R&D: ‘In Silico’, New Series of AI-Synthesised Personas (R&D)
For the last few months I’ve been focussing on generating sythesised images and video works using the ‘data exhaust’ from the GAN powered deepfake image making process. The series is entitled In Silico and each piece is comprised of a time-lapse of synthesized images generated by GANs (AI) during the deepfake computer vision process. For this series the personas of the founders of surveillance capitalism will be grown by AI ‘in silico’. These artworks will be forms of character studies comprised of AI-synthesised personas for Mark Zuckerberg, Steve Jobs, Bill Gates, Jack Dorsey, Jeff Bezos, Larry Page and Sergey Brin. All immensely powerful white, straight men.
In silico is a term that refers to experiments from the 1990s from a niche field of sociology called ‘artificial societies’, which involved attempts by crude multi agent systems to ‘grow’ society in silico. The concept is also found in Isaac Asimov’s ‘Foundation series; where scientists used large data sets about societies to not only predict the future but also control it.
The AI involved in the synthesisation of deepfake faces offer us a window into the wider tensions that exist concerning how image-making occurs today; for whose eyes they are created for; and as a result – how ‘ways of seeing’ are changing as a result of artificial silicon-based systems.
What happens when our identities are grown ‘in silico’ by single or multi-agent systems today? Billions of images, biometric personas and behavioral data are created by machines for machines. These immense archives (as of 2019, over 250 billion images have been uploaded to Facebook alone) are trawled by sophisticated algorithms searching for clues about the behaviors and tastes of the human race.
A select group of predominantly white men in Silicon Valley have immense white power, monopoly and control of these powerful systems that create so much of our culture and define our understandings of reality. They do so whilst resisting outside scrutiny, accountability and regulation.
‘In Silico’ subverts the power of new computational forms of image making by AI and turns the computer vision lens on to the founders of surveillance capitalism themselves to interrogate their personas and ask questions concerning white power, identity, privacy, truth and hyperreality.
Low Resolution Test #1:
1 – synthesised images from the ‘data exhaust’ of the deepfake process compiled to create the artwork:
View this post on Instagram
High Resolution Test #1:
1 – synthesised images from the ‘data exhaust’:
View this post on Instagram
🎧: Gorecki, Symph #3 Op.36 (Soprano: Zofia Kilanowicz)
The ‘Data Exhaust’: Iterations of faces of Mark Zuckerberg ‘grown’ by the AI during the deepfake process:
The deepfake computer vision process creates tens of thousands of iterations of a face that the GANs are attempting to synthesise with ever increasing accuracy during the modeling process. These images comprise the data exhaust of the deepfake machine learning process and strike a conceptual resonance with the ‘data exhaust’ that Google and other tech giants like Facebook, Twitter, Amazon et al base their economic model of surveillance capitalism on. For them, the data exhaust is human experience in aggregate that has been renditioned as behavioural data in order to sell advertising to us with ever increasing efficiency online. This new raw material – human experience converted to behavioural data and the economic logic of surveillanc capitalism have underpinned new technologies of power that have emerged online in the last 20 years. Technologies of power that have fundamentally reshaped how image-making occurs today – and for whose eyes (hint: machines).
- 10 iterations
- 6K iterations
- 20K interations
- 45K interations
- 70k iterations
- Looking through the eyes of the machine: a timelapse of the computer vision system that extracts and collates a face dataset needed as part of the deepfake image synthesisation process