[Un]seen Warsaw is a project that aims to explore Poland's capital architecture through Artificial Intelligence generated images. The project is an invitation to take a virtual walk on the streets of Warsaw that do not exist and were generated by training a generative Machine Learning model on real photographs.
The artificially created images are meant to dispense a familiar feeling to a viewer accustomed to Warsaw architecture. At the same time, they bring a sense of disorientation to the surface since the places observed cannot be precisely located as they do not stand in the real world. The project’s purpose is to provide a new aspect to recognizable sites and an invitation to crossroads and encounter corners that don’t truly exist within the city, but that can be perceived as simultaneously reliable and convincing. A closer look reveals how the fluid transition between reality and fiction can bring about exotic and alienating objects disclosed by the deformation of buildings as if a mad architect had given the city a whimsical composition.
The AI training model is based on 3000 still images shot by the artist along the streets of Warsaw in November 2020. Each of the works has been assigned an address in the virtual multidimensional space by using real Warsaw street names and numbers that were taken as input to generate those particular images. While such an address usually does not exist on the map of Warsaw, the project assumes the role of imagining an alternative augmented version of the city.