Page 366 - The Black Box Book

IN L IN E S
S H O W C A S E
Artists understandably try to make the presentation of their art on
the internet look appealing. Their websites therefore often conform to
contemporary ideas of an attractive web presence. My video is another
form of self-reflection on artistic practice. It is a dataset presentation in
which I situate myself in a kind of animated browser. In the language of
Youtube videos, I write and show the names of the artists represented on
camera with smiles. In this way I mimic the work of a ‘ring girl’ in a web
presentation environment (Figure 12).
Dataset analyses are another case of representing traditional and
emerging images of social reproduction in the online context. The online
space creates the illusion of a more democratic background for social
interaction. For instance, it makes it easier to perform under multiple
identities, and it liberates many forms of care that were previously unpaid
for (ASMR videos, etc.). Even through the practice of a recommendation
guide, which has been transformed in the online environment into an ac-
tivity of learning algorithms, is not immune to the pervasiveness of gender
stereotypes in our society. In addition, the gaze through the screen, or
the gaze of the camera, is always extremely aestheticized, spectacular,
and fetishizing. It imports an agenda derived from a cinematic language,
that remains a vehicle for the sublimation of a range of hitherto unre-
flected issues, and is closely linked to marketing strategies. The camera
exploits the soft tissue of the people and creates a pressure for visibil-
ity that escalates when combined with the apparent timelessness of
the online environment.
366
“In order for the AI to read the images and learn to recognize them, it
must first decompose them into rows, into individual pixels arranged
side by side. It then compares them to each other. If each image in our
dataset was decomposed into a single line that is 1 pixel high, and we
also wanted to rearrange the entire dataset in this way, print it out and
display it in a gallery, we would get a color print of 262,144 x 136,458
pixels, which is 53.33 GB. With current technology, however, it is not
possible to print such a huge image; it is not even possible to display
it on a regular computer. So, we can only approximate its visual
structure through a 100 times smaller preview to give you an idea.”
(Javůrek, T. & Meixnerová, M. & Trnková, B., 2022)
Another opportunity where we can visually evaluate one of
the technical processes associated with AI learning is the decomposition
of a dataset into rows. This step permanently breaks down the human-un-
derstandable content of the image and creates a pure abstraction, which
in turn begins to make sense to the machine on a mathematical level. As
artists, however, we were intrigued by the colorfulness of this abstract
result. We expected to see grey, as the dataset contains many black
and white photographs, typographies, and records of graphic sheets of
various graphic techniques and views of interiors where the predominant
tonality is white and grey. However, green is also significantly represented
in the result. This is probably due to the photographs of the artworks on
the exterior. In the spirit of the project’s hyperbole, however, we can now
declare that Czech art is on average olive green, even grey (Figure 13).
The color analyses of the datasets provide another field for possible cor-
relations between artificial and artistic perception.
367