Sunset, one of the first DeepDream images produced by its inventor, Alex Mordvintsev in May 2015 before its public release.

DeepDream is a computer vision program developed by Google engineer Alexander Mordvintsev, utilizing a convolutional neural network (CNN) to enhance patterns in images through algorithmic pareidolia, producing surreal, psychedelic visuals. The program popularized the term "deep dreaming" referring to the generation of images that activate specific features in a trained deep network. Originating from Google's "Inception" neural network for the 2014 ImageNet Challenge, DeepDream's open-source code was released in July 2015. The process involves feeding an image into the neural network and iteratively adjusting it based on the network’s recognition of various features, amplifying edges, and shapes to create hallucinatory effects where familiar objects morph into complex, often bizarre patterns. DeepDream showcases the power of neural networks in image processing and offers insight into the workings of machine learning models, merging technology and art.

Provenance:

“Sunset”, created in May 2015, was minted as a unique NFT in September 2024 as part of the group show "Automat und Mensch 2.0," a retrospective of AI history at Kate Vass Galerie. While DeepDream was released in July 2015 as an open source for public use, “Sunset” was made before Google's public launch and only became allowed for sale since June 2024.

DeepDreams

More from Alexander Mordvintsev

More from Alexander Mordvintsev