Page 368 - The Black Box Book

M O C K U M E N T

“Andreas Gajdošík and Vilém Duha have uploaded the works con-
tained in the dataset into the Google Open Images crowdsourced
dataset to tag them as art. Before that, this tag had contained just
an insignificant number of items. As a result, neural networks which
will be taught on this popular dataset in the future shall perceive
the notion of art in favor of the Czech visual art. A documentary is
devoted to the potential impact of this action.”
Aimee’s monologue – digital guide to the exhibition AI: All Idiots
(Figure 14)

It is an artistic gesture expressing the authors’ good-natured effort to
make Czech art visible not only on the world art scene, but even to confuse
what will be understood as art in the future with Czech art. The creators
uploaded a large part of the All Idiots dataset through Google’s Crowd-
source application and web interface to Google’s Open Images Dataset,
which contains nine million images and from which other learning algo-
rithms most often draw. The hypothesis and conceptual intent was that
by creating a surfeit of Czech art in the category/tag “art” in a major
dataset, a position of power will be gained that can potentially manifest
itself in the future – artificial intelligence of the future will understand
the category of art based on examples of Czech art, i.e., the category of
art for artificial intelligence will potentially fully merge with the category
of Czech art. However, it was more important to emphasize the important
role of dataset content in the development of learning networks.

Figure 12: Showcase,
Barbora Trnková, AI: All Idiots,
the MeetFactory Gallery, Prague, 2021,
photo: Katarína Hudačinová
368 369