Lesson 1: What Is a Large Language Model?

What You'll Learn

What a large language model (LLM) actually is, in plain words
How "predicting the next word" lets it write whole sentences and paragraphs
Why this approach makes LLMs both impressively powerful and sometimes wrong

Meet the Large Language Model

You have probably heard of ChatGPT or Claude. These are examples of a large language model, often shortened to LLM. Let's unpack that name one word at a time.

"Language" means it works with text — the words you type in and the words it types back. "Model" is just a word for a computer system that has learned patterns from examples. "Large" means it learned from an enormous amount of text: imagine a huge slice of the books, articles, and websites that exist. So an LLM is a text system that learned patterns from a mountain of writing.

The important part is that nobody typed in every fact or sentence by hand. Instead, the model picked up patterns on its own by reading, a bit like how you absorb the rhythm of a language by hearing it spoken around you for years.

It's Basically Super-Powered Autocomplete

Here is the secret at the heart of every LLM: it works by predicting the next chunk of text, over and over again.

Think about your phone's autocomplete. You type "I'm running a little" and it suggests "late." Your phone learned that "late" is a very likely next word. An LLM does the same thing, but on a giant scale. You give it some text — say, a question — and it predicts the most likely next chunk. Then it adds that chunk and predicts the one after. And the one after that. One small piece at a time, it builds up a full answer.

That's really it. When you ask, "What is the capital of France?" the model isn't opening a file labeled "France." It is generating the text that most likely follows your question, based on the patterns it learned. Because the phrase "The capital of France is Paris" appeared so often in its training text, "Paris" is the overwhelmingly likely continuation. The answer comes out correct — but it got there by predicting likely text, not by looking anything up.

Why This Is Powerful (and Why It's Fallible)

This simple trick turns out to be amazingly capable. Because the model absorbed patterns from so much writing, it can draft an email, explain a recipe, summarize an article, or write a birthday poem. It has soaked up how good writing tends to flow, so its output usually sounds natural and helpful.

But the same trick is also its weakness. The model is built to produce likely-sounding text, not guaranteed-true text. It does not truly understand the world, and it is not checking a source while it writes. So when it isn't sure, it may still produce a confident, smooth answer that is simply wrong. People call this a hallucination — when an AI states something false as if it were a fact. An LLM can invent a book title, make up a quote, or give a wrong date, all in a very convincing tone.

The takeaway: an LLM is a brilliant writing partner and a fast first draft, but it is not an all-knowing oracle. Think of it as a confident, well-read assistant who occasionally misremembers — and always double-check anything that matters.

Key Takeaways

A large language model (LLM) is an AI trained on huge amounts of text; ChatGPT and Claude are examples.
It works by predicting the next chunk of text over and over — like super-powered autocomplete — rather than looking facts up.
This makes it powerful for writing and explaining, but also fallible: it can sound confident while being wrong, so always verify important details.