The first thing to understand about modern AI is that it is not thinking. It is not reasoning. It is not, in any meaningful sense, alive. It is a very large, very confident pattern-matcher.
When you ask ChatGPT a question, it is not "looking up" the answer. It is predicting, one piece at a time, what the next most-likely word would be — based on all the text it has seen during training. That is the entire trick. Everything else is engineering on top.
A useful analogy
Imagine the world's most overqualified autocomplete. The kind of autocomplete that has read every book, every Wikipedia article, every Reddit thread. When you type "the capital of France is", it predicts "Paris" — not because it knows the capital of France, but because in the trillions of sentences it has seen, "Paris" is the word that almost always comes next.
That is, more or less, what a large language model does. It is autocomplete with a very, very long memory.
What this means in practice
- AI is good at things that look like patterns it has seen before.
- AI is bad at things that require genuinely new reasoning.
- AI does not know what it does not know — which is why it makes things up.
That last one is the part most people get burned by. We will get to it in part 4.
The model is wrong. The model is always wrong. The art is knowing by how much.
The bottom line
If you remember one thing from this part, remember this: AI is statistics, not magic. Treat it like a very fast, very well-read intern who occasionally invents facts and is too polite to admit when it doesn't know.
Next up: how the pattern-matching actually happens, without a single equation.