Artificial Intelligence: Foundations of Computational Agents / Edition 2

by David L. Poole, Alan K. Mackworth

Overview

Fully revised and updated, this third edition includes three new chapters on neural networks and deep learning including generative AI, causality, and the social, ethical and regulatory impacts of artificial intelligence. All parts have been updated with the methods that have been proven to work. The book's novel agent design space provides a coherent framework for learning, reasoning and decision making. Numerous realistic applications and examples facilitate student understanding. Every concept or algorithm is presented in pseudocode and open source AIPython code, enabling students to experiment with and build on the implementations. Five larger case studies are developed throughout the book and connect the design approaches to the applications. Each chapter now has a social impact section, enabling students to understand the impact of the various techniques as they learn them. An invaluable teaching package for undergraduate and graduate AI courses, this comprehensive textbook is accompanied by lecture slides, solutions, and code.

Product Details

ISBN-13:	9781107195394
Publisher:	Cambridge University Press
Publication date:	09/25/2017
Edition description:	New Edition
Pages:	820
Product dimensions:	7.17(w) x 10.28(h) x 1.54(d)

About the Author

David L. Poole is Professor of Computer Science at the University of British Columbia. He is a former chair of the Association for Uncertainty in Artificial Intelligence, the winner of the Canadian AI Association (CAIAC) Lifetime Achievement Award, and is a Fellow of AAAI and CAIAC.

Alan K. Mackworth is a Professor Emeritus of Computer Science at the University of British Columbia, where he co-founded the pioneering UBC Cognitive Systems Program. He served as President of CAIAC, IJCAII, and AAAI, and now acts as a consultant, writer and lecturer. He is a Fellow of AAAI, CAIAC, CIFAR, AGE-WELL and the Royal Society of Canada.

Table of Contents

Preface; Part I. Agents in the World: 1. Artificial intelligence and agents; 2. Agent architectures and hierarchical control; Part II. Reasoning and Planning with Certainty: 3. Searching for solutions; 4. Reasoning with constraints; 5. Propositions and inference; 6. Deterministic planning; Part III. Learning and Reasoning with Uncertainty: 7. Supervised machine learning; 8. Neural networks and deep learning; 9. Reasoning with uncertainty; 10. Learning with uncertainty; 11. Causality; Part IV. Planning and Acting with Uncertainty; 12. Planning with uncertainty; 13. Reinforcement learning; 14. Multiagent systems; Part V. Representing Individuals and Relations: 15. Individuals and relations; 16. Knowledge graphs and ontologies; 17. Relational learning and probabilistic reasoning; Part VI. The Big Picture: 18. The social impact of artificial intelligence; 19. Retrospect and prospect; Appendices; References; Index of Algorithms; Index.

What People are Saying About This

From the Publisher

“This text is a modern and coherent introduction to the field of Artificial Intelligence that uses rational computational agents and logic as unifying threads in this vast field. Many fully worked out examples, a good collection of paper-and-pencil exercises at various levels of difficulty, programming assignments based on the custom-designed declarative AILog language, and well-integrated online support through the AISpace applets complement the presentation. If you plan to teach a course in Artificial Intelligence at the upper-division undergraduate level or beyond, you must give serious consideration to this thoroughly enjoyable book.”
Marco Valtorta, University of South Carolina

“This book fills a real gap in the AI literature. It is accessible for advanced undergraduate students, without compromising technical rigor. It is concise, but still gives a modern presentation of all major areas of AI. It is an eminently useful textbook for introductory courses to AI. Poole and Mackworth have made a valiant effort to impose some order on the wide and heterogeneous field of Artificial Intelligence. In this order, all of AI is placed in a design space for intelligent agents defined by dimensions of complexity.”
Manfred Jaeger, Aalborg University

“The clarity of this book is amazing! Material in each chapter is a perfect blend of accessible stuff for beginners, theory and challenges for advanced students, and reference materials for experts, organized into sections so you can split off the right bits for your students. Its like having three textbooks in one! Definitely the must-have textbook on AI for the 21st century. I know mine will be within reach for years to come.”
Jesse Hoey, University of Dundee

"This book, by two of the foremost researchers in Artificial Intelligence, marks the transition of the field from a miscellaneous assortment of unrelated techniques to a genuine scientific discipline. It presents the fundamental concepts of AI in a coherent structure, which shows how different techniques are related and complementary. The book is written in a clear and engaging manner, which makes it suitable both for the serious student and for the intellectually curious layperson."
Robert Kowalski, Imperial College London

"This book is a wonderful, well-written introduction to a field that is interesting to many, fascinating to some; a field that involves tremendous complexity. The authors manage the complexity by beginning with the simplest elements and building on these to progressively broaden and deepen the treatment. They provide a large number of references for those who wish to go beyond the text. [I] recommend this excellent work."
G. R. Mayforth, Computing Reviews

"It has been about 15 years since the last major history, and a lot has happened. And Nilsson has a unique viewpoint: he has been a key early AI researcher, an influential lab leader, a AAAI president, author of three very different AI textbooks, and a teacher and department chair at a leading AI department. Furthermore, he has the disposition of a careful scholar and is not inclined to push just one viewpoint. From the beginning, his work has spanned the logical and probabilistic approaches to AI—he could give a more balanced overview than someone who has worked in only one of these camps. I wanted to hear his take on the history of AI. The Quest for Artificial Intelligence is more personal yet more comprehensive, and presents a more nuanced appreciation of the place in history of each event. Make no mistake: this book is a history—a true quest."
P. Norvig, Artificial Intelligence (2011), doi: 10.1016/j.artint.2010.11.024

"There are several AI textbooks on the market at the moment with the same target audience but one of these books—by Russell and Norvig—is dominant, almost completely so, and is used in approximately 1100 universities in 100 countries. Over the years, I have taught from earlier editions of the texts by Luger and by Rich, Knight, and Nair. Like many, I switched to the text by Russell and Norvig (hereafter R&N) shortly after the first edition came out in 1995. R&N is an excellent and highly regarded text. Yet after more than a decade of teaching through three editions of R&N, I recently switched from R&N to the new text by Poole and Mackworth (hereafter P&M). Let me explain why. R&N has aimed at being comprehensive and is not as selective. The result, I believe, is that the book has become overly long and less integrated and less useful to the average student. Some topics and chapters in R&N contain much more material than can reasonably be covered in an introductory course that aims for breadth (as many AI courses are structured). The result is often unsatisfying for the instructor and students find it difficult to wade through the extra material to find what is relevant for their course. P&M, in contrast, is more selective in its coverage. In summary, I highly recommend this book to instructors of introductory AI courses and to those who wish to learn about the foundations of the field through self-study. As many will be aware, there is already an excellent textbook by Russell and Norvig with the same target audience that has dominated the field for more than ten years. However, all things considered—selective coverage, level of detail, quality of explanations, exercises, online materials, free availability, and so on—I believe the clear advantage goes to the newcomer. Certainly the book sets a new standard for AI textbooks with its supplementary online tools and tutorials."
Peter van Beek, University of Waterloo, AI Journal

Editorial Reviews

‘This is an important textbook. Based on their broad experience, the authors harmonize some of the most exciting recent developments in the field, such as generative AI, with more traditional methods, within a unified agent framework. This will broaden the perspective of those relatively new to the field, for whom AI and deep learning appear almost synonymous.’ Yoav Shoham, Stanford University and AI21 Labs

‘This book is a tour de force. It provides a comprehensive introduction to so many topics in modern AI. The clarity of the exposition and the ability to capture the intuition underlying complex concepts make this book compelling and appealing to a broad audience.’ Pascal Van Hentenryck, Georgia Institute of Technology

‘This new edition offers an up-to-date account of AI, presenting the field in an accessible and unified manner. I particularly like the ‘relations-late’ approach, in which first-order logic and relational AI are covered later, after thoroughly covering more basic, feature-based methods. The hybrid data-driven/model-based approach to agent design that the authors propose will be essential to the development of reliable and trustworthy intelligent systems.’ Kevin Patrick Murphy, Google Brain, author of Probabilistic Machine Learning

‘Poole and Mackworth's now classic textbook has guided my senior undergraduate AI class since its first edition. Coupled with online resources, the book presents a comprehensive overview, with technical substance and many pointers for further study, in a coherent structure that fosters learning of key interrelated concepts. The third edition updates the content to cover the massive recent AI advances.’ Jesse Hoey, University of Waterloo

‘Machine learning has undergone spectacular advances over the last few years, but to harvest the new capabilities one needs an engineering framework to build computational agents. This book teaches students about the concepts and techniques that make that possible.’ Rodney Brooks, MIT and Robust AI

‘Wide-ranging, well-organized, up-to-date, and in-depth coverage of the AI world. The numerous figures, algorithms, and extensive references make this a valuable resource that readers will return to repeatedly. Instructors and students will benefit from the well-crafted end-of-chapter exercises. The thought-provoking social impact sections in each chapter and the social impact chapter admirably address the positive and harmful impacts on people. These complement the strong technical descriptions, wisely encouraging researchers and practitioners to limit the risks by highlighting human-centred AI. Poole and Mackworth are highly acclaimed experts who eagerly present their subject with enthusiasm and thoroughness.’ Ben Shneiderman, University of Maryland, author of Human-Centered AI

‘This revised and extended edition of Artificial Intelligence: Foundations of Computational Agents should become the standard text of AI education. Computer science students will find in this volume a broad and uniquely coherent perspective on many computational models of learning, reasoning, and decision-making. Students of causal inference, in particular, will rejoice at viewing the causal revolution reconnected to its roots in formal logic and probabilistic decision-making, strengthened and reinforced by concrete algorithms, challenging exercises, and open source AIPython codes. Highly recommended.’ Judea Pearl, UCLA, Turing Award winner and author of Causality and The Book of Why

‘This textbook is impressively comprehensive, covering all the major AI paradigms that have been introduced and studied over the years. At the same time, it is up to date with the latest technical advances and interdisciplinary perspectives on social impacts. I expect it to be a valuable resource for both teachers and students.’ Peter Stone, University of Texas at Austin

‘Artificial Intelligence: Foundations of Computational Agents is a great AI textbook written by prominent leaders in the field. It covers everything you want to know about AI in a very accessible style, accompanied by a wide range of thoughtful and challenging exercises. I find this book to be an extremely valuable resource, not only for teaching, but even more so for offering an updated reference to a wide spectrum of foundational subjects at the current frontier of AI.’ Rina Dechter, University of California, Irvine, author of Constraint Programming

‘Poole and Mackworth's book has been my go-to resource for students who need an introduction to Artificial Intelligence. While the previous versions have provided a complete overview of the field, the newer version organizes this information in a crystal clear manner. The division of the topics based on what the agent knows, what is in the world, and what the effects of its actions are allow for a logical flow of topics inside AI. As a comprehensive textbook for AI that includes slides, solutions, and code, this book is a must-have on the bookshelf for AI instructors, students, researchers, and practitioners.’ Sriraam Natarajan, University of Texas at Dallas

‘This is a great foundational book on the science of AI, covering the main concepts and techniques using a simple structured approach. The extensive material on the social impact of AI provides much needed attention to the responsible design and use of AI. AI researchers can find here the indispensable foundational knowledge and the needed ethical attitude to create beneficial AI innovation.’ Francesca Rossi, IBM Fellow

‘The latest edition of Poole and Mackworth's book emphasizes the societal impacts of AI in every chapter, making it an essential read for anyone interested in AI, especially those who will shape its future to ensure these powerful technologies benefit society and minimize harms.’ Saleema Amershi, Microsoft Research

‘This textbook provides an amazing introduction to the field of AI. By bringing together learning, reasoning, and decision-making, it shows the rich interconnections across the various AI subfields. The writing is just at the right level to introduce students to the different facets of AI. The updated edition seamlessly integrates the exciting developments in deep learning into the broader AI context. The text also highlights the societal impact of AI, including AI ethics and computational sustainability.’ Carla Gomes, Cornell University

‘Poole and Mackworth - two pioneers of AI - present an admirably broad and complete introduction to the field, with a very useful focus on intelligent agents. From deep learning to causal reasoning, from Bayesian networks to knowledge graphs, from fundamental algorithms to effective heuristics, this book covers a wide range of important topics, each accompanied by a timely section on social impact. Highly recommended!’ Holger Hoos, RWTH Aachen

‘Poole and Mackworth's Artificial Intelligence: Foundations of Computational Agents 3e is a tour de force. This is a comprehensive and clearly written text that takes the reader through core concepts in symbolic AI and machine learning, providing pathways for broad introductory undergraduate courses, or focused graduate courses. It's an outstanding resource for student and instructor alike. Whether you're a seasoned AI researcher or a student entering the field, you'll learn a great deal from reading this book.’ Sheila McIlraith, University of Toronto

From the Publisher