Page 356 - The Black Box Book

tally clean up part of an artwork. The categorization of art is a delicate,
time-varying dialogue that runs parallel to its creation, involving artists,
theorists, curators, gallerists, critics, historians, and viewers, all of whom
could hardly agree not only on classification but also on categories.
The simple, accelerated labelling that would be required to prepare a bal-
anced dataset is, in our view, beyond (not only) art.
Matěj Smetana is an artist known for experimenting with the pos-
sibilities of animation and the idea of an algorithmically moderated data-
set intrigued him. By automating the ordering of images in the dataset
according to a certain key, four films were created with a specific narra-
tive based on the conflict of original and emergent contexts (Figure 7).
“Artificial intelligences excel in recognizing patterns and extraordi-
nary mechanical memory… They are geniuses and idiots at the same
time; idiot-savants. The dataset can be sorted in various ways. Half
a million images can be arranged according to their similarity, col-
our, composition of image elements; their size, symmetry, shapes
and so on… Sorting is the basics of Artificial Intelligence. Thanks to
sorting, AI can learn various things. We humans perceive the rapidly
changing images – in some way similar – to a movement; simple
animation; This effect is caused by the persistence of vision and
the processes in the human brain. By simple sequencing of images
connections are created perceivable only by human beings, that
are hidden to artificial intelligence. This is another big difference
between how the dataset is perceived by oneself and by artificial
intelligence. Sequencing 136 458 images according to one clue has
taken the artificial intelligence approximately fourteen days. There
are four sequences in the installation: created by visual similari-
ty, color, composition and surface and lines. Matěj Smetana has
placed the simple animations on top of four robot vacuums, and
let them cruise the Meet Factory. The vacuum cleaners represent
the seemingly utilitarian development of artificial intelligence devel-
opment and of technology in general. When installed in the gallery,
the vacuums steal away the job of cleaning workers, as artificial
intelligence is expected to do in a number of professions. But these
robots are just relatively stupid automatons. This is not the first time
robot vacuums have been used in a gallery installation. In relation
to this topic it can even be considered an ironic installation cliché.”
Aimee’s monologue – digital guide to the exhibition AI: All Idiots
356
R EP R E S EN T A T I V E
Through extractive, selective, and sorting processes, algorithms reduce
and trivialize the complexity of the original situations. Reduction and sub-
traction belong to the palette of techniques used for artistic expression,
for example in sculpture. In our project, we thematized reduction through
its hyperbolization: a single representative of the entire dataset was dis-
played in the gallery installation of the exhibition (Figure 8).
“One image has been randomly chosen from the collected data-
set - a digital reproduction of a painting, which was later loaned
for the exhibition to stand as a representative of the whole da-
tabase collection. A thumbnail image selected by the algorithm.”
Aimee’s monologue – digital guide to the exhibition AI: All Idiots
S TAT I S T I C D A TA
“The chart on the wall is based on the original complete dataset
of the Czech art scene created for the AI: All Idiots project. Who
are the heroes? Who are the outsiders? Who is the best Czech art-
ist? Only when you understand the data can you start the teach-
ing process. The dataset serves as a source material for artificial
intelligence I have trained, as well as for invited mediators from
the ranks of artists: Andreas Gajdošík, Vilém Duha, Matěj Smetana,
Jana Bernartová, Barbora Trnková, Tomáš Javůrek.
In Meet Factory, they present artistic outputs based both on
the dataset and on the outputs of the artificial intelligence that is
also processing this dataset. The crude language of numbers and
comparisons, rough cuts into the soft matter.” Aimee’s monologue
– digital guide to the exhibition AI: All Idiots
We have downloaded 800 names of artists from Artlist.cz. Almost
half of them did not have their own websites. 476 artists had websites,
but only 456 were active. 20 had a website, but it was not possible to
download images from them. Only 398 addresses could be used to fill
up our database. We used a bot programmed on the Nightmare library
to automatically aggregate image data from websites.
In order to show how the artificial intelligence software analysis
performed on the datasets, we distinguished female artists from male
artists. The simplest way to use the algorithm to separate these two
groups was to use the Czech spelling rule in our case, since the dataset
primarily contains Czech names, and identify the women in the dataset by
the surname hyphenation flag of the ending -ová. The statistical deviation,
i.e., the number of female artists who do not hyphenate their surnames,
appeared negligible in terms of pragmatic data. There were 565 male art-
ists in the dataset, but 192 female artists, or only one third (Figure 10).
Furthermore, the data show that women have proportionally more active
357