What Is a Large Language Model?

A large language model, or LLM, is a type of artificial intelligence system trained on very large collections of text to understand and generate human language. LLMs power chat assistants, writing tools, translation, coding helpers and many other applications. Students meet them in tools that can answer questions, summarize text and draft writing.

How they work, in brief

LLMs are built using neural networks, most commonly a design called the transformer. During training, the model reads huge amounts of text and learns to predict the next piece of text, called a token, given what came before. Repeating this over vast data teaches it patterns of grammar, facts and styles. Later steps, such as fine-tuning and training with human feedback, shape the model to follow instructions and be more helpful and safe.

When you type a prompt, the model produces a response by repeatedly predicting likely next tokens. Because the process involves probabilities, the same prompt can give different answers.

What they can do

• Explain concepts in different styles.
• Summarize and rewrite text.
• Brainstorm ideas and outlines.
• Translate and correct language.
• Write and explain code.
• Answer questions across many topics.
• Work with documents that you provide.

Limits and risks

• Errors and hallucinations: they can state false information confidently.
• Outdated or missing knowledge, depending on training and tools.
• Bias from training data.
• Lack of source tracing, so references may be invented.
• Privacy: inputs may be stored or used to improve systems.
• Over-reliance, which can weaken learning.
• Academic integrity risks if used against course rules.

Using LLMs as a student

• Check your course policy before using them for coursework.
• Use them to explain and practice, not to replace your thinking.
• Verify facts and sources in reliable references.
• Protect private information.
• Keep notes of how you used them, in case you need to disclose.
• Learn prompting skills, giving context and constraints.

Terms you will see

• Prompt: the instructions or question you give.
• Context window: how much text the model can consider at once.
• Fine-tuning: additional training for a purpose.
• Retrieval: connecting a model to documents or search to ground answers.

Related terms

See ai hallucination, cookies, algorithm, single sign-on and peer review.