Curatorial                                Contact                               



Through Machine & Darkness,
video installation (projection approx W2m), 35 minutes, looped, 

Installation view: Of Stars & Chasms, ArthouSE1. Credit. Benjamin Deakin

A video extract below comprises two 4 minute clips from
the beginning and the end of the machine learning.

Current development phase supported by Arts Council England Developing Your Creative Practice Grant (2019)
www.throughmachineanddarkness.co.uk



Responding to the increasing use of AI and machine learning in examining astronomical datasets, ‘Through Machine and Darkness’ uses a DCGAN/neural network trained on RAW images from the Hubble Space Telescope (HST). Attempting to conjure a total view of the cosmos, the imagery produced explores the latent space of the algorithmic imagination, bridging elements of magical thinking with science and technology.

The first development phase of the project has run on 45,000+ images from the ACS WFC camera on the HST. The RAW image data has been chosen as it represents as closely as possible what the camera ‘sees’ and has not been cleaned of artefacts such as cosmic rays.

Image outputs taken at increments from a DCGAN’s training are edited together to form a moving image piece. As it transitions, the image becomes more resolved, giving an insight into the how the machine learns to generate its image of the cosmos –
a computer generated parallel universe. Deep space objects, like galaxies and nebula seem to emerge from the darkness.

Programmer: Doug Neal. Scientific advice: Iva Momcheva &  Joshua Peek, STScI/NASA, USA.