The Thing That Isn’t


The Thing That Isn’t, 2016
electronics, software and drawings



Equipped with two tiny computers, electronics, speaker and software, ‘The Thing That Isn’t’ describes, reflects and comments on the making of robot drawings in process. A computerized voice reflects on the drawing process and adjusts it according to the drawing movement it finds in its code. The resulting robot drawing is a mixture of reflection on the code and of the movement created during its making. When the drawing is finished a booklet with the spoken text will accompany the completed drawing as part of the final product.

Example of transcript:

so what embodies information technology sampled, precisely?
decidedly frozen,
good there …

second level.
after this
Turn an unnecessary corner again, orient!
forget this now,
such and such
That appears to embody a democratic point…
turn leave here because my opinion be indecipherable .
equitable nuance.
belittled my undertake, she bequeath, not fail.
accurate sensuality completes the market?

No no no nord-west .
part of
each day be a little balcony .
Middle blow.
anything goes


Long description of the project:

The Thing That Isn’t by Jerry Galle combines a word-processing system with a drawing robot. The work is controlled by two computers, which use electronic components to maneuver the drawing arm and broadcast spoken sentences through the built in speakers. Thus the automated system executes a program that determines its drawing. The program also examines the drawing during the process, comments on it, and modifies its subsequent behaviour. Galle fed the learning machines with texts, discussing themes such as artificial intelligence or the CIA’s interview techniques, as well as essays about art and pedagogy.

Firstly, the texts are taken apart into smaller units, which are classified according to shared traits and arranged under various headings, such as art, language, or psychology. Secondly, the whole text is harvested for connections between individual words, for example synonyms, similar dictionary entries, etymological connections, and frequency of word use. Thirdly, the program generates a decision tree – a tree-like representation of the connections and correlations within the body of text. This decision tree creates vectors, which the movements of the drawing head are then aligned to. At the same time, an independent sentence is generated using words from the original texts and the decision tree. The sentence determines the actions of the drawing arm and are verbalized, so spectator can perceive the essence of this automated thought process.

In this never-ending process, ‘The Thing That Isn’t’ searches for the correlations between words and their possible meanings. This doesn’t lead to a conclusive interpretation, but rather to a network of many interrelated interpretations. Information is extracted and repeatedly altered. Does the loss of information in the translation process come about because the field of terms is constantly expanding? Or because the software’s abilities are expanding to include complex ‘thought’? Such questions arise concerning the abilities and limits of algorithmic perception. Would we trust a machine to have subjective, aesthetic experiences? How meaningful or absurd is it to have machines simulate human behaviour?