Module 2: How Does AI Work?

Imagine teaching a child to recognize dogs. You don't hand them a rulebook saying "four legs + fur + tail = dog." Instead, you show them hundreds of photos: big dogs, small dogs, cartoon dogs, even plush toys. After enough examples, something magical happens — the child stops memorizing and starts recognizing.

That's AI in a nutshell: learning from examples to spot patterns. No programmer sits down and writes "if tail wags, then happy dog." The machine figures it out by observing thousands of cases.

In this module, you'll use tools like ChatGPT and Teachable Machine to see AI learning in action—experiencing firsthand how machines recognize patterns and make predictions.

🧠 What AI Actually Does

At its core, AI works by:

1. Processing data — analyzing examples (images, text, numbers)
2. Finding patterns — detecting relationships and rules automatically
3. Making predictions — applying learned patterns to new situations
4. Improving over time — adjusting based on feedback

Unlike traditional software that follows pre-written rules, AI creates its own rules by observing data. It's less "programming" and more "teaching through examples."

👨‍🍳 The Chef Learning Analogy

Think of AI as a chef learning a recipe — not by reading instructions, but by tasting hundreds of dishes:

• After 10 pasta dishes, the chef notices: "Ah, salt makes it pop!"
• After 50 dishes: "More garlic = bolder flavor"
• After 200 dishes: "This combination of herbs creates magic"

The chef never saw a recipe book. They learned by pattern recognition across hundreds of samples. That's exactly how AI models learn — by "tasting" data until patterns emerge.

📧 Real-World Example: Your Spam Filter

Your email spam filter is a perfect AI example. Here's what it does NOT do:

Instead, here's what actually happens:

The genius? The spam filter never "knows" what spam is — it just recognizes patterns statistically associated with junk mail. When scammers change tactics, the filter learns new patterns.

🎯 Hands-On Exercise: Train Your Own AI

Let's make this concrete. You're going to train an AI to recognize emotions in text — without writing a single line of code.

Step-by-Step Instructions:

1. Go to: teachablemachine.withgoogle.com
2. Choose: "Text Project"
3. Create two classes:
   • Class 1: "Positive"
   • Class 2: "Negative"
4. Feed 10 examples each:
   • Positive: "I love this!", "Amazing experience!", "Best day ever!"
   • Negative: "This is terrible", "I'm so disappointed", "Worst purchase"
5. Click "Train Model" (watch it learn patterns)
6. Test it: Type "This made me smile" — does it guess correctly?

Extra Challenge:

Try confusing the AI with sarcasm: "Oh great, another Monday." Does it understand context? (Spoiler: probably not — that's a current limitation!)

⚠️ Common Mistakes About AI Learning

📝 Mini-Project: Movie Sentiment Predictor

Let's apply what you've learned in a practical project you can do right now:

What you'll discover: The AI identifies patterns you might not consciously notice — like "but" often signals a negative turn, or "exceeded expectations" indicates satisfaction.

📚 Summary: The AI Learning Process

Let's recap what we've learned about how AI works:

• ✅ AI learns from examples, not explicit programming
• ✅ Pattern recognition is the core mechanism
• ✅ Quality data matters more than quantity
• ✅ Feedback loops help AI improve over time
• ✅ AI doesn't understand — it detects statistical correlations

📝 Test Your Understanding

🚀 Next Step: Types of AI

Now that you understand how AI learns from data, let's explore the different types of AI that exist — from narrow specialists to hypothetical general intelligence.

📚 Continue Learning: Take Your AI Understanding Further

Want to dive deeper into how AI works? These courses and resources offer structured learning paths for serious students.