A large language model, often shortened to LLM, is an AI model trained on very large amounts of text so it can predict and write language. It powers tools that chat, draft letters, summarize reports, translate and write code. Say you ask a chat assistant to turn your meeting notes into a thank-you email for donors. A large language model writes that email by picking one likely word after another.
The key point is that it produces text that sounds right. It does not check that text against facts. That is why it can state false things with full confidence, a problem called hallucination.
Why it matters
Most chat assistants your staff might use are built on large language models, including the ones profiled in the tools section of this site. Treat anything they write as a first draft. Check names, numbers, dates and any legal or medical claims before the text goes out under your organization’s name. Before you sign up for a tool, ask the vendor three questions. Which model does the tool use? Is the text you type kept or used for training? Who at the vendor can read it?
Published Updated