Virtual Gallery: Knots
In order to draw a machine must first learn to see. I give machines the ability to create abstract physical paintings that represent their own inner worlds. The abstract representations evoke specific responses from computer vision systems by suggesting the paintings represent specific real world objects. By studying datasets of thousands of real photos, machines are able to capture the visual essence of a category in ways that are meaningful to other machines. Intelligent algorithms we have created are given the means to express themselves visually so that we can appreciate their own unique way of seeing the world.
Knots are a uniquely human concept. Knots can be a valuable tool or a nuisance depending on context. Visually knots present themselves in an endless variety of shapes, but is there a common visual representation that ties them together? This series of work explores the visual language of knots as seen through the eyes of a machine.
The room reflects my overall process in refining these shapes. Knot Cards capture the initial series of possible shapes in miniature. Successful versions are redone on paper reflecting further experimentation to find ink combinations that elicit the strongest machine responses. These most evocative of these are realized on canvas and grouped into a single artwork to showcase various facets of the algorithmic gaze. Nine Knots captures the core visual language of knots as seen through the eyes of a machine. Knot Three Ways reflects how training data bias different systems to prefer different colour combinations. Four Knots simultaneously presents structured variations in colour and shape as a visual confusion matrix. The set White Knots on Brown captures a broad set of possible abstract forms locked into a single colour palette.
