Core Components of AI

This chapter explains the foundational components of Artificial Intelligence, focusing on datasets, models, and optimization. It distinguishes between structured and unstructured data, as well as labeled and unlabeled formats, emphasizing how dataset quality directly determines the scope and reliability of AI systems. It introduces key learning paradigms—supervised, unsupervised, and reinforcement learning—illustrating their applications to architecture, from style classification to energy optimization. The text then details the training process, describing how models identify and generalize patterns from examples. Optimization and evaluation methods are also addressed, including cost functions, accuracy metrics, and the importance of generalizability. These mechanisms ensure that an AI system is not merely memorizing examples but can apply learning to new situations. The chapter frames these concepts within architectural practice, showing how curated and well-labeled datasets, coupled with rigorous model training and evaluation, are essential for producing reliable, context-sensitive applications of AI in design and building.