1. What is Machine Learning?
Traditionally, when we want a computer to do something, we write a list of rigid rules: Rules + Data = Answers.
But how do you write a rule for what a cat looks like? A cat's ears fold, tail bends, color changes. It's too complex!
Machine Learning flips the script: Data + Answers = Rules. We show the computer thousands of cat photos, and the computer figures out the rules itself.
2. Linear Classification
The simplest way computers learn is called Classification—drawing a boundary line to separate different classes of objects.
Suppose we operate a packing house sorting Apples and Oranges. We measure two variables:
• Weight (in grams)
• Redness (scale from 0 to 100)
Apples tend to be heavier and redder. Oranges are lighter and more orange. If we plot them on a graph, can we draw a line to separate them?
3. The Limits of Hardcoding
If we hardcoded this decision, we might write:
"If Redness > 50 AND Weight > 150g, then it's an Apple."
But what about a tiny, exceptionally red apple? Or a giant, heavy orange? Our static rule gets confused because real-world data overlaps.
A traditional system fails when things get fuzzy. This is why we need a system that adapts and draws a dynamic line.
4. Let's Train a Linear Separator!
Below is the Fruit Factory Inspector Widget. Apples are red dots; oranges are orange dots.
Your task is to act as the ML algorithm! Adjust the sliders to rotate and slide the decision boundary line until it perfectly separates the fruits. Then click "Run Test".
5. Perceptrons to Neural Networks
The single line you just adjusted is the fundamental building block of AI, created in 1957—the Perceptron.
A perceptron makes binary choices by checking which side of the boundary a data point falls on.
By linking **millions** of these lines together in layers, we create Neural Networks. This allows computers to understand images, recognize speech, and generate text.
6. Congratulations!
You have completed Lesson 1! You have learned:
• The core difference between programming and machine learning.
• How computers separate categories using linear classifiers.
• How perceptrons scale up to modern neural nets.
In the next lesson, we will move to Neural Networks and see how to train a car to drive itself on a virtual track!