It Knows Only What It Learns
by Anonymous · ⭐ 1 · 🍴 0
🧺 AI seems smart because it has “seen a lot.” In the first tab, teach it by putting cat, dog, and rabbit pictures into baskets yourself. Then empty the baskets completely—the AI can’t say anything, because no learning means no answers. Can you find a way to make it recognize all three animals with just three pictures? ⚖️ Next, fix it. With a basket holding 6 cat pictures and 1 dog picture, it calls a dog “cat.” You can add more dogs, but it’s fun to discover that removing cats fixes it too. The key isn’t “how many,” but “what it compares.” 🎯 Last, find the reason. Where should you look when AI gets it wrong? The answer is always in the basket. This app was made with code too · Remix it and try adding other animals!
Ready to remix — hit the 🍴 Remix button below. You'll sign up first, then this app is copied into your workspace.
About this app
It only knows what it saw is an interactive AI literacy game for children. Through three tabs, learners teach a simple simulated AI using cat, dog, and rabbit picture cards, correct an unbalanced training basket, and identify why the AI makes each prediction. The activities demonstrate limited training data, class imbalance, and how the examples in a dataset influence results.
Use cases
- Elementary students learning AI and machine learning basics
- Teachers introducing training data and data bias
- Families exploring how examples affect an AI prediction
- Computer science lessons about classification and datasets
Features
- Add or remove cat, dog, and rabbit cards from a training basket
- Test the simulated AI with selected animal pictures
- Complete four teaching missions, including an empty basket and a three-card challenge
- Correct biased datasets by adding or removing examples
- Choose explanations for predictions based on what is present or missing
- Track mission progress and reset or advance through activities