Definition
A neural network is a computing system made of layers of simple mathematical units, loosely inspired by neurons in the brain. Each unit receives numbers, multiplies them by weights, adds them up and passes the result forward. Learning means adjusting those weights, millions or billions of them, so the network's output gets closer to the right answer on training examples. A neural network is the building block of deep learning and of every large language model.
You will rarely design a neural network yourself, but you will buy products built on them: optical character recognition for invoices, voice assistants, translation tools, fraud detection, and the language models behind AI assistants. When a vendor says a feature is powered by AI, there is almost always a neural network inside. Understanding this helps you ask the right questions: what data was it trained on, how is it evaluated, and how often is it updated.
Modern neural networks are enormous, with frontier models estimated in the hundreds of billions of parameters, and they are trained on specialised hardware. A common misconception is that a neural network reasons like a brain. It is a very large function fitted to data. It generalises impressively, but it does not know when it is wrong unless you build checks around it, which is why evaluation and guardrails matter.
In practice
A bank's card fraud system is a neural network scoring every transaction in milliseconds. It learned from years of labelled fraud cases and blocks suspicious payments before the customer notices anything.
Why it matters
Neural networks are the reason software quality jumped in the last decade. Knowing that they are statistical, not logical, helps you judge where to trust them and where to keep a person in the loop.
Frequently asked questions
- How does a neural network learn?
- It makes a prediction, compares it with the correct answer, measures the error and nudges its internal weights to reduce that error. Repeated over millions of examples, this process, called training, produces a model that performs well on new data it has never seen.
- What is the difference between a neural network and an algorithm?
- An algorithm is a fixed set of steps written by a programmer. A neural network is a flexible structure whose behaviour is learned from data rather than written by hand. The training procedure is an algorithm; the resulting model is the neural network.