AI Literacy / Glossary

Neural network

A neural network is a kind of AI model made of many layers of simple connected units. Each connection has a strength (a weight) that is tuned during training until the network gives good answers.

Think of a neural network as a huge set of tiny dials arranged in layers. Information goes in at one end, such as the pixels of a photo, passes through the layers, and comes out as an answer, such as "cat".

During training, the network's answers are compared with the correct ones. The dials are nudged a little after each mistake, millions of times, until the errors get small. Nobody sets the dials by hand.

Neural networks are loosely inspired by the brain, but they are maths, not thinking cells. They power image recognition, speech recognition and language models, and it is often hard to explain exactly why a big network gave a particular answer.