The field of artificial intelligence (AI) emerged in the 1950s with ambitious goals of constructing a thinking machine that would rival the human brain. Very quickly it was clear that the research and the technology of the era were not at a level congruent with the goals of this endeavor. Research and results started appearing in what Kurzweil terms “narrow AI,” where artificial agents exhibit something of an intelligent behavior in a limited subdomain of human expertise. We witnessed several brilliant expert systems that AI advocates bragged about, but they were rooted heavily in Aristotelian mathematics and logic. By the 1980s, many proclaimed AI dead. It was reincarnated in a variety of post-AI disciplines, such as cognitive science, agency, multiagent systems, and developmental robotics, which continued to attack the challenges of mimicking the human brain/mind based on the new learning and technologies of the time.

The original goal of AI, however, has not been forgotten. AI appears to have returned with fields such as artificial general intelligence (AGI) or human-level intelligence. This book is an embodiment of these current efforts, and aims at “outlining a practical approach to engineering software systems with general intelligence at the human level, and ultimately beyond.”

Written primarily by Goertzel, with significant contributions from the other two coauthors, this book is a synthesis of the research of an army of scholars that have collaborated with the primary author over the years, and is the next-to-last volume of a set of six related volumes on the topic. This book outlines the framework on a higher level, and its subsequent volume [1] digs deeper into the details of CogPrime. This book uses this term for the first time, and in congruence with the open software project OpenCogPrime, which many graduate students and researchers in the field have contributed to.

This volume comprises 19 chapters organized into six general parts. Depending on the focus of the discussions, some chapters include high-level outlines and philosophical commentary while others go into the nitty-gritty of particular issues. The book should be read in a sequential order. Miscellaneous appendices are only available online and are not a part of the bound volume.

The introductory chapter reminds us of why AI as we used to know it came about. The first part of the book, “Overview of the CogPrime Architecture,” talks about the architecture, processes, and modules of the framework. The second section, “Artificial and Natural General Intelligence,” discusses human intelligence and existing cognitive architectures across a variety of schools of thought. Part 3, “Toward a General Theory of General Intelligence,” discusses what forms and dimensions these theories may take. Part 4, “Cognitive and Ethical Development,” talks about translating the Piagetian theory on the stages of human development and the development of morals in a human, and subsequently an artificial agent. In the fifth part, “Networks for Explicit and Implicit Knowledge Representation,” structures for representing knowledge and networks are discussed. The concluding part, “A Path to Human-Level AGI,” outlines how the goal of creating an artificial human and superhuman may be achieved.

Some ideas used in the framework are well proven and known, and others are based on a gut feeling, without any proofs. There is no need to judge this approach. The fact is, we still know very little about what intelligence is, how it comes about, and how our precious organ that embodies it functions. It is commendable that there are fellow researchers that are still enthusiastic about the original goal of the AI, and have not been discouraged by the many told and untold stories of failures of the narrow AI period.

One may feel disappointed by how little the thinking and the formalism used in this book resemble those of narrow, traditional AI. It seems that the major lesson learned by some from traditional AI, that formalisms are problematic, has not been experienced by all. Or maybe some are still so hopeful that the traditional formal, mathematical, and logical way of approaching the challenge will work now that we know more and have better technology. Or maybe they are not ready to give in. Or maybe, just maybe, it is hard to veer away from one’s early training.