Also called: artificial neural network, ANN · Last checked: 6 November 2026
In short: A neural network is a computer model made of many small connected maths units that pass numbers to each other. By adjusting the strength of those connections during training, it learns to turn inputs into useful outputs.
Table of Contents
How it works
- Inputs. Data such as words, pixels or numbers goes into the first layer of units.
- Weights. Each connection has a weight, a number that sets how much one unit affects the next. Training adjusts these weights until answers improve.
- Output. The last layer gives the result, such as a translated sentence or a yes or no.
Where you’ll meet it
Google Translate, ChatGPT and Microsoft Copilot all run on neural networks. So do the voice assistants and spam filters you use each day.
What it means for your business
- Language barriers shrink. Neural translation lets a small team answer customers and read contracts in other languages.
- Quality varies by task. Test a tool on your own documents before you trust it with customers.
- Choose where data goes. Paid business plans usually promise not to store or train on your text. Check before you upload.
Watch out for
- Fluent but wrong. Output can read well and still change the meaning, so legal and safety text needs a human check.
- No explanation. A neural network cannot say which rule it followed, because it has no written rules.
Example
DeepL. DeepL, a Cologne company founded in 2017, built its translation service on its own neural networks. In May 2024 it raised $300 million in a round led by Index Ventures, at a valuation of $2 billion. That doubled the $1 billion valuation from its January 2023 round. DeepL said it served more than 100,000 businesses and public bodies.
Neural network vs deep learning
| Neural network | Deep learning | |
|---|---|---|
| What it is | A model built from connected units | Training neural networks with many layers |
| Size | Can be small, with one or two layers | Has many layers |
| How they relate | The building block | The method that stacks the blocks deep |
Frequently asked questions
Does a neural network work like a human brain?
Only loosely. The idea came from brain cells, but a neural network is maths running on chips. It has no understanding, feelings or awareness.
Can a neural network explain its answers?
Not easily. Its knowledge is spread across millions of numbers, so tracing one answer is hard. Explainable AI tools give partial reasons, but they are estimates.
When should a business pick a neural network over simpler methods?
When the data is text, images or sound. For tidy tables of numbers, simpler models are often cheaper, faster and easier to explain to customers or regulators.