This book presents sequential decision theory from a novel algorithmic information theory perspective. While the former is suited for active agents in known environments, the latter is suited for passive prediction in unknown environments.

The book introduces these two well-known but very different ideas and removes the limitations by unifying them to one parameter-free theory of an optimal reinforcement learning agent embedded in an arbitrary unknown environment. Most if not all AI problems can easily be formulated within this theory, which reduces the conceptual problems to pure computational ones. Considered problem classes include sequence prediction, strategic games, function minimization, reinforcement and supervised learning. The discussion includes formal definitions of intelligence order relations, the horizon problem and relations to other approaches to AI. One intention of this book is to excite a broader AI audience about abstract algorithmic information theory concepts, and conversely to inform theorists about exciting applications to AI.

Auteur

Marcus Hutter received his masters in computer sciences in 1992 at the Technical University in Munich, Germany. After his PhD in theoretical particle physics he developed algorithms in a medical software company for 5 years. For four years he has been working as a researcher at the AI institute IDSIA in Lugano, Switzerland. His current interests are centered around reinforcement learning, algorithmic information theory and statistics, universal induction schemes, adaptive control theory, and related areas.

IDSIA (Istituto Dalle Molle di Studi sull'Intelligenza Artificiale) is a non-profit oriented research institute for artificial intelligence, affiliated with both the University of Lugano and SUPSI. It focusses on machine learning (artificial neural networks, reinforcement learning), optimal universal artificial intelligence and optimal rational agents, operations research, complexity theory, and robotics. In Business Week's "X-Lab Survey" IDSIA was ranked in fourth place in the category "Computer Science - Biologically Inspired", after much larger institutions. IDSIA also ranked in the top 10 of the broader category "Artificial Intelligence."

Texte du rabat

Personal motivation. The dream of creating artificial devices that reach or outperform human inteUigence is an old one. It is also one of the dreams of my youth, which have never left me. What makes this challenge so interesting? A solution would have enormous implications on our society, and there are reasons to believe that the AI problem can be solved in my expected lifetime. So, it's worth sticking to it for a lifetime, even if it takes 30 years or so to reap the benefits. The AI problem. The science of artificial intelligence (AI) may be defined as the construction of intelligent systems and their analysis. A natural definition of a system is anything that has an input and an output stream. Intelligence is more complicated. It can have many faces like creativity, solving prob lems, pattern recognition, classification, learning, induction, deduction, build ing analogies, optimization, surviving in an environment, language processing, and knowledge. A formal definition incorporating every aspect of intelligence, however, seems difficult. Most, if not all known facets of intelligence can be formulated as goal driven or, more precisely, as maximizing some utility func tion. It is, therefore, sufficient to study goal-driven AI; e. g. the (biological) goal of animals and humans is to survive and spread. The goal of AI systems should be to be useful to humans.

Contenu

Short Tour Through the Book.- Simplicity & Uncertainty.- Universal Sequence Prediction.- Agents in Known Probabilistics Environments.- The Universal Algorithmic Agent AIXI.- Important Environmental Classes.- Computational Aspects.- Discussion.