The chapter discusses the contrast between systems where the programmer designs the structure and those where the system learns from data, highlighting the limitations of static intelligence versus learning through data. It explores the potential for algorithmic creativity in unconstrained search spaces like genetic algorithms and delves into how Large Language Models (LLMs) map tokens into a vector space to determine semantic similarity. The chapter also covers Hebbian learning in neural networks like transformers and the use of gradient descent to fit curves and organize tokens on a manifold in LLMs.