In the field of AI, there is a range of often autonomous subfields. This occurred as a reaction to the extreme difficulty revealed by all approaches to the more global goal of complete AI. It is a classic divide-and-conquer approach to excessive complexity, but can it work for AI?

On the negative side, note that investigations within most (if not all) subfields seem to reveal that knowledge of the world and sophisticated contextual awareness are essential for high-quality AI. In other words, real progress in each subfield seems to require major progress in every other as a corequisite, if not a prerequisite.

On the positive side, there are some clear indications that intelligence is not all-or-nothing. There are many examples of intelligent humans who lack sight, speech, and hearing. In addition, there is a case to be made for the modularity of mind both from localization of functional components of intelligence in specific anatomical regions and from observed phenomena such as optical illusions. Many optical illusions remain compelling illusions even when the basis for the illusions is physically demonstrated or explained. This persistence might be interpreted as an absence of close integration between visual processing and reasoning abilities.

One pervasive conundrum is that a broad and deep knowledge of the world appears repeatedly in the subfields of AI as a necessity for intelligent activity, with an associated implication that more knowledge supports higher levels of intelligence. However, from a computational perspective more information (in a larger knowledge base) must lead to more searching for relevant pieces, which takes more time. Stated succinctly, the more you know, the slower you go. This is true for AI systems but not, it seems, for humans. In fact, the growth of a human's expertise is often associated with a faster performance rather than a slowdown. Why? For some, the answer is that classical programming on conventional computers is the wrong basis for intelligence. Therefore, other computational paradigms are being explored. For others, the answer is less radical but no less obscure: Information cannot just be added on, it must be properly integrated and indexed in sophisticated ways.

A final general point of difficulty is that intelligence in humans is characterized by tentative conclusions and sophisticated recovery from errors. We do not always get things right, but we can usually readjust everything appropriately once we accept the error. In AI systems, the compounding of tentative conclusions is difficult to manage and recovery from error is often no easier because we typically have to foresee each source of error and supply a recovery procedure that will always work.

One general conclusion to be drawn from this survey is that AI has not been as successful as is popularly believed. One rejoinder to this common charge is that as soon as an AI problem has been solved by the construction of a programmed system, that problem ceases to be called AI. There is some truth in this, but not much. The real AI problems of adaptivity, natural language understanding, image understanding, and robotics are still out there and waiting to be solved.

Work in AI does not undermine a belief in humanity by aspiring to mechanize it. Quite the contrary—the more one examines the problems of intelligence, the more one's admiration grows for a system that seems to solve all of them at once, often effortlessly.

See Also the Following Articles


All About Alzheimers

All About Alzheimers

The comprehensive new ebook All About Alzheimers puts everything into perspective. Youll gain insight and awareness into the disease. Learn how to maintain the patients emotional health. Discover tactics you can use to deal with constant life changes. Find out how counselors can help, and when they should intervene. Learn safety precautions that can protect you, your family and your loved one. All About Alzheimers will truly empower you.

Get My Free Ebook

Post a comment