Stock Images - A Phantasmagoria of The Apparatus. In: Towards Technosophy (2022)
“Images are mediations between the world and human beings. Human beings “ex-ist,” i.e. the world is not immediately accessible to them and therefore images are needed to make it comprehensible. However, as soon as this happens, images come between the world and human beings. They are supposed to be maps but they turn into screens: Instead of representing the world, they obscure it until human beings’ lives finally become a function of the images they create.”1)
As an artist, I am interested in the relationship between science, culture, and visual representation. I am interested in how representation affects norms of meaning-creation, questioning the consequences of the acceptance of photographic images as purveyors of truth as well as our relationship with and understanding of images in a post-photography historical context. Do we still trust images in this day and age? As Addis Goldman and Alex Langstaff point out, “we are trapped in the double bind of being increasingly suspicious of photographs, yet more and more reliant on them to organize our world.”2)
The digital transformation of images has augmented the transformation of images into pure data. We all live surrounded by data images. Images have shifted into a source of raw material that feeds sophisticated autonomous learning machines that are able to recognise, categorise and classify objects, humans, and spaces. This automatisation of the process, be it the photochemical process through which a photograph is produced, or the functioning of an algorithm, is the very basis for our at least relative trust in the objectivity of what we look at. The photographic image is no longer only a visual representation of the real world but rather a large pool of visual information transformed into readable, classifiable, and evaluable data: the evolution and transformation of the photographic image from an object indexically correlated to the real object photographed to strings of zeroes and ones stored in a computer.
As Tom Gunning points out, the “digital revolution” doesn’t really challenge indexicality per se: “An index need not (and frequently does not) resemble the thing it represents. The indexicality of a traditional photograph inheres in the effect of light on chemicals, not in the picture it produces”. […] The “truth claim” of photography […] relies on both indexicality and visual accuracy,” 3) the complex relationship between indexicality and iconicity.
Technical images seem to be ubiquitous in today’s world. For this project, I focused on stock photography, which has been gaining hold in the 1920s and has since become a specialty in its own right.
The stock industry has quickly transitioned to the digital realm, now fully on keywords for sorting and retrieving photographs to be identified, through their metadata. Microstock created a whole new field of photography that revolved around generic images and ideas that could fit a range of contexts. Generic photos were the basis of the industry, however, the field was full of clichés. In the last 10 years, the look of microstock photography has shifted, trying to move away from obvious clichés and embracing a certain authenticity: “In 2019, clients and photographers prioritize the mood and feeling in shots, and pay less attention to technically correct shots. It is unique perspectives and real-life imagery that define the market today, by representing diverse cultures and backgrounds.”4)
And yet, I believe the recent shift towards a self-proclaimed “authenticity” has happened only on a very surface level, as the metadata layer still exposes the limits and biases still very much present as a manifestation of the Flusserian’s apparatus at play.
In the context of stock images and image automatisation, image and text can’t be disentangled. It is safe to say that text, not images, is at the real frontier of our new cultural-political AI reckoning. Algorithmic means of image production and recognition rely on semantic vectors to ascribe emotion, intention, gender, and racial identity, although these may not be perceptible to the naked eye.
As a starting point, I took the Flusserian concept of “the gesture of photographing,” which he defines as “the movement of doubt…the philosophical gesture par excellence”5) and applied this on a stock image platform,6) selecting the first three most popular results.
Figure 1 Modern seniors taking picture of themselves
Figure 2 Multiracial group of young people taking selfie
Figure 3 Young woman taking picture with smartphone
Technical metadata can be seen as an inherently productive and creative means of translating between image and text and code. I wanted specifically to look at how the metadata expresses the relationships between the core components of the images I collected. I wondered about how the text of the metadata would describe:
- The person taking a picture
- The gesture of taking a picture
- The photographing device
- The object of the picture
Looking at how these layers of information take shape on the image text’s level, its metadata, we can easily see how the language reduces the interpretation of its visual clues into stereotypical keywords. I looked at each of these images’ metadata, isolating the terms that corresponded to the parameters I have outlined above, creating word clusters:
Figure 4 Metadata text for the image “Modern seniors taking picture of themselves”
Figure 5 Metadata text for the image “Multiracial group of young people taking selfie”
Figure 6 Metadata text for the image “Young woman taking picture with smartphone”
Is this “invisible” to the naked eye layer of information a way for the apparatus to shape the way we will be looking at the pictures? Can this biased process be deactivated at all, given the need for metadata to present short keyword descriptions that allow us to find these images on a search engine? How could we try to turn stock images into Flusserian’s impossible images? Maybe a way to hack the apparatus is to explore what would happen if we removed all the semantic clues that could suggest a biased interpretation of the image and focused on the pure “gesture of photographing”.
Figure 7 This is the metadata referring to the gesture of photographing, collected from the previous three images
By merging the previous three examples of stock images, deploying the layers of extra information that they provided, and keeping to the bare clues that point to the gesture of photography, I wanted to move a step closer to creating an impossible image. What can be interesting is the idea to shift the focus of image making as a generative rather than derivative approach from its metadata.
This project is still ongoing, as the impossible image can stretch itself out of the binary field of the visualtext field and move towards a sensuous exploration of reality.