Thesis Info

Thesis Title
Artificial Intelligence and Art
Penousal machado
2nd Author
3rd Author
Number of Pages
University of Coimbra
Thesis Supervisor
Amílcar Cardoso
Supervisor e-mail
Other Supervisor(s)
Language(s) of Thesis
Department / Discipline
Artificial Intelligence
Languages Familiar to Author
Portuguese, English, Spanish
URL where full thesis can be found
Artificial Intelligence, Evolutionary Art, Biology Inspired Art, Generative Art, Computational Creativity
Abstract: 200-500 words
The main motivation for this thesis is the desire to study the questions raised by the development of Artificial Artists – artificial autonomous systems of artistic creation with capabilities similar to the ones of human artists. The path taken frequently assumes an exploratory nature that conducts to different areas of knowledge and research. The first step of this journey consists of the analysis of the state of the art in the domain of the application of Artificial Intelligence techniques to artistic tasks. Based on this work, a set of relevant aspects for the development of an Artificial Artist is identified and a model for its development proposed. The second step encompasses the development of an interactive evolutionary art system, NEvAr, and its validation as a tool for artistic production. According to the presented model, this tool constitutes a generative engine that, once complemented with an evaluation module, can give rise to an Artificial Artists. The need to perform aesthetic judgments conducts to the study of the aesthetic and artistic phenomena, allowing us to propose a bio-evolutionary explanation for the artistic activity and establish a relation between visual perception and aesthetic. Complexity is seen as one relevant aesthetic principle, which motivates its use as a basis for aesthetic judgment formation. The attained experimental results confirm the relevance of complexity in aesthetics and the adequacy of the followed approach. The achieved success leads to the integration of the developed techniques into NEvAr, which is the first step towards the transformation of this tool into an Artificial Artist. As a consequence of the work pertaining the aesthetic evaluation of images, the concept of Artificial Art Critic arises. An architecture and validation strategy for this type of system is proposed. Following this methodology, Artificial Art Critics are developed for musical and visual domains, and experimental results in tasks of style and author identification and aesthetic judgment presented. Finally, the integration of these critics in NEvAr is discussed and the preliminary experimental results of the integration presented.