Neural Network Application
In collaboration with Adam Harvey
Thanks to the Creative Industries Fund NL
Courtesy: Constant Dullaart, Carroll / Fletcher, London, Future Gallery, Berlin & Upstream Gallery, Amsterdam
In his work DullDream, Constant Dullaart deals with Convolutional Neural Networks (CNN)—artificial neural networks that enable machine learning and pattern recognition through artificial intelligence. They are used for facial and speech recognition, and they can also determine objects in images based on their shape. Google’s Deep Dream exploits this capability: the program identifies specific elements based on form and then intensifies those elements in the image. Pattern recognition becomes pattern hallucination—the algorithm incorporates eyes or animals where there are none. Constant Dullart’s software does the opposite: while Deep Dream highlights and intensifies patterns, DullDream reduces the specific characteristics of forms. Deep Dream becomes a DullDream. The program enables users to upload images of themselves and have them returned devoid of individual characteristics—an impressive statement against increasing regulation through pattern recognition.