The short version
AI chat tools guess what word comes next. That’s it.
They read a huge amount of text and learned the patterns in it.
To guess the next word about tax, it had to sort of learn tax.
So it writes answers. It does not look them up.
That one fact explains everything else in this book.
Key idea
It’s a pattern machine, not a fact machine
It doesn’t “know” things like a filing cabinet does.
It makes text that looks like a good answer.
Usually that means it’s right. Sometimes it only looks right.
Words you’ll hear
- Model
- — the trained brain doing the work.
- Prompt
- — whatever you type.
- Token
- — about three quarters of a word. How AI counts text.
- Context window
- — how much it can hold in mind at once.
- Hallucination
- — when it makes something up. The key word in this book.
- Cutoff
- — the date it stopped learning. Anything newer needs web search.
- Agent
- — AI that does things, not just talks. Chapter 33.
How it differs from Google
Google finds pages. AI writes something new.
Google gives you links to read. AI gives you an answer to check.
Need one hard fact? Search. Need a draft, summary or plan? Use AI.
Most AI tools can now search the web too. Turn that on for anything factual.
The right way to think about it
Treat it as a smart new assistant who has read everything and knows nothing about you.
That means: give it context, check its work, don’t take the praise seriously.
It is not your friend, not awake, and remembers nothing unless the app saves it.
Watch out
Two ways people go wrong
Treating it as an oracle. Believing it because it sounds sure.
Treating it as a toy. Giving up after two bad tries with a four-word prompt.
The answer is in the middle.
Try this
Ten minutes
Ask any AI: “Explain what you are in plain words. Then list five things you’re bad at.”
Then ask: “Which parts of that are you least sure about?”
Explain what AI like you actually is, in plain words, using examples from the daily life of a [your job, e.g. kindergarten teacher in Ipoh]. No jargon. End with the one thing people most misunderstand about you.
Create a free account and you can edit this prompt and run it live, right here in the lesson. Sign in free →
I was told you don't actually 'know' things — you predict likely words. Show me what that means: answer 'What is nasi lemak?' and then explain, step by step, how you produced that answer.
Create a free account and you can edit this prompt and run it live, right here in the lesson. Sign in free →
The 30-second version
It guesses the next word, based on huge amounts of reading.
It writes answers rather than looking them up. That’s why it can invent things.
Google finds. AI drafts.
Think: a clever new assistant who knows nothing about your life.