The Quest for Artificial Intelligence [Hardcover]

Nils J. Nilsson's book begins with the story of how artificial intelligence originated in 1956 at a Dartmouth summer project that had the goal of "making a machine behave in ways that would be called intelligent if a human were so behaving." It relates how in the fifty-plus years that followed, AI has been the subject of overly-optimistic predictions, academic arguments that its goals are unachievable, funding excesses, and funding droughts. But the underlying reality is that AI has contributed key components to the technology foundations that shaped the modern world, and indeed has transformed our view of machines and of our relation to them.

The algorithms that compute your driving directions, and also compute the paths of characters in video games? They rely on results from AI research on mobile, intelligent robots. Those surprisingly high-quality voice response systems we encounter when we phone a customer-service number? They use results from AI research in speech recognition. The recommender systems ("You might also like") used by many web vendors? They use machine learning methods whose history is described by Nilsson. And AI technology is embedded in a host of less-apparent applications ranging from medical devices to automated securities trading systems.

Nils J. Nilsson's comprehensive account of the evolution of AI covers the field from its inception to recent times. All the major sub-fields of AI receive attention--from game playing to automatic problem solving, from computer vision to speech and language understanding, from expert systems to machine learning and probabilistic reasoning--all these and more are covered.

Nilsson enriches his account by viewing major developments through a multi-faceted prism. He describes AI's challenges, the approaches adopted and the landmark systems in just enough detail to give the reader real insight into the technical substance of the field. He also describes the funding issues and controversies that have swirled around AI since that very first Dartmouth meeting. And he introduces the reader to scores of brilliant, frequently colorful, characters whose contributions and opinions have influenced the course of developments.

For the AI practitioner, this book is a rare example of that often proclaimed, but seldom sighted species, the "essential volume" for your library. Your perspective on AI cannot help but be enhanced; you'll gain an increased appreciation for the time it takes for a good idea to mature and find a place in the world; and you may even be encouraged to revisit nearly-forgotten ideas that have relevance to current research issues.

But the book has appeal for the general reader as well. Nilsson is a masterful teacher and storyteller, and his description of timeless philosophical issues and intellectual challenges are as clear as you will find in as confined a space. Technical approaches are profusely illustrated and diagrammed, but remain accessible to any reader with an active curiosity. The tone of the book is straightforward and conversational, with neither the stuffiness of a self-important academic nor the breeziness of a science popularizer.

Predictions about AI have proven hazardous for 50 years, but I'll make one here: It will be a long time before any writer attempts a sequel to this unique and valuable volume.

This is an extremely literate, well written history of the first fifty years of AI by someone who fortuitously came in on the ground floor of this field. Nilsson's perspective is unique and invaluable for anyone interested in broadening their horizons, and in appreciating how many talented and driven individuals have contributed to AI's successes.

As a lay reader, I skipped the notes and many of the technical details and diagrams. I enjoyed the many interesting references scattered through the text. Just to give a flavor of these, in the first chapter alone there are references to Homer's "Iliad", Ovid's "Metamorphosis", The Talmud, opera (Offenbach's "Tales of Hoffman"), and theater, Capek's "R.U.R." I won't mention more of them here but leave them for you to discover, choice morsels all. Although this is a scholarly work, it's accessible to anyone who is interested in what AI is all about.
AI has already become an integral part of our lives. It's used for computing driving directions, interactive computer games, aircraft control, credit card fraud detection, vending machine currency recognition, robot control, speech recognition, and face identification, to name just some of the more prominent examples.

I came away marveling at how far this field has come in 50 years and convinced of the need for more basic research. Most of the important inventions were due to basic research. At the time, the results, to an untrained eye, looked stunningly simple. People thought, "What good is that?" We're now reaping the harvest of those years of early work, and one hopes that, along with applications, basic research in the field will continue.

This book is a significant contribution to the history of science.

Nils Nilsson is indeed one of the pioneers of AI (artificial intelligence), in addition to being the author of several good earlier texts in the field. As always true for this author, the book is carefully written.

The content of this 562-page book is probably the most careful and up-to-date documentation of who did what and when in AI, spiced with the unique personal recollections, and in that sense it is quite useful as a historical record.

However, the serious flaw of the book titled "The Quest for AI"--as with the whole AI enterprise--is the lack of a *critical* evaluation of the 'progress' made so far, which would have been so valuable for the development of this scientifically very unusual, pioneering, endeavor. Worse yet, completely missing from the book is an adequate appreciation of the strategic scientific role of AI among sciences. Indeed, since the fathers of the Scientific Revolution (mainly of the 17th-18th centuries) have intentionally excluded the mind from the modern scientific view, the successful development of AI will,of necessity, have enormous implications for the restructuring of *all natural sciences*. (Of course, such discussion would have vitiated any claims about the present 'successes' of AI.)

This poor 'tradition' of not adequately evaluating the 'progress' in AI goes back half a century and can be explained by the fear of loosing funding. So the main problem with the book is that the author does not get to a serious critical evaluation of the progress in the field in light of the monumental quest to bring the mind into the mainstream scientific view. I find it impossible to believe that the present incremental and haphazard 'progress' in AI will get us to the "promised land". It is enough to have sufficient respect for this unprecedented scientific undertaking in order not to expect that such bridge into the unknown territory can be built incrementally, out of the old pieces. Not only the original leap *must* be radically new, but all the pieces will also have to be new (see also the last two pages in von Neumann's book "The Computer and the Brain").

We should admit that the actual 'progress' described in the book is mainly due not to any breakthroughs in our understanding of intelligent information processing but to the ubiquitous trends in the miniaturization of hardware and the improvements in software.

To some extent such orientation of the book is not surprising since the author admits (on p. 515) that "predicting . . . where Ai's present momentum will take us is problematic." In other words, Who knows where the next fad will take us?

In addition to 3-4 pages in the last chapter devoted to "Controversies" and "How Do We Get It?", the book does include a small part VI--titled "Entr'acte", from French Entracte, meaning "between the acts"--which has two chapters "Speed Bumps" and "Controversies and Alternative Paradigms". But those 'innocent' controversies have not altered the triumphant march of AI, in which "its basic research workers produced a significant number of powerful new technical tools and sharpened others." (p.347)

Just imagine: you got decent education, you are a bright person, and you want to do AI. Given this and also the desire to survive in an academic environment, you, together with your peers, will have some ideas and will implement them as programs. The critical question is not whether your program can do something that only *appears* to be clever (you are bright enough to accomplish this), but whether your work is AI. How do you distinguish between some clever piece of software and AI? To be useful this software does not have to be AI at all.

So if you are one of those serious readers who are trying to get an idea of what the intelligent information processing is about you will be quite disappointed. This history extensively (but misleadingly) portrays AI as making steady progress towards the goal. Yet to some of us--including one of the founding fathers of AI, Marvin Minsky--it is quite clear that, unfortunately, despite more than adequate human and financial investments over the last half a century, we have barely moved towards the original AI's goal, despite the introduction of zillions of new terms to justify the funding. (Some of us do try completely new approaches.)