← AI Foundations

⏱ About 60 min✦ 100 XP

Train Your Own Smart Helper

You have done something amazing. You have been on a big journey through Tier 1 of the Owens AI Institute. You learned what AI is. You learned how machines find patterns. You learned that AI needs examples to get smarter. You learned that AI has no feelings — it is a powerful tool made by people, for people. Now it is time for the biggest step of all. This is your capstone project. That means you are going to take everything you know and use it to build something real. You are going to train your own AI helper. Let's go!

Your Mission

Your mission is to teach a computer to tell two or three different things apart — using only your drawings. You will use a special tool called the Teachable Machine lab. You can open it right now inside the Owens AI Institute at: The Teachable Machine Lab (go to /institute/lab/teachable-machine) In the lab, you will: Pick two or three groups of things you want the machine to learn. Draw lots of examples for each group. Press the Train button and watch the machine learn. Test it with new drawings you haven't shown it before. Make it smarter by adding even more examples. Then you will explain what your machine learned to someone else.

What success looks like

When you are done, your trained helper should be able to look at a new drawing it has never seen before and make a good guess about which group it belongs to. It will not be perfect — and that is completely normal. Even the best AI systems in the world still make mistakes sometimes. What matters is that YOUR examples taught it something real.

Step-by-Step: Train Your Helper

Build Your Teachable Machine

  1. Step 1: Open the lab.
  2. Go to the Teachable Machine Lab at /institute/lab/teachable-machine. You should see a drawing canvas and some group boxes.
  3. Step 2: Choose what your helper will learn.
  4. Decide on two or three groups. Good ideas: cats and dogs. Circles and stars. Apples and bananas. Happy face and sad face. Pick something you enjoy — this is YOUR project!
  5. Step 3: Name your groups.
  6. Click on each group box and give it a clear name, like Group 1: Cat and Group 2: Dog. Clear names help YOU remember what each group is for.
  7. Step 4: Draw your first group.
  8. Select Group 1. Draw at least 10 examples of that thing. Draw them from different angles. Draw some big and some small. Draw some simple and some with details. The machine needs VARIETY to learn well.
  9. Step 5: Draw your second group (and third if you have one).
  10. Switch to Group 2. Draw at least 10 examples of that group too. Make sure the drawings actually look different from Group 1 — otherwise the machine will get confused.
  11. Step 6: Press the Train button.
  12. Click Train. Watch what happens! The machine is now studying all of your drawings and looking for patterns. This might take a few seconds.
  13. Step 7: Test your helper.
  14. In the test area, draw something new that you have not used as a training example. What does the machine guess? Does it get it right?
  15. Step 8: Improve your helper.
  16. If the machine gets it wrong, that is a clue! Go back and add more examples. Try drawing clearer pictures or more variety. Then press Train again. Keep improving until you are proud of how it does.

What if your machine keeps guessing wrong? Do not worry — this happens to everyone, even professional AI engineers. Here is what to do: First, look at your examples. Are they clear? Are they very different from each other? Are there at least 10 per group? Second, add more examples. More examples almost always help. Try drawing the same thing in different ways — big, small, tilted, simple, detailed. Third, train again. After you add new examples, press Train again so the machine can learn from your new drawings too. Remember: the machine is not broken when it makes mistakes. It just needs more help from you. YOU are the teacher here.

The secret to a great training set

Give your machine VARIETY. If you are teaching it to recognize cats, do not draw 15 copies of the exact same cat. Draw a big cat. Draw a tiny cat. Draw a cat sitting. Draw a cat standing. Draw a striped cat and a spotted cat. The more different your examples look from each other, the better your machine will get at recognizing cats it has never seen before.

Show What You Learned

Training the machine is only half the project. The other half is sharing what you built. Find a grown-up, a family member, a friend, or your class. Show them your trained helper. Let THEM draw something and watch the machine make a guess. Watch their face when the machine gets it right! Then explain how it works. Use your own words. You know this — you just did it yourself.

Present and Reflect

  1. Step 1: Demo your helper.
  2. Let someone else draw a picture in the test area. Watch what the machine guesses. Did it get it right? Show them a few examples.
  3. Step 2: Explain how you taught it.
  4. Tell them: what groups you picked, how many examples you drew, and what happened when you first trained it.
  5. Step 3: Answer these reflection questions out loud (or write them down if you like):
  6. - What did your machine get right most of the time?
  7. - What kinds of drawings did it get wrong? Why do you think that happened?
  8. - What was the trickiest part of being the teacher?
  9. - If you could train it on one more group, what would you add?
  10. Step 4: Explain where the smarts came from.
  11. Can you explain to your audience that the machine did not just know this? It learned it from YOUR drawings? Try saying: My machine learned from the examples I gave it. The pattern it found came from me.

Flashcards — click each card to reveal the answer

You trained your Teachable Machine with 5 examples of circles and 5 examples of stars. It keeps getting new drawings wrong. What is the BEST next step?

Your friend says: My Teachable Machine is SO smart — it figured out what circles look like all by itself! Is your friend right?

You are officially an AI teacher!

Think about what you just did. You decided what to teach. You created the examples. You checked how well the machine learned. You improved it when it made mistakes. That is exactly what AI researchers do — just with bigger datasets and fancier tools. Your machine is smart because YOU made it smart. Every correct guess it makes came from something you drew. That is real AI, and you built it.

You have finished Tier 1 of the Owens AI Institute — and you did not just learn about AI. You became someone who TEACHES AI. Whenever you use a voice assistant, a photo app, or a recommendation engine from now on, you will know something most people do not: that behind every smart-seeming machine is a pile of examples someone gathered, organized, and checked — just like you did today. When you are ready for the next challenge, the Middle-school tier is waiting. You will go deeper: bigger datasets, real-world bias, and the big questions about how we make AI fair and responsible. But for now — celebrate. You earned it.

← Back to AI Foundations