Why do some creatures have fur while others have scales? Can you predict a species just from height and number of legs? What makes one rule better than another?
In How Machines Learn, we explore how machines turn data into decisions. Throughout the course, we study a mysterious collection of fictional creatures, each described by measurable characteristics. Using this same dataset, we gradually uncover the ideas that power modern AI.
We begin by looking at data as points in space: grouping similar creatures together and asking what “similar” really means. From there, we build decision rules and grow them into decision trees. We compare different algorithms on the same task and investigate how to measure performance — when is a model accurate, and when is it just lucky? We then explore learning through rewards before culminating in a simple neural network that combines weighted rules into a powerful predictive system.
Along the way, we’ll encounter nearest neighbors, clustering, decision trees, evaluation metrics, and perceptrons — not as black boxes, but as logical tools we can understand and analyze.
Prerequisites: No programming is required. Curiosity, logical thinking, and a willingness to question how decisions are made are strongly encouraged.