LLM

An LLM is a model trained on very large amounts of text to learn which word typically follows another. From this statistical pattern comes the ability to produce coherent text for almost any input, from summarizing a document to answering a question. There is no stored database of facts the model looks things up in — only a learned probability for how a text plausibly continues.

In practice, people encounter LLMs through a chat interface such as ChatGPT, Claude or Gemini, or embedded in software that drafts, translates or summarizes text in the background. You type an instruction, the model returns text, with no human checking anything in between. Different LLMs vary in size, training data and cost per request, not in how they fundamentally work.

A common misunderstanding is assuming an LLM knows or understands things the way a person does. It produces the statistically most plausible text regardless of whether it is factually correct, and states wrong information just as fluently as right information. Mistaking a confident phrasing for a checked answer is the actual risk.