Layers and depth

A neural network is organized into layers stacked on top of each other. Information enters the first layer, gets transformed, passes to the next, and so on. The number of layers is the model's depth, which is why people say "deep learning."

Why stack them? Each layer can build on the previous one's work. In a language model, early layers tend to catch simple patterns (basic grammar), while deeper layers capture abstract ones (the overall intent of a sentence). Depth lets the model understand things in stages rather than all at once.