The Unique Horrors of A.I. Food Slop

What do we hope to communicate when we film and photograph food? In his 1957 collection, “Mythologies,” Roland Barthes argued that food photography was often an exercise in whitewashing: the highly stylized pictures featured in glossy magazines, which he called “ornamental cookery,” precluded a person “from touching on the real problems concerning food,” such as the “primary nature of foodstuffs, the brutality of meat or the abruptness of sea-food.” These photos were “meant for the eye alone, since sight is a genteel sense,” Barthes wrote. Modern food television, meanwhile, engaged in a different kind of beautification. Bill Buford, writing for The New Yorker, in 2006, explained that cooking shows often filmed food “so close that you can see an ingredient’s ‘pores’ . . . which then triggers some kind of Neanderthal reflex.” (As Bob Tuschman, then the Food Network’s head of programming, told Buford: “If you’re flicking from channel to channel and come upon food that has been shot in this way, you will be hardwired as a human being to stop, look, and bring it back to your cave.”) The term “food porn” grew in popularity around this time, becoming a catchall for visuals of ornate dishes “divorced from [their] nutritive or taste qualities,” as the critic Richard Magee put it, in a journal article from 2007. Salivating over images of a blue-rare steak or a crystallized-apple crumble permitted one to feel as if culinary appreciation could be refined in the abstract, without needing to perform labor, expend capital, or actually eat anything.

The next frontier of food imagery has proven far less appealing than whatever food porn was. From Minneapolis to Hexentanzplatz, restaurants are plastering menus and promotional materials with A.I.-generated renditions of shrimp scampi and quiche, chicken nuggets and burritos, Italian subs and bacon, egg, and cheeses. Anyone with even the mildest case of trypophobia—the sense of revulsion from tiny clusters and holes—should avert their eyes. A nauseating, nightmarish quality defines A.I.’s insectified interpretation of human cuisine, as if food were just another surrealist goo to stochastically iterate upon. This may explain the categorical outrage over this latest misuse of generative A.I.: such renderings of food—genuinely, slop—offend the materiality of the living in a way that A.I.’s distortions of words and other inanimate objects somehow do not. These images violate science; they mock life itself. It is Barthes’ conception of ornamental cookery infested with maggots; it is the Food Network, aired from hell.

As A.I. weasels its way into every corner of daily life, it was only a matter of time before it came for our food. Unlike film and music and literature, there have hardly been any anxieties about the machines coming for cooking—though perhaps there should be. Beyond graphic design, restaurants have started integrating A.I. into their business operations to “optimize both topline sales and overall profitability,” according to Jean-Georges Vongerichten, a chef who owns and operates fine-dining establishments around the world. By inserting “sales metrics and menu design” into a large language model like ChatGPT, Vongerichten says that his restaurants can determine which dishes belong on a menu and which don’t, and which should be prioritized to maximize profits. Ben Triola, the executive chef at the Chloe, in New Orleans, told Food & Wine that he used A.I. for inspiration to produce “images and ideas foreign to classically trained chefs by pulling ingredient pairings and plating ideas from across the world,” an exercise that Vongerichten also confessed to implementing.

One could argue that automating administrative tasks—such as answering customer phone calls and handling reservations—may save a business money and time, in turn allowing its staff to devote their energies to food preparation and culinary experimentation. But employing A.I. to “collaborate” on recipes and menu development seems to eschew a fundamental element of preparing and sharing food. “The food we make for ourselves, and for our people, reflects where we come from and what we turn to for comfort,” the celebrity chef Marcus Samuelsson wrote, in a 2023 essay for the Times. “It shows love and respect. It is a language that is learned.” Such is a common, almost clichéd, refrain among professional and home cooks: the most nourishing, affirming food contains some ineffable concoction of soul and spirit and ingenuity. Although the essayist William Deresiewicz has argued against romanticizing food, writing that food cannot be considered art because it “does not speak” and houses no philosophical ambiguity, a dish can nonetheless communicate a whole slew of subterranean meanings: a family or geographical history; a long-forgotten memory; a feeling of almost inexorable exaltation or disgust. For this reason, Triola’s belief that a machine’s aggregation acumen presents chefs with an innovative tool for designing recipes strikes me as misguided. Just as an L.L.M. pulls “images and ideas” from literary texts yet produces vapid and meaningless “literature” of its own, a statistical model can pair flavors and ingredients without conveying any of the essential spirit or experience needed to make a dish sing. It is food not as culture, and not even as science, but as math.

Have visual artists ever been able to apprehend the spirit and sanctity of what we eat? In the mid-eighteen-hundreds, amid the advent of photography, the Victorians shot still-life arrangements of lizards crawling over fruit and dead game birds hanging from a nail, capturing with their new technology the savagery of hunger, the oneness between natural decay and human appetite. For much of the following century, food photography functioned as a sort of ethnographic art, a form of restrained realism preoccupied with light and shadow, shape and symbolism. But, in the nineteen-thirties, the introduction of color printing to advertising changed how food was shot on camera, and thus conceived of culturally. Dishes were now a thing to be drooled over and sold, products that emphasized the sensual, the luscious, the beautiful: thick triple-layered cakes and buttered biscuits, perfectly proportioned milkshakes and overflowing cartons of French fries. As advertising became more sophisticated, so did photo manipulation. Squares of cardboard were pinned under burger patties to reinforce plumpness; motor oil was poured over pancakes so as not to overly moisten them. The surreality of commercial food photography and its later evolution into food porn may have been an artificial and sensationalist representation of the things we ate, but it at least managed to seduce. Its concerns were aesthetic, its ends gluttony and lust.

Restaurants choosing to showcase computer-generated approximations of the dishes they serve inspire neither gluttony nor lust. A menu promising a glistening digital cheesesteak or an eerily geometrical turkey club evokes little more than horror in those who behold it. Why, then, are so many eateries and establishments relying on generative A.I. to create their promotional materials? It’s free, for starters—human photographers and graphic designers cost money, even bad ones. Most of the offenders, too, appear to be delis and bodegas and fast-casual food carts, venues where quality is secondary to moving product quickly and cheaply out the door. The A.I. images are a sort of shorthand for saying, Yes, we serve food, but it is slop, just like the photos we use to market it—likely frozen and factory-made, shoved into a grease-caked frier, and still somehow lukewarm and soggy when you take your first bite. It’s the opposite of ornamental cookery; these images, as gross as they are, invite us to confront the truth about what it is that we’re feeding ourselves. ♦

Source link