the early pages of Agapē Agape, and while the piano does continue to reappear throughout the text, Gaddis focuses next on existential consequences for art that the player piano represents.

Homo Ludens in the Age of Mechanical Reproduction

After tracing the history of the player piano, Gaddis stages an extended hypothetical dialogue between two titans of cultural history, John Huizinga and Walter Benjamin. Huizinga’s most famous work, Homo Ludens (also referenced by Flusser) suggests that play is a necessary condition of generating culture, and that in the 18th century the rise of art corresponded to the decline of religion; Benjamin’s most famous work, The Work of Art in the Age of Mechanical Reproduction, discusses, among other things, the way in which mechanical reproduction devalues the “aura” of an authentic original artwork, and shifts the social value of art form the private sphere of elite collectors to the public sphere of galleries and museums. The form of this conversation floats associatively between the authors’ voices, a back-and-forth of the major points of their work.

“God knows what it takes anyway my head’s splitting, falls right into line doesn’t it, collapse of authenticity collapse of religion collapse of values what Huizinga called one of the most important phases in the history of civilization, and Walter Benjamin picks it up in his Art in the Age of Mechanical Reproduction in the heap somewhere, the authentic work of art is based in ritual he says, and wait Mr. Benjamin, got to get in there the romantic mid-eighteenth century aesthetic pleasure in the worship of art was the privilege of the few. I was saying, Mr. Huizinga, that the authentic work of art had its base in ritual, and mass reproduction freed it from this parasitical dependence. Ah, quite so Mr. Benjamin quite so, turn of the century religion was losing its steam and art came in as its substitute would you say? Absolutely Mr. Huizinga, and I’d add that this massive technical reproduction of works of art could be manipulated, changed the way the masses looked at art and manipulated them. Inadvertently Mr. Benjamin, you might say that art now became public property, for the simply educated Mona Lisa and the Last Supper became calendar art to hang over the kitchen sink. Absolutely Mr. Huizinga, Paul Valéry saw it coming, visual and auditory images brought into homes from far away like water gas and electricity and finally, God help us all, the television. Positively Mr. Benjamin, with mechanization, advertising artworks made directly for a market what America’s all about. Always has been, Mr. Huizinga, Always has been, Mr. Benjamin. Everything becomes an item of commerce and the market names the price. And the price becomes the criterion for everything. Absolutely Mr. Huizinga! Authenticity’s wiped out when the uniqueness of every reality is overcome by the acceptance of its reproduction, so art is designed for its reproducability. Give them the choice, Mr. Benjamin, and the mass will always choose the fake. Choose the fake, Mr. Huizinga! Authenticity’s wiped out, it’s wiped out Mr. Benjamin. Wiped out, Mr. Huizinga. Choose the fake, Mr. Benjamin.”

Following this exchange, Gaddis includes a bit about Dolly the sheep, the first successful mammal clone (which occurred in 1996, two years before Gaddis completed Agapē Agape), leading into a passage that gets to the core of the implications of what is lost in this new universe of imitation:

“I’ve got to write this down before it’s lost, before it’s stolen, just to get the sequence right, what follows what, post hoc ergo this game you can’t win because that’s not why you play it trying to cultivate this whole swamp of chaos and chance, of paradox and perversity to wipe out the whole idea of cause and effect and, and, get my breath before i lose the, these belly-talkers and detached selves bred and cloned to be reproduced because that’s the heart of it, where the individual is lost, the unique is lost, where authenticity is lost not just authenticity but the whole concept of authenticity, that love for the beautiful before it’s created that that, it was Chesterton wasn’t it? that natural merging of created life in this creation in love that transcends it, a celebration of the love that created it they called agapē, that love feast in the early church, yes. That’s what’s lost, what you don’t find in these products of the imitative arts that are made for reproduction on a grand scale got to find some paper, piece of blank paper i’ve finally got the pencil now, now.”

What is lost is not just authenticity, but the concept of authenticity, of love—agapē—God’s love, love without reciprocation, a deep and profound sacrificial love. Agapē has been torn wide open, agape. In passages like this, the sincerity beneath Gaddis’s bitterness and nihilism and irony shines through. Those prickly layers of postmodern armor are a defense mechanism against the tenderness of the loss at the heart of his work, the loss of love. His intensity of feeling for this loss accounts for the seriousness with which he rails against it, the engine that drove his desire to continue to make art, even throughout a life where his talents were mostly under recognized and underappreciated.

Imitation of Life: Phantom Hands

It’s sensible that Gaddis, who started his career with a novel on art forgery, and ended it with a treatise on the player piano’s fabrication of artistic creation, speaks to the current conversation on AI-generated art. As AI art’s fidelity improves, its ability to mimic artists’ styles getting better and better, it marches closer to the fulfillment of his worst fears. Here the narrator quotes from advertisements on piano rolls:

“Here’s one, here’s what they wanted. ‘you can play better by roll than many who played by hand' you see? ‘and you can play all pieces while they can play but a few. and now even untrained persons can do it,’ breaks your heart.”

Sounds familiar. He goes on to describe the “piano player” that came before the player piano, an automata in the vein of Vaucanson’s flute player, that was advertised as being able to simulate the great masters, like Bach, Haydn, Beethoven, and Chopin.

“Here’s Debussy and Grieg giving a testimonial. ‘Many of the artists will never play again, but the phantom hands will live forever’ there that’s what it’s about, no more wooden fingers but phantom hands.”

Phantom hands becomes a recurring phrase, representing the trace of the performer whose work is being mechanically reproduced—the ghost of the artist as a living vital phenomenon resurrected, through binary necromancy, as entertainment.

“where did it go from participating even in these cockeyed embraces with Beethoven and Wagner and, and Hofmann and Grieg and these ghostly hands on the, what took it from entertaining to being entertained? From this phantom entertainer to this bleary stupefied pleasure seeking, what breaks your heart.”

AI is the ultimate set of phantom hands, the master of simulation that takes it beyond what Gaddis could have imagined. Not only can they play an existing song in the style of a master, but they can apply the master’s style to entirely new songs. All art is curation, yes but AI is trained on the collected images of human art history, an endless pool of influences to draw from that expands far beyond the limits of a single human memory. While this infinity of influences means that the AI can easily make new images that could have never been imagined previously, it’s far more common that AI artists default to well-trodden paths. It’s common practice that adding “in the style of [X artist]” will give images the patina of a canonical artist, living or dead, and there’s a shortlist of artists or styles that many AI artists draw from (Greg Rutowski, Moebius, HR Giger, trending on artstation, 8K unreal engine, etc). More directly related to the music metaphor, there are also dynamic performance models based around the human voice, like Holly Herndon’s Holly+, which will soon make any artist’s voice available for everyone to try on. As an inevitable result of its datasets source in the cumulative fruits of humanity’s labor, the ultimate job of AI is imitation.

But on some level, isn’t all art imitation? In the previous essay, I defined curation as the defining act in artmaking, and curation is all about synthesizing received influences, the same thing that humans and AI both do. And for all his blustering about authenticity, Gaddis’s own artistic practice frequently weaves imitation, plagiarism, and mimicry into its fabric, deconstructing the relationship between copy and original. The Recognitions in particular includes extensive unattributed quotes and allusions, everything from advertisements to Flemish art history to medical texts and all types of literature and poetry, the origins of which can be traced page by page on the website The Gaddis Annotations. In some ways, his novels could be considered proto-hypertext, collections of overdetermined word clusters pointing outwards to a decentralized bibliography in a thousand other directions. They’re novels of communication, of systems, networks, speaking in the language of their time, holding the realities of 20th century American society that Gaddis loathed up to a funhouse mirror, while simultaneously indulging in aspects of them, interpolating them into the form. This is all to say, he himself was a master of imitation, as a gesture to pay tribute to his influences, draw attention to his themes of mimicry and reproduction, and as a way to participate in the global dialogue of art.

Networks of Meaning

So much of aesthetic theory focuses on a work’s internal qualities—its form and content, its dialogue or characters or images or colors, structure, composition, style. But equally important is everything external to the work itself, the intertextual dialogues it bridges with art history and history in general—everything outside the boundaries of the frame. Tapping into this layer opens up a channel of infinitely forking paths. You never know when one word, one image, will instigate a decades long quest, or how far it will take you. Being assigned an article on the player piano, finding The Recognitions in the library, reading the word “agape” in an article online… The avenues that open up from chance encounters, weaving through works across time and space, are an essential side effect of engaging with art, and a core aspect of curation.

Ironically for a medium directly conjured from such a vast well of art history, AI art rarely has this contextual quality. Or at least, it’s harder to pinpoint. AI images, encountered across disparate social media channels, mostly feel disembodied from a singular, recognizable creator, or any wider context—creations of phantom hands. This could be attributed to a number of factors: the absence of spatial and temporal grounding, being created and displayed primarily online with few centralized galleries, the lack of an artists’ signature or built-in metadata, artists’ styles being drawn from without rather than within, the sheer volume of images and the speed at which they’re created. More often than not, the type of AI model being used is more recognizable than the style of an individual AI artist. These are all consequences of the emerging phase that AI art is still in, but none of them are inherent, and the potential to cultivate a richer context is not only there, but necessary, if AI art is to evolve beyond a novelty, a mere pair of phantom hands.

Arguably the most recognized and recognizable working figure in AI is Turkish American artist Refik Anadol, who is about to have his first solo exhibition at the MoMA, and the first major AI artwork in the MoMA. The piece emerged from an online collaboration with the MoMA last year, and is based on AI trained with over 138,000 images from the MoMA’s art archive. Though it’s trained on existing art, the work itself is highly abstracted, leaning into the hallucinatory uncanny quality that AIs can produce. In an interview with the MoMA, Anadol describes the way in which AI machines mirror the human capacity to synthesize influences and memories in a work:

“I think every single location in latent space resonates with how we perceive what happens in our lives. As humans, when we think about how we learn, and recollect our memories, or where our consciousness comes from, we realize that the machine’s process is not so different from our thinking or imagining. And I think machines help us define this powerful feeling of being able to touch and reconstruct memories.”

Anadol’s work is recognizable in part because his aim is not to recreate existing artworks or styles, even when drawing explicitly from one of the richest repositories of Western art. There’s a transfiguration happening, a dialogue with influences rather than just an imitation. When asked about the role of AI artists, he responds:

“For me, art reflects humanity’s capacity for imagination. And if I push my compass to the edge of imagination, I find myself well connected with the machines, with the archives, with knowledge, and the collective memories of humanity. So, like many creators, artists, designers, I feel like I am trying to find ways to connect memories with the future. And it’s not too hard to imagine, for example, that this specific data universe is the entire representation of every single data point that ever existed in the MoMA archive. And they’re all connected, based on their similarities, by the most cutting-edge, recent algorithms.”

This answer resonates with a forward-looking approach to AI, where the source of the AI’s data is a means to make something new, rather than towards simply recreating the past. Even as the publicly accessible models veer closer and closer to coherence and photorealism, the most successful AI art will likely be the pieces that draw out the more distinctive and uncanny qualities specific to the medium. With Anadol’s work soon to be on display at the MoMA, we’re clearly at a turning point for the canonization and distribution of AI art. It’s still uncertain how this will translate to other AI artists, and how future AI art will be cataloged, distributed, and valued.

Art Against the Algorithm

Gaddis’s career fixated on a body of themes that reached their apex in Agapē Agape, at the turn of the millennium. The despair he felt at the devaluation of art in favor of mindless disposable entertainment, or “content,” is easy to feel today, when those forces have only continued to accelerate exponentially. In the face of such a bleak situation, the question emerges: why is art still worth making? Filled with so much negativity, why did Gaddis himself continue to create? The work itself is the answer. It still matters. And it matters more than ever that we take seriously the idea of making art, that the elevation of anyone to the status of artist doesn’t mean the demotion of all art to “content” churned through the all-or-none machine, lost in a sea of indistinguishable data.

Despite his fight against technology, Gaddis’s intertextual approach validates that, on some level, art is data—nodes in a decentralized thought network created through dialogue over the course of thousands of years of history. We can acknowledge this lens, but at the same time, believe that encounters with art can be meaningful in a way that can’t be reduced to a mere exchange of information, and that there is still a case for humans over algorithms. With the proper effort from both artists and engineers, AI art can continue to push art history forward rather than becoming a closed loop of nostalgia, like a player piano roll on endless repeat. As long as there are humans, there will be a place for art, for agapē, for the feeling of discovery when you stumble across something that someone made and released—against all odds—and it opens up another world.

In recent months, AI text-to-image artworks have been flooding the internet, releasing a deluge of discourse around the role of artists in a rapidly changing world. I’ve been revisiting the 1985 book Into the Universe of Technical Images by Vilém Flusser, a philosopher and media theorist who I first encountered in the context of film, but turns out to be shockingly relevant to the current wave of AI image models and the questions they raise about creativity, art, and labor. A close reading of Flusser’s prophetic text can help to answer some of these questions, and to clarify the role of artists in the fast-approaching future.

Traditional Images, Technical Images

In broad strokes, Flusser’s account of cultural history can be summed up:

From traditional images (2D art images made by hand, like cave paintings),

to linear texts (i.e. written language works, like the Bible),

to technical images (images created mechanically by an apparatus, like a photograph).

Flusser lays this out using the model of a ladder as the historical progression between these steps:

“First rung: Animals and “primitive people are immersed in an animate world, a four-dimensional space-time continuum of animals and primitive peoples. It is the level of concrete experience.”

Here we can imagine the apes in 2001: A Space Odyssey, proto-humans fighting for survival in a new world.

“Second rung: The kinds of human beings that preceded us (approximately two million to forty thousand years ago) stood as subjects facing an objective situation, a three dimensional situation comprising graspable objects. This is the level of grasping and shaping, characterized by objects such as stone blades and carved figures.”

The apes have developed tools, giving them an edge over the beasts.

“Third rung: Homo sapiens sapiens slipped into an imaginary two-dimensional mediation zone between itself and its environment. This is the level of observation and imagining characterized by traditional pictures such as cave paintings.”

Some stability has set in, the origins of civilization, the dawn of human expression. Traditional images will reign for thousands of years. These are mythical images, in which the makers of images tried to “reduce their subjectivity to a minimum,” and stick to established aesthetic codes within a community. Cave paintings and tribal masks all share a similar form–the aim is maintaining a tradition, not originality. They reflect a mythic consciousness, based on a cyclical conception of time.

“Fourth rung: About four thousand years ago, another mediation zone, that of linear texts, was introduced between human beings and their images, a zone which human beings henceforth owe most of their insights. This is the level of understanding and explanation, the historical level. Linear texts, such as Homer and the Bible, are at this level.”

Linear texts mark the beginning of civilization and history as we know it, “for only then did imagination begin to serve (and oppose) conceptual thinking, and only then did image makers concern themselves with being original, with deliberately introducing new symbols, with generating new information.” The one-directional form instigated a linear conception of history–time progressing forward towards the future.

“Fifth rung: Texts have recently shown themselves to be inaccessible. They don’t permit any further pictorial mediation. They have become unclear. They collapse into particles that must be gathered up. This is the level of calculation and computation, the level of technical images.”

It’s the final, fifth rung where we find ourselves, where advances in science and math have scattered linear text into points, bits of abstract data, mathematical formulas and binary code. No longer flowing linearly, they form an interconnected web that scatters in all directions.

“In my terminology I say that before the invention of writing, people thought in a prehistoric way, and after the invention of the alphabet, historical consciousness was elaborated. And now we are in the process of elaborating a post-historical, structural way of thinking.”

AI text-to-image models bring the dialectical relationship between text and image into a full circle synthesis, with the input of text leading directly to the output of technical images.

“Images of our time are infected with texts; they visualize texts.”

Particles & Pixels

While traditional painted images have a cohesive physicality, technical images are ephemeral arrangements of particles: whether it’s the photons captured by the photosensitive chemicals or sensor of a camera, the pixels lighting up on a monitor, or the gaussian noise patterns that AI images emerge from, technical images are reorganizations of particulate data that can only ever achieve an illusory wholeness. They are always reducible to the particles that constitute them.

“With technical images, it is about first programming the computation of particles, then deprogramming them to convert them into informative situations. It is about a gesture that takes place in a particle universe, with fingertips touching keys, and the structure of this gesture is as particulate of the structure of the universe, that is, it consists of clear and distinct mini-gestures.”

Flusser here is referring to cameras and other technology of his era, but the same principle applies pretty directly to the way that an AI is trained on a dataset of images, creating a network of trained “particles” floating in the cloud that are then reassembled into new images via prompts. So we can see AI text-to-image models not as a brand new phenomenon, but the natural apogee of the technical image as particle reassemblage. The difference, as compared to other technical images like photographs, is not of kind, but degree–the scale of access, ease of use, and number of possible images (infinite).

“Producers of technical images, those who envision (photographers, cameramen, video makers), are literally at the end of history. And in the future, everyone will envision. Everyone will be able to use keys that will permit them, together, with everyone else, to synthesize images on the computer screen.”

AI image synthesis is still in its infancy, but it is advancing with remarkable speed, and we’re not far off from a future where everyone will be using AI to generate images, all the time. If that’s the future, or even the present, where can we situate the role of existing artists now and going forward?

The Myth of the Artist

“Images that can be telematically manipulated could give rise to an art that is still inconceivable, a pictorial dialogue infinitely richer than linear, historical dialogue could ever have been. Such a society, in dialogue through images, would be a society of artists. It would dialogically envision, in images, situations that have never been seen and could not be predicted.”

The imminent future, where anyone has the ability to generate high-quality images or videos with a few keystrokes, threatens the status of the “Artist” as a singular creator, an individual who has put in the hours to develop their technical skills and unique point of view. This existential disruption to the concept of the artist is at the center of controversy around text-to-image models. Dialogues around the implications of this shift are passionately unfolding right now online, but Flusser shows us that the shift was already well underway at the time of his writing nearly 40 years ago. Motivational structures of capitalism, technological progress, and the internet have been shifting the roles of artists and their livelihoods for decades, and AI art is just the latest manifestation of these same forces. But, being the inflection point of this epochal shift, it could actually provide an opening towards something genuinely revolutionary, rather than another bandaid on the shaky foundations of institutions crumbling beneath our feet.

“The social structure that is now appearing represents a synchronization of radiating images with the dispersed, lonely, depersonalized people who sit at the terminals of these rays. Revolutionary visualization tries to replace this structure with another in such a way that the images bring new interpersonal relationships into being and lead to new social configurations, the names of which remain unknown for now.”

Flusser is talking here about the current (in the 80s) state of centralized distribution channels broadcasting externally in one direction (i.e. TV, film, radio), giving way to something more decentralized and participatory. Writing on the cusp of the invention of the internet, he could already see that technical images would be the new means of communication, eventually superseding linear text. As we can see with emojis and memes, declining literacy rates, the rise of video platforms like Youtube and Tik Tok over written content, and the prominence of image-sharing platforms like Instagram, Tumblr, and Pinterest, its inarguable that the internet has brought the global populous into dialogue through technical images.

These new platforms have been disrupting long-held values about where and how meaning is found and shared in art, particularly for new tech-literate generations. The era of “Great Artists" with huge audiences has been declining for some time–the internet has fostered a culture of interactive, interconnected niche communities, rather than a one-directional monoculture. And the old institutions that have until now determined and regulated value have been slow to fully adapt. AI will accelerate this fragmentation and make the future even more granular–rather than a list of Great Artists peppered across a linear timeline, as in the 20th century, every moment will be exploding with Artists whose work resonates and cross-pollinates in all directions to smaller and smaller niches. With the models for how artists make money, how audiences spend money, and the role of outside organizations in a state of flux, the constant that remains is art itself, and its ability to generate meaning for practitioners and audiences.

“The central problem to be discussed with regard to a dialogic society is that of generating information. It is this problem that was called ‘creativity’ in former times. How do we get information that is unpredictable and improbable? It looks as though it suddenly appears from nowhere, as if it were a miracle. Hence the concept creation ex nihilo; hence the belief in a creator god; and once the veneration of creative people, above all so-called artists. The problem of generating information must be lifted out of this mythologizing concept to grasp the revolutionary possibilities of a telematic society, a true information society” (87)

Online, the question is often framed as an existential battle between traditional (albeit frequently digital) artists and AI users, humans and machines. It’s understandable, but Flusser allows for a more optimistic framing, reversing the existing paradigm of artists as isolated, singular individuals uniquely qualified to do creative work. Contesting this account of art history, Flusser proposes that abandoning this myth of the artist will liberate everyone to have access to the same power of creation previously limited to “Artists.” While we, artists, can lament the loss of our special status earned through decades of hard work, Flusser pushes us to focus on the flipside; the liberation of everyone gaining the ability to create images. Seeing it not an act of subtraction, erasing artists and their role in society, but of addition. To fully understand the stakes of this shift, we should first get to the bottom of what makes an Artist, and the core of what defines artmaking.

Artist As Curator

“For from now on, human freedom no longer consists in being able to shape the world to one’s own desires (apparatuses do this better) but to instruct (program) the apparatus as to the desired form and to stop (control) it when this form has been produced.”

More than any other medium, AI text-to-image synthesis lays bare the truth that all art is really curation. An artist absorbs influences from the world–from experiences, memories, other artworks–and synthesizes them into a new from via a medium.

Traditional images embody this translation of mental labor into physical form most apparently: when we look at a painting, we perceive the direct results of a human hand, we are apprehended with a record of concerted physical effort that occurred in a specific time and place. Formal choices are immediately apparent in the final image.

Technical images like photographs also reveal the hand of the artist, but more abstractly and indirectly: when we look at a photograph, we perceive a vision where the human hand is absent, because the mediums of technical images are black boxes, where programs happen invisibly. Rather than brush strokes applied over time, a photograph is created in an instant, so the human touch is felt retroactively in every choice made by the artist/operator leading up to the moment they press the shutter button–at which point the program inside the black box of the camera runs automatically.

AI images work in the same way–an operator makes choices in the construction of their prompt, and an image is produced automatically with a final keystroke. The technical mechanics are different than photography, but the process is structurally the same–the sum total of human decisions leading up to the moment that a program is run in a black box.

We can define these decisions in the creation of images as curation–the conscious narrowing of infinite choices towards an intentional desired outcome. Curation happens simultaneously in both the artist and their medium, and both feed off each other. An artist curates their inner self to shape the moment of creation, and the medium’s intrinsic qualities curate the qualities of the image that is produced. An artist limits themselves by their choice of medium, and the medium limits them in turn by its constraints. Art is what emerges in the space where these two levels of curation harmonize.

Obviously, on a certain level we’re all artists on a daily basis every time we idly doodle or snap a picture with our phones. But what we generally recognize as an “Artist” is someone who has, with time and practice, curated a depth in both their inner self and their medium of choice to produce a special harmonious result that generates meaning and emotion in an audience. The most significant artists will do so in a way that is unique to them, curating outcomes that could recognizably only come from them.

The barrier to entry is not identical across mediums–anyone picking up a camera for the first time could quickly press a button to take a photo, but not anyone could pick up a violin and produce a sound, let alone a song. In either case, the endgame of Artistry is never guaranteed by mastery of a medium, because achieving harmony relies on a self that is curated to meet it. Someone studying photography for decades could still lack the inner depth to produce images that resonate with an audience, or, someone who has only a basic grasp of the violin could produce music that brings an entire concert hall to tears.

Every medium has a different resonance, a different ratio of the depth they require a person to bring to them to reach harmony. AI text-to-image art has perhaps the smallest gap between effort and quality of outcome, essentially putting the mastery of medium at your fingertips–but in turn, by cheapening the time to curate and produce, it raises the bar for the depth of self required to express something meaningful that can stand out among millions of other images. As with most art, the vast majority of AI images right now are uninspired, derivative, and repetitive. But there are exceptions, and they have the potential for much more.

The constant dance between the curation of self, medium, and harmony–and their relativity across time and culture–is what makes Art elusive, exciting, and difficult. It’s this intangibility and rarity that has led to the veneration of Artists, the few who truly find meaningful harmony–the group which Flusser says will disappear. And on some level it may–but it won’t be by eliminating Artists. It will be via the rise of a medium that, by its intuitive and unlimited possibilities, expands the potential of everyone to be Artists, if they choose to rise to the occasion. From the standpoint of art as a creative practice, this is a great thing. But it does raise questions for those who hope to make a living on their arts, potentially giving a competitive playing field a lot more players.

Art and Capital

The definition of Artist gets murkier when we consider artists working within the context of capitalism; because art is not strictly speaking “useful,” there has always been an uneasy partnership between artists and capital, and most working artists’ labor is notoriously undervalued. We should acknowledge first that the system is already broken–AI isn’t barreling in out of nowhere to wreck a well-oiled machine that’s working great for everyone. The role of the working artist is always precarious when it comes to technological change–for example, the rise of CGI artists in Hollywood replacing practical special effects artists, matte painters, and stop motion animators. But while it’s true that some jobs were lost, it’s not so black and white–many of the artists of the old ways either adapted to the new technologies, or were able to leverage their experience to supervisory roles: their hard-earned knowledge, their depth of curation, still translated to the new context. Since the beginning, art has never been a stable career–change is the only constant.

There are a myriad of ways that people make money from artistic work, and there’s no one-size fits all answer to how AI will affect different industries. Many have argued that AI art in its current form directly threatens some categories of artistic jobs, like concept artists or storyboard artists–but businesses trying to save money are already outsourcing these roles overseas on Fiverr or cutting corners in other ways, and companies that care about art and people will find a way to use them. A human artist could become a luxury for the businesses who can afford it, an in-house prompt artist could soon be a job title, or the situation could remain relatively similar with existing artists just incorporating AI into their workflow as a tool for ideating or iterating. The exact mechanics of how things will shake out is unclear, but it’s certainly not going to be the case that overnight, trained artists can no longer find work because every company starts using AI for everything.

Whether it’s working artists trying to make a living, fine artists exhibiting in galleries, or hobbyists making art for its own sake, the core function of art remains; but wherever implemented, AI will make the speed at which it is produced and shared exponentially faster, making art more social, less precious and scarce, and endlessly modifiable. It’s the full transition from the one-way model of centralized broadcasting distribution channels, to a decentralized social network of communication, and plugging in is less like the lonely introspection of past artistic production, and more like playing a game.

Art as Play: The Rise of Homo Ludens

“The person of the future, playing at the keyboard, will be ecstatic about the creation of durable information that is nevertheless constantly available for a new synthesis… the person of the future will be absorbed in the creative process to the point of self-forgetfulness. He will rise up to play with others by means of the apparatuses. It is therefore wrong to see this forgetting of self in play as a loss of self. On the contrary, the future being will find himself, substantiate himself, through play. The “I” that eidetic reduction (and neuropsychological, psychological, and informatics analysis) has shown to be an abstract concept, to be nothing, will be realized for the first time through creative play.”

When speculating on the future of technical images, a word Flusser returns to again and again is “play.” Anyone who has practiced art can recognize that in its purest form, creativity is play–an unselfconscious, childlike engagement with the unlimited possibilities of a medium, rearranging elements to make something new. Anyone who has used a text-to-image model can recognize the same feeling, amplified and distilled. The models offer a striking balance of simplicity and depth, offering a much more instant gratification than other mediums. It’s incredibly fun, putting the play at the forefront and minimizing the frustration that often plagues the creative process. In combination with the networking possibilities of the internet, Flusser’s vision of the future is a collective ego-death where the social play of image-making supersedes the outdated isolated process of artistic creation.

“Future images will be at a high level because they will owe their production to the dialectic between theory embedded in the apparatuses and the intuitive hallucinatory power of the envisions… creation there will not be limited to a few ‘great people’ who produce informative works empirically by means of a lonely inner dialogue. .. instead everyone will participate in the creative process and test their intuitions and inspirations against the theories embodied in apparatuses, of whose riches we as yet have no inkling.”

It’s important that play is often social. Most people have had fun with the interactive nature of social media in some form or another. The way that many artists reach an audience now is by distributing their content on social media channels among various communities. The groupchat structure of Midjourney’s Discord channel is a fairly literal example of Flusser’s vision, with users sharing and interacting with each others’ images in real time. The web app DALL-E Mini (renamed Craiyon) had a viral moment on social media with users sharing 3x3 grids of low-quality generated images, often meme-y or pop-culture related. It was a perfect example of the kind of dialogic exchanges that Flusser predicted. The authors anonymous, anyone could participate in the play of creating and sharing images, and the stakes weren’t monetary. It was purely for fun.

Looking closer at the success of Craiyon, it’s important to understand why these images in particular resonated with such a wide audience, despite being “lower quality” than many of the shinier, cleaner DALL-E or Midjourney images that often just gather dust. The technical limitations of Craiyon force more emphasis on the curation of the self to create harmony. The lo-fi, uncanny nature of the images makes them better suited to funny prompts, as a vehicle for personality rather than beauty. Craiyon’s meme-like images prove that it’s not solely the technical mastery of images that makes them resonate, and sometimes it’s the opposite.

Flusser underscores the revolutionary potential of collaboration for creativity with the example of chess:

“The inner dialogue that was once so exciting can easily be simulated with chess. One sits alone at the board and alternatively moves the white and black pieces. Interesting, informative situations can result. But as soon as a second player joins in, it immediately becomes clear how limited the initial situation… under pretelematic conditions, including the present, the singular way of playing was responsible for almost all information (scientific, philosophical, artistic, or political). Telematics, on the other hand, will involve very many players in the game, and the playing competence will expand exponentially. All the information generated until now by great individuals (our entire cultural inheritance) will be regarded as relatively space in the future. Compared to synthetically produced information of the future, compared above all with future images, the culture of the past will appear as a mere starting point. It will become clear that a systematic, conscious creativity really begins with telematics.”

It’s in some ways a utopian vision, this transition of humankind towards homo ludens, (man the playful) but Flusser is clear that the utopian outcome is not guaranteed; he acknowledges the potential downsides of such a future, the anxieties of homogeneity and stagnation that are already front and center in the current state of culture. AI has the potential to either break open this stagnation, or be its ultimate fulfillment.

The Future of Images

“In their current first phase, technical images can still constantly renew themselves by feeding on history. But history is about to dry up, and this exactly because images are feeding on it, because they sit on historical threads like parasites, recoding them into circles. As soon as these circles are closed, the interaction between image and person will, in fact, become a closed feedback loop. Images will always show the same thing, and people will always want to see the same thing…”

Flusser’s thesis offers two possible outcomes from technical images–one utopian and one dystopian. We’re already in the midst of the dystopian descent into the closed feedback loop, and have been for a while; just look at the current state of Hollywood as an endless recycling of nostalgic franchise IPs, or streaming sites as data-driven, algorithmically-curated “content.” AI can either be the final nail in the coffin, or a way out.

“The current interaction between images and human beings will lead to a loss of historical consciousness in those who receive the images and, as a result, also to a loss of any historical action that could result from the reception of the images. But this current interaction is not yet leading to the development of a new consciousness, unless it changes radically, unless the feedback is interrupted and images begin to mediate between people. Such a rupture of the magical circle between image and person is the task we face, and this rupture is not only technically but above all existentially possible.”

Flusser’s prognosis evokes the stultifying content fatigue that we already face on the internet, which is about to escalate with the inevitable flood of AI content on the horizon. We risk drowning in it, but at the current inflection point, we still have the chance to steer things in a more utopian direction if we can see the situation clearly. The key is to see AI images not as a historical aberration, but an inevitable step in the revolution brought about by technical images. From this perspective, there is an opening for artists to embrace these new tools, push their unlimited potential, and break free from creative stagnation through new ways of envisioning, new ways of collectively imagining value and meaning.

A Spiritual Revolution

“Every revolution, be it political, economic, social, or aesthetic, is in the last analysis a technical revolution. If you look at the big revolution through which mankind has gone, let’s say the Neolithic Revolution or the revolution of Bronze Age, or the Iron Age, or the Industrial Revolution, every revolution is, in fact, a technical revolution. So is the present one.

But there is one difference. So far, techniques have always simulated the body. For the first time, our new techniques simulate the nervous system. So that this is for the first time, a really, if you want to say so, a really immaterial, and to use an older term, spiritual revolution”

-Vilem Flusser, 1988 interview

We are standing on the cusp of a revolution, a position that is by its nature difficult to predict. Such uncertainty requires tools for navigation, and I hope this brief introduction to Vilém Flusser can serve as a compass to help guide the conversation by adding some broader theoretical context to the AI art discourse, and offer an antidote to the despair and disdain so many artists are projecting onto the situation. We need to look head-on at the reality that AI art is here, and it’s not going away–but that doesn’t mean “art is dead.” Art has survived millennia of technical advances, and is arguably most vital and alive in these moments of transition. As long as there are humans there will be art. If we are to trust Flusser (and I think we should, given the accuracy of his vision), then we can be optimistic for the future of art and AI’s role in it, if we approach it with the right mix of thoughtfulness, caution, and hope.