AI glossary
What is a neural network?
A neural network is a model made of layers of simple connected units, with weights that are adjusted during training. It is the basis of deep learning and of LLMs.
It is loosely inspired by the brain: each unit, or artificial neuron, takes in numbers, combines them according to its weights and passes the result to the next layer. No single neuron does anything interesting; the behavior comes from millions of them working together.
Training a network means adjusting those weights. You give it an example, measure how wrong it was and nudge the weights in the direction that reduces the error. Repeated millions of times, the network ends up capturing patterns nobody explained to it.
When a network has many layers it is called deep learning. LLMs are giant neural networks with a particular architecture, the transformer.
An example
A network that spots cats in photos has no "if it has whiskers" rule: by adjusting its weights over thousands of photos, it learned which pixel combinations tend to show up in a cat.