A first course in how machines sense, reason, and learn — from first principles.
AI is the science of building systems that perform tasks we'd call intelligent — recognizing a face, understanding a sentence, or choosing a winning move.
The key idea
Instead of being told every rule by a programmer, a modern AI system learns patterns from examples — much like a student learns from practice.
Before a system can recognize a cat or translate a sentence, it has to learn from data. This module is about what "learning" actually means for a machine.
The system studies labeled examples — inputs paired with the right answers — and learns to predict the answer for new inputs.
No labels at all. The system explores raw data and discovers its own structure — groupings, patterns, and outliers.
The system learns by doing — trying actions, receiving rewards or penalties, and improving its strategy over time.
Information flows from an input layer, through hidden layers that combine and weigh signals, to an output — the network's answer. Training nudges those weights until the answers get better.
Train your first model to read handwritten numbers — and see exactly why it works.
Teach a system to tell the difference between reviews that are positive and negative.
Investigate where these systems fail — and who is responsible when they do.