Guide
What is a large language model? A plain-English explanation
If you have used an AI chatbot, you have used a large language model, or LLM. This guide explains how one works without any maths.
The core idea: predict the next piece of text
An LLM is a very large neural network trained on enormous amounts of text. During training it plays one game billions of times: given some text, guess the next small chunk (called a token). Each wrong guess nudges millions or billions of internal numbers (parameters) so the next guess is slightly better. After enough rounds, predicting the next token well requires picking up grammar, facts, writing styles, and even patterns of reasoning.
From text predictor to helpful assistant
A raw model just continues text. To turn it into an assistant, developers add steps: training on example conversations, and reinforcement learning from human feedback, where people rank answers and the model learns to prefer helpful, honest and safe ones. Hidden instructions (the system prompt) then set its role and limits.
What LLMs are good at
- Drafting, rewriting and summarising text
- Explaining ideas at different levels
- Translating between languages
- Writing and explaining code
- Brainstorming and organising information
Where they go wrong
- Hallucinations: they can state invented facts or fake sources fluently, because they generate plausible text, not verified truth.
- Out-of-date knowledge: a model only knows what was in its training data unless it can search the web or use documents.
- Maths and counting: they can slip on precise calculation unless they use a calculator tool.
- Bias: they can reflect biases in their training data.
- Manipulation: clever prompts can sometimes bypass safety rules (jailbreaks) or hide instructions in documents (prompt injection).
How to use them wisely
Treat an LLM as a fast, knowledgeable but sometimes overconfident assistant. Check important facts against reliable sources, do not paste secrets, and keep a human in charge of decisions that affect health, money, law or safety.
Want the vocabulary? See our AI glossary.