Learning to See is an ongoing series of works that use the latest machine learning algorithms to reflect on how we understand the world. What people see is a reconstruction based on our expectations and previously held beliefs. Learning to See is an artificial neural network loosely inspired by the human visual cortex. It looks through cameras and also tries to understand what it sees. Of course it can only see what it already knows — the same as us. This work is part of a broader line of research about the difficulty of seeing the world through the eyes of others. Learning to See: Gloomy Sunday is a video and an interactive installation where the recordings taken by a live camera aimed at a table covered with objects are analyzed by a series of neural networks trained on different data sets (ocean, fi re, clouds, and flowers).

Credit: vog.photo 2019