Re-Generation: Uncanny Density

Through Generative Adversarial Networks (GAN) of colour-coded plans, these houses are programtically mixed to counter the "rooms first" logic that drives architecture today (according to Andrew Holder's "On Sufficient Density"). Real estate and developers focus on rooms, GFA, and programs; Our Gan-generated mixed program plans blur the lines between interior and exterior, creating walls that enclose and expose. Your toilet overlooks the coffee table, your bed lies on a bed of grass, and you can no longer seperate one room from another. Doubt is provoked, and the 3D version of the Zen koan is born.
 
A GAN is a machine learning AI. It takes a training set of images (hundreds to thousands of photos) that get regenerated into new photos that appear real to a human observer. Using a GAN is a literal recycling of old projects. Where the essense of previous projects are usually left behind and only assets are reused, the GAN combines and examines previous projects. The only bias it holds is from what images are inputed.
 
Class: ARC480 (Lazy Computing)
Professor: Andrew Bako
Project Members: Rick Schutte and Zhenxiao Yang