The winters did not end AI. They made the limits of its existing methods harder to ignore.

Search and written rulebooks still worked inside certain kinds of problems. But neither had provided a reliable path to general intelligence.

Learning from examples offered another direction. Instead of asking people to describe every decision in advance, a machine could adjust itself by working through examples.

The next module begins with the simplest version of that idea: one artificial neuron and one adjustable connection at a time.