The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) Creative Adversarial Networks for Art Generation with Ahmed Elgammal - TWiML Talk #265
May 13, 2019
Join Ahmed Elgammal, a Rutgers University professor and director of the Art and Artificial Intelligence Lab, as he takes us through the fascinating world of AI-generated art. Discover the innovative AICAN project, which creates original portraits using a blend of AI and centuries of artistic tradition. Elgammal discusses the delicate balance of aesthetics in Generative Adversarial Networks and how AI mirrors the evolution of artistic styles. Dive into the future of creativity, where machines collaborate with human artists to redefine artistic boundaries!
AI Snips
Chapters
Transcript
Episode notes
Art's Deeper Meaning
- Ahmed Elgammal's career revolves around AI, especially computer vision, combined with a passion for art.
- He finds the emotional impact and layered understanding of art intriguing, a challenge for AI.
AICAN: Creative Adversarial Network
- AICAN, a creative adversarial network (CAN), generates original portraits.
- It was trained on over 500 years of European canonical art, focusing on innovation in art.
Balancing Aesthetics and Innovation
- AICAN uses style ambiguity to push innovation while adhering to aesthetic principles.
- It generates art fitting general aesthetics but not existing styles, fostering novelty.