Learning about machine learning from kids
summary
Kids use Scratch blocks wired to ImageNet and simple face, sound and text classifiers to build supervised models for chatbots, Pac‑Man agents, rock‑paper‑scissors, postcode OCR and a book‑shelf sorter; many projects run successfully but failures reveal lessons. Training more examples steadily improves the Pac‑Man controller, a crowdsourced tic‑tac‑toe model reaches draws with hundreds of samples, a newspaper headline project teaches train/test separation, and experiments with a confidence threshold (kids call it “shy”) expose calibration, bias and how stacked models combine.
AI-written summary · about
chapters (15)
- 0:00Intro
- 0:42Scratch
- 3:28Training models
- 5:24Pacman game
- 6:56Knots and crosses
- 8:30Media literacy
- 10:07Machine learning with text
- 11:57Alexa example
- 14:35Confidence scores
- 16:58Other ways of using confidence scores
- 21:20Rock Paper Scissors
- 23:12Recommendation Engine
- 27:30Sound Recognition
- 29:05Fruit Recognition
- 33:20Books
