Scott Eaton’s “Contemplating Motion” is an exploration of human motion, abstracted in time. The piece combines video shot by the artist with a custom neural network. The neural network driving the aesthetic of this piece began as an experiment attempting to teach an AI about reflections—could it learn to reflect things outside of its immediate view? This task requires a type of ‘visual imagination’ that artists cultivate when training to draw or paint. The resulting network, though often unpredictable, dynamically renders form from silhouette, imagining sophisticated plane changes inside the figure with highlights continuously rising and falling from these surfaces. The work was modified to allow the neural network to fade the motion through time, continuously moving and then forgetting, in some ways reflecting the mind of the dancer as she improvised the movement.