Teach a Machine
Kenndu vél
The computer doesn't know — until they teach it.
Book a free trialWhat they'll learn
- AI starts knowing nothing Discover that a model knows nothing until they show it examples — they are the teacher.
- Better lessons, smarter AI See how more and better examples make a model noticeably smarter.
- Good data in kid language Understand labels, categories and what 'good data' really means, explained simply.
- Why too little confuses it Watch a model trained on too few examples get confused — and learn why.
A 'smart machine' the child trained to sort the class's things — demoed live.
The 13-week journey
-
Sprint 1 — The empty-headed computer
Acting games where the class 'is' the model; the first lesson — a computer that knows nothing until shown examples. Ends with a two-category model trained from scratch.
-
Sprint 2 — Better lessons, smarter machine
Adding examples and watching accuracy climb; clear vs confusing examples; the 'too few examples' trap. Ends with a model the class improves together.
-
Sprint 3 — Teach your own
Each child picks something to teach the computer to recognize, gathers examples, and writes its 'lesson plan'. Honesty thread: the model is only as good as its lessons.
-
Week 13 — Showcase
Live demos of each trained machine, plus the audience trying to stump them.
What we cover
Every topic, unit by unit — so you know exactly what your child builds and learns.
The empty-headed computer
- Acting out 'being the model' — learning only from examples
- A computer knows nothing until it's shown things
- Training a first two-category image model
- Categories and labels in plain language
Better lessons, smarter machine
- Adding examples and watching accuracy climb
- Clear examples versus confusing ones
- The 'too few examples' trap
- Improving a class model together
Teach your own
- Choosing something to teach the computer
- Gathering good, varied examples
- Writing the model's 'lesson plan'
- Honesty: a model is only as good as its lessons
Showcase & testing
- Testing a model to find its weak spots
- Polishing the demo
- Live demos and audience stump-the-machine
A child-trained classifier sorting real objects in front of the audience, with its 'lesson plan' on the wall.
The clearest possible first lesson in how AI works: the computer is a student that knows nothing, and the child is the teacher. Train it, test it, improve it — and feel in your hands why data quality is everything.
The hooks
Kid hook: “I’m the teacher and the computer is my student — and it only knows what I taught it.” Parent hook: “They learn the single most important idea about AI: it’s only as good as the examples it’s given.”
Who it’s for
Curious 7–9s; no experience needed. Thrives: the explainers, the sorters, and any child who loves being the one who knows.
Outcomes — by the end, students can
Explain that AI learns only from examples; train and improve an image classifier; tell a clear example from a confusing one; predict why a model fails; describe in their own words what ‘good data’ means.
Tools & compliance
Teachable Machine on school tablets/laptops (no accounts), instructor-led demos. No chatbots at this age.
Where this course fits
Level 2 of the Explorers path: follows How Machines Learn, leads to Make with AI — and on to AI Trainers at 10–12.
Parent questions
How is this different from Level 1 (How Machines Learn)?
Level 1 introduces how AI learns and telling real from AI; Teach a Machine goes deeper into training, testing and improving models of their own.
Does my child need to read well?
No — it works with pictures, sounds and spoken instruction.
Any chatbots?
None. Account-free training tools only at this age.
The first lesson is a free trial.
Book a no-commitment trial — pay nothing if it's not a fit.
Book a free trial