1 Artificial Intelligence and Agents

1.7 Overview of the Book

The rest of the book explores the design space defined by the dimensions of complexity. It considers each dimension separately, where this can be done sensibly.

Chapter 2 analyzes what is inside the black box of Figure 1.3 and discusses the modular and hierarchical decomposition of intelligent agents.

Chapter 3 considers the simplest case of determining what to do in the case of a single agent that reasons with explicit states, no uncertainty, and has goals to be achieved, but with an indefinite horizon. In this case, the task of solving the goal can be abstracted to searching for a path in a graph. It is shown how extra knowledge of the domain can help the search.

Chapters 4 and 5 show how to exploit features. In particular, Chapter 4 considers how to find possible states given constraints on the assignments of values to features represented as variables. Chapter 5 presents reasoning with propositions in various forms.

Chapter 6 considers the task of planning, in particular representing and reasoning with feature-based representations of states and actions.

Chapter 7 shows how an agent can learn from past experiences and data. It covers the most common case of learning, namely supervised learning with features, where a set of observed target features are being learned.

Chapter 8 shows how to reason with uncertainty, in particular with probability and graphical models of independence.

Chapter 9 considers the task of planning with uncertainty, and Chapter 11 expands the case to multiple agents.

Chapter 10 introduces learning with uncertainty, and Chapter 12 deals with reinforcement learning.

Chapter 13 shows how to reason in terms of individuals and relations. Chapter 14 discusses how to enable semantic interoperability using what are called ontologies, and how to build knowledge-based systems. Chapter 15 shows how reasoning about individuals and relations can be combined with planning, learning, and probabilistic reasoning.

Chapter 16 reviews the design space of AI and shows how the material presented can fit into that design space. It also presents some ethical considerations involved in building intelligent agents.