Mario Klingemann is a German artist who uses neural networks, codes, and AI to create generative art. He put images of portrait paintings from the 17th and 19th centuries into GAN (Generative Adversarial Networks), which then produces replica or newly generated portraits based on its own interpretations. GAN is a deep neural net architecture that consists of two nets, going one against the other; the generator and the discriminator. The generator takes in random numbers in order to create images. Then, those generated images are sent to the discriminator with a stream of images from a given dataset (i.e. images of paintings that were put in). The discriminator’s job is then to produce probabilities, a number from 0 to 1, where 0 represents authenticity and 1 represents fake.
In Klingemann’s artwork, GAN isn’t simply duplicating old paintings (input). GAN decides its own aesthetic using a Tinder-like selection of what it finds aesthetic or not, and after studying its own interpretation of the basic features of a face, it produces an infinite stream of portraits that are rather uncanny and creepy. Personally, I always believed that art was a field AI could never achieve to learn, since there are no set of rules on how to express and create art. Creativity is unique to individuals since we all have different interpretation of the world and different ways of self-expression. Mario Klingemann’s pioneering artworks proved me wrong, and this is why they caught my eye. I am excited, at the same time, terrified to see what the future holds for the AI art.