Essay by Lindsay Comstock
Pondering the Liminal Space of AI in Art
Generation and discrimination are crucial to the artmaking process. Art is most often generative, even if it seeks to destroy in form or concept; artists must learn to discriminate between noise and truth to bring forth a clear vision.
David Young works with artificial intelligence as a tool for doing precisely these two things. Using an open source Generative Adversarial Network (GAN) program, Young trains the machines to recognize and generate images based on photographs he takes of flowers — in this case, dandelions — at his home in the Upstate, New York town of Bovina.
As the two-part program — the discriminator and the generator, which act as adversaries in order to create meaning — is shown images, it gradually builds a network of understanding, not unlike neural pathways in the brain. The discriminator is “shown” the images and begins a process of interpreting the binary data set of ones and zeroes. Then it communicates the information to the generator, which creates images. Through a weeklong process of back-and-forth communication and learning, the machines produce images that slowly morph from digital noise to recognizable form. Young then curates a series from these images, selecting for classical considerations of beauty.
Calling it “Little AI,” as opposed to Big Tech, Young — who studied design and AI at MIT and has always championed creative uses of new technologies — works with an intentionally small set of source images, just a couple dozen (instead of millions). By constraining the technology, the machine struggles to create something accurate; sometimes there’s a breakdown in learning and the output becomes unrecognizable. He breaks the technology to understand how it works. In the end, he prioritizes beauty and aesthetics over corporate focus on optimization, efficiency, and consumption.
As a conceptual artist, Young’s intent is to interrogate perception, asking viewers to suspend their knowledge of the plant and meditate on the new forms produced by emerging technologies. The images seem to ask: Can we look at a common dandelion without a frame of reference? Can we see beyond our associations with dandelions as harbingers of spring or childhood wish-making? Can we look beyond the plant’s role as weed or medicine?
The resulting images are otherworldly amalgamations of man, nature, and machine. A single image holds both realistic interpretations of the dandelion, in transition from flower to seed, and machine-hewn geometric patterns. The ambiguous focus of the machine-generated images might have only been achieved in the past through several images digitally composited in Photoshop. Printed using a dye-sublimation process on aluminum, the resulting works have a luminous, almost holographic quality. They seem to exist in a liminal space that is neither here nor there, external or internal.
Sometimes the machine renders patterns that existed in the data of the original image, but were not visible to the naked eye, and therefore interprets the dandelion in a way that appears foreign. These patterns reveal the limitations of human perception and observation. The way we see the world is not the whole truth of the world; it’s only one way we experience it through our senses.
What began as Young’s desire to gain a new understanding about the machine and how it learns became a deep devotion to the unfolding beauty within the process of art-making. He is concerned foremost with aesthetics in AI Art.
The tension between art and technology has existed since the advent of the camera and methods for reproduction. Like photography, which has long provoked debate by scholars and critics about its relevance as a fine-art form, people are slow to accept AI as art. There’s still the “shock of the new” with AI, even though artists have been working with it in various forms for the last fifty years.
Art critic Jerry Saltz wants us to judge AI Art through one question: Could this have been made by a human? If the answer is yes, he says, it’s not art.
While some artists instruct AI to produce art autonomously, others see it as a collaboration with the machine. Young envisions his place in the canon as part of a post-photography and “proto-something else” movement. What that something else is is yet to be defined. He resists anthropomorphizing the machine, seeing it as a tool rather than a collaborator.
Words become a trap, Young says, and it’s easy to give too much authority to technology, including facial recognition, which leads to various types of incongruencies and discriminations if we don’t use the emotional human filter through which to judge the information given to us by AI. Human discernment must enter in order to override machine discrimination. The works Young makes couldn’t exist without his data input. He creates and selects the images he uses to teach the computer programs. His eye is in the curation and selection process of the final works.
Like the Dutch Masters, who sometimes included objects in their still lifes that could not have existed together in reality, or the Hudson River School painters, who played up the sublime in nature, or photographic artists who use tricks of the camera or the computer to manufacture scenes, art made with AI also asks viewers to suspend their ideas about truth in form, even if the image appears photorealistic.
One could argue that every creative act, including this piece of nonfiction writing, includes elements of fiction. There is no truly objective point of view; “truths” are often conditioned perceptions.
It follows that the images the machines are being trained upon aren’t real in the first place — a photo is never a true capture of reality because it prioritizes qualities of light and shadow; it adjusts for exposure and white balance; the shape itself is a bias — what the artist chooses to leave out of the frame can be equally important as what he or she endeavors to include.
Machine learning as art punctuates a common trope in popular culture: we’re in a “post-fact” world. Rather than the idea that we are no longer wedded to truth in society as our last president made clear, we might relax our insistence upon arbitrary “truths” and begin to move through the world in ways that are new and beyond boundaries and more aligned with the universal Truths underlying existence.
Young’s art asks questions that artists have been asking for millenia through different mediums: What is understanding? What is representation? What does it mean for a machine to understand nature? And how can we better understand nature through the filter of the machine?
If the machines can show us anything, it’s that we’re constantly in process, constantly evolving, and that evolution isn’t necessarily linear. The work speaks to the ephemeral nature of reality and the present moment: Just as we think we’ve grasped it, it becomes the past. The dandelion, in its transition from flower to seed, becomes a totemic image of impermanence and the point of tension right before everything explodes and the seeds of change are disseminated. Just as the AI technology with which the images were made will soon become obsolete, the dandelion will wither. It’s in these small deaths that something new can emerge.