Content Aware, Cursed Image, 2021-2024
Artist's Book
This project began in the early months of 2021, just prior to the precipice of the frenzied introduction of publicly available “Artificial Intelligence” text-to-image taking hold, and a permanent but subtle shift in the way we interact with images, or more specifically, photographs. As these developments evolved it seemed critical to put the project on hold and continue thinking through its implications the past few years. This time has been consumed by a vacuous and irrational discourse surrounding the maturation of these technologies; acting as if they are something wholly new rather than a final or penultimate step in a project embarked upon from the beginning of linear text or human-kind’s descent into abstraction: our development and addiction to images through their various stages and the evermore difficult problem of communication. To make sense of these developments one can look to the theorization of Vilém Flusser and the timeline laid out from traditional images, to linear text, to technical images. These text-to-image generators are the literalization of this trajectory, making the human operator a functionary of the apparatus (black box) which encodes linear text into technical images.
There are two primary developments in this technology which have pushed me to complete this project. First is the transition and evolution in “GAN” to “diffusion” models of image generation, popularized by the likes of DALL-E and Midjourney, and Stable Diffusion. Second is the more recent introduction of Photoshop’s “Generative Fill” in summer of 2023. The premise for this publication began with linking earlier GAN image generation with Photoshop’s earlier “Content Aware” fill tool.
In 2019 an image began circulating which spurred this link, which I attempt to locate while writing this text. After a series of keyword combinations, I finally entered the proper search terms: “photo where nothing is recognizable.” I can trace it back to a 2019 Twitter post with the caption, “Name one thing in this photo.” It looks like a photograph, but it is not. An early example of GAN generated imagery it is the amalgamation of source material used to create a familiar but unrecognizable facsimile. Psychedelic and at the edge of the uncanny valley, its exact origin is still uncertain. Speculation regarding the image’s origin claim it was created to mimic the effects of a stroke, however there seems to be no evidence supporting this claim. These early uses of GAN appear to mimic output generated by Photoshop’s “Content Aware” fill tool, likely because they use similar technologies. Introduced in 2010, Content Aware Fill analyzes the surroundings of a selected area, filling the selection with texture and color found nearby. Early imitations of similar technology were propagated by the 1987 film, Predator. Lurking out of sight, the titular character is obscured using cloaking camouflage. The Predator’s cloaking technology samples and fractures its surroundings—with a quick pass its presence is almost imperceptible, but with a long gaze an aberration in the real becomes evident.
As generative artificial intelligence images become pervasive, our sloppy terminology and categorization of images needs reconsideration, as the distinction between “image” and “photograph” become further askew. As these new means of furthering abstraction in communication are realized one might consider an “uncanny valley” between an image and a photograph. This uncanny valley consists of a scale from cave painting to unaltered photograph with “artificial intelligence” generated images lying near the bottom of the valley. Within this scale sits the cursed image. “A cursed image refers to a picture (usually a photograph) that is perceived as mysterious or disturbing due to its content, poor quality, or a combination of the two. A cursed image is intended to make a person question the reason for the image's existence in the first place.” Online sources speculate the phenomena of “cursed images'' originated on cursed.images.tumblr.com, on October 28th, 2015. The first image posted by the account shows an elderly farmer surrounded by crates of red tomatoes in a wood-paneled room. In the past, cursed images were photographs, wholly indexical representations of actual physical phenomena. Moving forward this is unlikely to be the case. During the first moments of consumer oriented “AI” image generators cursed images were amongst the first output. The results of prompts such as “Fisher Price guillotine” would circulate online and in eventuality artists such as Jon Rafman would lean into the technology’s propensity for psychedelia and cursed qualities.
My interest in the Content Aware tool dates back to the 2016 controversy surrounding Kelley Walker’s “Black Star Press” and “schema” works at St. Louis’ Contemporary Art Museum. Working as an attendant at the neighboring museum, I was in close proximity. Walker’s work was and is openly continuing Warhol’s project, using some of the same images as a point of departure. Walker was criticized for his cavalier use of civil rights imagery and Black cultural icons, and when questioned about this at an artist talk accompanying the exhibition, he failed to address these concerns seriously. The recording of the artist talk and following Q&A would never be posted online. Attending every event regarding the exhibition and reading every piece of text surrounding the work and the exhibition I was fully engaged with this work like nothing else before. It would be some of my first exposure to contemporary image based practices outside of photography. At the time I was experimenting with the content aware tool in my work, a professor sent me an email with the subject line “I made Kelley Walker’s work more content aware:” with an attached image, one of Walker’s King magazine cover works, with cover model Regina Hall’s body now shrouded by the tool. Within the exhibition, tucked away in a back room, were new works made by Walker for the exhibition. These works seemed to be some of the most crucial, and likely went almost entirely unnoticed. I have never spoken with anyone who seems to be aware of them. Taking on the formal qualities of Walker’s Volkswagen ad series, the work is nine MDF panels hung in a grid, Pantone 032 U as a background, with the New York Times National Section from Sunday June 25, 2015 as their subject. For those unfamiliar, Walker would be one of the primary artists working with 3D modeling during its early years. In these works Walker scans both sides of printed material, UV-wraps a plane of equal proportion, allowing him to manipulate the “page” in digital space. He warps and folds the page over itself, as well as digitally cutting holes through the pages revealing what is behind. The visible headlines read “A Day of Determined Hope as Charleston Mourns 3 More” – “Police Begin Stressing De-Escalation Tactics, Despite Skepticism in the Ranks” as well as something indistinguishable about the Supreme Court’s new ruling allowing same-sex marriage all alongside a plethora of advertisements. Where Walker’s other work falls flat, this one shines, flattening the tragedies, joys, and perversions of American life into one image, all while exploring the then-novel slippage of digital-virtual abstraction.
Sometime prior to the exhibition I became independently aware of the images that surrounded the Charleston murders, and they have haunted me since my first exposure. On June 17, 2015 white supremacist and Neo Nazi, Dylann Roof entered Emanuel African Methodist Episcopal Church in Charleston, South Carolina and murdered 9 Black parishioners with the hope of igniting a race war. Roof was self-radicalized through online platforms and three days after the murders, his website “The Last Rhodesian” containing a manifesto and zip file of images was discovered. 36 of the 60 images are self portraits, taken at his home and various significant transatlantic slave trade-related sites. These are cursed images.
Considering the continuum of the cursed image as well as the development of “Artificial Intelligence” generative imaging technologies, this publication contains the 36 self portraits which have been procedurally altered with the following actions in Photoshop.
> Open File
> Select Subject
> Expand Selection: 15 pixels
> Fill Using: Content Aware, With Color Adaptation, Opacity 100%, Mode: Normal
> Save As: .psd
In the production of this work the removal of the hand and extraneous “expressive” gestures became critically important given the subject matter. For this reason the form and design language of this research is informed by the digital existence of these images. The cover of the book takes the form of Mac OS Finder folder icons and the type is set in SF Pro, Apple’s system font. The book block is the US standard paper size of 8.5x11”, bound with a prong fastener, a common mechanism used for file folders and legal purposes. Opposite each altered image, in white and gray inks, is the subject selection, revealing Roof’s location within the image.
This publication is made available on the basis of an indefinite loan. By loaning you support the costs materializing this research into its physical form.