The big picture

"AI" has become a catch-all word. On any given day it might mean a chatbot, a self-driving car, a movie recommendation, or a sci-fi robot. To use these tools well, it helps to separate three ideas that sit inside one another like nesting dolls:

• Artificial Intelligence (AI) is the broad goal: machines doing things that normally require human intelligence — understanding language, recognizing images, making decisions.
• Machine Learning (ML) is the approach behind almost all modern AI: instead of a human writing step-by-step rules, we show a program many examples and let it learn the patterns itself.
• Large Language Models (LLMs) — the engines behind ChatGPT, Claude, and Gemini — are one specific kind of ML model, trained on an enormous amount of written text.

So every LLM is ML, and all ML is AI — but not the other way around. A spam filter is ML but not an LLM; a hand-coded chess program is AI but not ML.

What "trained on text" really means

An LLM learned by reading a huge slice of the internet, books, and articles, and playing one simple game billions of times: predict the next word. Given "The capital of France is ___", it learns that "Paris" is far more likely than "banana".

Scale that idea up to billions of examples and the model becomes startlingly good at continuing text in ways that are coherent, relevant, and often correct. That's the surprising punchline of modern AI:

Much of what feels like "thinking" is an extremely sophisticated version of predicting what word should come next.

Why this matters for you

Keeping this mental model in mind explains the tools' quirks:

• They sound confident even when wrong, because fluent text is what they were optimized for — not truth.
• They shine at language tasks like rewriting, summarizing, and explaining, because that is literally what they practiced.
• They have no memory of you between sessions unless the product deliberately adds one.

Try this: Open any chatbot and ask it the same question twice in slightly different words. Notice how the answer shifts — proof you're seeing fresh predictions, not a lookup in a fixed database.