AI-generated art fools viewers more than Art Basel works in comparative study

Creative Adversarial Networks (CAN), a collaboration between Rutgers University, the College of Charleston, and Facebook AI Labs, autonomously generates artwork. In a 2016 study, judges on Amazon's Mechanical Turk were able to differentiate CAN's images from 25 abstract expressionist masterpieces, yet they deemed AI creations more innovative (53%) compared to traditional artworks (41%). The CAN framework consists of a generator and a discriminator that assess images against a collection of 81,000 paintings by 1,100 artists spanning the 15th to 20th centuries. This initiative is inspired by Colin Martindale's 1990 work, 'The Clockwork Muse,' which posits that creativity adheres to deterministic patterns. CAN aims for an optimal level of novelty and follows previous AI art endeavors like Google's DeepDream (July 2015) and Magenta, with input from Facebook's Mohamed Elhoseiny.