Google’s Deep Dream software works based on a neural network algorithm, a program that simulates a biological brain’s working in a digitally and simplified manner. Originally created to find patterns in images to classify them automatically, this neural network is especially able to detect animals, human faces, architectures and other common patterns found in images that circulate in the Internet.

Deep Dream inversely applies this pattern recognition, reinforcing everything that the neural network is originally trained to find, but on images that do not necessarily contain these patterns. The result is dream-like or nightmarish images that disturbingly express the patterns stored in complex software such as this.

On display at the exhibition is a video produced by the Itaú Cultural team with urban scenes from Avenida Paulista, in São Paulo, which were then processed using the Deep Dream software.