AI Trainers
Gervigreindarþjálfararnir
They don't just use AI. They train it.
What they'll learn
- Train real models Build working models from images, sounds and movement using their own examples.
- Models as controllers Wire a trained model into a project so it becomes a live game controller.
- Measure and improve Check how accurate a model is and make it better with more and better data.
- Bias, proven hands-on See for themselves how biased data produces biased AI — and how to fix it.
A project controlled by a model they trained — with the training story on the wall beside it.
The 13-week journey
-
Sprint 1 — Training camp
Classifiers from scratch: images, sounds, poses; the garbage-in-garbage-out lab. Ends with a sorting app trained on the class's own data.
-
Sprint 2 — Models as controllers
Wiring a trained model into a project: jump on a hand-raise, steer with a lean, fire on a clap; accuracy tuning as gameplay tuning. Ends with a gesture-controlled build.
-
Sprint 3 — Fair or fooled?
The bias experiments: train narrow, watch it fail on the wider world; the real-versus-generated lab; mapping where AI already lives in their lives. Ends with the showcase build plus its training poster.
-
Week 13 — Showcase
The audience plays the gesture builds — then tries to fool the models, and learns why they mostly can't.
What we cover
Every topic, unit by unit — so you know exactly what your child builds and learns.
Training camp
- Building classifiers from scratch: images, sounds, poses
- Training data, examples and labels explained
- The garbage-in-garbage-out lab
- A sorting app trained on the class's own data
Models as controllers
- Wiring a trained model into a project
- Gesture controls: jump on a hand-raise, steer with a lean, fire on a clap
- Measuring a model's accuracy honestly
- Accuracy tuning as gameplay tuning
Fair or fooled?
- Bias experiments: train narrow, watch it fail on the wider world
- The real-versus-generated lab
- Improving a model with better, broader data
- Mapping where AI already lives in their daily feeds
Showcase & the training story
- Making a 'how I trained it' poster
- Presenting the model and its project
- The audience plays — then tries (and fails) to fool the models
A gesture- or voice-controlled project driven by a model the child trained, plus the 'how I trained it' poster.
The band’s signature differentiator. Students train real classifiers, wire them into their own projects as living controllers, then deliberately break them to learn about data, fairness and failure. Still entirely account-free — they build understanding before they’re ever handed a chatbot.
The hooks
Kid hook: “I trained an AI to recognize my moves — it’s the controller for my project.” Parent hook: “They don’t just use AI — they train it, test it, and see exactly where it fails. That’s real understanding.”
Who it’s for
No prerequisites; graduates of the Explorers path accelerate. Thrives: tinkerers, question-askers, and any child whose parents want substance behind the AI buzzword.
Outcomes — by the end, students can
Train image, sound and pose models with Teachable Machine and Machine Learning for Kids; explain training data, examples and labels in their own words; connect a trained model to a project as a live controller; measure accuracy honestly and improve it with better data; demonstrate how biased data produces biased results.
Tools & compliance
Teachable Machine and Machine Learning for Kids (account-free / school-managed), block-based project tools, school laptops with cameras and mics, consent-covered in-class camera use, nothing uploaded beyond the tools’ in-session processing.
Where this course fits
The Explorers path (How Machines Learn → AI All Around Us) is the natural prequel; AI Makers at 13–15 is the direct continuation.
Parent questions
Is this using ChatGPT?
No — under-13s never use chatbots here; they train their own models with account-free tools.
Is machine learning really teachable at 11?
Training-by-example is exactly how the field works, and children grasp it faster than adults.
What does my child come home with?
A working AI-controlled project and the ability to explain how it learned.
The first lesson is a free trial.
Book a no-commitment trial — pay nothing if it's not a fit.
Book a free trial