Both human and machine cognition are based on compression. We perceive information using categories and types. However, there is a significant difference between how compression works in the arts and in AI. While the artworks frequently depict human characters, symbols, or scenes that "compress" human experience or idealize real life, they also contain concrete and distinct details. This combination of general and concrete, typical and unique, is particularly important in modern art (19th-21st centuries).
When we train AI models, the training data is repeatedly compressed. Indeed, learning "patterns" entails removing outliers and unique details while retaining only the most common and frequently occurring data and associations. Given the fundamental mechanism of generative AI, how can we expect AI tools to create new artifacts with sufficient concrete and unique details?
In my talk, I will elaborate on the idea and provide examples of creative and subjective compression in various art forms. I will also show examples of my recent generative artworks and discuss the strategies I use to overcome this significant limitation of AI media.
Lev Manovich is an artist, writer, and one of the most influential theorists of digital culture worldwide. He is currently a Presidential Professor of Computer Science at the City University of New York's Graduate Center and the Director of the Cultural Analytics Lab. After studying painting, architecture, and filmmaking, Manovich began using computers to create digital art in 1984. He has played a key role in creating four new research fields: new media studies (1991-), software studies (2001-), cultural analytics (2007-) and AI aesthetics (2017-). Since 1991, he has published 190 articles that have been translated into 35 different languages and reprinted over 800 times. He authored and edited 16 books, including Artificial Aesthetics, Cultural Analytics, Instagram and Contemporary Image, Software Takes Command, and The Language of New Media, which has been called "the most provocative and comprehensive media history since Marshall McLuhan." His projects have been exhibited in 14 solo and 122 international group exhibitions at many prestigious institutions, such as the Institute of Contemporary Art (London), the Centre Pompidou, The Shanghai Biennale, and The ZKM | Center for Art and Media.
Keynote Speech: Compression / Generation: Averages and Details in Art and in AI
Back
International Conference DCAC
Tsirigoti Square 7, 49100 Corfu
Ionian Islands Region, Greece
av-dcac@ionio.gr
