Artificial Intelligence: A Modern Approach, by Stuart Russell and Peter Norvig, has appeared in four English editions: 1995, 2003, and 2009 from Prentice Hall, and 2020 from Pearson. The authors' editions page lists 22 translated editions in 14 other languages, and the companion site also hosts a collection of published reviews that is almost entirely favorable. That makes a fair review of the book mostly a matter of reading the review record carefully, not of adding another verdict.
Key takeaways
- The English editions are dated 1995, 2003, 2009, and 2020, and the 2003 edition drew the most translations (11).
- The newest edition has one listed translation, into French in 2021, so readers of other languages often rely on the 2009 or 2003 edition.
- The authors' reviews page is hosted by the authors and is nearly all praise; its few criticisms concern thin coverage, errors in the first edition, and pseudocode clarity.
- A textbook dated 2020 predates the large language model era, so it explains foundations, not current products.
- The exercises, code, and pseudocode are free on the companion site; the book itself is a commercial publication.
Four editions, listed by the authors
The editions and translations page on the University of California, Berkeley companion site gives the following English record. Years and publishers are as the authors list them.
| Edition | Year | Publisher | Listed translations |
|---|---|---|---|
| First | 1995 | Prentice Hall | 3 (Hungarian, Italian, Spanish) |
| Second | 2003 | Prentice Hall | 11, including Chinese, French, German, Japanese, and Russian |
| Third | 2009 | Prentice Hall | 7, including Chinese, German, Korean, and Kazakh |
| Fourth | 2020 | Pearson | 1 (French, 2021) |
The page also lists Global Editions of the third and fourth editions from Pearson. A separate companion page for the fourth edition describes 28 chapters in seven parts and two appendices, and it carries no notice of a fifth edition as of October 2026. Its last-modified date is August 22, 2022. Another article on this site covers what the fourth edition contains, so this one concentrates on the record around it.
Who wrote the reviews, and who hosts them
The Berkeley site keeps a reviews page that sorts comments into instructors, students, published reviewers, authors of other AI texts, and other professionals, with chapter-by-chapter remarks at the end. The page was last modified on April 22, 2020, and most entries are undated. Published reviewers are named with a venue: the Usenet AI newsgroup in November 1994, AI Expert in January 1995, AI Magazine in Fall 1995, The Knowledge Engineering Review in 1996, Computing Reviews in May 1997, and the Palo Alto Networks Cybersecurity Canon in 2017.
Because the authors host and select the page, it works as a record of reception, not as an independent evaluation. A reader who wants a neutral comparison with other textbooks has to look elsewhere, and this article does not rank the book against any.
What the critical remarks say
Nearly everything on the page is praise, which makes the exceptions informative. Gerd Brewka, writing in The Knowledge Engineering Review in 1996, called the book likely the best general AI textbook of its time while listing a few errors, including an incorrect statement about first-order logic, which he attributed to the book being a first edition. His main complaint was topic coverage: he quoted the authors' dismissal of nonmonotonic reasoning as "theoretically interesting but practically irrelevant" and argued it deserved more than a page. Philip Chapnick, in AI Expert in 1995, found natural language processing and computer vision lighter than the treatment of logic. An anonymous student remark on the page says the pseudocode is sometimes too vague to implement, though the online Lisp code often fills the gap.
Two other entries show who the book was written for. Martha Pollack, reviewing the first edition for AI Magazine in 1995, recommended it for teaching, including courses organized around agents, and said even a reader who skips the agent-oriented material gets a strong education in the field. Adrian Culley, in 2017, rated it highly recommended for readers at the master's and PhD level and praised its online Common Lisp and Python implementations.
Where a 2020 textbook stops
Reviews from the 1990s judged a book about symbolic reasoning, search, and early learning methods. A reader in 2026 meets a different question: the fourth edition is dated 2020, and it cannot describe systems released afterward. The practical reading is to treat the book as a map of foundations (agents, search, logic, probability, learning) and to consult current papers and documentation for chat assistants and image generators.
The companion site helps close part of the gap. Exercises live on a separate website, the code is published in a GitHub organization, and pseudocode and figures are downloadable as PDFs. The site also maintains an errata page, which is worth checking before relying on a specific passage.
FAQ
Is there a fifth edition of the book?
The companion site showed no notice of one when checked in October 2026. The newest edition on the authors' list is the 2020 Pearson edition.
How many languages has it been translated into?
The authors' editions page lists translations into 14 languages other than English, spread over 22 translated editions. Most of them are of the second and third editions.
Are the exercises and code free?
The companion site hosts the exercises, code, figures, and pseudocode without charge. The textbook is sold by Pearson, and the site does not state a price.
Is the book written for beginners?
It is written as a university text. Published comments differ on level, from recommendations for teaching to a 2017 note aimed at master's and PhD readers, so a prospective reader should look at the contents and a sample chapter first.
Related reading
- Melanie Mitchell's guide for general readers
- the AI For Dummies series by edition
- how to weigh preprints, peer review, and press coverage
More in AI Basics.
Last verified: October 2026
Sources
- All Editions and Translations of AI: A Modern Approach, companion website, University of California, Berkeley
- Reviews and comments on AIMA, companion website, University of California, Berkeley, modified April 22, 2020
- Artificial Intelligence: A Modern Approach, US edition, companion website, University of California, Berkeley, modified August 22, 2022