The Artifice of the Image : Language and Image in a Networked Environment
Produced by: Alexander Velderman
This grid of images was generated using a custom-trained implementation of Style Gan-2 by Nvidia. The Generative Adversarial Network (GAN) synthesizes unique images that are generated based on a set of training data heavily inspired by the work of Anna Ridler and her work Mosaic Virus (2019), which explores imagery and abundance. In this case, I trained the network on images of various monkeys to test whether the network would be able to generalize a primate or if it would simply make monstrous, soupy, three-eyed creatures. The end result was a halfway point of recognizability and abstraction from the original content. I thought monkeys would make the interesting subject matter as StyleGan was originally trained to generate human faces. Additionally, monkeys frequently appear as both a relatable form with images of them displaying very humanoid seeming behavior and their presence as a meme similar to cats and dogs. I was interested in the way Machine Learning (ML) would not only interpret the images, but process them. This piece considers several questions. What are the ways that these images might be interpreted? How can these interpretations extend further into how we understand images and what “monkey” truly means in our imagery? Finally, do these images have utility or will they simply be forgotten? In my research and artistic practice, I have chosen to focus on updating the ontology of the digital image.
What role do images play in the way that we communicate and curate our personal and cultural memories? How does the unlimited bandwidth and resolution of the present digital world change our self-understanding? Faced with the sheer ubiquity of images, the task of reading and interpreting images on a day-to-day basis is relegated to a position of pure reference to images as objects. Subsequently, we have created a lockstep between imagery and traditional language, wherein they are both modified by the networked and technological structures through which we communicate. In the following paper, I consider the relationship between machine language and Natural Language via Images. Specifically, I examine how, in the place of natural language, images have become not only a means of technology, but an emergent form of natural language themselves. In doing so, I argue that the amorphous role of the networked image, which functions somewhere between evidence and communication, keeps us stuck in the present, whilst continually reanimating our past. This line of reasoning harmonizes with Mayer-Schöenberger 2019 regarding the consequences of keeping data indefinitely. He positions the selective process of memory and forgetting as a key element of our humanity, which allows us to respond to each other and our world in the moment (Mayer-Schöenberger 2019).
I have encountered a divide in how images are thought of by those who were born with and without access to high-speed internet. Younger people are more likely to interact with the online world as something innate. In turn, this comfort and ease of engaging in online space become a place in which images are shared at a rapid rate, and proliferate inheriting the traits and memories of their viewers. Although this is not unique to internet culture, the difference is the ease, speed, and development of visual lexicons that are appropriated or spawned out of unique spaces.
From this foundation, I am interested in the ever-changing role that archives play in image and network culture. One work that is particularly formative to my practice regarding image archives and the evolving conception of the image is Aarti Akkapeddi’s Ancestral Apparitions. They explore their own familial and migrant archives through the hazy lens of images generated by a neural network in the style of old family photographs. The work explores this generational change in looking at photographs and images as modes rather than moments. They generate images, not of their family in particular, but of a family that has been trained on their own family. The result is an archive that fits into the mode of a family archive, which is familiar in look and feel, but is yet uncanny and impersonal.
Artificially generated images bring this divide clearly into view. In a networked environment, where the ability to create and share images is unrestrained, images begin to take on forms analogous to language by being able to convey essence holistically without definitive clarity (Heidegger 140). As machines require the highest degree of fidelity in linguistic tasks for the sake of clarity with little room for interpretation, they lose the very essence of language and become information (Heidegger 141). This is why, oftentimes, the distinction between natural language (spoken by you and I) and machine language ( designed for clarity and execution) becomes clear when discussing the discipline of technology itself. As Heidegger cites Humboldt: “Language is not merely a means of exchange for mutual understanding, but is a true world that the spirit must set between itself and objects” (Heidegger 139). If we consider the language used by the machine, we find that it stops at the first half in its mutual understanding, but cannot give us further depth into how essence is created relationally. To further unpack this notion, Heidegger notes that “everyone incessantly speaks and their speaking still says nothing” (ibid 140). In effect, this notion posits that the form and context of language precedes the importance of content in creating relation and gesture. This same phenomena of speaking without saying anything can be extrapolated into the way images are circulated, evolved, and grown online. Their origin is a mirage, unlike the referential photo of yesteryear, They are not representative of time and place, but of their own self reference.
While drawing the direct line between images and traditional forms of language may be tempting, I do not think that images supplant the role of language, but rather, act as a crude meta-language. I believe that the best way to pull this apart is to look at the work of Tacita Dean, entitled The Green Ray, in correlation to Walter Benjamin's The Translator’s Task. Dean’s 16mm film is a short narrative about capturing the green ray, a rare phenomena of refraction usually observed by sailors at sunset. The narrator contrasts her experience and subsequent capture of the green ray on her film camera to a couple next to her filming the sunset digitally. Dean states that, while filming, she believed she saw the ray but was dissuaded upon watching the couples’ digital playback, which did not contain the green ray. Upon developing her film, her mind was once again changed; she claimed that she saw the ray on the frames, despite it never appearing in the film. While the digital is immediately available to us, with cold indifference, the finite process of film processing proves to be the very process of viewing itself. The example of Dean’s work represents the awkward place that networked images occupy, existing between a reference to its own image and language. Although her work was more a love letter to film, it is a perfect example of the oscillation that any networked image occupies. This kind of in-between manifests in the very place she sets herself in opposition to in the film. Had she remained true to her sentiment she would have in fact refused to have her film scanned.
Looking to Walter Benjamin’s Task of the Translator, his insights are applicable between this awkward misstep that networked images have created. It is in this space that there is a loss of sorts - not of meaning in the literal sense, but the very essence of the meaning itself. Benjamin uses the example of translating the word bread, the German word brot, and the French word pain ( Benjamin 257). These words mean the same thing: bread. However, a hearty german bread is not the same as a light french bread, which presents the problem plainly. Benjamin goes on to argue that if separated for long enough, languages may evolve to the point where pain and brot would no longer come to signify the same thing. Similarly, with images, the relationship between reproducing, documenting, and generating new images entails potentially infinite translations.
The possibility of endless and lossless translations of images begs the question. How do we deal with faultless memory? In the development of our technologies, data is assumed to last forever (Pitsillides 11). Does this mean that we not only hope to preserve all images, but also reanimate them? Is having both perfect memory and infinite language to articulate that memory of benefit to how we communicate? There must be some consequence to being stuck with this vast plethora of extraneous information. Viktor Mayer-Schönberger argues that what allows humans to move forward is the ability to forget, without which we would be “unable to generalize the abstract” (Schönberger 2009). As we lose the ability to forget, due to the perpetuity of the networked image, does culture become stuck in a quagmire of the past, doomed to eternal reflection? If our memories have become technology and our very essence and interaction is technology, then we have become relegated to a hyper-reality that we cannot escape. This issue is brought up by Stacey Pitsillides et al. 2012 in reference to their notion of the Museum of Self, through which they question the value of remembering everything and what societal norms this shift might have. If we continue to generate more extraneous data, will our conception of communication simply be subsumed by technology itself? It seems that the answer would be yes. We build our communication atop the very catacombs that we construct with data that we determine to be necrosed. However, those catacombs are built on the false assumption that they cease to exist. Rather the death of a disused image becomes one of purgatory, stuck between its original and disuse.
To communicate the constantly changing role that contemporary images take on requires a broader survey of not only viewing practises but linguistic practises as well. In this paper, I have described the role that images take in my artistic practice. I have also shown how images function as a meta-language that evolves through networked communication streams. Finally, I have considered the potential ramifications that come about when imaging shifts towards being used as language and to our ability to culturally process this change. The digital does not forget. In the absence of a hard override, we must contend with our newfound abilities to constantly live reality in the past. Through this work, I have delineated my conception of the image and reified the importance of forgetting that is embodied by non-digital artistic practices.
Akkapeddi, Aarati. “Ancestral Apparitions.” Aarati Akkapeddi, 2020, aarati.me/#aa.
Benjamin, Walter. Task of the Translator. The Belknap Press of Harvard University Press, 1996.
“Chapter One .” Delete: The Virtue of Forgetting in the Digital Age, by Mayer-Schönberger Viktor, Princeton University Press, 2009.
Dean , Tacita, director. The Green Ray . YouTube, www.youtube.com/watch?v=A9meDXPhKIo.
Gregory, Wanda Torres. “Heidegger on Traditional Language and Technological Language.” The Paideia Archive: Twentieth World Congress of Philosophy, 1998, pp. 119–130., doi:10.5840/wcp20-paideia19986142.
“Mosaic Virus, 2019.” ANNA RIDLER, 2019, annaridler.com/mosaic-virus.
Pitsillides, Stacey, et al. “Museums of the Self and Digital Death: An Emerging Curatorial Dilemma for Digital Heritage.” Design Department , 2012.
“The Iceberg.” Giorgio Di Noto, www.giorgiodinoto.com/4222149-the-iceberg.
Mario. “10 Monkey Species.” Kaggle, 28 June 2018, www.kaggle.com/slothkong/10-monkey-species.
NVlabs. “NVlabs/stylegan2: StyleGAN2 - Official Tensorflow Implementation.” GitHub, github.com/NVlabs/stylegan2.