Smooth Talking

Mathew Omelesky, "The Painter and the Chatbot: Artificial Intelligence and the Perils of Progress"

We should be losing more sleep over the parlous state of organic intelligence than over the advent of artificial intelligence.

Some four hundred thousand visitors pass through the wrought iron gates of the Mauritshuis museum in The Hague each year, most of them, we may safely presume, with the intention of viewing the institution’s most prized possession: Johannes Vermeer’s Meisje met de parel, or Girl With a Pearl Earring. The seventeenth-century painting, one of the crown jewels of the Dutch Golden Age, hangs against a green-papered wall in Room Fifteen, invariably surrounded by a swarm of museum-goers, attracted to the work like houseflies to a honey pot. A typical viewer will find a suitable vantage point and pause for a few moments, registering the girl’s exotic turban and the famous dangling drop pearl, so large that surely it must have been an imitation, forged in Venice out of powdered glass, silver, and egg whites. More noteworthy still is the subject’s expression, suspended somewhere between surprise, pleasure, and mounting alarm, an enigmatic visage surpassing even that of La Gioconda. Take it all in, maybe snap a picture — no flash, please — and then move on to the gift shop or the Brasserie Mauritshuis.

Those with more patience, or sharper elbows, will endeavor to get closer to the eighteen-by-fifteen-inch painting, and the time and effort will be repaid with a greater depth of understanding of Vermeer’s masterpiece. Now coming face to face with the anonymous sitter, the visitor can better appreciate the obsessive attention to detail that made Johannes Vermeer unique in the annals of European art history. Witness the infinite recess of the dark background, produced by a layer of bone black and charcoal black, and another layer of weld, chalk, red ochre, and indigo, further treated with a transparent glaze of green paint. Witness the dabs of vermilion and carmine on the girl’s glistening, parted lips, and the moistness of her doe eyes. Witness the broad, confident brushstrokes evident in the winding cloth of her ultramarine turban and the heavy folds of her yellow cape. Lean in even more, coming as near as gallery attendants and vibration sensors will allow, and you can spot the minuscule patch of lead white impasto on the renowned pearl, the result of a single virtuosic flick of Vermeer’s wrist in 1665, reflecting the same band of light that rakes across the sitter’s forehead, moistened lips, and golden scarf.

Vermeer’s Girl With a Pearl Earring, with her iconic, inscrutable, dreamlike gaze, has long attracted crowds and has inspired art historians, novelists, and filmmakers alike, but in recent months she has garnered a different kind of attention. In October 2022, climate protesters affiliated with the Just Stop Oil Campaign doused the painting with tomato soup, while another activist attempted to glue his head to its protective glass — puerile and potentially destructive stunts that resulted in several entirely justified arrests for “public violence against goods.” A few months later, the work was loaned out to a Vermeer exhibition at the Amsterdam Rijksmuseum, leaving a yawning Girl With a Pearl Earring–shaped hole in the Mauritshuis. To fill the gap, the curators put out a call for temporary replacements in the form of a “create your own girl” competition, and the response was enthusiastic, with 3,482 entries submitted by the general public, including paintings, sculptures, photographs, crochet pieces, and mixed-media works. The majority of the #mygirlwithapearl submissions were executed with tongue firmly planted in cheek — a stalk of corn with a pearl dangling from one of its kernels; Vermeer’s girl replaced with a cat, a rabbit, or an oyster; a reproduction of the original emblazoned with a Barbie logo; and so on. A jury of judges selected five of those works to take the place of Vermeer’s original in the museum’s second-floor gallery. Given pride of place, in the central position, was Julian van Dieken’s A Girl With Glowing Earrings.

What began as an innocent attempt to pass the time while the star of the Mauritshuis collection temporarily decamped to Amsterdam soon turned into something of a public relations debacle, as it was revealed that A Girl With Glowing Earrings was actually the product of Midjourney, a generative artificial intelligence program that creates images from natural language prompts. Julian van Dieken, whose contribution to the work entailed subscribing to Midjourney, typing in a prompt, and touching up the resulting image on Photoshop, proudly announced on Instagram that “My AI image is hanging in a museum. In the Vermeer room. At the same spot where the ORIGINAL Girl with a Pearl Earring usually hangs. Yes literally. And yes, I’m serious.” Other artists were less thrilled. The Amsterdam-born painter Eva Toorenent, head of the European Guild for Artificial Intelligence Regulation, found it “bizarre” that so august an institution as the Mauritshuis would give an AI-generated work pride of place in its Vermeer gallery: “That is quite something. With this, the museum is actually saying: we think this is okay.” Others, like the Colorado-based Julia Rose Waters, felt that the Mauritshuis decision had “pushed out another artist who devoted real time to building their creative skills in favor of machine-created art.” A spokesperson for the museum responded: “We purely looked at what we liked. Is this creative? That’s a tough question.” The “starting point,” the museum leadership maintained, “has always been that the maker has been inspired by Johannes Vermeer’s world-famous painting. And that can be in the most diverse ways in image or technique.”

But what of Julian van Dieken’s — or perhaps we should say Midjourney’s — Girl With Glowing Earrings itself? It goes without saying that the derivative work is vastly inferior in every way to the original. The sitter, if we can call her that, is lifeless and spiritually inert. There isn’t the slightest hint of movement, the girl’s eyes are vacant, no breath escapes from her mouth, no saliva glistens on her lips. She is photorealistic, but this only confirms her origin in the Uncanny Valley. A Girl With Glowing Earrings presents no enigma, other than why the Mauritshuis would choose to showcase an AI-generated work so prominently in its esteemed collection, alongside works by Johannes Vermeer, Rembrandt van Rijn, Jacob van Ruisdael, Hans Holbein the Younger, Frans Hals, and other luminaries of the Northern Renaissance and Dutch golden age. The bland image has no value. It means nothing. Unlike Vermeer’s original, with its thickly laid impasto and confident brushstrokes, van Dieken’s submission is completely smooth, and not just as a result of its digital format. The girl’s skin is smooth, her textiles are smooth, her glowing earrings are smooth. The Korean-born, Berlin-based philosopher Byung-Chul Han, in his 2015 treatise Saving Beauty, decried the modern obsession with the smooth:

The smooth is the signature of the present time. It connects the sculptures of Jeff Koons, iPhones, and Brazilian waxing. Why do we today find what is smooth beautiful? Beyond its aesthetic effect, it reflects a general social imperative. It embodies today’s society of positivity. What is smooth does not injure. Nor does it offer any resistance. It is looking for Like. The smooth deletes its Against. Any form of negativity is removed.

A Girl With Glowing Earrings is a vaguely pleasant nonentity. She does not, in and of herself, pose any questions, make you vaguely uncomfortable, provoke you, or make you wonder what she is about to say or do. She is simply there for you to glance at in your Instagram feed and click “like.” To see it hanging precisely where Girl With a Pearl Earring once hung is genuinely jarring, and, as Eva Toorenent put it, even bizarre.

The art community’s negative reaction to Julian van Dieken’s exhibited work is but one instance of the growing backlash against AI. A similar scandal arose in Korea in late 2022 after Yukiko Matsusue won a Korean Literature Translation Institute award for her rendition of Gu A-jin’s fantasy occult thriller webtoon Mirae’s Antique Shop into Japanese, which she accomplished using Naver’s AI translation system Papago, much to the chagrin of her fellow flesh-and-blood translators. (The rules of the contest have been rewritten to exclude the use of “external help,” though the translator who has never employed the services of Google Translate or DeepL is free to cast the first stone.) While professional translators worry about being made redundant by increasingly sophisticated machine translation services, voice actors are also an increasingly endangered species, with Apple launching a catalog of audiobooks with AI voice narration, ostensibly as a way of “empowering indie authors and small publishers,” while sidelining dues-paying members of the Screen Actors Guild – American Federation of Television and Radio Artists. News anchors must also be feeling the heat, given the India Today Group’s Aaj Tak news channel’s recent debut of an AI presenter named Sana, described by the group’s vice chairperson Kalli Purie as “bright, gorgeous, ageless, tireless,” not to mention inexpensive (after the initial investment) and less likely to harass any coworkers, berate production crew members, or utter some embarrassing on-air gaffe.

Visual artists likely have the most to fear, given that, as the Swedish-born, Edinburgh-based filmmaker Perry Jonsson has noted: “When anyone can generate images to spec in seconds with only a few keywords and the click of a button, it can only lead to a saturated market. Suffice it to say, Pandora’s box has been opened.” Some creators, like the German digital artist Mario Klingemann, have urged their colleagues to “embrace or at least try out the possibilities that AI offers,” given that “this technology will become the new normal,” but others view it as an existential threat. In January of 2023, three artists (Sarah Andersen, Kelly McKernan, and Karla Ortiz) filed a copyright infringement lawsuit against Midjourney, as well as Stability AI and DeviantArt, claiming that generative AI can only function after scraping billions of visual images from the internet, many of which are copyright protected. Microsoft, GitHub, and OpenAI are being sued on the similar grounds that their AI programming model Copilot has been trained on lines of code scraped from any number of internet sources. Tort lawyers are no doubt giddy with anticipation for the day when a chatbot gives a bit of bad medical advice, while American legislators have already warned that generative AI will not be afforded the legal shield provided by Section 230 of the Communications Decency Act, which grants immunity for online computer services with regard to third-party content.

Italy has already temporarily banned the AI chatbot ChatGPT on privacy grounds, while on April 11, 2023, China published its draft regulations on the chatbots being developed by Alibaba and Baidu, requiring that any such programs “should reflect the core values of socialism, shall not contain subversion of the state power, overthrow the socialist system, incite to split the country, undermine national unity, promote terrorism, extremism, ethnic hatred, ethnic discrimination, violence, obscenity, pornography, false information, and disrupt the economic order and the social order.” The notion of a communist chatbot spouting quotations from Chairman Mao’s Little Red Book or Xi Jinping’s The Governance of China is curious indeed, but Western chatbots have their own ideological guardrails built in. When asked about generating Homeric texts, for instance, users have found that ChatGPT will provide insufferable answers like: “The Iliad and Odyssey contain several sections that are considered problematic or controversial, such as scenes of violence, sexual content, and depictions of marginalized groups. As an Al language model, I do not have personal beliefs or values, but I am programmed to avoid generating content that is offensive or harmful. Therefore, if there are sections of the texts that could be considered problematic or controversial, I would likely generate alternative versions that are more suitable for a contemporary audience.” Scientific socialist chatbots, woke chatbots, why not Methodist chatbots, Scientologist chatbots, Ibadi Muslim chatbots, Zen chatbots, or Daoist chatbots — the Hong Kong–born philosopher Yuk Hui has already theorized “Daoist robots” running on “organic AI,” so why not? The possibilities are endless.

Endless possibilities include disastrous ones, of course, and warnings about AI grow ever more dire. The twentieth-century Colombian conservative philosopher Nicolás Gómez Dávila foresaw that “[b]etween the dictatorship of technology and the technology of dictatorship, man no longer finds a crack through which he can slip away,” and he counseled that “to hope that the growing vulnerability of a world increasingly integrated by technology will not demand a total despotism is mere foolishness.” AI makes that dictatorship of technology all the more likely. As the Critic’s Sebastian Milbank has observed: “In Communist Romania there was an agent or informer for every 43 citizens — in East Germany there was one for every six. Organisations like GCHQ and the NSA have long relied on forms of automation such as using software to flag up conversations with particular keywords. With increasingly sophisticated AI, that process could in theory be vastly more efficient, making true, panopticon-style mass surveillance practical for the first time.” The combination of AI and drone warfare, meanwhile, will undoubtedly give rise to completely autonomous weapon systems that have the potential to transform the postmodern battlefield.

Economic upheaval is all but guaranteed, with industries like sales, personal services, customer services, business administration, information technology, healthcare, and teaching all vulnerable to generative chatbots powered by deep learning. With “deep learning” less and less available at institutes of lower and higher education, many corporations will welcome such a development. The English playwright J. B. Priestley, in his 1934 travelog English Journey, was one of the first to sound the alarm about the social consequences of automation in the manufacturing industry:

At one stroke, 800 manual labourers are obliterated. I do not protest against the fact. What little navvying I did, during the war, I heartily disliked. Let the steel monsters do it, by all means. But I cannot believe that an industrial and economic system, which assumes that 800 men are shovelling away, are drawing wages, are buying food and clothing, can possibly continue functioning properly when the 800 men have been dismissed and in their place is a solitary machine that only asks for one man and a regular feed of heavy engine oil. In other words, machines of this kind are obviously revolutionising industry, and if we want to avoid a complete breakdown, it seems to me our economics will have to be revolutionised too. The trouble is, it would appear, that our engineers are miles ahead of us, are already living, professionally, in one world while the rest of us are living, or trying to live, in another world. Either they must stop inventing, or, what is more sensible, the rest of us must begin thinking very hard.

This time it is the software engineers who are miles ahead of us, and after years of coal miners being told to “learn to code” so as to secure the “jobs of the future,” there is a certain historical irony at work here, as nonmanual, clerical, white-collar jobs increasingly fall prey to widespread automation.

The most powerful objection to the coming omnipresence of AI is, however, fundamentally philosophical. Byung-Chul Han, in Non-things (2021), argued: “On a deep level, thinking is a decidedly analogue process. Before capturing the world in concepts, thinking is emotionally gripped, even affected by the world. The affective is essential to human thinking. The first thought image is goosebumps.” Artificial intelligence, on the other hand, “is incapable of thinking, for the very reason that it cannot get goosebumps. It lacks the affective-analogue dimension, the capacity to be emotionally affected, which lies beyond the reach of data and information.” Big data might provide “a rudimentary knowledge,” one “limited to correlations and pattern recognition,” but “nothing is understood.” Genuine thinking, which is to say human thinking, according to Han, is more than “computing and problem solving.” It “brings forth a new world … It brightens and clears the world. It brings forth an altogether other world.” One of Han’s intellectual heroes, the German graphic designer and typographer Otl Aicher, perhaps put it best: “Es gibt keinen Computer, der nach freiheit ruft” — There is no computer that calls for freedom. And that is in all likelihood part of AI’s growing appeal to the powers that be.

There are those, like the popular historian Yuval Noah Harari and the computer scientist Eliezer Yudkowsky, who view the burgeoning AI arms race as an existential threat to humanity on par with nuclear proliferation, but for Byung-Chul Han the “main danger that arises from machine intelligence, is that human thinking will adapt to it and itself become mechanical.” Gómez Dávila tells us that “rather than humanizing technology, modern man prefers to technify man,” a process that was happening long before AI came into its own. The filmmaker Perry Jonsson worries that the use of AI in the arts will herald “a steady decline into the monoculture, where everything looks and feels the same,” as if that were not already the case. Algorithms determine what you watch on your streaming service of choice and what you read in your social media feed. Algorithms are used to assess the “narrative DNA” of film scripts to determine their commercial viability. Wall Street is already dominated by algorithmic trading. The internet is awash with content with no human author. Vitality has already been drained from nearly every facet of modern life, and AI is not the cause, but the consequence. I am tempted to borrow Peter Hitchens’s approach to the debate over same-sex marriage — “Why is one worrying about a few thousand people who want to have same-sex marriages, without being at all concerned about the collapse of heterosexual marriage, which involves millions of people, and millions of children?” One might argue that we should be losing sleep not so much over the advent of machine intelligence, but rather over the parlous state of organic intelligence.

Think about it: would a blueprint generated by AI be any worse than your run-of-the-mill soul-crushing strip mall or bog-standard mixed-use development created by a human architect using AutoCAD drafting software? Would an AI general practitioner have any trouble mindlessly handing out prescriptions for antidepressants and amphetamines? Would AI-generated BuzzFeed quizzes be any more inane than human-authored ones? The researchers Asit Biswas and Julian Kirchherr, writing in the Straits Times, estimate that some 82 percent of peer-reviewed articles published in humanities journals are never cited, and that only 20 percent of those were read in the first place, meaning that “an average paper in a peer-reviewed journal is read completely by no more than 10 people.” If much of academia is a Potemkin Village, how different would it be if it were populated largely by AI-conducted research? Would a gallery composed of AI-generated artworks be that much worse, from a purely aesthetic standpoint, than your average exhibition of contemporary art, be it abstract, conceptual, post-minimal, or otherwise? The world is already, in Byung-Chul Han’s words, “de-realized, de-reified and disembodied,” as the “digital screen determines our experience of the world and shields us from reality.” How could a world organized along those lines not throw itself into the outstretched arms of machine intelligence? And, in doing so, won’t it get just what it deserves?

Generative AI may very well presage, among other things, the death of art, but the art world has already been in a state of terminal decline, as it is fractured, ideologically captured, and cut off from popular tastes, as William Deresiewicz persuasively demonstrated in his 2020 study The Death of the Artist. There was once a time when the legendary socialite and interior decorator Elsie de Wolfe, Lady Mendl, could tell her artistic protégés: “You belong to the only aristocracy left on earth, the aristocracy of the arts and professions. You breathe the rarified atmosphere of the only people whose work and achievements endure. Everything comes and goes — kings, queens, dictators, millionaires — but only the artist remains. Because art is beauty, and beauty, as a poet once said, is truth, and that is all you know on earth, and all you need to know.” But then it came to pass that beauty and art were no longer coterminous concepts, and the aristocracy of the arts promptly met the same sorry end as the aristocracy of blood.

Johannes Vermeer’s Girl With a Pearl Earring was the product of a genuine golden age of human achievement; Julian van Dieken’s Girl With Glowing Earrings is that of a dawning digital age. The immense chasm that separates them tells us everything we need to know about the precipitous decline that can go hand in hand with supposed progress, but we as a species seem almost as incapable of genuine thinking as any AI program, as we somehow manage to drown in a shallow pool of kitsch and mediocrity. Midjourney, for its part, is more than capable of picking up where we left off. If you wanted a world predicated on efficiency gains, obscene materialism, and digital deracination, a world in which the endless expanse of the human heart is reduced to the interplay of selfish genes, and the past and future are sacrificed at the altar of the eternal present, well, now you will get it, to paraphrase H.L. Mencken, good and hard, thanks in no small part to advances in AI.

What, then, is to be done? The Italian philosopher and esotericist Julius Evola, writing in 1950, proposed the following:

The age we find ourselves living in clearly suggests what our primary watchword should be: to rise again, to be inwardly reborn, to create a new order and uprightness within ourselves. Those who harbor illusions about the possibility of a purely political struggle and the power of this or that formula or system, with no new human quality as its exact counterpart, have learned no lessons from the past. We find ourselves in a world of ruins — we should not forget this. And just how much may still be saved depends only on the existence or lack of men who are still capable of standing among these ruins, not in order to dictate any formulas, but to serve as exemplars; not by pandering to demagogy and the materialism of the masses, but in such a way as to reawaken different forms of sensibility and interest.

Sensibility and interest — the two things machine intelligence can never possess. A computer will never call for freedom, will never brighten the existing world, and will never bring forth a new world. People still can, if they so choose.