Having established a definition of Large Language Models, this chapter examines their internal workings at a high level. We will look at how these models process and generate text.

You will learn about:

• How text is broken down into units called tokens and represented numerically through embeddings.
• The fundamental concept of predicting the next word as the basis for text generation.
• The significance of training data size and model parameters ($P$) in determining an LLM's capabilities.
• A simplified overview of the Transformer architecture, a common structure for these models.
• How the preceding text, or context, influences the model's output.

This chapter provides a conceptual foundation for how LLMs operate without requiring deep mathematical or programming knowledge.