de Propios Martínez, Cristina "Art and artificial intelligence: machine learning techniques in the contemporary generative art." PhD
, University Complutense of Madrid, 2022
Keywords/Fields of Study : Art and Artificial Intelligence, Artificial Neural Networks, Machine Learning, Artificial Life, ALife Art, Generative Art
564
pages, written in Spain
, copyright © Cristina de Propios Martínez, 2022
Abstract: Throughout the following research, Art and artificial intelligence: machine learning techniques in the contemporary generative art, the recent artistic trend «AI Art (Artificial Intelligence Art)» understood as the art produced by means of artificial intelligence (AI), will be studied. The field of artificial intelligence has been studying computational models that are able to carry on human-like activities since mid-20th century, nevertheless, it wouldn’t be until the development of the latest advancements in machine learning — a branch of the AI in charge of studying and developing algorithms that enable computers to learn —, when it experienced a legitimate upturn and started to have an impact in its use in the arts. The following investigation aims to bring the field of machine learning closer to the public from an artistic point of view, in order to explain the techniques that are inherent to it and therefore democratize an art originated in a technical area.
In order to contextualize and locate this recent genre, the first part of the thesis corresponds to precedents and context of the investigation, in which a tour around the antecedents of artificial intelligence framed in a wider scope will be taken: artificial life, where we find a human intention to imagine and design artificial beings. These antecedents began at literature and devices from ancient times, went through the influence of avant-garde science, artistic practices in the cybernetic period resulting in robotics, systems theory, and computational generative art, to culminate in evolutive art, considered the closest precedent to AI Art. The works cited in this context are classified as «hard» (hardwarebased), «wet» (biochemical-based) and «soft» (software-based), corresponding to one of the most popular classifications of artificial life.
Once set the basis of a preliminary context, the second part of the thesis starts with a brief introduction to the field of artificial intelligence, highlighting those events that made it possible for the development of the machine learning field to occur. On the basis of the little available information on specific bibliography, a chapter is dedicated, in a more comprehensive way, to the explanation of the most employed machine learning techniques in the AI Art trend, with the purpose of better understanding the new genre, being it axiomatic that the reason of ist upturn is indeed the development of said techniques. Additionally, a taxonomy of the most used machine learning techniques of the trend will be undertaken, as one of the main objectives of the investigation and with the specific objective of its dissemination.
After explaning the techniques of machine learning, the investigation will focus on artificial intelligence from an artistic point of view, establishing three parallel lines of research: first of all, we find those works in which machine learning techniques are used as a tool for analysis in the field of art; secondly, those that are used for studying computational creativity; and lastly, the artistic practices that consecrated AI Art. This last line is explicitly addressed in different sections that analyse the origins, definition, and their corresponding aesthetic tendencies. Drawn from the premise that this art is originated in technical research laboratories, the reason for the praising of the artistic trend created by the usage of machine learning techniques is researched, the artworks and artist that belong to the trend, as well as the study of the state of play of the artificial intelligence art in Spain, as the last research point in this line of investigation.
The second part of the following research will conclude on the ninth chapter, dedicated to an explanation of a personal research work using techniques inherent to the field of artificial intelligence, specifically an innovative technique called Neural Cellular Automata (NCA), presented as a new technique and artistic proposal that joins in the artistic circuit of AI Art. This technique applies to two parallel research lines: on one hand, as the study of computational creativity and, on the other, as an artistic digital tool. Furthermore, the result of the research process using said technique in two production phases will be presented, the first phase focusing on experimentation and resulting in the video artwork «Growing pictures», that ensued in the second phase: the final completion of a digital sculpture, «Artificial Roots».
After presenting the body of the research following the stated objectives, as well as the raised hypotheses, conclusions bring the thesis to an end by mainly focusing on insisting on the importance and significance of machine learning techniques in the development of AI Art, resulting in different aesthetic tendencies determined by the artificial intelligence technique employed. Overall, as it is shown throughout the investigation, artwork and tecnique are linked, and this is, precisely, the cornerstone of this new form of art.
Department: , University Complutense of Madrid
Advisor(s): María Jesús Romero Palomino
