Art & AI

This project brings together histories and concepts of visual culture and image making with explorations of technology and artificial intelligence. We are an artist and AI researcher exploring approaches to applying machine learning to create artistic images. Our team is analyzing the application of machine learning to the creative act of image making. Working collaboratively in the areas of art, research, digital technology, and popular culture, we will lead experiments in a range of visual renderings starting with 2 dimensional and eventually 3 dimensional and animated results.

The project is unique in the way that it merges artistic knowledge and technique with machine based methods and production and also raises questions about the impetus and drive of image making. We challenge and complicate notions of traditional systems such as portrait making at the same time as we show the synergies between contemporary technologies and artistic histories. We also explore the nature of chance and examine how machines and AI shift our notions of creativity and productivity.

The basic approach is based on a 19th-century French atelier method, which posited that if an artist could master a range of facial and anatomical features she could create lifelike fictional faces and figures. The approach also recognizes 20th-century games like Changeable Charlie, which marketed to parents and children 4,194,304 possible unique faces. We are looking at both knowledge-based and statistical approaches that can be applied to generating artistic images. We are surveying other work in this area, primarily work done at Rutgers in AI and Art, which applies deep learning to this problem.