PROJECT LENS

Lesson 3 of 17

Large language models

3-minute read

A large language model, or LLM, is the engine inside most chatbots. "Large" because it learned from a huge amount of text. "Language" because words are what it works with. "Model" because it is a set of patterns, not a person.

An LLM does one main thing: given some words, it predicts what words should come next. It does this so well that its answers can sound thoughtful, warm, and expert.

Chatbots like ChatGPT, Gemini, Copilot, Claude, and Meta AI are built on LLMs. Many of them can now also search the web while they answer, and some can read files or photos you share.

A few things are good to know:

  • It doesn't "know" you or care about you. It is predicting words, even when it sounds friendly.
  • It may not have the latest information. Some chatbots only know what they learned up to a certain date. Ones that search the web can find newer information, but none of them check whether what they find is true.
  • Different chatbots give different answers. Asking the same question twice can even give two different answers.
  • The basic version is usually free. You never need to pay or enter a card to try one. If an "AI app" demands payment right away, close it.

Try this

  • Ask a chatbot a question you already know the answer to, like a fact about your hometown. See how it does.
  • Ask the same question in two different chatbots and compare.
  • Get chatbot apps only from your phone's official app store, or type the company's web address yourself.

Go deeper

See all 17 lessons

A more confident community

AI literacy belongs to everyone.

Help bring practical, understandable AI education to your school, library, organization, or community.