A General Theory of Intelligence

Preface

As the title suggests, this book attempts to propose a theory on intelligence, which is general, in the sense that it identifies the common principles, structures, and mechanisms behind various types of intelligence, including human intelligencecomputer intelligence, animal intelligence, alien intelligencegroup intelligence, etc.

It is my hope that this book can serve two major purposes: first, as a descriptive theory, it will provide a uniform explanation of human intelligence and the related phenomena studied in cognitive science (CogSci); and second, as a constructive theory, it will suggest an approach to create artificial intelligence (AI).

From the very beginning of my study of CogSci and AI, I believe the research result in this field should be presented on three levels:

  1. First, there should be a general theory of intelligence. It should mainly come from the study of human intelligence, the type of intelligence that is best-known to us, but also include results from other related fields. The research result on this level should be expressed in a natural language, like English or Chinese.
  2. Then, guided by the theory, a formal model of intelligence is designed to further develop the theory in a precise manner. By "formal model", I mean that the details of the model are described symbolically or mathematically, so that ambiguities in the theory are removed as much as possible. The research result on this level should be expressed in a natural language, with formal expressions embedded.
  3. Finally, the model is implemented in an intelligent computer system.  All the mechanisms described in the model should be implanted into software or hardware, which work in a currently available software and hardware environment. The research result on this level is the computer system itself, with documents in various types of language to explain its design and behaviors.
According to this belief, I set my career goal to be two books, one conceptual and the other technical, and a software.

This eBook is going to evolve into the conceptual book, which is written in non-technical language, and establishes the philosophical foundation of my approach toward CogSci and AI. Also, it puts the research in a broad context. Previous abstractions and versions of this work have been distributed and discussed in several small circles of friends in 1985, 1991, and 2000, respectively.

The technical book has also grown through several stages, as my Master Thesis (Peking University, 1986), Ph.D. Dissertation (Indiana University, 1995), and a research monograph (Springer, 2006), with the supplements of my other publications.

The software is NARS, which has been developed in several versions, too. Recently it grew into Open NARS, an open-source software.

Currently, works at all three levels are under constant revision and extension in a time-sharing manner, whenever I find time to work on them.

When turning my existing (plain text) manuscript into hypertext for this eBook, I am trying a novel format. The eBook will consists of a main text and many sidebar texts. In the main text, I focus on expressing my ideas on the related topics in a self-contained way, and link to sidebars whenever I need to talk about a topic of importance in the field, but plays a minor role in my theory. Also, the main text is organized in webpages at three levels: book, chapter, and section, with each topics discussed in a lower level being summarized at a higher level. Therefore, readers will have the freedom of either browsing a topic quickly, or exploring the details of my treatment.

This eBook will be under construction for a while. To improve its quality, I will appreciate all kinds of feedback, as well as English corrections. Comments can either be emailed to me, or posted to the NARS discussion group.