ABSTRACT
This chapter explores the concept of fine-tuning in large language models (LLMs) and draws parallels between this process and urban design. Fine-tuning is essential to render LLM responses socially acceptable, addressing biases and adapting to specialist domains like urban planning and architecture. I connect this concept of tuning to my previous work on urban environments, where subtle adjustments shape interactions and spaces. The chapter highlights the balance between pre-training and fine-tuning, emphasising the role of human oversight in mitigating certain biases in the training data and enhancing LLM performance. The chapter critiques both LLM fine-tuning and bias as necessary in rational thought, inviting further exploration of AI’s impact on urban life.
