Your First AI Model

Time to Get Hands-On! 🎉

You've learned what AI is and how machines learn. Now it's time to build one yourself.

Don't worry — you won't need to install anything or write any code. Everything happens right here in your browser.

---

What We'll Build

We're going to train an AI that recognises drawings. You'll:

1. Draw examples of different shapes (circle, square, triangle)
2. Label them ("this is a circle", "this is a square")
3. Watch the AI learn from your examples
4. Test it with new drawings to see if it learned correctly

This is real machine learning — the same concept used by Google, Apple, and every major tech company.

---

Try It: Go to the Playground

Head over to our AI Playground to try the Drawing Recogniser!

Here's what to do:

1. Click on the drawing canvas
2. Draw a simple circle
3. The AI will try to guess what you drew
4. Try different shapes — squares, triangles, stars
5. See how confident the AI is in its guess

---

What's Happening Behind the Scenes?

When you draw and the AI guesses, here's what's happening in simple terms:

1. Your drawing becomes numbers — The computer doesn't "see" your drawing. It converts it into a grid of numbers (bright pixels = high numbers, dark pixels = low numbers)

2. It compares to known patterns — The AI has been trained on thousands of similar drawings. It knows that circles tend to have certain patterns of numbers, squares have others

3. It calculates confidence — For each shape category, it calculates a percentage: "72% circle, 18% oval, 10% square"

4. It gives its best guess — The highest percentage wins!

This is the same basic process used in medical imaging — doctors' AI tools look at X-rays and compare them to millions of previous X-rays to detect problems.

---

Understanding the Results

When you see the AI's guess:

• High confidence (80%+) — The AI is quite sure. Your drawing closely matches patterns it has seen before.
• Medium confidence (50-80%) — The AI is somewhat sure, but the drawing could match multiple shapes.
• Low confidence (below 50%) — The AI is guessing. Your drawing doesn't clearly match any pattern it knows.

---

Experiment Ideas

Try these experiments to understand AI better:

Experiment 1: Draw the Same Shape Differently

Draw a circle 5 different ways — big, small, thick, thin, wobbly. Does the AI still recognise all of them?

Experiment 2: Try to Trick It

Draw something between a circle and a square (a rounded square). What does the AI say? This shows you how AI handles ambiguity.

Experiment 3: Draw Something Unknown

Draw a star or a heart. The AI wasn't trained on these — what happens? This demonstrates the limits of training data.

---

What You've Just Done

Congratulations! 🎉 You've just experienced the core concepts of AI:

• Training data — The examples the AI learned from
• Pattern recognition — How the AI identifies shapes
• Prediction — Making a guess based on learned patterns
• Confidence scores — How certain the AI is
• Limitations — AI can only recognise what it was trained on

These same concepts power everything from self-driving cars to medical diagnosis to voice assistants.

---

Quick Recap 🎯

1. AI converts your drawing into numbers and compares it to known patterns
2. Higher confidence = the drawing more closely matches known examples
3. AI can only recognise what it's been trained on — it can't handle truly unknown things
4. The same basic process powers real-world AI applications

---

What's Next? 🚀

Coming soon: AI in the Real World — where we'll explore how AI is used in healthcare, agriculture, education, and more. Plus, we'll discuss the important topic of AI ethics and bias.

Stay curious! 🌟