Term explained
What is a neural network?
A neural network is a mathematical structure, loosely inspired by brain cells, made of layers of simple units whose connection strengths are tuned during training to recognise patterns.
In plain English
A neural network is the basic machinery behind most modern AI. Picture a large grid of simple calculating units, arranged in layers. Each unit receives numbers from the layer before, multiplies them by its own set of weights, adds the results, applies a simple rule, and passes the outcome onward. Information flows in at one end — the pixels of a photo, the words of a sentence converted into numbers — and comes out the other end as a prediction.
The network learns by comparing its output with the right answer and nudging every weight slightly in the direction that reduces the error. Repeat that millions of times and the connections settle into a configuration that handles the task well.
The name comes from a loose analogy with neurons in the brain. The analogy is historical and quite rough; a neural network is really just a very large pile of arithmetic.
Why it matters
Neural networks are flexible enough to approximate almost any pattern, given enough examples and enough units. That generality is why the same basic idea now underpins image recognition, translation, speech, recommendation systems and chatbots, rather than each needing a bespoke method.
An example
A bank’s cheque-reading system takes a scanned digit as a grid of light and dark pixels. Early layers respond to edges and curves, later layers combine those into shapes like loops and strokes, and the final layer outputs which digit from zero to nine is most likely.
What to watch out for
The brain comparison misleads people. Neural networks do not think, remember experiences or hold beliefs; they compute an output from an input. They can also be surprisingly brittle — small changes to an input that a person would never notice can flip the answer entirely.