The development of artificial intelligence today is focused on machine learning. Machines learn by themselves to carry out tasks using examples that we teach them. The aim of these developments is to automate the maximum possible number of processes and to apply them to vast databases: to classify, to identify patterns, to predict behaviour and to carry out mass monitoring. The worst side of automatic learning is the one that is in keeping with our world of constant surveillance on a large scale, a world in which mass data are regarded as equivalent to natural resources and their exploitation, termed ‘data mining’.
If machines learn and do so in this context, what we need to champion is the bad pupil: everything that sidesteps the norm. If the world of artificial intelligence uses the metaphor of learning, what we need is to formulate a critical pedagogy. If the intention is for artificial intelligence to replicate that of humans on inhuman scales, what is required is to champion non-mimetic artificial intelligence that gives rise to unexpected relations and images. If visual culture today is expanding on its invisible side, the one on which machines generate images that only other machines will see, we need to consider how we can embody those images in order to undo their ghostly action in our surroundings.
This website contains the research we did on artificial vision and image generation using deep learning neural nets between June 2017 and April 2018.