Skip to main content
notice

September PhD defense: Sofian Audry

September 1, 2016
|


Sofian Audry, Thursday, September 15, 2016, in the Milieux Resource Centre, EV-11.705, 1515 Rue St. Catherine W. at 10:00 a.m.

Thesis Title: Machines That Learn: Aesthetics of Adaptive Behaviors in Agent-based Art

Examining Committee:

Thesis Supervisor: Chris Salter, Design & Computation Arts

External Examiner: Dr. Peter Cariani, Biomedical Engineering, Boston University

Concordia External Examiner: Lynn Hughes, Studio Arts

Internal Examiners: Bart Simon, Sociology & Anthropology; Jean Dubois, Fine Arts (UQAM)

Chair: Debbie Folaron, Translation Studies                       

ABSTRACT

Since the post-war era, artists have been exploring the use of embodied, artificial agents. This artistic activity runs parallel to research in Computer Science, in domains such as Cybernetics, Artificial Intelligence and Artificial Life. This thesis offers an account of a particular facet of this broader work — namely, a study of the artistic practice of agent-based, adaptive computational artistic installations that make use of Machine Learning methods. Machine Learning is a sub-field of the Computer Science area of Artificial Intelligence that employs mathematical models to classify and make predictions based on data or experience rather than on logical rules.

These artworks that integrate Machine Learning into their structures raise a number of important questions: (1) What new forms of aesthetic experience do Machine Learning methods enable or make possible when utilized outside of their intended context, and are instead carried over into artistic works? (2) What characterizes the practice of using adaptive computational methods in agent-based artworks? And finally, (3) What kind of worldview are these works fostering?

To address these questions, I examine the history of Machine Learning in both art and science, illustrating how artists and engineers alike have made use of these methods historically. I also analyze the defining scientific characteristics of Machine Learning through a practitioner’s lens, concretely articulating how properties of Machine Learning interplay in media artworks that behave and evolve in real time. I later develop a framework for understanding machine behaviors based on the morphological aspects of the temporal unfolding of agent behaviors as a tool for comprehending both adaptive and non-adaptive behaviors in works of art. Finally, I expose how adaptive technologies suggest a new worldview for art that accounts for the performative engagement of agents adapting to one another, which implies a certain way of losing control in the face of the indeterminacy and the unintelligibility of alien agencies and their behaviors.




Back to top

© Concordia University