The Swamp Intelligence is a system based on an AI algorithm – sngan_projection – that uses white noise to generate hybrids of natural, architectural and technological imagery, uncanny for human perception. The project engages a concept of ecosystem where natural forms organically merge with digital life forms.
The Swamp Intelligence is based on a NoPlace Neural Network (NNN), a variant of a conditional AI (artificial intelligence), used to unveil neural network abilities to generate novel visual imagery. The NNN has two components that work in opposition: a generator that is learning how to generate an image from a noise pattern and a discriminator that tries to guess which image is real and which is generated. In its constant balance between the real and the imagined, NNN mimics the dampening counterforce effect of a biological swamp.
Unlike earlier architectural statements — such as Superstudio’s Continuous Monument, which conceived architecture to be independent of its environment — the NNN works to find ways to organically combine its architectural proposals into existing landscapes. Being based on completely different foundations, subject to strict logic, and lacking context, the machine quite paradoxically becomes less restricted to imagine than the artist who sets it up, producing rather unsettling imagery for an observer to interpret.
Nomeda & Gediminas Urbonas, in collaboration with Jonas Kubilius. Swamp Intelligence (installation view), 2019, AI installation, sngan_projection, diorama, 205×40×242 cm. Animal – Human – Robot exhibition, Mo Museum, Vilnius. Courtesy of the Artists