RNN, robotter, brush, stone, water 
Dimensions variable
Together with the programmer Nikolai Nikolaev, we developed an Artificial Intelligence which is trained with a set of data based exclusively on works of art. I started experimenting with a Recurrent Neural Network (RNN) for training from character-level language models. The RNN can be used to generate text character by character that will look like the original training data. If we consider art to have the potential to show things from a different perspective, would this AI then operate with shifted meanings beyond the traditional categories? Would it stick to the traditional gender division if it ‘grows up’ with Picasso’s portraits or Robert Morris’ installations?

The installation keeps hallucinating new concepts for art works and writes them with water on a stone. Each proposal evaporates soon after, before the process starts again.


Installation view, Viafarini, Milan 2018

To make the Artificial Intelligence produce its own sentences it needs a seed-text. The seed-text for this work is the question: “What do you see”.

© Gabriel Hensche 2024