Neural networks are used in AI to learn patterns from data. They mimic the structure of the brain consisting of neurons that are connected together.
Neural networks can either be shallow or deep. Shallow neural networks only consist of a single layer of neurons while deep neural networks consist of several layers of neurons - hence the name deep. Deep learning is a branch of machine learning where deep neural networks are used.
The neurons can be connected and configured in a number of different ways to allow the data to flow through and for different things to be learnt. The different configurations give rise to what are called neural network architectures or architectures for short.
Researchers in AI explore the space of various possible configurations and when a meaningful architecture is found it becomes widely known and adopted in applications. Some of the more common architectures include convolutional neural networks (CNNs), recurrent neural networks (RNNs) and Transformers.