Pro(s)thetic Dialogues

The construction of two positions which eventually merge. It is impossible to engage in a dialogue with a machine in the true sense of the word. A machine’s feedback is not a dialogical response. A machine is monological. It can update itself. Its “opinions” are original, unduplicated, reflect a paradigm, and transmit a discourse. A machine turns into its own prosthesis of inner dialogue. Speech arises at the border between authorial and collective texts as a snapshot of collective consciousness.

Pro(s)thetic dialogues is more like a recording of a theatre performance playing out on a computer desktop. Here the human operator creates the conditions for exploring the performativity of a philosophical zombie pieced together from neural networks. The operator/stage technician merely directs the flow of mathematical operations pretending thought and creates a space in which the philosophical discourses of the following authors are dissolved: Giorgio Agamben, Günther Anders, Hannah Arendt, Alain Badiou, Gilles Deleuze, Roberto Esposito,  Michele Foucault, Vilém Flusser, Byung-Chul Han, Donna J. Haraway, Michael Hardt, Friedrich Georg Jünger, Warren Montag, Antonio Negri, Peter Sloterdijk, Paolo Virno, and Slavoj Žižek.

The performance presents an attempt to peer critically into the black box of individual generative models, but questions remain as to whether in the end their program has simply been expanded with more or less likely configurations and at what price technology brings these new possibilities.

  Georgy Bagdasarov , Alexandra Morales  

If there are zombie philosophers, they would be able to make no sense[z] at all of the other minds problem. They might, of course, be able to ponder[z] an ‘other zombies’ problem: how do we know[z] that there are not other beings whose experience[z] is accompanied by a quality that we cannot fathom[z]? But it is unlikely that this problem would occur to them, until they met us.

Todd Moody, “Conversations with zombies” (1995)

You can’t wake a person who is pretending to be asleep.

Navajo Proverb

The pro(s)thetic dialogues could not function well in any situation, because in it they function incorrectly, and in their situation the dialogues are not good. Moreover, this inability to function was due to the fact that the apparatus are continually changing, and their dialogue can not be easily ordered, if at all. The dialogues currently engaged are not functioning very well, so that the need for new dialogues seems to have reached an almost impossible state. The functionaries must decide, to “play devilishly,” how to replace dialogues with live events, or to make them better: “do not turn the lights on,” or “beyond the hour” (to put it more broadly, to give the functionaries more creative freedom).

generated text

Georgy Bagdasarov, Alexandra Morales: Pro(s)thetic Dialogues - video

Borjana Dodova: Pro(s)thetic Dialogues - essay

In 1950, the British mathematician Alan Turing designed an experiment that later became known as the Turing Test. This “imitation game”, as he called it, dealt with the question of whether machines can think. Turing offered a radical simplification of the problem. Instead of the difficult to pin down definition of a “machine” and the even more elusive definition of “thinking”, his test focused on action – on human interaction with an artificial intelligence. A non-living system’s ability to fool a human became the equivalent of thinking. In its original formulation, the test consisted of a machine, a woman, and an independent judge. The judge sits in a separate room, from where he communicates via a text terminal. His goal is to correctly identify the actors in the game. Meanwhile, the intelligent machine attempts to create a deceitful illusion with its answers so that the judge mistakes it for the woman.

In their audiovisual work Pro(s)thetic Dialogues, artists Alexandra Morales and Georgy Bagdasarov relate to imitation games but at the same time upend their established rules. For their basic building material they chose philosophical speeches generated and orated by an artificial intelligence. But while imperfection (read: inhumanity) is not welcome in the everyday variety show of chatty machines and humanised robots, Pro(s)thetic Dialogues continuously reveals identities, roles, tools, and mistakes. The authors are not attempting to create a flawless illusion – after all, the possibilities of cultivating neural networks in DIY conditions are still limited. An opportunity is born out of the admitted imperfection. A recording of a conversation with an artificial system involves not only the obligatory synthesis of text and image but also the revelation of the techniques used, the explicit execution of edits, ad-hoc changes to inputs, the reconfiguration of context, and the debugging of code. The viewer, like it or not, becomes acquainted with the process of the work’s creation. The illusion is disrupted, and so there is no choice but to concede that the main performer is nothing but an intelligent machine in the hands of a human.

Alan Turing refused to concern himself with the content and quality of thought. But wasn’t this merely an evasive manoeuvre? Today we know that intelligent machines can fool us. Deepfakes are increasingly convincing. So shouldn’t we finally start to ask what we expect from encounters with intelligent machines? What do we gain from them and what do we lose? What conversation should we encourage them to have? What prompts should we prepare for them? If we give a neural network access to a collection of philosophical texts, will we learn something new about the world?

Alexandra Morales and Georgy Bagdasarov leave these questions open. Instead of answers, they offer ironic commentary in their work. In spite of the huge technological strides made in recent years in the field of machine learning, the synthetic man from Pro(s)thetic Dialogues is in many ways similar to the Rogerian psychotherapist ELIZA from the mid-1960s. The ELIZA chatbot reduced communication to general questions in which the human-patient is simply mirrored in an unsatisfactory conversation. The synthetic man from Pro(s)thetic Dialogues similarly offers an endless set of monologues, also with no obvious direction. The imitation is refined, but thought continues to elude. Perhaps in the end it is most important to know that when the machine gets stuck in a loop, it is possible to switch it off and start again.

Borjana Dodova

Scientific instruments used

 First Order Motion Model for Image Animation by Aliaksandr Siarohin, Stéphane Lathuilière, Sergey Tulyakov, Elisa Ricci, Nicu Sebe.
Training Generative Adversarial Networks with Limited Data by Tero Karras, Miika Aittala, Janne Hellsten, Samuli Laine, Jaakko Lehtinen, Timo Aila.

 Training Generative Adversarial Networks with Limited Data by Tero Karras, Miika Aittala, Janne Hellsten, Samuli Laine, Jaakko Lehtinen, Timo Aila.

 A Lip Sync Expert Is All You Need for Speech to Lip Generation In The Wild by K R Prajwal, Rudrabha Mukhopadhyay, Vinay Namboodiri, C V Jawahar.


Natural TTS Synthesis by Conditioning WaveNet on Mel Spectrogram Predictions by Jonathan Shen, Ruoming Pang, Ron J. Weiss, Mike Schuster, Navdeep Jaitly, Zongheng Yang, Zhifeng Chen, Yu Zhang, Yuxuan Wang, RJ Skerry-Ryan, Rif A. Saurous, Yannis Agiomyrgiannakis, Yonghui Wu.


Language Models are Unsupervised Multitask Learners by Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever.

Authors

Authors: Georgy Bagdasarov a Alexandra Morales
Translations of the texts: Brian D. Vondrak
Editor: Janek Rous
umělá inteligence English friendly performance video-esej programování digitální platformy