RGBFAQ – From Bell Labs to synthetic datasets and the new ‘technical debt’

Latest in the series of video essays by an artist and researcher Alan Warburton, is ‘RGBFAQ‘, tracing the trajectory of computer graphics from WW2 to Bell Labs in the 1960s, from the visual effects studios of the 1990s to the GPU-assisted algorithms of the latest machine learning models. The story culminates with the emergence of the synthetic dataset: computer-generated images used as ‘ground truth’ for training computer vision algorithms.

Synthetic data is increasingly sought after as a ’clean’ alternative to real world data sets, which are often biased, unethically sourced or expensive to create. And while CGI data seems to avoid many of these pitfalls, my argument aims from the outset to consider whether the virtual world is as clean and steady as we think. I try to catalogue the ‘hacks’ used to construct the foundations of simulated worlds and suggest that the solutions of early computer graphics create a technical debt that might be less than ideal material on which to build the foundations of yet another generation of technology.  

Alan Warburton

The concept of the exploded image (or the alternative ‘hyperimage’) articulates the slipperiness inherent in all discussion of digital aesthetics. The image discussed in RGBFAQ bridges XYZ and RGB, space and colour, data and aesthetics, machine and human, weapon and tool. Alan’s tries to make it clear that while many might mistake a contemporary image for a plain, traditional photograph, it has long been something far more than that; a decoy, a classically Baudrillardian simulacra. 

If RGBFAQ concludes with a sense of the power of digital imaging technologies (and perhaps the feeling of an accident waiting to happen) it is also an attempt to speak of wonder and empowerment: the kaleidoscopic possibilities for interpretation, intervention and synthesis that the exploded image allows. 

Alan Warburton

For this and more, learn about Alan’s work on → alanwarburton.co.uk

Project Page