Artificial intelligence

This textbook teaches artificial intelligence through one organizing idea: an intelligent agent is one that acts rationally, picking the action expected to best serve its goals given what it knows. Stuart Russell and Peter Norvig use that idea to tie the whole field together.

The topics begin with search and planning, then turn to logic and how knowledge is represented. Probabilistic reasoning and machine learning follow, and the book closes out with language, robotics and vision. Each topic is built on the ones that come before it, so the book reads best in sequence, though it also works for looking up a single subject.

It first appeared in 1995 and is used in more than 1,500 universities across 135 countries, which makes it the most widely used AI textbook. The fourth edition, from 2021, expands its treatment of deep learning, reinforcement learning, fairness and AI safety, reflecting how far the field has moved since the first edition.

The approach is that of a university course text: systematic, cumulative and aimed at students meeting the field for the first time. It also serves working practitioners who want a reference covering the breadth of AI. Many AI researchers cite it as the place where they first learned the subject.