AI Mistakes and Limitations
Machines make mistakes; they only know what they've been shown; bad training data leads to bad results; AI is not magic — just maths on data; showing edge cases and failures
Share and support
Teaching Table · Plan outline
Set out your lesson
A reviewed step-by-step guide is not available yet. This is an outline from the plan, not a fully authored guide.
No teaching instructions are available for this lesson. Choose another topic or return to your plan.
Family records and proof options
Keep a record of your learning
Save photos, notes, and attendance. Export a proof packet for compliance or portfolio reviews.
Learning resources
Reviewed external links that reinforce this concept. They do not replace Gakuva’s teaching approach.
Code.org · Practice link — not the full lesson
AI for Oceans
Deliberately probe mistakes: train a limited model, then test odd examples to see confident wrong answers. Discuss that AI can fail without “understanding.”
Free · no login required to play · account optional to save · no ads observed
Review and open practice
Save a screenshot of a wrong or uncertain classification and a parent note of why the learner thinks the model slipped.
Opening this link does not save learning or assess mastery.
Open practice (new tab)Source, access and review notes
Deliberately probe mistakes: train a limited model, then test odd examples to see confident wrong answers. Discuss that AI can fail without “understanding.”
Save a screenshot of a wrong or uncertain classification and a parent note of why the learner thinks the model slipped.
Free · no login required to play · account optional to save · no ads observed
Unknown — link only
Google Creative Lab · Practice link — not the full lesson
Teachable Machine · Image
Adult-guided mistake lab: train with unbalanced or similar-looking classes, then show a misclassification and improve the examples.
Free · no login · no ads observed · webcam optional
Review and open practice
Save a misclassification screenshot and a second screenshot after adding better training examples.
Opening this link does not save learning or assess mastery.
Open practice (new tab)Source, access and review notes
Adult-guided mistake lab: train with unbalanced or similar-looking classes, then show a misclassification and improve the examples.
Save a misclassification screenshot and a second screenshot after adding better training examples.
Free · no login · no ads observed · webcam optional
Unknown — link only
Check understanding
- Give an example of AI making a mistake (voice assistant mishearing, auto-correct error, wrong recommendation)
- Explain that AI mistakes happen because of gaps or errors in training data
- Describe why AI is not magic — it follows mathematical rules applied to data
“If an AI incorrectly identified a picture of a muffin as a chihuahua, could your child explain why that kind of mistake happens?”
Curriculum record
- Type
- Conceptual
- Subject
- Computing
- Domain
- Artificial Intelligence
- Age range
- Ages 7–9