ABSTRACT
This chapter explores the significance of sequences and order in both urban and linguistic contexts, connecting these concepts to neural network techniques in AI language models, specifically the GPT architecture. I emphasise the importance of word order in sentences and how LLMs use advanced methods such as positional encoding to process sequences efficiently. Through practical demonstrations, including path generation through a grid and reconstructing the alphabet, the chapter illustrates how AI systems handle spatial and temporal sequences, and explains the mathematical foundations of positional encoding, highlighting its role in enhancing AI performance and its parallels to the cyclical nature of human experience.
