elen.xynogala@gmail.com bio


installation, performance, machine learning

This Research conducted during British Council’s Venice Fellowship programme. The purpose of this project to explore different ways of collaboration between human and machine entities through the generation of art interpretations applied to Cathy Wilkes’s Exhibition in the British Pavilion for the 58th International Art Exhibition - La Biennale di Venezia.

How can machines interpret art?

Even if something seems absent from vision, there is still an underlying presence that we try to understand. Humans, tend to search for a message behind artworks decoding the different perspectives of them.
Inspired by Cathy Wilkes’s exhibition, two trained machines create a dialogue like art critics, trying to interpret her work. The algorithms generate stories and images which refer to subjects of death, pain, loss, motherhood, childhood and faith.
A human actor is called to interpret these generations into drawings. These successive interpretations by human and machine actors, lead to the loss of the initial message.
It looks like the Chinese telephone game, in which the message is getting garbled along the way. The audience can compare the original idea with the final version and is being questioned about the notion of open interpretation in art but also the unreliability of the computational systems.

Sound: generated by Music Transformer - Recurrent Neural Networks
Images: generated by Attentional Generative Adversarial Network
Captions: generated by im2txt model and a Fully Convolutional Localization Network
Stories: generated by GPT-2 model


The performance created in collaboration with the artist Theodora Kanelli in the frame of the ARTWORKS Fellowship. The Results of the whole experiment can be found here.