In these varying essays, we have stalked the digital image. We began by exploring cameras, by looking at how the Mavica as a digital camera changed the kinds of images made at Abu Ghraib. We looked at the cameras of smartphones and found a continuum of the image in camera apps and in filters an implicit belief that the digital image holds truth of the ‘real’. We  examined the GAN, the ultimate camera, which in its performance of the camera exposes the image's loaded state. Then we turned to people in front and behind of cameras in ‘Performance’. Looking first at Abu Ghraib as a clear demonstration of how the camera morphs the photographed and how the digital allows for a lightness, making it easier somehow to smile over a dead body. The Lynndie pose was traced through divergent spaces, demonstrating not only how the network fundamentally changes the kinds of images we make, how we make them, why, and how they look, but also how the network functions as a complexifying device, innovating and adding layers atop images. Both features that position the digital camera as practically further from the ‘real’ than ever before. Looking forwards again we asked what the GAN performs, and how the GAN sees us. It was established here that the GAN is convincing because it makes photographs and we trust photographs above all else; that we have become referent to the digital image. Turning to data we briefly considered images as two states of data, both states serving to hide the image itself. Let's turn for one last time to the GAN and see if it has any answers as to what this means. 

 

Every image generated by the BigGAN is a networked image as every image within the ImageNET dataset is a networked image. The state of networked images as instantaneous and also part of an incomplete assemblage creates an image that is ‘self-referential, recursive, and undecidable’ embodying a ‘viral logic of intensity, multiplicity and incompleteness in which the image refers only to itself’.(1) The GAN performs this state of the image. It desperately tries to make a singular image but sneezes and makes 1,000, removing the possibility of any image being singular. Images generated by the GAN are precisely ‘self-referential, recursive and undecidable’ as are, it seems, all digital photographs. The combination of this state with the digital image’s constant positioning next to and as the real, with digital images that present themselves as a continuum of existence, with digital images that seek at every turn to hide the fact that they are photographs from us, is a heady and dangerous cocktail for us who have over the last century or so become so innately reliant on camera produced imagery.

​

Let’s look to the GAN performing for us. Watch it take a recursive set of networked images; twenty thousand photographs of apples. Watch it try desperately, fight with itself, to backtrack, to remove from the network and remake the individual representational image, to reintroduce singularity to the networked image rather than accept the network as singularity. Watch the GAN muddle around on the floor scraping desperately to hold on to and make something singular and definite, an absolute generic apple. It doesn't know, with its dirty knees, like the education of Rousseau’s hypothetical Emile, it has all been a construct.(2) Its knowledge of apples is from digital photographs of apples, which is from the network’s social knowledge of apples, how we perceive apples, the audience, the family, up and down and forever. 

​

The GAN in attempting to produce singularity instead models the realities of the digital image in its absurdity. After first indulging in the image, only truly wanting to memorise, the GAN then gets into a representational fight. It encounters the sliding scale of real embodied in an image and suddenly images it only wanted to remember become patterns, averages, as it attempts to recreate that platonic image. 

 

The tragedy of the GAN is that even when it is ‘successful’, when it can consistently produce images that trick people, it is only ever making photographs. It only ever sees through photographs. With a digital camera taped over its eyes, the GAN is a mirror that exposes us as we clutch our digital images, cramming them into our eyes to pass the blood-brain barrier and exist as us. Perhaps the GAN can allow us to look into the mirror and see our cameras and maybe start removing them, rebuilding our memories apart from the image, and allowing the photograph to finally exist on its own terms, at a distance from us.