Prediction is the superpower behind almost every AI you use, and last week's time series lab was your first taste. This week you make predictions yourself, measure how wrong you are, and learn that "being less wrong each round" is literally how models train. STEAM focus: mathematics.
Use creativity to solve problems with AI: music, design, and storytelling powered by models. Caribbean creativity plus AI is an unfair advantage.
Solve clues, use AI tools, win as a team. Every clue teaches you a different way to prompt and verify AI answers.
Make predictions with data and machine learning. We build a simple predictor together and test it against real numbers.
Here is monthly rainfall for a Kingston-area station from January to September. October is hidden. Your prediction error will be measured the exact same way a model's error is measured. Hint: think about what hurricane season does to rainfall.
Models are pattern-finding machines. Prove you can out-pattern one.
1. Home Data Logger (Tuesday to Thursday). Pick one thing you can measure daily: time your bus or taxi arrives, the afternoon temperature, or how long the light stays red at one crossing. Log it for three days, then predict day four before it happens. Bring your prediction and your error to Friday's session.
2. Prediction spotting (weekend). List three predictions AI made for you this week (weather app, "you might like" suggestions, traffic ETA). For each, name which of the five Week 2 model types is probably behind it.