An Introduction to Universal Artificial Intelligence

by Marcus Hutter, Elliot Catt, David Quarel

An Introduction to Universal Artificial Intelligence provides the formal underpinning of what it means for an agent to act intelligently in an unknown environment. First presented in Universal Algorithmic Intelligence (Hutter, 2000), UAI offers a framework in which virtually all AI problems can be formulated, and a theory of how to solve them. UAI unifies ideas from sequential decision theory, Bayesian inference, and algorithmic information theory to construct AIXI, an optimal reinforcement learning agent that learns to act optimally in unknown environments. AIXI is the theoretical gold standard for intelligent behavior.

The book covers both the theoretical and practical aspects of UAI. Bayesian updating can be done efficiently with context tree weighting, and planning can be approximated by sampling with Monte Carlo tree search. It provides algorithms for the reader to implement, and experimental results to compare against. These algorithms are used to approximate AIXI. The book ends with a philosophical discussion of Artificial General Can super-intelligent agents even be constructed? Is it inevitable that they will be constructed, and what are the potential consequences?

This text is suitable for late undergraduate students. It provides an extensive chapter to fill in the required mathematics, probability, information, and computability theory background. You can also visit the author

Genres: Computer Science, Artificial Intelligence

516 pages, Paperback

Published February 29, 2024

Reviews

[Tim · Rating 5 out of 5]

This is a greatly expanded and updated successor to Hutter's original Universal AI textbook (published in 2005). Anyone interested in a formalization of AI (and AGI, and ASI) should read this book.

The book clearly lays out the core ideas underlying Universal AI, and supports them with rigorous formalism and proofs. Compared to the 2005 textbook, important scientific advances and proofs developed in the last 2 decades were added. Similarly, multiple chapters were added that discuss computable approximations to AIXI (including the exemplary MC-CTW AIXI). The background section has been expanded to make the book more self-contained, and difficulty-graded exercises have been added to every chapter. Finally, the book has a chapter on Feature Reinforcement Learning, and a discussion of AI Safety and the Philosophy of AI.

The title 'an introduction to' is an understatement, perhaps following Li & Vitanyi whose famous standard reference bears a similar title. Besides the central theory, sophisticated approximations to AIXI and some state-of-the-art theoretical results and open questions are discussed. While the book is self contained, and introduces computation theory, coding, Kolmogorov complexity, and Algorithmic probability, it is aimed at advanced undergrad or PhD level and is a great resource to teach a comprehensive course on Universal AI. I am convinced that this will be the standard reference for the foreseeable future, and that the book will contribute to further popularising the core theory behind universal sequential prediction and decision-making and how to combine them into Universal AI. I cannot think of any other source where the AIXI framework is laid out so comprehensively and accessible.

[Roman · Rating 5 out of 5]

It is very rare when deep philosophical insights are supported by strong formal mathematics. If you want to understand probability, information theory, computation theory, Bayesian reasoning, and how all of them interrelate with one another, while also being motivated by the deepest questions of how thinking works in general, this is a book for you.

[Jan Farrell · Rating 5 out of 5]

I love the cover!