Writing Photographs
Stafford Hiroshi Smith (Artisanal Photographer : )
Grand Valley State University
The camera preceded photography by several centuries serving as an astronomer's aid and an artist's tracing device. But it may be the camera that is bowing out first, while photography lives on. This is all thanks to artificial intelligence. The recent spread of easy-to-use, software like DALL-E, Stable Diffusion and Midjourney has changed the debate from whether digital photographs can be trusted to whether or not photography exists at all. The verb, "to take" a picture may need to be changed to "to write" a picture, as the tiresome task of working with a living subject and getting them to pose and crack a smile becomes a thing of the past. How this impacts working photographers and photography curricula remains to be seen but cannot be ignored. The postmodern dream of semioticians has at last been realized as photography has truly become language, a literal one, as the camera is replaced by the keyboard and memory blends with imagination.
Over the course of photography's history different verbs have had their day. We are most familiar with the phrase "taking a picture," which certainly has had the longest run. But in photography's misty early years people said, "drawing a photograph." That may seem awkward now, but it made sense in the 1840's when photography was largely seen as nature drawing herself with light. It was the introduction of the point and shoot camera in the 1880's and the spread of the snapshot that truly gave meaning to the phrase, "take a picture." And this was in spite of Kodak's best efforts to make a verb out of their company name.
However, "taking a picture" almost sounds like theft. It's like the camera gave its users license to take things of little value that didn't belong to them, the way we pick up seashells at the beach or pluck flowers from a field. The potential images were always there, floating around like dry leaves in the wind. We just needed a device to reach out and grab them. But not everything was valueless. Taking someone's likeness on the sly, that did seem like theft, and a new outrage in the late 19th century emerged over voyeurs hiding in the bushes with their Kodak Brownies waiting for assignations and romantic interludes.
In the 1970's and 80's postmodernism suggested an exhaustion of all poses, locations, and scenarios. Many believed all possible photographs had been taken, and to take anymore would be redundant. All that was left to do was to use existing photographs to fulfill any future needs for image making. This is not very different from how AI synthesizes photographs. Examples of this can be seen in the work of Richard Prince, Sherri Levine and Barbara Kruger. Other photographers deliberately staged their photographs to create unique subjects. This acknowledged the conscious construction of reality inherent in all photos.
A term coined by legendary Ansel Adams was repurposed to describe this new direction, "make a photograph." For examples of this we can think of Sandy Skoglund, Cindy Sherman and Laurie Simmons. How naive the postmodern conceit of all the photos being already taken sounds today in the age of the ubiquitous smartphone camera.
Consider that if the phrase "taking a photo" implies theft, does "taking a selfie" imply stealing from oneself? Certainly, the amount of information made available to data miners would corroborate this idea. Maybe the phrase "discarding a likeness" would be more appropriate. The ability to alter oneself to the point where one's "photos" no longer resemble the original, might warrant a return to the postmodern verb "make" rather than "take." But we may already be past that point to bother discussing it.
While AI maybe easy and fun to use generating dogs snorting coke and 3-armed people covered in bees in Times Square it is something of a black box in that hardly anyone understands how it works. This is another shared dimension with traditional photography. When Kodak unveiled the point and shoot camera back in 1888 it was both literally and metaphorically a black box. With the brilliant slogan, "You push the button, we do the rest," Kodak changed what had been a difficult, dirty, and dangerous medium into something quick, clean and easy overnight. What the consumer friendly, easy-to-use, and fun-to-play-with Kodak no. 1 hid was an army of laborers in a factory in Rochester, NY, doing all the complex work of developing, printing and reloading the little cameras. It's ironic that if anyone wanted to do all this on their own, they had to enter a different black box called the darkroom. What Kodak prepared America for was the hiding away of the innerworkings of technology and encasing it in a smooth and featureless container. And although Kodak is largely out of the photography business, their legacy of trading knowledge and understanding for convenience lives on. The black box, while small and hip on the outside has grown exponentially on the inside, like a tent in Harry Potter or the Tardis in Dr. Who. It masks data mining, surveillance, and market research. While we marvel at how quickly we can learn how to make the prompts to get AI to make entertaining images, we never ask ourselves what AI is learning about us.
The debut of artificial intelligence programs like Open AI's DALL-E 2, and a quickly growing field of competitors may be the biggest change to the medium of photography since the introduction of digital technology. These programs take a few words entered via keyboard to produce an image in seconds. These images can resemble any visual medium from photography to graphic design, to painting. As of this writing hands and faces are still a problem for AI but that issue is rapidly improving.
The media has predictably fixated on stirring up fear among the general populace with dire warnings about AI taking over and not being able to know whether one is conversing with a real human or not. Images from Invasion of the Body Snatchers and John Carpenter's The Thing come to mind. Nothing crosses the uncanny valley without sinister motives apparently. In Invasion of the Body Snatchers people are replaced with exact copies grown in pods. The replicants may resemble the originals on the outside, but inside there is only a soulless vacuum. The pod people eventually take over the earth, triggering the fear of replacement theory so well cherished by the right wing today. Change is scary and the fear mongering media thrives on it. It's easy to forget that we are kept in a constant state of fear to keep our industries and economy humming. One new boogie man replaces the last.
We have only to look back at other innovations and inventions that struck terror in the hearts of the general populace to put things back in perspective. When digital photography began to rival film photography in quality and popularity, people were worried that they wouldn't know whether photographs were real or not. People also fretted over the power of Photoshop for the same reasons and yet somehow, we survived, and photography is more popular than ever. In James Thurber's "My Life and Hard Times" from 1933 an elderly lady is terrified of electricity in her house. She worries that her Victrola will blow up now that it isn't hand cranked and that she will be electrocuted by electrons leaking out from sockets in her house. Will our fear of AI today be that risible in the future? As someone who grew up during the Cold War with the threat of a nuclear annihilation that never came, this fear mongering of the "AI-take-over" seems overblown.
What is more of a challenge is how AI perpetuates stereotypes and augments misrepresentations in terms of gender and skin color. According to Bloomberg Technology: "The world according to Stable Diffusion is run by white male CEO's. Women are rarely doctors, lawyers, or judges. Men with dark skin commit crimes, while women with dark skin flip burgers." With AI programs drawing upon pre-existing pools of stock photos, which have already been accused of gender and racial bias, they are exacerbating the problem of misrepresentation and stereotypes and could trap us in a spiral of regurgitated entrenched imagery. When we use AI to generate an image, we are not creating anything new from scratch, we are remixing and repurposing existing materials. One way to move forward would be to make the effort to create new imagery, thereby saving photography and making representation more equitable. However, another AI platform, DALL-E 2, is making efforts to impose racial and gender diversity, sometimes with strange results. I had prompted DALL-E 2 for an image of a Japanese gangster, known as a Yakuza, in a sushi restaurant. Yakuza are primarily Japanese men, and while it spit back 3 menacing looking East Asian males, it also gave me a black woman in that role, presumably because of a directive to encourage diversity.
Another odd parameter that I've run up against is a reluctance to depict women in Iran without head coverings. I specifically asked for women in Tehran without head coverings but out of 16 images, I managed to get only one partial result that matched what I had asked for. This makes me think that working with AI is like a child playing in a sandbox as opposed to being let loose on the beach. Parameters that have been set up give the illusion of freedom, but we are confined in a predetermined safe space.
While most traditional photographs distort or pervert reality through idealization, these new creations have no original to which they refer. They are a composite simulation based on an idea of a situation or event that never existed. These AI creations very neatly fit Beaudrillard's definition of a 3rd order simulation. Once the software improves to the point where they become indistinguishable from real photographs we will no longer know if the events they signify happened or not. Control of the narrative of history will become even more unstable than it currently is with the ease of generating these very complex and convincing images that masquerade as evidence. What this means to a culture already descending down the path of misinformation and conspiracy theories will be revealed in the 2024 US presidential election.
The photographic community is divided over this new development. Many fear for their careers or dismiss it as soulless. Some want to scurry into their darkrooms and lock the doors. But others, like postmodernist Laurie Simmons, have embraced it and use it as a tool to enhance their creative output. (Aperture) Her technique involves initially creating an image with AI and then refining it in a program like Adobe Photoshop. Speaking about her new series In and Around the House, Laurie Simmons said, "I feel as deep a connection to language as I do to images, though they've always lived in separate compartments of my brain." (Fellowship) The term "synthography" has been coined to describe hybrid imagery like this. I share Ms. Simmons' adaptation of AI by working to integrate it into my photography classes. Synthography is now part of our curriculum, albeit at an experimental level.
One of the first things one discovers is the difference in results depending on the prompts one enters into the software. A level of conceptual understanding is necessary in addition to descriptive powers to get the best results. You become something of an AI whisperer. This is the beginning of a new skill, to be able to get the AI to do what you want it to and not merely accept what it gives you. This is the point where we either are defined by AI or we define it. My students and I are using this software in our history of photography class to create photographs both in the visual style of a photographer and to reflect the concepts of that person's work and the era or movement they were part of. Rather than forging a fraudulent work we are using AI to deepen our understanding of their thoughts, concepts, and visual strategies.
Original Images by Sebastiao Salgado on left, AI generated by Kiara Henery using Stable Diffusion on right
Getting students to think conceptually has always been paramount. Replacing the lens with the keyboard supports this initiative. It takes a while to get good enough at coaxing the AI to generate an image that resembles that of the photographer. But do the results merely mimic the visuals or do they convey something of the concept and essence of the original as well. This encourages students to think and research their photographers more deeply before using AI to generate an image. This may replace the essay in the future now that Chat GPT is being widely used by students to do their writing for them. We can pit one form of AI against another. Below is an image that succeeds in conveying a film noir look of the crime ridden underbelly of New York in the 1940's but fails as one that channels the spirit of WeeGee. It lacks the voyeuristic qualities and the raw display of violent crime.
Original Wee Gee crime scene photo Fake Wee Gee generated with DALL-E2
In my production classes I find that AI is a powerful sketching tool helping create scenarios that might not have occurred to students themselves. We use it as a starting point to build upon. For a project in response to readings from Simulacra and Simulation by Jean Beaudrillard, the text that inspired films like the Matrix and the Truman Show, students began by generating their images in AI as sketches. Some moved on to recreate or adapt the images through photography. Some used pieces from the AI in combination with their own photography. Others composited or refined the AI generated images into their final images. One student created a parody of concerns over AI in an attention getting campaign to address environmental issues. She requested imagery based on her message of ceasing to breathe to save the earth. It has a dystopian feel about the worst that AI might end up doing to us.
Images by Annabelle Robinson, Advanced Digital Photography, Winter 2023
Another student used AI to create elements that she arranged in a down-the-rabbit-hole sequence that challenged our basic acceptance of what is real.
Images by Paige Metcalf, Advanced Digital Photography, Winter 2023
In conversations with students none of them felt that they had created the AI generated portions of their images even though they had called them into existence. There was little feeling of accomplishment or satisfaction that comes from a job well done. Although they were fascinated with the possibilities and most of them enjoyed the exercise. But I did wonder if this had to do with the novelty of it. One student did reject it entirely and went out and took her own photos because she felt that the actual existence of her subjects was necessary for her work to carry any weight.
A new word has been coined for this type of human/tech collaboration, para-creativity. And what is generated through para-creativity is the aforementioned synthography. I'm sure new words will be generated in the near future. Perhaps one should even ask AI to come up with the name for what it does. While AI is getting good at mimicking lighting, costuming, props, and locations, where it fails is with human interactions and a compelling narrative about the creation of the work. We want to hear about the challenges faced by the photographer, whether they be whizzing bullets, over-inflated egos, or environmental disasters. We also value the human interaction involved in getting a portrait. The most interesting photographers are not just technically skilled and brilliant conceptualists, they are people who engage with their subjects. They talk people into posing, they talk their way past emotional and physical barriers. They embody a spirit of adventure, caring, and justice as they become witnesses to world events. They dare to do what no one else will. And this means a lot in this era of paralyzing anxiety. They show us something unique that tells us something about ourselves and what some humans are capable of. We live vicariously through them. From Robert Capa's gripping account of landing with the first wave of soldiers on the beaches of Normandy on D-Day to Yousef Karsh bolding yanking the cigar out of Churchill's mouth just before pressing the shutter we thrill to the tales told of people who went beyond pale.
Nikon has launched a campaign to argue that the earth is beautiful enough without having to generate fantastical scenes. (PetaPixel) They ironically pair actual places on earth with prompts suggesting alien or fantasy-landscapes out of Star Wars or Tolkein. But maybe this is the point. The places they show are far flung, expensive and time consuming to get to. For anxiety ridden youth who are afraid to leave the house or interact with people face to face, travel by AI might be the solution.
Back in March, my wife and I went to New York City and we took this photo by the East River. When we learned that we'd be going to France to present at this conference we decided to make a representation that preceded reality, a postmodern conceit, and used AI to place us by the Seine.
The phrase, "Pix or it didn't happen," is headed for the dust bin.
Aperture Magazine, How Will AI Transform Photography? Charolette Kent, March 16, 2023
https://aperture.org/editorial/how-will-ai-transform-photography/
Bloomberg.com, Humans Are Biased. Generative AI is Even Worse, Leonardo Nicoletti and Dina Bass, June 15, 2023
https://www.bloomberg.com/graphics/2023-generative-ai-bias/
Fellowship: Laurie Simmons, In And Around the House, Claire Silver, June 15, 2023
https://fellowship.xyz/collections/laurie-simmons-in-and-around-the-house-ii/
PetaPixel: Nikon Fights Back Against AI Images, Michael Zang, June 15, 2023