This work draws parallels between biological diversity and the diversity of generative images, examining themes of life, death, and evolution within the realm of the imaginary. The AI-generated images resemble body parts, clothing, microorganisms, jewelry, mineral fragments, fruit, or imaginary creatures that can appear attractive at one moment and repulsive the next. They simultaneously evoke familiar objects and defy complete recognition. The assembly and disassembly of digital textures mimic biological patterns, sharing the same material but differing in organization and appearance.
Generators, Discriminators and In-Betweeners
Ulya Soley, 2021
How does it feel to look at the objects, plants, and animals in the world, to learn and distinguish them one by one when you have not learned anything yet? We experience this at the beginning of our lives, and unfortunately, we cannot remember what it feels like. Kerem Ozan Bayraktar's four-channel animated video titled Modes (2020) shows a few hours of images that are familiar in one aspect but appear in unfamiliar forms when viewed in their entirety. It offers the viewer an experience of what it might feel like to exist in a world where images are not yet familiar.
The four screens, mounted on a metal axis, are divided into square areas in a grid arrangement. In these areas, each image appears and disappears for a while, and then reappears, sometimes slightly changed, in another square. Bayraktar produces these images using the GAN algorithm. GAN stands for Generative Adversarial Network. It is activated through two networks working in opposition to each other: the Generator network and the Discriminator network. While the Generator generates data that resembles reality, the Discriminator learns to distinguish false data from real data. The more data entered into the system, the more realistic the images produced by the Generator and the higher the discrimination capacity of the Discriminator. Using this algorithm, the artist brings together different objects, creatures, plants, and animals, performing inter-species matching in the digital environment.
Although it is possible to establish an analogy between the Generative and Discriminative actors of GAN and the artist and the viewer, Bayraktar is neither trying to produce data that resembles reality like the Generative network, nor does he expect the viewer to distinguish these data accurately like a well-trained Discriminator. The resulting visualization is complex and confuses our minds. The forms are very familiar yet unidentifiable—shapes we have not encountered before, but which we somehow recognize, giving the viewer an uncanny feeling with their coexistence. As we watch, this hypnotic visuality evokes different associations and familiar possibilities, but our efforts to define, categorize, and understand them are always fruitless.
At first glance, these images seem to suggest a biological aesthetic, giving us an idea of what the inside, the microscopic, looks like. The language we have learned about the functioning of biology and the diversity of species—the reproduction of species, the copying errors that occur during reproduction, and the new species that arise from these errors—also resembles the method of the algorithm in the video. The new species that emerge from the combination of different objects or organisms continue to evolve into each other, and this is the transformation we experience in the video. The same data appears again and again in different forms. Each moment we witness appears as a new formation. The transformation we encounter in this digital structure mimics biological scenarios when viewed from a distance. A form is born, merges with other forms, transforms, and disappears. This evokes the cycle of birth and death. However, this process of appearance and disappearance does not follow linear time. Bayraktar divides a nearly three-hour-long video into frames, eliminating the distinction between before and after, and creating a timeless plane that is the scene of transformations. We witness a cross-section of an image's life; each time it appears, it is slightly different from the previous one, but it is not possible to distinguish between old and new.
Another element that defines the boundaries of the video and encourages the viewer to persist in identifying these objects is the division of the screens into squares in a grid pattern. Inspired by natural history museums, Bayraktar reconstructs the aesthetics of a traditional collection exhibition in a much more dynamic way. Since this new construction does not allow for a static reading, it renders the rigid and categorical grid layout dysfunctional.
Each time the images appear, both their forms and their locations change. This reinforces the dichotomy between the defined and the undefined, and between the regular and the irregular, placing the viewer in an uncanny position, caught in the middle. Just as biological (natural) processes and digital (artificial) processes, which are thought to be opposites, work in a very similar way, could it be that these dualities are actually more transitive and transformable concepts than we think? Can we forget the oppositions we set up in order to unlearn and look for the answers in the transitive relations or in-betweens that things establish with each other?
GÜNCEL
08.02.2021 • SAYI 2 6
Dosya: Müstesna Kadavra
Galeri Nev
Orijinal Türkçe Metin