The Road to General Intelligence

Abstract

This book extensively discusses additional capabilities required for general intelligence—for example, abduction, analogy, and hypothesis generation. Capturing such abilities in AI systems in a general and humanlike way has been the subject of much research but little progress, partly due to the lack of progress in capturing the causal knowledge and reasoning underlying them. Here the authors describe how such abilities have been implemented in a reference system that exhibits Semantically Closed Learning (inspired by the concept of semantic closure in open-ended evolution, and incorporating further ideas from category theory). In short, this book provides an intriguing and provocative framework for thinking about what general intelligence is, and how its essential abilities might be attained able by machines in an economically viable manner. The philosophy behind both programming-language theory and category theory plays key roles in the formalization and development of the main ideas. The authors also provide pointers to what research challenges lie open. Given the complexity and intricacy of the destination, the road to general intelligence will be a bumpy one. This book gives a thought-provoking view of one pragmatic direction toward this goal