More and more the way we communicate and live our daily lives is influenced and led by technology.

Again and again we need to learn how to understand and use the new forms of communication, at the same time we have already started to entrust the technical apparatus and technologies behind them to interpret information.

What happens when we transfer the crucial part of a communication flow to machines and let them filter and steer it?

Sure "someone is getting the donuts" - but who?

"Talking Machines" takes this issue up and calls for the interpretation of the observer.
A recording of John F. Kennedy's speech from the “Cuba Missile Crisis”, which was recorded in an endless loop, interpreted and given out by a speech module, is used as a starting point of the installation. The text develops iteratively, similarly to the game "Chinese whispers". In the conflict of a permanent deconstruction between announcement and interpretation, an absurd, almost philosophical word braid is being developed.

On the flowing, constantly reforming speech, another technical procedure is applied. A real-time computer-based Sentiment analysis figures out the emotional development of each newly created text section. This course is documented on a scale of +10 to -10 by a drum recorder- first of all digitally and then analogically. The interpretation of the conversation on an emotion level adds on a very human touch to the "machinery texts".

Just as the text becomes more and more abstract to the observer, fragments in it seem to make the meaning more recognisable. With the emotional evaluation of these texts, a measure is applied that is not applicable as an so called orientation for people on machines "produced" texts but that can nevertheless open a further assessment dimension.
The decision with regard to the external shape of this installation is not so much due to aesthetic considerations, but results from the choice of the source text. The gramophone horns reinforce the historical reference to the content. Moreover the form visualises the inherent mechanisms of compression and distraction of data and their meanings of the outlined process.

The course of the Cuban missile crisis was in large parts determined by the communication of the U.S. and Russia that is still regarded as a lesson for prudent diplomacy. Here too much depended on the successful and unsuccessful interpretation of what was said, what can be caused by "hissing" like it is described in Shannon’s sender- transceiver-model. To provoke this hissing, directional microphones are used to for the installation to record the texts that allow the addition of "hissing" or even of a new text from the observer. The drum recorder recorded separately the fluctuations in the measured value of emotion, even though the motors that move the arms to mark the top and bottom, work inaccurately at the present state. Therefore, in the future the reproduction of the curve has to be represented purely digitally.

In nearly every test it was noted that after about five runs the emotional development of the text fell into the negative scale. The not entirely serious question should be: Can computers have an evil being?
The entire structure is implemented with Processing and Arduino, which exhausts the performance of the two settings strongly, but also allows to work exclusively with open libraries. The basis is the Speech-To-Text Library of Florian Schulz and the Text-To-Speech Library of Nikolaus Gradwohl. The starting text is spoken or played from a device, so that computer one can automatically go on recording. The soundtrack is then analysed and transcribed and given out as a written text. At this point the text is passed via HTTP POST to emotion analysis ( and returned as a numerical value according to its polarity (neutral, positive or negative) in the form of a JSON object. This data is formatted and rewritten so it is readable for processing. Based on the calculated data, the graph is recorded digitally and the newly interpreted text is spoken by the tts Library. In an infinite loop this process is alternately performed between the two computers.
Jonas Loh
Steffen Fiedler
Jurij Tschabanow
Andreas Grycz
Florian Schulz
Ceri Mitchel
Talking Machines

Talking Machines

A recording of John F. Kennedy's speech from the “Cuba Missile Crisis”, which was recorded in an endless loop, interpreted and given out by a spe Read More