(use the arrows to navigate through the images)
The human form, in particular the unadorned nude, has been a feature of artistic expression since we made our first art, 30,000 years ago, (possibly even earlier).
Summoning such forms from an artificial neural network seems absurd – how can a machine, trained on pictures from the internet, ever understand the human body? The answer, of course, is that it cannot. These systems do not ‘understand’ anything, in the human sense, they can only ever give us nuanced reflections of the data we feed them.
However, when looking at these images we “bring a lot to the table”.
Perception is a two way street. What we “see” is only partially drawn from the external world – the vast majority is constructed by the brain. More often than not, what we see is what we expect to see.
Art is the same, but moreso – part of the pleasure in art is playing with this ‘expectation’ system.
We want to see things. We are hard-wired to expect to find form rather than noise. Art plays with this expectation, good art feeds us truth through misdirection.
The Golem is a human-like creature from Jewish folklore, hewn from clay and mud.
In Modern Hebrew, golem is used to mean “dumb” or “helpless”. Similarly, it is often used today as a metaphor for a mindless lunk or entity who serves a man under controlled conditions, but is hostile to him under other conditions.
Pulling human forms from an AI feels like a similar process. By performing the right kind of incantations, shapes can be drawn out of the mire, but these shapes can never be truly human.