The one paragraph that explains it
An AI chatbot — ChatGPT, Claude, Gemini, whichever — is what's called a large language model. It was shown an almost unimaginable amount of human writing: books, articles, forums, code, conversations. From that, it learned patterns in how words follow other words. When you type a question, it isn't looking anything up. It's predicting, one piece of a word at a time, what a good answer would probably look like — based on everything it's read. That's the whole trick. It's a very sophisticated autocomplete, not a mind and not a database.
That one fact explains almost everything you'll notice in the next four lessons: why it can write a warm, specific paragraph about your business in nine seconds (it has read millions of paragraphs like it), and why it can state something completely false with total confidence (it's predicting a plausible-sounding answer, not consulting a fact-checker). Same mechanism, two very different outcomes.
What it's genuinely good at
- Language, in almost any direction. Writing, rewriting, summarizing, translating, explaining, brainstorming. This is its home turf.
- Holding a lot of text in mind at once. Paste in a long document and ask questions about it — it can do that far better than it can guess an obscure fact from memory.
- Being a thinking partner. Talking through a messy decision, playing devil's advocate, asking you questions you hadn't thought to ask yourself.
What it's genuinely bad at
- Being right about obscure facts. Specific dates, statistics, names, and citations get invented with total confidence. More on this in Lesson 4.
- Knowing what happened recently, unless it's specifically searching the web for you (some tools do this now — you'll usually see a "searching" indicator).
- Knowing anything about you that you haven't told it in this conversation (or saved to its memory, where that feature exists).
Think of it less like a librarian who knows facts, and more like a brilliant, extremely well-read intern who's fast, tireless, and occasionally makes things up with a completely straight face. You wouldn't publish an intern's first draft unread — same rule here.
So is it "intelligent"?
Depends what you mean. It can do things that look startlingly like reasoning — solve a logic puzzle, spot a flaw in an argument, write code that runs. It cannot want anything, feel anything, or know that it exists. The honest, useful stance: judge it by what it produces, not by what it might "be." You don't need to resolve the philosophy to use the tool well.
Open any AI chatbot you have access to (or wait until Lesson 3, where we'll set one up together) and ask it one real question about something you actually care about. Notice how it answers — the shape, the tone, the confidence — more than whether the answer is "right." That's the skill this whole course is building.