Generative Adversarial Networks Experimentation Lab

Description

Convenors: Leonardo Impett, Geoff Cox, David Berry and Anne Alexander
Please note change of venue to Alison Richard Building
This workshop will introduce and explore prospective methods and approaches for critically engaging with the images of people created through what are called Generative Adversarial Networks or GANs. The workshop will create design experiments using GANs as provocations to expand debate about taken-for-granted notions of interpretation, realism and authenticity. As human and machine vision are ever more systematically intertwined in an age of machine-learning reproduction we aim to reflect on the consequences for research practice. In one recent famous example, a GAN-generated image by the Obvious collective sold for $432,500 at Christie’s raising debates about authorship and the new medium of machine-learning image production.
Unlike many AI methods, GANs are unsupervised (or self-supervised); they don’t inherit the bias of generic computer-vision pre-training datasets like ImageNet. Instead, two neural networks (a ‘generator’ and a ‘discriminator’) attempt to fool each other, continuously closing the gap between the generated fakes and the original dataset. Images generated by GANs are a mathematical representation of the set of images fed into the GAN, giving us a machine-window into a large set of images: a computational distancing-effect from a direct visual reading of the material. In effect, machine learning as distant reading.
Unlike the image at Christie’s, GANs do not produce a single ‘fake’ image, but an infinite manifold of possible solutions. The challenge of reading an infinite manifold – in terms of typicality, pattern and pastiche – raises important theoretical questions about the epistemological role of images in the humanities and across other many disciplines.
Click here to book your place.
A sandwich lunch will be provided. Please help us to cater accurately by cancelling your ticket if you book and can no longer attend. Please email specific dietary requirements to Michelle Maciejewska, mm405@cam.ac.uk by 4 December
 
Bibliography
Fabian Offert. Notes on the Curatorial Challenges of an Emerging Media Art Form. 2018
T. J. Clark. Art History in an Age of Image-Machines. 2008
Melvin Wevers, Thomas Smits, Leonardo Impett. ImageTexts. 2018