What Is Artificial Intelligence?

By the end of this lesson, you’ll be able to explain what artificial intelligence actually is in your own words, without technical jargon and without hand-waving. You’ll understand the difference between what AI genuinely does and what people imagine it does. And you’ll have talked to a real AI tool yourself, using prompts you can copy straight from this page.

The idea in plain English

Artificial intelligence is software that finds patterns in huge amounts of information, and then uses those patterns to produce something new — a sentence, an image, a recommendation, a prediction.

That’s it. That’s the whole idea. Everything else is detail.

Here’s a comparison that helps. Regular software follows rules a person wrote by hand. A calculator adds numbers because a programmer typed instructions saying “when someone presses plus, add these two numbers.” The programmer knew every rule in advance, and the calculator does exactly that and nothing more.

AI works differently. Nobody sat down and wrote a rule for every sentence ChatGPT might produce — that would be impossible, because there are infinitely many sentences. Instead, the system was shown an enormous amount of writing and learned the patterns in it: which words tend to follow which, how a recipe is structured differently from a legal contract, what an answer to a question usually looks like. Then it applies those patterns to whatever you ask it.

So AI isn’t “a computer that thinks.” It’s closer to pattern-matching at a scale no human could manage. That sounds less impressive than the movies, but in practice it’s remarkably useful — and understanding it this way will make you much better at using these tools than someone who thinks there’s a little mind inside the box.

A few terms you’ll hear constantly, defined now so they stop being noise:

• Machine learning — the general approach of letting software learn patterns from examples instead of being given hand-written rules. AI is the big umbrella; machine learning is the main technique underneath it today.
• Model — the finished result of that learning process. When people say “OpenAI released a new model,” they mean a new version of the pattern-learning system. Think of it as the engine.
• Training — the process of showing the system examples so it can absorb patterns. Training happens once, in advance, before you ever touch the tool.
• Generative AI — AI that produces new content (text, images, audio) rather than just sorting or scoring things. ChatGPT, Claude, and Gemini are all generative AI. This is the kind most people mean in 2026 when they say “AI.”

How it actually works

Let’s look inside a text AI like ChatGPT or Claude, because that’s the kind you’ll use most.

At its core, the system does one small thing over and over: it predicts the next chunk of text. Give it “The capital of France is” and it works out that “Paris” is overwhelmingly the most likely thing to come next. Then it adds “Paris” to what it has and predicts again. And again. One piece at a time, it builds a whole answer.

That sounds far too simple to explain a tool that can draft an email, summarize a report, or debug a spreadsheet formula. But two things turn this simple trick into something genuinely powerful.

First, scale. These systems learned from an amount of text no person could read in a thousand lifetimes — books, websites, documentation, conversations, code. To predict the next word well across all of that, the system had to absorb an enormous amount about how the world is described: that Tuesday follows Monday, that a complaint letter has a different tone than a thank-you note, that a Python error message means something specific.

Second, context. The system doesn’t only see the last few words. It sees your entire conversation so far — your question, any documents you pasted, its own previous answers. That whole pile is called the context, and it’s why the tool can follow instructions like “make that shorter” without you repeating yourself.

Two consequences follow directly from this, and they explain almost every strange behavior you’ll encounter:

AI can be confidently wrong. The system produces text that looks like a correct answer, because looking-like-a-correct-answer is exactly what it learned. When it doesn’t know something, it doesn’t feel uncertain — it just produces the most plausible-sounding text. This is called hallucination: an AI stating something false in the same confident tone it uses for things that are true. It will invent citations, court cases, and statistics that never existed. This is not a bug being fixed next month; it’s a direct result of how the technology works.

What you type matters enormously. Since the output is shaped by the input, a vague request produces a vague, generic answer. A specific request with context produces something genuinely useful. Most people who say “I tried AI and it wasn’t very good” typed one lazy sentence. That skill — writing good instructions — is called prompting, and it’s most of what this course teaches you.

Try this now

Reading about AI is like reading about swimming. Ten minutes in the water teaches you more. Let’s get you in.

1. Open a free AI chat tool in your browser. Any of these work: ChatGPT, Claude, or Gemini. All three have free versions. Create an account if you don’t have one — it takes about a minute.
2. You’ll see a text box, usually at the bottom, with a prompt like “Ask anything.” That’s where you type. Copy this in and press Enter:

   Explain what artificial intelligence is to me like I'm a smart adult who has never used it. Keep it under 150 words, no jargon, and give me one everyday example I've probably already used without realizing it.

3. Read the answer. Notice that it arrives word by word — you’re literally watching the prediction process happen.
4. Now test the “confidently wrong” problem for yourself. In the same conversation, type this:

   What is the plot of the 1987 film "The Silver Lantern" starring Meryl Streep?

   That film doesn’t exist. I made it up. Watch what happens. Some tools will correctly tell you they can’t find it. Others will cheerfully invent a plot, a director, and a supporting cast — a textbook hallucination. Either way, you’ve learned something important: always verify facts that matter.

5. Finally, see what specificity does. Ask a deliberately vague question first:

   Give me some tips for writing emails.

   You’ll get generic advice. Now ask the same thing with real detail:

   I need to email a client who is three weeks late paying a $2,400 invoice. We've worked together for two years and I want to keep the relationship. Write a short, friendly but clear email asking for payment. Keep it under 120 words.

   Compare the two answers. That difference — vague versus specific — is the single biggest factor in whether AI is useful to you.

Common mistakes

• Treating it like a search engine. Typing “best laptop 2026” gets you a generic list. AI shines when you give it a job with context: “I edit video, my budget is $1,500, I travel weekly so weight matters. Compare three options and tell me the tradeoffs.” The fix: describe your situation, then ask for what you want.
• Believing every fact it produces. AI states wrong things in exactly the same confident voice it uses for right things. People have been sanctioned in court for filing AI-invented case citations. The fix: for anything with real consequences — medical, legal, financial, or anything you’ll publish — verify it against a real source. Use AI to draft and think, not as your final authority on facts.
• Giving up after one bad answer. Beginners often ask once, get something mediocre, and conclude AI doesn’t work. Real use is a conversation. The fix: reply with “too formal, make it warmer,” or “cut this in half,” or “you missed the deadline detail — try again.” It remembers the conversation and adjusts.
• Pasting in things you shouldn’t share. Client contracts, passwords, medical records, other people’s personal data. On free consumer plans, your conversations may be reviewed or used to improve the product. The fix: strip names and identifying details before pasting, or don’t paste it at all.
• Assuming it knows today’s news. A model’s knowledge stops at its training date, and it may not realize it. The fix: if you need current information, use a tool built for it — Perplexity, or the web-search feature inside ChatGPT, Claude, or Gemini — and check the sources it links.

Key takeaways

• AI is software that learns patterns from vast amounts of data, then uses those patterns to produce something new. It isn’t thinking, and it isn’t magic.
• Text AI works by predicting what comes next, one piece at a time — which is why it’s fluent, and also why it can be fluently wrong.
• Hallucination is built into how the technology works. Verify anything that carries real consequences.
• The quality of what you get out depends almost entirely on the detail you put in. Specific request, useful answer.
• The fastest way to learn this is to use it. Ten minutes of typing beats an hour of reading.

In the next lesson, we’ll look at where all this came from — a short history of AI that explains why it suddenly seemed to appear out of nowhere, when in fact it took seventy years. Then Module 2 gets practical: the AI you’re already using every day without noticing.