Browse Books

Reviewer: John D. Sankey

This book has strong parallels to another by Servan-Schreiber [1]. Servan-Schreiber argued that France, even Europe, was doomed to lose its future unless its managers abandoned French norms to adopt the methods of the Harvard Business School, and its society abandoned European traditions in favor of an American style of education. Servan-Schreiber was wrong. Despite his own later personal involvement in French government, France refined existing social and governmental structures and, as we all know, has remained very much a power in world affairs. I recently saw several copies of his book, rejected in the 25 cents boxes outside a second-hand bookstore. The Fifth Generation, at heart, has a similar message. The traditions of the United States hide a potentially fatal weakness that will allow Japan to seize the future, unless America adopts Japanese methods, at least with regard to research into computer intelligence, and formally organizes a massive centralized AI program. It is not possible to discuss the book in detail without first dealing with these fundamental assumptions for, in my view, they are wrong on two counts. First, research is the study of what we do not understand, and we certainly do not at present understand how we understand, which is what AI is about. Research is done by thinking in a way that no one has before, a notion inherently antithetical to centralist direction. Second, as an outsider, it seems to me that the strength of the United States has always been in a seething mass of ideas backed up, not led, by the business strength of large corporations and government. Now, when the software business so easily provides financial rewards adequate for small peer-group business, seems the worst of all times to lose faith in the American way with ideas. I suggest that readers skip the polemics, and start with Part 3, on expert systems. As one might expect from these authors, they write with precision and clarity on the history of AI, and the structure and commercial possibilities of expert systems. In Part 4, they sketch the basic details of the plan of the Japanese institute directing the work, and present a very sympathetic view of the difficulties, both cultural and technical, that the Japanese themselves foresee. Their view of British plans, in Part 5, is considerably less sympathetic, and Russian possibilities are dismissed in a page. It is in Part 6, which details all the reasons why the United State is going to fail in AI, where I feel the authors have misjudged most the strength of their own society. It is not on the factory floors that AI will flourish. To a foreigner, to hear an American wailing that “we have no national vision to sustain us” seems inappropriate, when a third of all research is funded by one agency, the Pentagon. To claim, as the authors do, that the US is in a “dismal national tailspin” because children are not studying mathematics and physics any more, is reminiscent of the British belief that no one could govern without the guidance of the Odes of Pindar. Truly useful computer software, including AI, depends on neither. The authors' despair, over the distrust of many Americans of intellect, denies the evident fact that intellect is prized by the tiny peer-group businesses that are turning out some of the best software in the world. I suspect that this book, like Servan-Schreiber's, will end up in the 25 cents bin. The United States seems perfectly capable of solving AI problems in its own way, as are the French, of refining social and governmental structures to defend itself from crisis.

Reviewer: Oscar Firschein

“Generations” of computers have been counted as follows: (1) vacuum tube computers, (2) transistorized computers, (3) integrated circuit computers, and (4) Very Large Scale Integrated (VLSI) circuit computers. Knowledge information processing systems are considered to be the Fifth Generation. Japan's Fifth Generation project began three years ago, to much fanfare, with the setting up of the Institute for New Generation Computing Technology (ICOT) and a projected budget of 500 million dollars over a ten-year period. This book is a passionate warning about what will happen to the US if the importance of the Fifth Generation is ignored. The Japanese efforts at ICOT are used as the exemplar of what should be done. Feigenbaum and McCorduck are our modern day Isaiahs; like Old Testament prophets, afraid that the people will not listen, their voices are sometimes strident. The major parts of the book, with the catchy titles of the authors, are as follows: (1)The New Wealth of Nations: “Knowledge is power. Machines that can amplify human knowledge will amplify every dimension of power.” (2)It's Not Just the Second Computer Revolution—It's the Important One: “The computer, even in its early awkward, exasperating childhood, is changing our lives as we hoped, we knew, it would.” (3)Experts in Silicon: “Expert systems are agents of the second computer revolution.” This is a really good chapter describing expert systems to the lay person. (4)The Japanese Fifth Generation: “. . . the Japanese are the farthest ahead in perceiving where the new wealth of nations lies.” This section is an interesting one, first describing the Japanese Fifth Generation project and then examining myths about the Japanese. (5)The Nations: “. . . nobody can be unduly optimistic that the UK can reverse decades of bickering and carry out a coordinated national plan. . . . [T]he French, like the British, show that they understand where the future of their country's economic survival lies. Whether they can overcome the enervating arguments that seem to accompany almost any French undertaking remains to be seen.” (6)The American Response: “The Fifth Generation and what it represents compel us to confront an enduring theme in American life. It is anti-intellectu- alism. . . . [In contrast to the Japanese] we are currently betting that uncoordinated development of the information processing industry is a luxury we can still afford. . . . We [the authors] believe that Americans should mount a large-scale concentrated project of our own; that not only is it in the national interest to do so, but it is essential to the national defense.” (7)Epilogue, or It is Hard to Predict, Especially the Future: “. . . - reasoning machines will change things from how we know them in vastly unpredictable ways.” It is unfortunate that this book is being reviewed three years after publication since it is a call for immediate action. However, the late review provides an opportunity to examine what has happened in response to the Fifth Generation during this period. Datamation [1] reports that “Japan's fifth generation computer project moves into its ambitious middle phase with waning support from industry participants. . . . The budget of $20 million for fiscal 1985 is down 8.3 percent from 1984.” Rigas et al. [2] indicate that the tool-building phase is now complete at ICOT with the development of the Personal Sequential Inference (PSI) machine and a relational database machine known as the Delta machine. The main hope of this book, a National Center for Knowledge Technology, has not materialized. Instead, the major US efforts are those of the Defense Department's DARPA and the private-sector MCC, described below. There has also been an explosion of interest in the private sector as people find that a project incorporating the name “expert system” smells more sweetly to funding agencies than projects not so named. DARPA began the Strategic Computing Initiative (SCI) in late 1983, “to provide the United States with a broad line of machine intelligence technology and to demonstrate applications of the technology to critical problems in defense.” SCI was originally projects to be a $100 to $150 million/year project, but will be impacted by DoD budget cuts. SCI has a technology base development effort ranging from microelectronics to AI areas, such as expert systems, vision, and natural language, and culminating in three demonstration systems: an Autonomous Land Vehicle (ALV), a Pilot's Associate, and an Airland Battle Management System. The ALV is the farthest along, with the initial field test of the vehicle running autonomously on a paved road at 5 km/hr; this project was completed in the Fall of 1985. The Pilot's Associate project started in the Spring of 1986, and the Airland Battle Management project will begin shortly thereafter; see [3] for more details. The Microelectronics and Computer Technology Corp (MCC) was formed in 1983 by ten otherwise competing computer and semiconductor companies. (IBM is not part of MCC.) Davis [3] reports that three years later there are 21 companies, each of which pays one million dollars to join, along with yearly fees that support each research program in which it participates. The 1986 budget is about $65 million. MCC's work will not culminate in a specific Fifth Generation computer. Rather, it will advance computer technology to prototypes, and it is up to the shareholders to convert this into marketable products. The chairman, Bob R. Inman, sounds very much like Feigenbaum and McCorduck when he says, “In the 1990's we aren't talking about a computer industry, or a semiconductor industry, or a telecommunications industry. We're talking about an information-handling industry.” Fischetti [4] reports on the European work in AI, including multinational efforts such as the European Economic Community's ESPRIT and national efforts such as the French-initiated Eureka technology advancement program. Much of the work involves expert systems. The British Alvey program launched in 1983 projected a five-year, $500 million budget. So, in retrospect, it appears that the world is now moving in a direction desired by the authors: Attention has indeed been paid to the importance of the Fifth Generation. In the US, the DARPA and the MCC programs provide the only significant coordinated efforts, with private industry efforts being carried out the way the authors predicted: “a lot of short-term . . . research and development, spurred by nothing but immediate market considerations." One doesn't often find passion and commitment in the computing literature. (Besides the present book, only Weizenbaum [5] and Papert [6] come to mind.) The authors feel strongly about the dangers of ignoring the Fifth Generation and do their best to warn us, sometimes using Time/Newsweek-like prose for dramatic effect. Except for the third part, which describes expert systems, the book is more concerned with political, economic, and social issues than with technical exposition. Much of the rhetoric concerning the potential of expert systems should not be interpreted as scientifically based evaluation. Thus, the reader can expect to encounter fanciful gems such as, “. . . who can say how universal access to machine intelligence—faster, deeper, better than human intelligence—will change science, economics, and warfare, and the whole intellectual and sociological development of mankind__ __”