Quick answer
A large language model (LLM) is an artificial intelligence system trained on very large amounts of text to understand and generate human language. It works by predicting the most likely next piece of text, one small step at a time. Chat assistants such as ChatGPT, Claude and Gemini are built on LLMs.
LLMs can answer questions, summarise documents, translate, write drafts and help with code. They are powerful tools, but they do not “know” things the way people do, and understanding how they work helps you use them well.
Also readHow to Spot a Phishing Message: 9 Red Flags and What to Do
How an LLM works, in simple terms
1. Text is broken into tokens
The model reads text as tokens, which are words or pieces of words. A long word may be split into two or three tokens.
2. It learns patterns from huge amounts of text
During training, the model reads a vast collection of text from books, websites and other sources, and learns to predict the next token in a sentence. Doing this billions of times teaches it grammar, facts, styles and patterns of reasoning.
3. It is refined to be helpful
After this first stage, developers fine-tune the model with examples of good answers and human feedback, so it follows instructions and responds safely.
4. It generates answers step by step
When you ask a question, the model predicts one token after another to build its reply. The “context window” is how much text it can consider at once, including your conversation and any documents you share.
What does “large” mean?
“Large” refers to the number of parameters, the internal values the model adjusts during training, and to the size of its training data. Modern LLMs have billions of parameters and need powerful computers in data centres to train and run.
What LLMs are good at
- Drafting emails, articles and summaries
- Explaining concepts in simple language
- Translating between languages, including many Indian languages
- Helping programmers write and debug code
- Answering questions about documents you provide
Where LLMs go wrong
- Hallucinations: an LLM can state wrong information confidently. Always check important facts.
- Out-of-date knowledge: a model’s training data stops at a certain date unless it is connected to search.
- Bias: models can repeat biases found in their training data.
- Privacy: avoid sharing sensitive personal or company information unless you understand how the service handles it.
LLMs and Indian languages
Global models increasingly support Hindi, Tamil, Bengali and other Indian languages. Indian companies and research groups, including startups such as Sarvam AI, are also building models focused on Indian languages and use cases.
How to get better answers from an LLM
- Be specific about what you want, who it is for and the format you need.
- Give examples and relevant context.
- Ask it to show sources or reasoning, and verify key facts yourself.
Frequently asked questions
Is ChatGPT a large language model?
ChatGPT is a chat assistant built on large language models developed by OpenAI. Claude from Anthropic and Gemini from Google are other examples.
What is an AI hallucination?
A hallucination is when an AI model gives an answer that sounds confident but is false or made up. Always verify important facts.
Do LLMs understand what they write?
LLMs generate text by predicting likely next words based on patterns learned in training. They do not understand meaning in the way humans do, which is why their answers need checking.
Sources and further reading
Featured image: Photo: BalticServers.com / Wikimedia Commons (CC BY-SA 3.0)
