In a plastic-chaired restaurant in Manhattan’s Chinatown, huddled over the lukewarm remains of mu shu pork and chen cun fen, I watched my friend type the words the Queen’s corgis, in the style of Monet into his smartphone. There was something of an awkward pause while the machine got to work – maybe 15 or 20 seconds of something happening in the background; in the phone; or most likely, rather, in the cloud. I picked at my food with cheap chopsticks while an elderly couple bickered in Mandarin beside us, and while, on the screen, a progress bar marched inexorably to 100%.
Soon the something was done: the phone having transmitted the request via a series of local way-stations to a satellite orbiting somewhere above us, before then being sent on to one piece of a decentralized ecosystem of computing power, that piece located perhaps in Columbus, Ohio, or in Jakarta, Indonesia, or in Hyderabad, India – at which point a relatively new innovation in machine learning called DALL-E took the sentence the Queen’s corgis, in the style of Monet and did said something, before sending the result of said something back, such that my friend, triumphant, turned his phone to show three impressionistic, happy little guys, panting in the sun at Giverny.
The piece of art today known as “The Queen’s Corgis, in the Style of Monet” came into the world only weeks after another AI-generated image made many people angry for winning an art division at the Colorado State Fair. Jason M. Allen used Midjourney, a similar machine learning image generator, to create an image he called “Théâtre D’opéra Spatial”, and took first place in the Fair’s “digital art/digitally manipulated photography” division. In that Chinatown restaurant, “The Queen’s Corgis” had produced wry smiles and momentary delight, but in the days following the Colorado State Fair “Théâtre D’opéra Spatial” got a different response.
When the public learned that the winning entry was created by entering a few words into a computer, there was outrage. Some of that fury, according to the New York Times, came from people who felt deceived; who felt that Allen had cheated. Others felt Midjourney’s processes constituted plagiarism, or that redundancy for artists was on its way. Take this tweet from digital artist RJ Palmer: “What makes this AI different is that it’s explicitly trained on current working artists… This thing wants our jobs, it’s actively anti-artist.”
Palmer was reacting to the raw fact that an AI image generator can today take just about any snippet of text and turn it into an image that would take a human days, months, even years to create. Even more problematic: the generator can do that only because it has been fed art made by people – people who never had the possibility of informed consent. Such concerns have even triggered lawsuits. AI image-generation platforms including Midjourney, Stable Diffusion, and DreamUp are all currently facing class action cases. And these developments aren’t limited to images, either. Put any bit of text into the AI generator Chat GPT-3 and the program will write for you the rest of the paragraph. Open AI’s Jukebox will create for you a full music sample, vocals included, if you just give it a genre, artist, and some lyrics to use as a guide.
There are no shortage of examples, and there is no shortage of panic. I, for one, am convinced: by the end of the decade, or perhaps the end of this year, computers will be able to produce novels of equal quality, and in the style of, every writer from Murasaki Shikibu to Sally Rooney. Want to know what film Kubrick would have made had he been alive today? AI will soon generate a believable approximation. We should expect that AI will soon generate the vast majority of television scripts. Pop music, too. In every field of human expression we should not ask if AI will replicate human capability, but when.
These fears of plagiarism and redundancy mask a deeper, inchoate, but still more profound anxiety about the future. This is not merely the fear that artists are being cheated, or that they won’t be able to find jobs (that process is (has always been) well underway). This is the fear that the acceptance of AI-generated art represents the end of art itself. In a world where a machine can produce anything, and it can do so instantaneously, our conceptions of value and appraisal begin to disintegrate. “An article written by GPT-4 that is indistinguishable from one written by a writer at The New Yorker is of equal value” one friend told me a few months ago. He describes himself as “a reductionist about value”, and thinks, at the end of the day, everyone else is, too.
He is hardly alone. For many, art has simply reached its end. Some even seemed reconciled to it. Defending his actions to the New York Times, Jason M. Allen was almost glib. “This isn’t going to stop. Art is dead, dude. It’s over. A.I. won. Humans lost.”
Nearly a decade ago, as an undergraduate on the far side of this earth, I wrote a paper about a spam Twitter account called Horse_ebooks. Horse_ebooks was a spambot created to sell ebooks about horses, and, also, was everyone’s favorite Twitter account in the early 2010s. To achieve its goal, the bot trawled the internet, copied random bits of text, and tweeted them. Occasionally a tweet would include a link to an ebook, and some poor soul, hoping to learn about horses, might click on it.
If that was all Horse_ebooks was, I wouldn’t be writing about it again a decade later. The internet was, and is, full of spambots. But Horse_ebooks was a particularly bad spambot. The text it tweeted out was entirely random, mostly fragmentary, and often close to gibberish. It was also beautiful. One such tweet is now an iconic artifact of the internet. “Everything”, Horse_ebooks lamented on June 28, 2012, “happens so much”.
And so it was that for a number of years a corner of the internet fell in love with the poetry of a poorly written algorithm. I wrote then:
Equal parts cultural phenomenon, artwork and historical text, Horse_ebooks has the potential to radically reshape our assumptions of textuality, meaning and authorship, and is interesting as a text precisely because of its context-specific, avant-garde nature. It is to be expected, then, that in twenty years both Horse_ebooks and any analysis of it will be found uninteresting, in the same way that the digital scholarship of the late twentieth century lacks the shock value it once demanded.
The artificial equus felt, at first glance, like a vindication of Barthes’ Death of the Author, published more than 40 years earlier. Barthes wanted to free criticism from a suffocating over-reliance on authorial intent and biography, and to give up the search for the “ultimate meaning” of the text.
‘The modern writer’, Barthes wrote, ‘is born simultaneously with his text; he is in no way supplied with a being which precedes or transcends his writing, he is in no way the subject of which his book is the predicate.’ Modernity demands something like the birth of the reader, Barthes wants to say, a time in which we turn over to the reader the freedom and responsibility of the creation of meaning. By rejecting the notion of an “ultimate meaning”, we open ourselves to the many interpretive possibilities suggested by a text. The reader is not there to decipher, but to explore, to create.
So Horse_ebooks felt something like the apotheosis of literary postmodernism. We had a text, but no author, and without an author, no one bothered themselves with a search for Horse_ebooks’ “ultimate meaning”.
And yet, readers found meaning all the same. For those of us on Twitter at the time, the experience felt both liberating and communal. We were exhilarated by the solidarity of the thing. We collectively and organically came together to declare something random, something accidental, something emergent, art. And the whole episode happened without a dispute as to what was “really going on”. There are no debates to be had about authorial intent when you know the author doesn’t exist.
Instead something like a collective authorship emerged. Horse_ebooks’ author was us. It was like joining a crowd of people who have turned to watch the sunset – or like stumbling upon a group of birders captivated by an owl in Central Park. One quietly and curiously looks in the same direction, sees the beautiful creature, and smiles.
In 2011, Jacob Bakkila, a then-employee of Buzzfeed, managed to secretly get hold of Alexei Kouznestov, a 30-something Russian web developer and the programmer behind Horse_ebooks, and bought the beautiful dilapidated spambot off him.
Once he had access, Bakkila turned off the machine. Then, he began tweeting in the voice of Horse_ebooks. He told no one, and no one noticed. In the end he kept it up for two years: logging on each day, at all hours of the morning, venturing onto the internet, copying random pieces of text, and putting them out into the world. He could easily have automated the job for himself, all over again. He didn’t. As he told Susan Orlean in 2014, ‘I wanted to preserve the integrity of bespoke spam.’
‘Many people,’ wrote Orlean, ‘were angry not only because a favorite Twitter account was ending but also because they felt cheated that Horse_ebooks wasn’t what it had seemed.’ In Horse_ebooks there had been no need for an author. Now, suddenly, we had one – whether we wanted one or not.
What sort of art do we want from this world, from this time? What sort of piece would bring us back to each other? What sort of experience could return us to the old days, the days when we could imagine something different?
In 2015, a poem titled ‘The Bees, the Flowers, Jesus, Ancient Tigers, Poseidon, Adam and Eve,’ was selected for The Best American Poetry after it had been published by the Prairie Schooner. The poem had been submitted under the name Yi-Fen Chou, but, when informed of the poem’s acceptance, a man named Michael Derrick Hudson wrote to the editor of Best American Poetry to explain that Yi-Fen Chou was a pseudonym. Explaining his reasoning, Hudson wrote:
After a poem of mine has been rejected a multitude of times under my real name, I put Yi-Fen's name on it and send it out again. As a strategy for "placing" poems this has been quite successful for me. The poem in question ... was rejected under my real name forty (40) times before I sent it out as Yi-Fen Chou (I keep detailed submission records). As Yi-Fen the poem was rejected nine (9) times before Prairie Schooner took it. If indeed this is one of the best American poems of 2015, it took quite a bit of effort to get it into print, but I'm nothing if not persistent.
Hudson’s actions, like Allen and Bakkila before him, provoked fury. Many considered the tactic fraudulent; others found it racist. Save for a handful of reactive elements, who thought Hudson justified in fighting back against a culture of affirmative action, nearly everyone was angry.
The section you just read was written by ChatGPT 3.
‘In the gimmick,’ says Sianne Ngai in Theory of the Gimmick, ‘our spontaneously affective, explicitly aesthetic appraisal of an object’s form as unsatisfyingly compromised triggers and comes to overlap with economic and ethical evaluations of it as cheap and fraudulent.’ The gimmick works too hard and too little. It demands our attention and cheats to get it. It undermines our expectations of what effort, what labor, ought to go into a thing. When we discover how cheaply our attention can be bought, we respond with irritation, attraction, ambivalence. It is, Ngai says, capitalism’s most successful aesthetic form.
Call it The Reveal. The audience has one experience with a piece of art. Then, the author reveals, or is revealed, to be contrary to expectations. He is male, rather than female, or man, rather than machine, or machine, rather than man. His race may have changed. He may be straight, rather than gay. Perhaps he is a Palestinian, rather than an Israeli. We experience a bait and switch. It is a trick one can play with any axis of identification relevant to the audience’s experience with a text. The Reveal need not be a gimmick. And sometimes it is.
The trick only works because, despite Barthes’ insistence that we kill off the author, the author can never die. In our construction of meaning the origins of a work are inescapably relevant – we couldn’t ignore them if we wanted to. Remember: the next Lear will be written by AI.
Authentic Artists, a start-up that uses AI to produce music, creates fictional characters to stand in as artists. To be attractive to listeners, AI-generated music needs a face, Authentic Artists’ founder Chris McGarry told Kyle Chayka. “We wanted to answer the question, what is the source of the music? A semiconductor or cloud-based server or ones and zeros didn’t seem to be a terribly interesting answer to that question,” he said.
In a previous life I worked for a startup. We worked outside of an immense, extravagant New York City co-working space. Around us companies designed electric motorcycles; indoor food farms; photovoltaic cells. Across from us, a team of men built a robot that could make coffee.
The robot consisted of a long metal arm and an espresso machine. To test the robot, the men let us have free coffee. Every morning I would walk to the machine, enter my order on an iPad, and watch as this extraordinary piece of technology picked up a cup, placed it under the espresso spout, pulled a shot, frothed some milk, and returned to me a perfectly serviceable flat white.
The gimmick dominates. The things that demand our attention feel ever cheaper – too easily made, too easily consumed. We are seduced; we are turned off.
Because the gimmick is everywhere, we see the gimmick everywhere. Also: when I told you the second section was written by ChatGPT3, I lied.
Not long after the short story “I Sexually Identify as an Attack Helicopter” was published in 2021 by Clarkesworld, a science fiction and fantasy magazine, people started to get upset. Readers took the story, which made reference to a homophobic meme, to be a mockery of the experience of trans people. Later that month, under the weight of fierce online backlash, the author self-admitted to a psychiatric ward for thoughts of suicide and self harm.
The title of the story derives from a meme that emerged about the same time as Horse_ebooks. In 2014 the phrase “I Sexually Identify as an Attack Helicopter” was copypasta (a piece of text that proliferates across the internet) designed to parody posts by trans people. The short story, published years later, takes that copypasta meme seriously, depicting a world in which a narrator’s gender identity is reassigned to that of “attack helicopter” in order to make her a better pilot. Signing off the story was a single detail of authorial biography: ‘Isabel Fall was born in 1988’. With so little information about the author available, readers began to worry that Isabel Fall was not who she said she was.
‘To give an Author to a text,’ wrote Barthes, ‘is to impose upon that text a stop clause, to furnish it with a final signification, to close the writing….once the Author is discovered, the text is “explained”.’
In The Wizard of Oz, Dorothy and her merry band are overawed by the magnificent Wizard – a massive floating face with a booming voice. It is only when the lively dog, Toto, pulls back the curtain that the Wizard is revealed to be an ordinary man manipulating the face, and them all, via a series of levers and pulleys.
The readers of “I Sexually Identify as an Attack Helicopter” were desperate to pull back the curtain on Isabel Fall. Many were sure they knew what they would find: that Fall was a cis man conducting some kind of literary troll. Without concrete biographical knowledge of the author, readers began to look for evidence in the text. And they found, according to their own standards, ample material to adjudicate on Fall’s gender. One common criticism, noted Emily St. John while reporting for Vox: ‘Fall must be a cis man, because no woman would ever write in the way she did.’
After a month of free coffee, I found myself back to my old routine. Each morning I would walk past the robotic arm, out the door, and across the street. There, a bored university student in an apron would scald my milk and charge me five dollars. I forgot about the robot.
Isabel Fall was not a cis man; she was a trans woman, and the claim that she could not possibly be a woman triggered profound dysphoria. ‘I don’t know what I meant to do as Isabel,’ she told St. John. ‘I know [that publishing “Attack Helicopter”] was an important test for myself, sort of a peer review of my own womaness. I think I tried to open a door and it was closed from the other side because I did not look the right shape to pass through it.’
We live on edge. Eve Sedgwick calls it paranoia: ‘In a world where no one need be delusional to find evidence of systematic oppression, to theorize out of anything but a paranoid critical stance has come to seem naive, pious, or complaisant,’ she says. ‘[B]ut it seems to me a great loss,’ she continues, ‘when paranoid inquiry comes to seem entirely coextensive with critical theoretical inquiry rather than being viewed as one kind of cognitive/affective theoretical practice among other, alternative kinds.’
Do I need to tell you that the democratizing surge of the internet is one side of a coin? That the tide that sweeps away hierarchy and authority leaves us without hierarchy; authority? In place of undemocratic order we have, at least for a time, polyphonic chaos. The old masters are dead. All that is solid melts into air. The concrete is now fragmented; the coherent now inchoate. The disorder is dizzying. You feel panic, distrust, defensiveness, doubt. Too much is coming at you. You are holding on for dear life. You are holding on for dear life. You are holding on for dear life and yet the ride is not slowing down. You are holding on for dear life and yet the ride goes ever faster. You are holding on for dear life and also the world itself is accelerating.
Once we had Modernism. Then we said we were past, Post.
We live today in Internet Modernism.
No wonder we have this, our defensive crouch, our anticipatory stance that seeks to ward off the possibility of surprise. Having discovered the man behind the curtain (the man behind the machine? the (white) man behind the poem?) we never want to be shocked again. So we live inside an approach to art, to reading, to life, that anticipates, expects, pursues bad vibes. The only positive affect paranoia ever seeks, as the theorist Silvan Tompkins puts it, ‘is the shield for which it promises against humiliation.’ Expect the worst and never be disappointed.
AI can only produce kitsch. Here I lean on Clement Greenberg’s definition: the mass produced simulacra of high art that can only exist through the advances of industrialisation. Kitsch replaces and obliterates folk culture. It is a homogenizing force – turning all of modern society into the one proletarian blob. It is an inculcating form that leads the subject into a pliant relationship with capital. It is the product of Adorno and Horkheimer’s culture industry.
What it lacks, therefore, is what Adorno wanted from art: the ability to burst through the haze of capitalist hegemony and make society’s contradictions unignorable. An author is necessary. For the reader to feel the piece work upon them in this way, they need either to feel the intentionality of the work or, as in the case of Horse_ebooks, to impose a collective intentionality upon it. And I think that AI – by standing in for the author – denies the reader that possibility. All that remains is kitsch; gimmick; artifact.
Behind every artwork we want a hero or a villain. But when our expectations turn out to be wrong, we are thrown back upon our own power as readers – and that is a position of responsibility we are unready to bear. Reliant on the identity of the author in our interpretation of the text, we suddenly find ourselves floundering to make sense of the work in its new authorial form. Instead we kick back, and claim to have been betrayed by the artists – as though some unspoken, pre-existing compact were in play.
Play! What could be play is treated as peril. What Bakkila did had joy in it, too.
In a now-viral video, the master director of anime films Hayao Miyazaki attends the Dwango Artificial Intelligence Laboratory in Tokyo. It is 2016, and the staff show him a rudimentary AI video graphic generator designed to replicate human movement. The AI is not yet successful, so the designers, for whom Miyazaki is a hero, exhibit a kind of body-horror video. Their AI has generated an uncanny zombie figure who limps across the screen in a manner no human body could reproduce. They look to Miyazaki for his approval. It does not come.
Instead: ‘I strongly feel that this is an insult to life itself.’
Later, he is heard to mutter, ‘I feel that we are nearing the end times.’
What so many of us want, have wanted, is kitsch.
(What we are made* to want / what we think* we want / what we are made* to think* we want).
The reparative view is our alternative, says Sedgwick. It requires hope. It is enabled by having the ‘room to realize the future may be different from the present.’ But, without art that helps us break free from the ideological shackles of modern capitalism, we lack such room. We instead find ourselves unable to conceptualize forms of life beyond the narrow one we inhabit; we have lost the ability to imagine new futures.
It is not merely that capitalism is today the sole organizing principle on the planet. It is not merely that the gimmick dominates. The situation is much worse than that. We are the doleful inhabitants of a suffocated culture. We have no alternative but to live inside a ‘pervasive atmosphere… acting as a kind of invisible barrier constraining thought and action.’
That’s what Mark Fisher wrote in Capitalist Realism, anyway. Fisher wants to say it’s all material: that neoliberalism stripped the arts of state funding; that art became exposed to the market; that marketised art is mere commodity; that commoditized art can only look backwards, never forwards. Maybe. Either way the starting claim is true. We live in a nostalgia trip. Remixed and recycled. Dead CGI Jedi, revitalized 1950’s comic book heroes. We see zombies everywhere. Even our zombies, as in Dawn of the Dead (2004), are brought back to life for our consumption. Our lives today are all past, no future. We can barely imagine things ever being different. We cannot escape our paranoia.
In capitalism we want a man behind the curtain. We want a man pulling the strings. We want a villain to blame. But when we pull back the curtain we find no one standing there. Only history.
Sitting in that Chinese restaurant, my friend passed around his phone so that each of us could inspect “The Queen’s Corgis”. “Nice,” said each of us, in turn, before ordering more rice for the table.
There probably will be a future.