What IS Machine Learning?

Rules vs learning — and why that difference changes everything

← Course Home Module 1 of 8 Module 2: Your First Dataset →

⏳ Loading Python… (first load ~15 seconds)

📌 Before You Start

What you need for this module:

Basic Python knowledge (variables, if-statements, functions)
A curious mind — no ML experience required

Estimated time: ~45 minutes

What you’ll learn: The core idea behind ML, the 3 types of ML, and the workflow every ML project follows.

💡 The Big Idea

Traditional programming: You write the rules. The computer follows them.

Machine learning: You show the computer examples. The computer finds the rules.

Think about spam email. A traditional programmer might write: "if the email contains 'free money' or 'click here', mark it spam." This works — until spammers change their wording.

A machine learning approach shows the computer thousands of labeled examples (spam / not spam) and lets it figure out the patterns. When spammers adapt, you retrain with new examples. The rules update themselves.

This is the core shift: from explicit rules to learned patterns.

🧠 How It Works

The 3 Types of Machine Learning

🏫 Supervised Learning

You provide labeled examples (inputs + correct answers). The model learns to map inputs to outputs.

Examples: spam detection, image classification, predicting house prices.

🔍 Unsupervised Learning

You provide data without labels. The model finds hidden structure and patterns on its own.

Examples: customer segmentation, topic discovery in documents, anomaly detection.

🎲 Reinforcement Learning

An agent learns by trial and error, receiving rewards for good actions and penalties for bad ones.

Examples: game-playing AI (chess, Go), self-driving car navigation.

The ML Workflow

Collect Data — Gather labeled examples relevant to your problem. More diverse, high-quality data = better model.

Prepare Features — Clean missing values, encode categories, scale numbers. Raw data is rarely model-ready.

Train the Model — Show the model your training examples. It adjusts its internal parameters to minimize errors.

Evaluate on Test Data — Check how well the model performs on data it has never seen. This is the true measure.

Predict & Iterate — Use the model on new data. Collect feedback, improve, retrain. ML is a cycle, not a finish line.

ML You Already Use

Netflix recommendations — supervised + collaborative filtering (what do similar users like?)
Autocorrect & autocomplete — language models trained on billions of text examples
Spam filter — supervised classification trained on labeled emails
Face unlock on your phone — trained on images of your face vs. others
Credit card fraud detection — flags transactions that don’t match your normal pattern

▶️ See It In Code

This example shows the conceptual difference: a traditional rule-based classifier vs what the ML approach looks like. Run it and observe the output.

# ============================================ # TRADITIONAL PROGRAMMING: You write the rules # ============================================ def classify_email_traditional(email): email = email.lower() spam_words = ["free money", "click here", "winner", "congratulations you've won"] for word in spam_words: if word in email: return "SPAM" return "not spam" # Test it emails = [ "Win free money now! Click here!", "Meeting at 3pm tomorrow in Room 204", "Congratulations you've won a prize!", "Can you review my pull request?" ] print("=== Traditional Rule-Based Classifier ===") for email in emails: result = classify_email_traditional(email) print(f" [{result:^10}] {email[:50]}") print() print("=== The ML Approach (coming in Module 4!) ===") print(" Instead of writing rules, we'll show the model") print(" thousands of labeled examples like these:") print() training_examples = [ ("Win free money now!", "spam"), ("Meeting at 3pm tomorrow", "not spam"), ("Click here to claim your prize", "spam"), ("Your invoice is attached", "not spam"), ] for text, label in training_examples: print(f" Label: {label:^10} | Text: {text}") print() print(" The model learns the pattern. We'll build this in Module 4!")

This is a static display — the interactive exercise below lets you modify the code.

👋 Your Turn

The classify_email_traditional function currently only catches 4 types of spam phrases. Add 3 more spam phrases to the spam_words list and test them.

Try phrases like: "act now", "limited time offer", "you have been selected" — or make up your own!

def classify_email_traditional(email):
    email = email.lower()
    # Add 3 more spam phrases to this list!
    spam_words = [
        "free money",
        "click here",
        "winner",
        "congratulations you've won",
        # YOUR CODE HERE: add 3 more spam phrases
    ]
    for word in spam_words:
        if word in email:
            return "SPAM"
    return "not spam"

# Test with your new phrases
test_emails = [
    "Act now — limited time offer expires tonight!",
    "You have been selected for an exclusive deal",
    "Hey, did you finish the homework for Tuesday?",
    # Add one more test email of your own!
]

print("=== Testing My Enhanced Spam Classifier ===")
for email in test_emails:
    result = classify_email_traditional(email)
    print(f"  [{result:^10}]  {email[:60]}")

Output will appear here after you click Run…

💡 Hint: Add strings to the list like "act now". The .lower() call means you don’t need to worry about capitalization — just use lowercase in your list.

☕ Brain Break — 2 Minutes

Think about 3 apps you use regularly. For each one, ask yourself:

Does it get better over time, or learn from your behavior?
Does it make personalized recommendations?
Does it predict something (next word, next song, next product)?

If yes to any of the above — it almost certainly uses ML.

Examples to get you started: Spotify, TikTok, Google Maps, Gmail, Instagram, Duolingo, your phone’s keyboard. What does each one learn about you?

✅ Key Takeaways

Traditional programming = you define the rules. ML = the model learns the rules from data.
The 3 types of ML: Supervised (labeled data), Unsupervised (find patterns), Reinforcement (rewards & penalties).
This course focuses on supervised learning — the most common type in industry and research.
The ML workflow is always: Data → Features → Train → Evaluate → Predict → Iterate.
ML is already everywhere in your daily life — you’re learning to build the systems you already use.

🎉 Module 1 Complete!

You now understand the core idea behind machine learning. In the next module, we’ll work with a real dataset and learn how to explore it before building any model.

Continue to Module 2: Your First Dataset →

← Course Home Module 1 of 8 Module 2: Your First Dataset →