AI Experiment Explores Human Creativity with Modern VLMs and LLMs

Researchers are building upon a project from over a decade ago, which involved implementing CPPN-NEAT to study neural networks and human creativity through abstract art generation. This original project, a clone of 'Picbreeder', aimed to understand the mechanics of neural nets and the creative process. Today, the team is leveraging advanced AI technologies, specifically Vision-Language Models (VLMs) and frontier Large Language Model agents. They are integrating these into open-ended exploration algorithms. The primary objective is to computationally model and understand the fundamental aspects of human creativity, such as serendipity, memory, and the balance between exploration and exploitation. This initiative seeks to determine if contemporary AI systems can truly emulate the open-ended nature of human thought and innovation.