|
|
8 of 8 people found the following review helpful:
4.0 out of 5 stars
An elementary but interesting introduction to A.I., September 14, 2002
Is research and thinking on artificial intelligence stuck in a local minimum? Those in the field have attested to major advances in the last decade, but are these advances merely a renaming of approaches that were taken decades ago? This book does not address these questions as its major goal, but instead attempts to give a broad overview of how A.I. got started and where it is now, and where it might be going. The reader is lead to ask the questions above though after reading the book, for the author seems to ask them implicitly. Its validity as a science are questioned, and the future of A.I. is addressed in detail. The author though is fair in his representation of both sides of the A.I. debate. After a short introduction and a brief "A.I. debate timeline", the author begins chapter 1 with what could be considered to be the mathematical origins of the subject, due to the mathematicians David Hilbert and Alan Turing. Hilbert was essentially the originator of the formalist school of mathematics and proof theory, but his ideas were countered by Turing, and the mathematicians Kurt Godel and Alzono Church. These counterarguments are taken to be final by the author, and he follows the lead of many others in asserting this. But the "unprovability" results of Godel do not show up in the normal practice of mathematical research though, with the self-referential statements having to be artificially constructed. There are no examples in the everyday practice of mathematical research where these kinds of statements arise when engaging in the activity of making definitions and proving theorems. Empirically and practically speaking therefore, the Godel counter to the Hilbert formalism is weak. As far as any negative ramifications to A.I. are concerned, the author does mention the assertion of Douglas Hofstadter that Godel self-referential statements would be a (positive) sign of machine intelligence and self-awareness. The next chapter discusses the A.I. contributions of the LISP pioneer John McCarthy. A brief biography is given of McCarthy and how he got started in A.I. This chapter gives much insight into the "giddy" optimisim that surrounded the A.I. community in the 1950's, an optimism that has grown beyond all bounds at the present time. McCarthy's time scale for having machine intelligence is on the order of 500 years, but, as the author reports him saying: "The breakthrough could come this or next year." In addition, and I think correctly, McCarthy believes that computational power is not enough for advances in A.I., but some new ideas. When viewing the status of A.I. research, with fairness one could say that it is trapped in a local minimum, and some radically new ideas are needed to force it out of equilibrium. Computational power will certainly help in testing out these new ideas of course. In chapter 3, the author overviews the contributions and attitudes of Ray Kurzweil to A.I. Called the "optimist" by the author, and this is indeed an understatement. Kurzweil predicts the onset of thinking machines way beyond the capabilities of human intelligence by the year 2030. His contributions to A.I. and his technological ingenuity justify though this optimism. His attitude that computational power is the sole issue is not really justified, at least from current levels of knowledge. But increased computer performance may result in more innovative ideas to be developed, resulting in a kind of self-fulfilled prophecy for the rise of intelligent machines by the year predicted. Chapter 4 discusses an idea that you don't hear much about anymore: virtual reality. The author overviews the work of Jaron Lanier, the leading innovator of virtual reality software. Critical of the claims of A.I. researchers, Lanier has much to say about the future of both A.I. and software development. As reported by the author, his motivation for developing virtual reality is very intriguing, as he wanted to build an interactive computer-graphics program that would give mathematicians the power to express their ideas in graphical form. Software development though, according to Lanier, has taken a turn for the worse, with bug-ridden progams the norm rather than the exception, all written, he says, to take advantage of increasing microprocessor speeds. The future in 2030 is one where software maintenance is the predominant activity, according to Lanier. Lanier though omits the fact that software engineering is one of the main applications of A.I. at present, and shows every sign of increasing. Intelligent debugging, intelligent software maintenance, and even intelligent software development are acting as testing grounds and financial justification for A.I. In chapter 5, the (pessimistic) ideas of Bill Joy are discussed by the author. Joy is clearly very concerned that the future may result in a terrifying one for all of humanity, if indeed A.I. is realized to the point of autonomous, thinking machines. He believes that A.I. will reach such a status, but he is not optimistic as to its consequences. It is interesting to compare his ideas on software development with those of Lanier. Joy, as reported by the author, believes that it will not be like anything we currently understand. In addition, strong A.I., or a conscious thinking machine, does not have to be realized in order for it to be dangerous, Joy argues. The last chapter, entitled "Fact Versus Fiction" is an attempt by the author to wrap things up and assess just where we all are in A.I. research. As in most books of this kind, the arch-villan HAL of 2001: A Space Odyssey makes its appearance. HAL has turned into a sort of benchmark for A.I., in both popular and professional circles. And, interestingly, the movie "A.I." is mentioned also, it being held as an example of the current thinking in many A.I. circles that a machine must interact with the environment in order for it to become intelligent. But more A.I. is coming....
|
See all Editorial Reviews
