Your 8–10 Year Old’s First AI Project: A “Smart” Pet Classifier with Teachable Machine

Why this is a perfect first AI project (especially ages 8–10)

If your child is 8–10, they’re at a sweet spot: old enough to follow a clear sequence of steps, young enough to be amazed that a computer can “learn” from their examples.

A teachable machine project for kids works so well because it’s:

• Visual and immediate: They see results right away (no long setup).
• Low-code: Kids focus on the concept—training data—not syntax.
• Concrete: “Is it a cat or a dog?” makes sense in a way “optimize a model” doesn’t.
• A gateway skill: It quietly teaches the core of how to teach machine learning to kids: examples → patterns → predictions.

In this guide, you’ll help your child build a simple AI image classifier for beginners (kids) that recognizes pets (like cats vs. dogs, or “my stuffed bunny” vs. “my stuffed bear”). It’s also a great AI project for a 9 year old because it feels like magic—but it’s actually understandable.

What your child will learn:

• What “training” means in machine learning
• Why data quality matters more than “being smart”
• How to test and improve a model
• The basics of fairness (why your AI might be biased)

What you’ll need (and a 10-minute setup)

You don’t need special hardware. Keep it simple.

Recommended supplies:

• A laptop or tablet with a camera (laptop is easiest)
• Internet access
• 2–3 types of “pets” to classify:
  • Real pets (if they’ll sit still)
  • Stuffed animals
  • Printed photos of cats/dogs (works great)

Parent checklist before you start:

• Choose a workspace with good lighting
• Clear the background (busy backgrounds confuse the model)
• Pick two classes to start (you can add a third later)

Here’s a realistic time plan you can follow.

Step	What your child does	Parent role	Time
1	Pick categories (e.g., Cat vs Dog)	Help choose simple, clear labels	2 min
2	Collect examples with camera	Coach consistency (lighting, angle)	8–12 min
3	Train the model	Encourage patience; explain “training”	1 min
4	Test with new images	Ask questions; note mistakes	5–10 min
5	Improve (add better examples)	Help diagnose why it failed	10–15 min

Where to build it: Google’s Teachable Machine (search “Teachable Machine” in your browser). You’ll choose an Image Project and train it with your camera.

Build the “Smart” Pet Classifier (step-by-step)

This is the heart of the project. Read the steps once, then do them together.

1) Define the classes (labels)

Start with two categories:

• Class 1: Cat
• Class 2: Dog

If you don’t have pets nearby, use:

• Stuffed Cat vs Stuffed Dog
• Or even Animal vs Not Animal (toy car, book, etc.)

Parent tip: Labels should be mutually exclusive. If your child tries “Cute” vs “Not Cute,” you’ll end up debating feelings, not training a model.

2) Collect training examples (this is the “teaching”)

Teachable Machine will let you record examples for each class.

Aim for 30–60 images per class to start. Variety matters.

Collect examples like a pro:

• Change angles: front, left, right
• Change distance: close-up, medium, far
• Change background slightly: desk, couch, floor
• Change lighting slightly: brighter, dimmer

But keep these consistent (especially for the first run):

• Keep the pet centered in the frame
• Avoid extra objects that “give away” the answer (like a cat always next to a specific pillow)

A simple rule you can tell your child:

• “If the AI is guessing wrong, it might be because we taught it with confusing examples.”

3) Train the model

Click Train Model. This is usually fast.

Explain it like this:

• “The computer is looking at all the cat pictures and finding patterns.”
• “It’s not thinking like a human; it’s comparing visual clues.”

This is a great moment to teach a big idea in kid-friendly language: the AI is only as good as its examples.

4) Test it with brand-new images

Now the fun part: try “trick” tests.

Have your child test:

• A new cat/dog picture it hasn’t seen
• The pet from a new angle
• A stuffed animal that looks similar
• A photo on a phone screen (often harder)

As you test, ask:

• “What do you think it will guess?”
• “How confident does it look?”
• “What might be confusing it?”

5) Improve it (the most important learning moment)

When it makes mistakes (it will), don’t rush past it. This is where your child learns real machine learning.

Common fixes:

• If it mistakes cat for dog: add more cat examples with similar angles
• If it reacts to the background: add examples with new backgrounds
• If it guesses randomly: you may need clearer framing or more examples

Quick debugging questions (kid-friendly):

• “Did we show it enough kinds of cats?”
• “Did we accidentally teach it that the couch means ‘dog’?”
• “Do we have the same number of examples in each class?”

Make it work like a real scientist: quality, fairness, and safety

This section turns a fun activity into a meaningful learning guide.

A simple “data quality” checklist

Your child doesn’t need statistics—they need habits.

Use these practical rules:

• Balance your classes: try to keep cat and dog examples similar in count.
• Vary the examples: different colors, sizes, angles.
• Avoid shortcuts: don’t always hold the cat in your left hand and the dog in your right.
• Keep backgrounds mixed: otherwise it learns “living room” instead of “cat.”

Teach “bias” without making it scary

Kids can understand this:

• “If we only show the AI orange cats, it might not recognize a black cat.”

Try a mini challenge:

• Train with only one type of cat photo first.
• Test on a very different cat photo.
• Then add more variety and watch it improve.

That’s a gentle, concrete way to show why fairness and diversity in data matters.

Keep it safe and parent-friendly

If you use your child’s camera, keep privacy simple:

• Use stuffed animals or printed pictures if you prefer
• Avoid showing faces or personal information in the background
• Save or share the project only if you’re comfortable with where it’s stored

Next Steps: turn this into a mini AI “portfolio” in one weekend

Once your child has a working classifier, you can stretch it into a bigger learning win—without making it complicated.

Level up challenges (pick one)

• Add a third class: Cat vs Dog vs “Other” (like a teddy bear)
• Make it harder: classify “Sitting Cat” vs “Standing Cat” (pose recognition)
• Speed test: how fast can they gather 20 high-quality examples?
• Robustness test: try different lighting (lamp on vs off) and see what breaks

A simple reflection page (great for confidence)

Have your child write or tell you:

• “My AI can…”
• “My AI gets confused when…”
• “To make it better, I will add…”

How to get started today (10-minute plan)

1. Pick two labels: Cat and Dog (or two stuffed animals).
2. Collect 30 examples each with good lighting.
3. Train and test with 5 new items.
4. Improve by adding 10 better examples where it struggled.
5. Celebrate like a builder: name the model (e.g., “Pet Vision 1.0”).

If you want a more guided path, Intellect Council’s interactive lessons can help your child connect this project to the bigger ideas—like patterns, prediction, and responsible AI—through small, game-like missions.