What it means

A large language model (LLM) is a statistical software system trained on huge collections of written text and code. Underneath the chat window, it does one core job very fast: given a sequence of words and symbols, it calculates which piece of text is most plausible next.

Because human writing encodes grammar, domain jargon, reasoning steps and formatting habits, predicting the next chunk well allows an LLM to draft memos, summarise transcripts, translate languages and generate working code. Still, it is calculating probabilities over patterns, not consulting a verified database of truth.

Why it matters if you're a student

Almost every workplace AI tool you meet on an internship sits on top of an LLM. Understanding how it works stops you treating it like an oracle:

  1. Fluency is cheap: clean prose no longer proves you know your subject.
  2. Inputs decide outputs: vague instructions pull the model towards generic internet averages.
  3. Verification is the real skill: checking facts, logic and edge cases is what teams pay graduates to own.

An everyday example

Suppose you paste rough lecture notes on contract law into an LLM and ask for a revision table. It organises your bullet points into neat columns in seconds. If one concept was missing from your notes, the model fills that gap using statistical guesswork rather than your lecturer's actual reading list.

Common misconception

People often say an LLM "understands" or "looks up" facts like a search engine. On its own, a base language model does not browse live records or check whether a sentence is true. Unless paired with external search or trusted documents, it generates whatever phrasing fits the pattern best, even when that means inventing a plausible-sounding citation from scratch.

  • A token is the basic chunk of text, code or data an AI model processes, often equal to a word, part of a word, a number or a punctuation character.

  • A context window is the maximum amount of text, code or files an AI model can hold and process at one time across both your input and its response.

  • A transformer is a neural network architecture using attention mechanisms to weigh relationships across sequences, powering modern language AI models.

  • An AI hallucination happens when a model generates fluent, confident output that is factually wrong, unsupported by sources or completely invented.

  • Generative AI refers to models that produce new text, code, images, audio or video from prompts by learning statistical patterns in training data.

Where it matters

Explore how this concept shows up inside real graduate workflows and sector briefings:

Articles that mention it

Get the fortnightly briefing

theVANTAGEd examines AI across real industries with critical thinking, every two weeks. Written by students.