In this episode, I explore the concept of "Trained Creators" - systems like AI, humans, or nature that generate novel, functional patterns by leveraging deep learned correlations. Building on ideas from episode #61, I discuss how these systems, after training, act as probabilistic generative models, creating outputs that combine past knowledge in surprising and innovative ways. Creativity, I argue, exists on a spectrum, from small variations to groundbreaking ideas, and emerges through hierarchical processes within these models.
Using examples from physics, jazz improvisation, and AI like ChatGPT, I show how compressing known data into generative systems opens up vast possibilities, often surprising even their creators. This dynamic, where humans feed models and draw insights from them, forms a feedback loop that could drive a new era of collective innovation.
