"The Stranger Inside" by Meredith Westgate
on Art, Internal Landscapes, and Artificial Intelligence
———
Show me how it looks, this feeling.
Show me what its colors are; how it curves or even resists shape at all. I imagine it against a canvas, splayed, splattered. The lingering sort of grief that no one wants to hear about. Shaped into plaster, cast in metal, towering overhead. Grief is expected to dissipate with time, its fading grip another loss. How to articulate its duality, the way it’s pleasure and its opposite?
If I tell you, maybe I can see it, too.
My husband and I lay in bed, eyes fixed on his iPhone. We spend an hour summoning the right words, then watch as our images come to life. Are they though, ours? This is the work of DALL-E, a machine learning program that uses a neural network to create art, or something like it, based on the language we feed it. DALL-E’s process of creating, its way of “thinking,” has been trained on hundreds of millions of images on the internet, and yes I know what you’re thinking — haven’t we all?
The potential uses are endless, effectual; building a specific image rather than combing stock photography; mocking-up animation for a pitch deck; a stylized poster design; album art. And yet, as a writer obsessed with memory, my mind immediately goes to summoning the invisible, to the sand slipping between my fingers.
At Spring Street and Greene, I watch as a woman in an athleisure unitard slowly bends down to her knees in the middle of the crosswalk. My first instinct is urgency. As I rush closer, I notice another woman crouching on the sidewalk, camera perched—getting the shot—as the first arches her back and lifts her hands overhead in a sort of prayer. Beside the photographer is a mess of equipment and outfit changes, the entire thing staged.
I look for someone to trade a subtle head shake, commiserating over the world we live in being one of the few joys we have left. The only person I see passes swiftly, lifting his phone to get his own shot. He leaves the crouching photographer out altogether.
It comes as no surprise to me or my husband that without hesitating my first request on DALL-E is for a pbgv, a quirky hound, and the breed of our dog who recently passed away. Olive’s death is the kind of loss that will forever be recent, though in doing the math now, it’s been two years and five months.
I want to say that I dutifully hate this program, this proxy for art, creativity, craftsmanship. What I find is more of an experiment in my own memory, excavating some internal landscape, watching the invisible appear. We lay there, staring at his tiny, cracked screen for hours, conjuring images from a life I miss breathlessly, in most moments, most days. Soon the hairline fractures in the glass disappear, and I see only her, through the veil of different styles and settings that I learn to control with language, like a magician learning sleight of hand.
Yes, we are still staring at a screen, but this feels different — like a bridge to something elusive and so often lacking, especially on our devices. The images are so well executed that my impulse is to play the con game — to insert canonized artists to steer the direction towards the taste of my longing, the tone cast across my memory.
“Bonnard painting of pbgv running through a field of snow”
For better or worse, Bonnard’s style comes to mind immediately, with its emphasis on the everyday and the wistful, ephemeral quality of his brushstrokes that perfectly conveys the sense of a moment, blurred in memory. The colors come back muted, hinting at the familiar dustiness of pastel despite being entirely digital.
Still recovering from a bout of pneumonia, this set of results — again, I resist calling it art — sparks such joy that it launches a coughing fit, and I can barely breathe for minutes but still cannot look away, cannot stop beating my husband’s chest, both of us delirious with seeing the thing we want, right there, outside of us again.
She isn’t breathing, but it feels alive.
Next we try Alex Katz. I realize only later that whenever I include him, the moment I’m conjuring is based on photographs that I haven’t looked at in years. The flatness of Katz’s style feels appropriate for recreating that photograph that exists in memory, where I cannot pinpoint how I was feeling that day, or what we had been doing just before I hoisted Olive upright until she, low and long-bodied, was my height. I don’t remember how the beat-up white Roland Garros baseball hat ended up on her head, only that it sat at the perfect angle to make her look like any human trying to go under the radar.
“Pbgv in baseball hat sitting upright on a couch in style of Alex Katz painting”
The image comes back appropriately flat, her expression cool and caught off guard. The colors have the perfection of a Katz, too; bold but cohesive. A tuft of hair on her head slightly hides one eye, the way it often did. It’s impossible for DALL-E to know this, but it feels equally impossible to have this image in front of us if it didn’t.
We attempt more rounds, adding details from the interior of our funky Los Angeles a-frame, its slanted floating staircase, colorful stained-glass windows, rickety beams, our lease now long expired. The more we insert beyond that first prompt, the further it strays, and the more distorted the images become, feeling like loss all over again as the slice of life we summon skews into something grotesque and unfamiliar.
Perhaps a place like Los Angeles is too loaded. The tiny sliver that we want to depict — cascading jasmine in the hills, its winding roads, the stark shade of cypress trees — is not the Los Angeles the algorithm has been most trained on. The images it returns most closely resemble the wide, flat concrete of Venice, or the harsh sunlight against a strip mall, the cliché of looming palms. The more we try to correct the direction, and the more details we give, the more complicated and overworked the image becomes. As if by giving too much detail we are in fact burying our own perspective, the essence of the moment becoming muddled, overwhelmed, impossible. The writing comparisons are glaring and everywhere.
We go quiet, then pivot to Olive in the style of Klimt, a flurry of gold brushstrokes around and among her. Even her pose is regal, as if the algorithm once more understands the assignment; an icon of power, our own dog in gold.
“pbgv in the style of Klimt’s Woman in Gold”
Then we get cocky, send her out to space, with a cone on.
“pbgv in space, wearing a cone, in the style of Klimt”
A couple years before she died, Olive needed surgery to remove two tumors. They were benign, but one had become the size of a baseball, and both were increasingly dangerous because of their size and location. The baseball was under and around her tail — even with surgery, it risked lifelong incontinence — and the smaller one was on her cheek, a golf ball pressed against her optic nerve. Post-surgery, half of her face was shaved along with her entire lower body, so she looked like a hairless dog wearing a scruffy sweater, sporting Frankenstein-like stitches across her cheek and circling her tail.
The last few years with her felt like stolen time, and it started with a full month of her wearing a cone. I had always meant to paint her then in the style of a religious icon, cone as halo.
“Dutch Masters pbgv wearing a cone on her head”; “Dutch Masters pbgv wearing a cone as halo”; “Dutch Masters pbgv wearing a medical cone on her head, like a halo”
This takes several variations. One quick discovery is that, depending on the word order, “cone on her head” comes out alarmingly KKK, the algorithm lacking our learned aversion; the four panes swiftly deleted, as we learn our blind spots.
We try more artists; Renoir, Pissarro, Frankenthaler, even Chuck Close. It is not lost on me that most of the artists I choose are male, white, plucked from an established canon; so little faith do I have in the depth or diversity of DALL-E’s training, and so impatient am I to use their familiarity like a tool.
Some of my favorite memories are of my husband giving Olive baths — or standing in the shower with her and a tub of peanut butter — and then, to my surprise, blow-drying her after. She hated showers, yet I watched him pull a round brush neither of us ever used, possibly left by previous tenants, through her sopping, scruffy hair until it poofed, doubling her size. In my previous nine years of living alone with Olive, every similar attempt ended in my chasing her, both of us wet, shampoo everywhere. Her dampness lasting hours, everything dirtier than before.
“A tall blonde man, blow-drying a pbgv dog, in the style of Matisse painting.”
Somehow, it’s all there; Matisse’s familiar composition style, the colorful tiles, the odd bathroom that had a door to outside. Yet the tiles are yellow, not the turquoise in my memory, and Olive appears smaller, cartoonlike. The four versions of my husband each capture something entirely different but true; one wavy and amorphous, another the mere hint of a profile, a person.
Finally, we try one with me.
“Young blonde woman walking pbgv in the West Village in the style of Hopper”
The dogs aren’t Olive, the women aren’t me, the style is hardly even Hopper, but there’s something in the tone that captures the feeling of so many years, weekends alone and exploring together with nowhere to be. The shadows, the brick buildings and their awnings, the vague suggestion of us both. I feel it deep in my body, the way Hopper can capture an ordinary moment, charge it with so much more. It’s a memory. It’s so many in one.
After Olive died, I ordered a View Master, the same red plastic case from childhood, and chose ten photos for each of the reels. I hoped it would create a sort of space, the sense of going somewhere to visit her. The old-fashioned reels were the antidote to a Live Photo, whose hidden action can surprise you, pulling you so deeply into the black hole of missing someone or some time so fully that you can smell their breath, feel their fur against your nose. I wanted the distance; moreover, that it might conjure a memory and not its digital proxy.
Then a few months ago, a generous reader of my novel sent me a painting he had made off one of my favorite photos of Olive, taken while waiting for a sandwich in our old neighborhood, discarded Coke can in the foreground, public library in the back. When I opened the box, carefully packaged from its trip across the country, I felt a wave of her presence, that calm from her next to me again after so long. Every time I enter the room where it now hangs, I feel her there, the canvas so large as if to scale. As if she stands, watching from a distance.
Another program my husband shows me, Midjourney, generates its images from scratch, and as the dog from my text appears, I’m struck by the gaping hole around the cheek in the side of its head, its eyes at slightly different levels, a deranged sort of stare. It takes so much to make something look kind, calm. Sentient. That lack of awareness, no human hand at the helm, is a chilling reminder of the machine.
The power to evoke the invisible is validating for a writer, but the question of ownership with a program like DALL-E feels especially fraught for anyone who places exceptional value on words. When we look at our phones, we’re mostly presented with things, fixed and flat. We’re bombarded with a curation that culls us into conformity. It feels remarkably different, then, to summon familiar scenes, to describe the thing you remember and watch something non-human try to construct it for you. The more we feed the program, the more I feel compelled to refine and rethink the potential impact of each word, to pour over every connotation, or how even the order might direct the algorithm.
In short, I want to control the magic. And of course, the process begs the question of who is doing the creating? Are we, by feeding the algorithm a carefully crafted set of language, playing them towards our own desired result? Or is that like adding finesse to throw a tennis ball, which, regardless of style, will be brought back by your dog exactly as it pleases.
There is something to this conjuring. The brief wait while the images load recreates the feeling of taking a polaroid, of being in the moment with something new, where you turn to someone and show them, look at this moment I just captured, look at this magic.
Any photographs we hold afterwards have lost that, the magic long gone, and though I bristle and cringe at the idea of this program ever replacing art— a painter’s hand crisscrossing their canvas, the texture of the surface, the spontaneity, the interplay of light — I can’t help but feel this is an exercise that opens me, when so many today seem to close.
Several years ago, working for a literary magazine, I received a call about an AI service for reading and sorting fiction submissions, especially the slush pile. I didn’t understand the premise, but they said their service could trim the fat, weeding out submissions that had no merit.
My question to them was how they took style into account. Wouldn’t they present only the most middle-of-the-ground submissions, having discarded writing that plays with sentence structure, with simplicity, voice, intentionally fragmented thoughts? With an algorithm trained on aggregation and being objectively correct, wouldn’t they potentially always weed out something truly genius, truly fresh?
They had no answer for that. This was years ago, maybe now they do. But it speaks to the idea of ever replacing art with some sort of AI proxy, in that it doesn’t seem like a true threat, not for anyone who loves art. So rarely is it only about the image, but rather about the process, the imperfections, the context; the human touch conveyed through material, light, structure, sound. Machines creating the same make decor, maybe. But we have plenty of that already.
What is most interesting to me about a program like DALL-E is what it exposes about the discrepancies in communication, in connection — the artist’s eye present in exploiting those gaps, in revealing something else about language and humanity, about the role of an artist, even in its absence.
Back to Spring and Greene, to the middle-aged man in a hurry, iPhone cocked nevertheless. His rush to capture something remarkable, even if he misses the reality, is more notable than the silly moment of marketing we know all too well. To be human in our modern world is to be under the impression that to witness something — anything — and to remember it, means to record it.
We capture enough photos to train an algorithm, even though our minds are naturally doing the same. We are an accumulation of impressionist moments, not the digital snapshots we carry in our pockets, nor what we post to train others on what we witnessed, who we are. One of the most miraculous things we carry are these memories that build and layer to shape our experience of life, our identities, our perception. And we just give it away.
Within this program, of course, is the lurking terror under our modern, human existence. That the plethora of images we bombard ourselves with — social media, advertising, anything on our devices — slowly shapes our brains and the connections we make. A program like DALL-E can produce original works significant to us, functioning solely off that same process. If there’s an uncanniness to these images, capable of moving us while not being human-made, it is unmistakably that the same process seems to be sucking the humanity — or at least, the originality — out of us.
Certainly this is how advertising and language has functioned for ages; our wants and needs being dictated to us. What’s unsettling now is that these algorithms once built for us are capable of outpacing us, outmatching us, and outperforming us, using the same simple programming that feels, at times, like it might break our brains.
I wonder if an artist and DALL-E were given the same words to interpret, whether we could guess which result was human-made. The pit in my stomach says no, but I still believe that we would find the human art more moving, impactful. My hope is that we stop treating our brains as programmable — assaulting them with images, trends, memes, until we all think the same — and leave that to the machines that otherwise rest, empty.
We have the advantage of experiencing the world alive before us. Licking the blood off a paper cut while reading, tasting pennies, feeling the thin sting. But only if we let ourselves back into our own bodies long enough to linger.
Meredith Westgate is the author of The Shimmering State. Her work has appeared in Joyland, LitHub, No Tokens, and more. She is a graduate of Dartmouth College and has an MFA in Fiction from The New School. She is also a mentor with Girls Write Now. Visit her at meredithwestgate.com and on Instagram @meredithwestgate.
She is also the prose judge for this year’s Breakout Writers Prize.
