Arguing A.I.: The Battle for Twenty-first-Century Science Paperback – January 22, 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....

At 80 small format pages, this is just an evening's read. Although its ambitions are modest, it succeeds admirably as an entertainly written Cliff Notes to the great AI debates of our time. Williams is well informed and clear-headed in this wooliest of areas, and the result is a perfect introduction to the issues and people who dominate the debate.

Innovation in the field of artificial intelligence has had its ups and downs over the last half century. While many scientists in the fifties thought we would most definitely have solid AI applications in common use by the end of the twentieth century, reality has proved otherwise limiting AI success to only modest applications. It's not that there haven't been innovations in the field, but that the goal of computer replicated human intelligence is just much more complicated than originally thought.Well the end of the twentieth century has brought with it a new round of optimistic predictions about AI use in the next fifty years. "Arguing AI", written by Sam Williams, attempts to take an objective look at these predictions and the scientists or futurists that have made them. While short in length, the book presents a good overview of the AI `prediction' debate.Williams presents the material by looking at the visionaries involved in the debate.The first two chapters review the history of AI research by looking at the work of David Hilbert, Alan Turing and John McCarthy. Hilbert is credited with laying the groundwork for twentieth century mathematics and Turing is probably most known for the "as yet not testable" test of artificial intelligence, the Turing test. John McCarthy pioneered the term "artificial intelligence" in the 1950's.The following chapters discuss the current AI debate with optimist Ray Kurzweil, humanist Jaron Lanier and pessimist Bill Joy. While much of the information in these chapters highlights work presented in books and articles written by these men, most notably "The Age of Spiritual Machines" by Kurzweil, and the WIRED article "Why the Future Doesn't Need Us" by Joy, Williams manages to frame the arguments fairly well, thus giving readers a good overview of the current debate.One of the best features of the book comes at the end where readers are presented with a website resources directory that lists dozens of related online articles, websites and other materials that are mentioned in the book. It's a great resource for a more in depth look at the AI debate, as well as more detailed information about the current state of artificial intelligence research.For those looking for detailed information about artificial intelligence, this book may not be appropriate. But for readers interested in an overview of the challenges, possibilities and potential nightmares that artificial intelligence may have in store over the next fifty years, it is a worthy read.

In this book's favor, it does not purport to dig deep. It is a light read, easy in a sitting between meals, perhaps a step up from a full-length Wired piece.If you are new to AI and looking to get your bearings, this book can help. If you've read other popular accounts in the field, this book will seem breezy in comparison. For the moment, it is more up-do-date than most extant popular works, but that advantage is ever only temporarily conferred.If you are a student of AI, this book won't bring you much, including perspective. The `arguments' treated, while certainly part of the history of the field and part of the backdrop, just aren't front-and-center for most ongoing research. This deficit is likely as much an artifact of the author's approach to the subject through high profile personalities as any attempt to simplify which goes too far.In all, I don't think this book does much harm, but unless you're really new to the field, it doesn't do much good beyond entertainment. If that's all you're after, then this book has my hearty recommendation.