Paradigms of Artificial Intelligence

About the book

Paradigms of Artificial Intelligence teaches its subject by reconstructing programs that run. Peter Norvig works in Common Lisp, and the method is consistent: present a complete system, then examine the design decisions behind it. Readers see how a search routine, a rule interpreter or a pattern matcher is assembled, and what each choice costs in clarity and performance.

The book draws on established AI traditions, including symbolic reasoning, search, knowledge representation and natural language processing, and treats Common Lisp as the medium for expressing them. Norvig assumes some programming maturity but little or no Lisp experience, so the early material spends time on the language itself. Later sections move between general techniques and specific systems, showing how a small core of ideas recurs across very different tasks.

Engineering judgment runs through the book alongside the algorithms. It asks readers to notice why a program is organized one way and not another, and it treats code as something to read critically. That emphasis gives the book a longer shelf life than a survey of techniques alone.

How to read it

Read with a Lisp implementation open. The text rewards typing in and running the programs, and many points about design become clear only when a system is modified and breaks. Work through the language chapters first if Common Lisp is unfamiliar; skipping them makes the later systems harder to follow than they need to be.

The book is long and dense, and its examples are substantial. Pace yourself by treating each system as a self-contained study instead of reading straight through. Expect code listings that require careful attention, and expect some material to reflect the state of AI practice at the time of writing. The prose is direct, but the technical load is real.

Who it suits

Programmers who want to understand how symbolic AI systems are built will get the most from this book, especially those comfortable with another language and willing to learn Common Lisp. Students of AI who find abstract treatments thin will value the complete, runnable examples. Readers interested in software design and in the history of AI methods will also find useful material, provided they are ready for close reading of code.

Reading considerations

The book dates from 1991, so its examples and framing reflect AI research of that period; it is not a guide to current machine learning practice. Common Lisp is central rather than incidental, and readers unwilling to engage with it will struggle. Test the code against a current Lisp implementation rather than assuming it runs unchanged.