What is AI, actually?

The one-sentence version

AI is software that has learned from enormous amounts of human-created text, images, and data — and can now generate new text, images, or answers that feel surprisingly human.

That’s it. No robots. No science fiction. Just software that got very good at patterns.

A better analogy than “robot brain”

Most people picture AI as a thinking machine. It’s closer to a very well-read assistant who has absorbed millions of books, articles, and conversations — and can now answer questions, write things, and solve problems based on everything they’ve absorbed.

They don’t truly “understand” the way you do. But they’re fast, available at any time, and surprisingly capable.

Think of it like this

Imagine you could hire someone who had read every cookbook ever written. They've never tasted food — but they can suggest recipes, explain techniques, and adapt a dish to your dietary restrictions. That's roughly what AI does with text.

What can it actually do today?

The tools you can use right now — ChatGPT, Claude, Gemini — are called large language models (LLMs for short). Don’t worry about the name. What matters is what they can do for you:

WriteEmails, essays, social posts, summaries — from a short instruction

ExplainAny concept, at whatever depth you want, in plain English

Think with youBrainstorm ideas, work through decisions, get a second opinion

SummarisePaste a long document and get the key points in seconds

CodeWrite, fix, or explain code — even if you've never programmed

TranslateBetween languages, but also between "technical" and "simple"

What it genuinely cannot do

This matters just as much. AI cannot reliably:

Browse the internet in real time, unless the tool specifically says it can. Remember your previous conversations by default — each new chat starts fresh. Know what happened after its training data ended. Do things in the real world, like book a flight or send an email, without extra tools built around it.

And most importantly: it can be confidently wrong. AI doesn’t know when it doesn’t know something. It will sometimes invent facts that sound completely plausible. This is called a “hallucination” and it’s the main reason you should verify anything important it tells you.

The honest version

AI is a brilliant, fast, tireless assistant who occasionally makes things up and never admits it. Incredibly useful — but always worth a second check on anything that matters.

▶ Try it yourself — takes 2 minutes

Go to claude.ai or chatgpt.com (both have free accounts). Type this:

"Explain what a large language model is, like I'm 12 years old."

Notice how it adjusts its language based on your instruction. That's the core of how prompting works — which is exactly what the next page covers.