Articles

Human-Level AI: The Logical Road

An incomplete draft of a book incorporating concepts of logical AI, philosophy, common sense reasoning, nonmonotonic reasoning, situation calculus, elaboration tolerance, approximate objects, consciousness and problem solving.

This book proposes a path to human-level logical AI. To realize this path, or some other, conceptual advances are required. Ultimate success is likely, because understanding human level intelligence is a scientific problem like many others that scientific research has solved and is solving. How long it will take cannot presently be predicted. I have offered the range of estimates - five years to five hundred years. New scientific ideas are probably required - not just scaling up present approaches.

The present state of AI research permits many applications, but some of the applications that have attempted require aspects of intelligence that are not understood well enough to be programmed.

Contrast logical AI with approaches based on simulated evolution. The evolutionists propose to make a system that can evolve into one that behaves intelligently. The designers of the system need not even be able to understand how it works - any more than we have full understanding of how human intelligence works. This approach should also succeed eventually. Logical AI is more ambitious. We propose to understand the facts of the common sense world well enough to make a system that can act intelligently and can learn explicitly from its experience. The designers of logical AI systems will understand the intelligent behavior our programs exhibit. This approach will also succeed.

It's a race.

Download the article in PDF.