How Machines Learn
Hvernig vélar læra
They teach a computer — and can tell real from AI.
What they'll learn
- Learning from examples They teach a model themselves and watch it get smarter as they add better examples — the core idea behind all AI.
- Training real models Build image and sound classifiers in Teachable Machine that recognise their own drawings, objects and voices.
- Why AI gets it wrong Discover how too few or one-sided examples break a model — and trigger those mistakes on purpose.
- Spotting AI fakes Learn the tell-tale signs that separate a real photo from an AI-generated one.
A trained model the audience tries to fool, and a Real-or-AI quiz for the family.
The 13-week journey
-
Sprint 1 — How machines learn
Pattern games, 'train your classmate' (a child acts as the model, fed only examples), and a first Teachable Machine model: the class pet that recognizes their drawings.
-
Sprint 2 — Real or AI?
Sound models, the 'confuse the model' lab — starving it of examples to watch it fail — and spot-the-fake training on AI images. Ends with a how-to-tell report.
-
Sprint 3 — The Real-or-AI challenge
Teams build a Real-or-AI quiz for their families and a simple guessing game built around their model, with an online-safety thread throughout.
-
Week 13 — Showcase
The audience tries to fool the class model live, then takes the family quiz. Kids are graded on how hard their model was to trick.
What we cover
Every topic, unit by unit — so you know exactly what your child builds and learns.
How machines learn
- Pattern games and 'train your classmate' (a child acts as the model)
- Learning from examples — the core idea of machine learning
- Training a first image model in Teachable Machine
- A class pet model that recognizes their drawings
Real or AI?
- Training a sound-recognition model
- The 'confuse the model' lab — starving it of examples to watch it fail
- Why AI makes mistakes: too few or one-sided examples
- Spot-the-fake training on AI-generated images
The Real-or-AI challenge
- Building a Real-or-AI quiz for families
- A simple guessing game built around their model
- Online-safety habit: when something feels off, ask a grown-up
- Testing a model instead of trusting it
Showcase & telling real from AI
- Making the model hard to fool
- Spotting common signs of AI-generated pictures
- Running the live 'fool the model' challenge and family quiz
A trained model the audience tries to fool — and usually can't — plus a Real-or-AI family quiz.
The course every parent asks about, and the front door to an AI-only school. Children learn how machines learn — by teaching them — and turn deepfake worry into a calm, capable skill: telling real from AI. Everything runs through games, examples and account-free tools; nobody this age touches a chatbot.
The hooks
Kid hook: “I taught a computer to recognize my drawings. And I can spot fake pictures.” Parent hook: “My eight-year-old can explain how AI learns — and spot a deepfake. Can yours?”
Who it’s for
No prerequisites; pure curiosity required. Thrives: the why-askers, the pattern-spotters, and any child whose parents want them AI-literate before they’re AI-exposed.
Outcomes — by the end, students can
Explain in their own words that AI learns from examples; train an image and a sound classifier with Teachable Machine; predict when and why a model will fail; spot common signs of AI-generated images; play examiner — testing a model instead of trusting it; follow the family rule: when something online feels off, ask a grown-up.
Tools & compliance
Teachable Machine and Quick, Draw! (no accounts required), school tablets/laptops, instructor-driven demos for anything generative. No chatbot access at this age — stated proudly in parent communication, not buried.
Where this course fits
Level 1 of the Explorers path; Teach a Machine is Level 2, and AI Trainers at 10–12 is the direct sequel.
Parent questions
Will my child use ChatGPT?
No — under-13s never touch chatbots here; they train account-free models and learn the mechanics instead.
Is AI a real subject for a 7-year-old?
Learning-from-examples is — and it's the single most useful media-literacy skill of their generation.
Is it scary?
Not at all — it's playful and hands-on; the tone is empowerment, never fear.
The first lesson is a free trial.
Book a no-commitment trial — pay nothing if it's not a fit.
Book a free trial