Artificial neural networks are modeled after the human brain and nervous system. They contain interconnected nodes that mimic biological neurons. Information is passed through the network via weighted connections. The network learns by adjusting the weights based on examples provided in a training process. Common network architectures include single and multi-layer feedforward networks and recurrent networks. Backpropagation is a widely used training algorithm that propagates error backwards from output to input nodes to update weights. Neural networks can learn complex patterns from large datasets and are used for applications like classification, prediction, and data processing.