Get Access to Print and Digital for $23.99 per year.
Subscribe for Full Access
Adjust

After New York City experienced record rainfall this summer, my basement flooded. Water destroyed boxes of books and manuscripts, sitting several inches deep in a plastic crate full of photographs. I spent days trying to rescue as many as I could, peeling old prints apart and laying them out to dry. It was strange and sad, sifting through these fragments of my pre-digital life. I had taken the earliest when I was eight or nine, using a 1950s Brownie box camera that belonged to my mother when she was a girl. For some reason I refused to take pictures of people. The grainy prints of archaeological remains and tourist sites were frustrating and dull, bearing no trace of my family’s experiences in those places. I think I had the idea that photographs were documentation, that I had to take them for a scientific or historical purpose. People would have gotten in the way.

I owned cameras intermittently through my teens and early twenties, but for most of that time I believed that it was better to “be in the moment” than to look at the world through a lens—a fear of alienation that left most of my student life mercifully undocumented. I found a handful of pictures from those years in the basement, all taken by other people. Each one that I could save now seemed precious and meaningful, even those that weren’t technically “good” photographs. They connected me to a part of my past—and a version of myself—that has begun to feel very distant.

My ambivalence about photography wasn’t the only reason for the gaps. People just didn’t take as many pictures in an analog world. While the cost of pressing a button on my phone is basically zero, film came in rolls of twenty-four or thirty-six exposures, to be processed and printed in a laboratory. It was not only expensive but also bulky, and could be damaged by heat or light. I remember traveling in Benin carrying a battered SLR camera and a couple of lenses, rationing each exposure, painfully aware that every time I clicked the shutter, one fewer potential picture remained. Was this shot more important than some possible shot to come? Image-rationing pushed me toward my journal. I filled up pages with description instead.

The craft of pre-digital photography has begun to feel almost alchemical, which is to say that the practice of relying on chemistry to make images has come to seem esoteric, imbued with aura. As I grew more interested in photography, I learned basic black-and-white printing, standing under a red light in a darkroom, pouring chemicals into trays, handling damp sheets of coated paper with tongs. Digital photography has suppressed—or at least marginalized—this knowledge, along with a particular experience of image-making, the wondrous directness of light passing through a negative to react with sensitive chemicals on paper, the magical emergence of an image in a bath of developer, the skill of waving wands and perforated cards to dodge and burn sections of a print that are too dark or too bright.

I own photographs that seem overwhelmingly material: for example, a glass-plate image of a severe Victorian lady that I found in a junk shop. Judging by her clothes, she probably sat for her portrait some time in the 1860s. She poses beside a vase of hand-tinted flowers, and has slightly alarming hand-tinted salmon-pink lips. She’s set in a little tin frame stamped with fancy patterns. It seems incomplete to call this an image. It is an object, a thing with texture and weight.

Even my limited adventures in the darkroom were enough to teach me that there’s nothing immediate or straightforwardly truthful about photographs, that they’re constructed by all sorts of technical and aesthetic processes. Yet in thinking about what I do when I lift up my cell phone, I find I habitually fall back into the hackneyed vocabulary of “capturing a moment,” as if I’m bringing a net down over reality like it’s a rare butterfly, the way critics used to talk about photography. “Photographs really are experience captured, and the camera is the ideal arm of consciousness in its acquisitive mood,” wrote Susan Sontag in 1973. “Photographed images do not seem to be statements about the world so much as pieces of it, miniatures of reality that anyone can make or acquire.” That “seem” is doing a lot of work. Sontag would love to say that photographs have a direct connection to the world, that they have taken its stamp like a piece of wax. But she is too scrupulous. Even in the time before Photoshop and Instagram filters, before phone cameras that detect objects and automatically adjust for lighting conditions, she knows it’s not quite true.

Despite our ancient and cultivated distrust of images, we still speak in terms of “capturing” a moment, “taking” a picture, echoing a particular romantic notion of what it is to be a photographer. Henri Cartier-Bresson, whose famous 1952 book Images à la Sauvette (“pictures on the run”) was published in English as The Decisive Moment, embodies the romance of the Leica-toting flaneur, the street photographer with the compact camera, exquisitely alive to aesthetic possibility. Explaining such canonical pictures as his 1932 shot of a man jumping over a puddle (“Derrière la gare Saint-Lazare”), Cartier-Bresson wrote that

photography is the simultaneous recognition, in a fraction of a second, of the significance of an event as well as of a precise organization of forms which give that event its proper expression.

This is the photographer as gunslinger, recognizing fleeting perfection (look, he’s reflected in the water!) and instantaneously pulling the trigger.

In Cartier-Bresson’s analog era, even a professional photographer had to ration clicks. These days, while the experience of pressing the shutter at just the right time is still crucial to photography, most of us are not so much sharpshooters as firefighters, “hosing down” our subjects—as the evocative paparazzo slang has it—to take home and edit later. We’ve soaked the whole world in images. The vast majority of all the photographs ever taken have been taken in the past two decades. About 90 percent of pictures are now created using smartphones. For the most part, these images will never be printed, never have any existence as objects. They are uploaded to social media, sent as attachments, viewed on screens at various resolutions, or perhaps not seen at all. Many will be copied and compressed, increasingly degraded into jumbles of pixels, what the artist Hito Steyerl has called “poor” images. “The poor image is a rag or a rip; an AVI or a JPEG, a lumpen proletariat in the class society of appearances,” Steyerl writes. “The poor image has been uploaded, downloaded, shared, reformatted, and reedited. It transforms quality into accessibility.”

Against the proliferation of memes and GIFs and screenshots, consumers are invited to luxuriate in the rich image, to buy absurdly large flatscreen TVs, high-resolution projectors, Blu-ray discs of classic movies, all the expensive toys of home cinema. We’re encouraged to construct personal arenas for the consumption of HD blockbusters, hyperrealistic imagery that I (and I suspect anyone else whose aesthetic baseline was calibrated by 35-millimeter film) still find slightly uncanny, its resolution somehow excessive, a surplus that leaves the most banal rom-com teetering queasily on the edge of the transhuman.

There is a trade-off between image quality and accessibility. Visual materials once imprisoned in archives now circulate, albeit often in illicit or degraded forms. What is lost in quality is gained in velocity and reach. Online you can find rare videos, raw newsreel footage, digitized photo libraries. You can also find niche pornography, and cell phone recordings of every kind of unpleasant human interaction, from racist confrontations in suburban stores to war crimes. The democratization of image-making means that every jihadi bouncing through the desert in the back of a truck also has a phone in his pocket. Every bystander at an arrest can record and transmit pictures and videos. A Syrian military defector can smuggle thousands of photographs to a human rights group, providing evidence of mass torture. A team of citizen journalists can use publicly available images to prove that a particular Russian missile launcher shot down a commercial airliner over Ukraine. If you doubt the power of such poor images, you have only to say the name George Floyd, and consider the laws subsequently proposed by Republican politicians in an attempt to curtail filming or photographing the police.

As the image has liquefied, and the ability to manipulate it has become as simple as toying with a few sliders on our phones, the question of its relation to reality has become ever more urgent. We innocently make the sea a little bluer, the summer afternoon a little hazier. But we also make deepfakes of politicians, and photoshop the faces of movie stars onto BDSM pictures. Forensic image analysis is now an important public service. Analysts consider whether the objects in a photo appear to be lit by the same light source and whether the metadata corresponds to how the image looks. Repeated patterns of pixels may indicate that something has been removed and another part of the image cloned to replace it. If two people apparently sitting next to each other in a portrait have different-size glints in their eyes, it may suggest that they were lit differently, and so were never together at all. Even the distribution of information in a compressed image can tip off an analyst that it has been saved more than once since it was taken.

Ultimately, another development in photography may prove even more consequential than the transition from analog to digital: the word “photography,” with all its ancient Greek baggage about drawing with light, might have to be abandoned altogether. The artist Trevor Paglen has called for an expanded understanding of photography to include what he calls “seeing machines,” a category that “embraces everything from iPhones to airport security backscatter-imaging devices, from electro-optical reconnaissance satellites in low-earth orbit, to QR code readers.”

Machines are seeing in all these ways, and now many images are produced by the machines themselves for other machines to look at, without any human involvement at all. Vision is moving beyond us, beyond human scales and capacities. The photographs in the plastic crate in my basement are (or were) material objects. All I needed to see them were a light source and functioning eyes. The most recent were taken in 2006 or perhaps 2007. After that I bought a digital camera, and stopped dropping off rolls of film at the lab. My photos became data stored on hard drives. Right now, they’re a few clicks away from the document I created to write this essay, but if there was a power failure, I would have no way of viewing a single one.

The data produced by seeing machines may not look like anything we’d understand as a photo, but as vision recedes into a sort of hermetic mystery—a capability that must be mediated through technology before it is granted to us—it’s all the more vital that we persist in our efforts to understand it. We must ask ourselves what the machines want. That’s not to claim that our technologies have independent agency (though that time may come), but to say that we should pay close attention to the ways in which these systems see, the solutions they find to the tasks we set them.

In September 2020, a Twitter algorithm had the job of cropping pictures into the right format for preview images, something people would see before they clicked on a post. A white user noticed that in a picture of himself and a black colleague, the algorithm always cropped to focus on him. The post he made describing his observation went viral. The experiment was tried with other pairs of black and white faces. Barack Obama and Mitch McConnell: same result. Black faces were erased and white ones chosen. The Twitter executive Dantley Davis admitted the issue was real. “It’s 100% our fault,” he tweeted. “No one should say otherwise.” The technical reasons for the preference were complex. As the Princeton computer scientist Arvind Narayanan pointed out, at its heart was the use of argmax, a mathematical function that selects the most likely output from a probability distribution. Argmax is known to amplify biases, and to double down on small discrepancies in inputs until they become glaring. Twitter’s machine-learning algorithm had been trained to determine the “saliency” or interestingness of each part of an image, including factors such as brightness and color saturation. Its decision process was concluding that darker faces were less interesting than lighter ones. Ultimately, Twitter solved the problem by turning off automatic cropping entirely.

Our seeing machines are often attached to learning machines, and we are only beginning to witness the power of this combination. Much of the new corporate space race is about vision, putting lots of small, cheap satellites into orbit to provide a more complete and continuous view of the earth. Satellites are being equipped with multispectral cameras, seeing machines that can operate at wavelengths and frequencies the human eye can’t. As inputs to so-called deep-learning algorithms, this new generation of images can be used to build 3D models of structures on the ground, assess future crop yields, track moving objects, and predict the weather.

We are building a megastructure of vision, what the design theorist Benjamin Bratton has termed “planetary-scale computation.” The panopticon is here, and it does no good to shudder and invoke Orwell. Nothing is inevitable about the form of our creation, or about the concerns that should animate it. Asking what our machines want is also asking what we want. The information produced by space-based imaging has value to the military and the markets, of course, but it may also form a core part of what Bratton terms the “planetary competence” required for us to mount an effective response to climate change.

The visual culture that produced the crate of prints in my basement now seems as remote as the glass-plate image of the Victorian lady. Yet the break has not been complete, or even a break at all. Family photographers still tell everyone to say cheese. We still shuffle sideways to fit into the frame. Our human-scale visual culture persists, even as it forms an increasingly marginal part of photography, or whatever we choose to call what comes next.


More from

| View All Issues |

November 2023

Close
“An unexpectedly excellent magazine that stands out amid a homogenized media landscape.” —the New York Times
Subscribe now

Debug