2 Chapter 1: What Is Intelligence?
Before you write a line of code, you need to answer a question that most AI courses skip: what are you actually trying to build?
2.1 The problem with “intelligence”
The word gets used to mean a dozen different things. Passing a test. Winning a game. Generating text that sounds human. Driving a car without crashing. These are all called “AI,” but they have almost nothing in common technically.
If you start building without a clear target, you end up with a system that’s impressive in demos and useless in practice.
2.2 A working definition
For this book, intelligence means: a system that takes in information, builds a useful internal representation, and uses that representation to make decisions that achieve a goal.
That’s it. No consciousness required. No general reasoning required. Just: perceive, represent, decide.
This definition is useful because it tells you exactly what you need to build:
- An input pipeline that takes raw data and turns it into something structured
- A model that learns representations from that data
- A decision layer that maps representations to actions
Every AI system, from a spam filter to GPT, fits this pattern. They differ in how complex each piece is, but the structure is the same.
2.3 Why “general intelligence” is the wrong starting point
A lot of people want to jump straight to AGI. That’s like learning carpentry by trying to build a skyscraper. You’ll learn more by building a chair first.
This book starts with narrow, concrete problems. You’ll build systems that solve specific tasks. By chapter 8, the systems get complex. By the end, you’ll understand why “general” intelligence is hard, not because you read about it, but because you ran into the walls yourself.
2.4 What you need to know going in
You should be comfortable with:
- Python (not expert level, but you should know classes, functions, and numpy)
- Basic linear algebra (vectors, matrices, dot products)
- Basic calculus (derivatives, chain rule)
If you’re shaky on the math, chapter 2 covers what you need. No more, no less.
2.5 What’s next
Chapter 2 builds the mathematical toolkit. You’ll implement forward passes, loss functions, and gradient descent from scratch before touching any framework.