THE AGE OF INTELLIGENT MACHINES | Intelligent Knowledge-Based Systems–AI in the U.K.
September 24, 2001
- Ray Kurzweil
For reasons that will be made clear, the term “artificial intelligence” is not much used in the U.K. “IKBS” (intelligent knowledge-based systems) tends to be the term employed to cover work in expert systems, natural language, and logic languages. There is a long history of AI work in the U.K., dating back to Alan Turing’s pioneering work just before and after World War II. That strange genius foresaw the use of stored-program machines for much more than mathematical calculation, and in the early 1950s he enabled us to compare human and machine intelligence by postulating a test by which one could determine whether true artificial simulation of human intelligence had been achieved. He proposed that an observer in one room should try to tell if he is conversing with a man or a machine in another room. And he also predicted that this test for AI would be achieved by the year 2000. Well, there are still some years to go, and it remains an open question whether his prediction will be proved right; most of his other predictions have been amply fulfilled. One might add that some people believe Colby’s program PARRY has achieved AI, because it fooled psychiatrists into believing they were dealing with the output from paranoid human beings. If some people are not impressed with this, it may be because they are not impressed with psychiatrists.
Alan Turing worked in various places in the U.K., including the National Physical Laboratory, Bletchley Park, and the University of Manchester. After his time the center of progress in AI work shifted to the University of Edinburgh, where Donald Michie built up a considerable team and formed one of the first centers in the world for the study of AI. The university has produced excellent work in the language field, though perhaps its most renowned output has been its students, who have gone on to populate AI centers throughout the world. Donald Michie himself has moved on to found the Turing Institute at Strathclyde University, which provides an advisory and training service to industry. But the very broadly based team at Edinburgh, with groups ranging from linguistics through logic programming to speech recognition and computer science, still continues as perhaps the dominant center for AI work in the U.K.
Another much smaller center for AI work in the U.K. academic world is at Sussex University, where the POPLOG language environment is being developed under Professor Aaron Sloman. POPLOG contains compilers for PROLOG and Common LISP as well as one for POP, a simulation language that has its adherents in the U.K. community. After Edinburgh, perhaps the main centers for AI work are at Imperial College, London and Cambridge. Bob Kowalski heads the team at Imperial, where work is in progress on logic programming, expert systems, and Declarative Language architectures. One of the interesting recent applications of AI techniques there has been to the construction of a rule base for the interpretation of legislation, the chosen example being the British Nationality Act, perhaps not one of the most logically constructed pieces of British legislation.
Computer architecture work flourishes at many university centers. Imperial College and Manchester University are cooperating to produce a general-purpose parallel architecture, and a development stage of this machine, called Alice, has been installed at Imperial College and also at the computer center at Edinburgh University, where it is possible to make comparisons with various other machines. Other work in parallel architectures is in progress at universities at Cambridge, Reading, Bath, Southampton, East Anglia, Glasgow, and St. Andrews as well as at University College and Queen Mary College, London.
Work in natural language is relatively weak in the U.K., in comparison with the amount of work on logic programming and advanced architectures. But there is some work being done at Cambridge and the Open University, and as a part of speech recognition, work is in progress at various centers, most notably at Edinburgh.
It is generally believed that the Lighthill Study of Artificial Intelligence, which was commissioned by the Science Research Council toward the end of the 1950s, was a major setback for AI work in the U.K. In this report Sir James Lighthill suggested that one needed a major breakthrough before it would be possible to gain much by tackling the subject of AI as a coherent whole. He recommended that robotic developments should be pursued in their own right, just as the study of human intelligence should be, but that it would not be very profitable, for the moment, to expect an understanding of the way the mind operated to lead to practical applications of so-called artificial intelligence. In a sense, time has supported his contention, for with the exception of low-level vision, there is little evidence that the development of AI practical applications have received much benefit from the study of how the human mind works. Interestingly, it would appear that the study of human behavior has benefited rather more from developments in computer science.
It was probably not intended that the Lighthill report would result in a cutback in work on the component parts of artificial intelligence in the U.K., but this is generally said to have been the result. It probably also intensified the feeling in the U.K. that the term “artificial intelligence” is unfortunate, with its implications that the products are a replacement or rival to human intelligence. For this reason, when work on AI was intensified in the early 1980s, the term “knowledge-based systems” was employed, usually preceded by the word “intelligent” as a sop to the AI community.
If the Lighthill Report of the early 1970s was paradise lost for the AI community, the Alvey Report of the early 1980s was paradise regained. The Alvey Report was triggered by the Japanese announcement of their fifth-generation program at a conference in Tokyo in the autumn of 1981. In practice, this program is not very large by Japanese standards. The Japanese AI research center, ICOT, still does not number more than 60 workers. But the fact that Japan announced this belief in the importance of AI work to the future of the Japanese economy was enough to launch renewed support for AI in the U.S., Europe, and elsewhere. In the U.K. the growth of interest in AI, and indeed in information technology generally, actually preceded the Japanese announcement.
The Science and Engineering Research Council (SERC), the body in the U.K. that has the responsibility for funding academic research in the physical and biological sciences and in engineering, had already initiated a specially promoted program for distributed computing, which led to much of the flowering of work in parallel computing and declarative and logic languages. A study of the computer-architecture requirements for IKBS was initiated, and this in practice formed the background for the strategy for the IKBS or AI part of the Alvey program.
The Alvey Committee, which planned the program, saw it as a way of marshaling all the research resources of the U.K. to an attack on what were then seen, and are still largely seen today, as the enabling or underlying technologies required to support the whole of information technology. The committee’s report, accepted by the government in May 1983, called for a program costing 350 million pounds over 5 years, of which industry would pay for about half, or 150 million pounds. The government contribution of 200 million pounds would come from the Department of Trade and Industry, the Ministry of Defense, and the SERC.
The program was seen as one of cooperative research, and in practice the work is being carried out by some 200 project teams drawn from industry, the universities and polytechnics, and the research institutions, which are largely government establishments. Typically, teams from two or three firms and one or two universities take part in each project. Universities are involved in over 85 percent of the projects, and there is a special “uncle” project class for work of a long-range or speculative nature in the universities supported by some company that takes an avuncular interest in the work. Altogether, the program employs some 2,300 research workers and has probably doubled the effort in the fields covered. Sixteen clubs exist to bring together all the workers in a common field; for example, the Speech Club covers workers in the 10 projects in that field.
AI work largely falls into the IKBS and man-machine interface (MMI) parts of the program, the other parts being VLSI and software engineering. MMI covers human factors and speech and pattern recognition. There are also considerable developments in expert systems in the Large Demonstrator parts of the program, which are designed to pull the work together into some practically orientated major projects. Overall, AI work probably constitutes nearly half of the whole program, appearing in all parts. For example, the software-engineering program includes a set of integrated project-support environments, software-project tool sets. The most advanced of these will be based on the use of AI techniques. But it is only fair to add that the AI and software-engineering communities in the U.K. are not really converging as fast as one might like to see.
Much of the AI work in the U.K. has suffered from a lack of suitable experts. So part of the Alvey program is devoted to building up the skilled manpower. When the research-trained postgraduate workers from the Alvey projects soon start becoming available for employment in the general community, this will make a very considerable impact on the available skilled manpower. The program has initiated a Journeyman scheme under which good people from industry with skills in information technology but lacking in AI expertise go to sit at the feet of masters in the academic centers for six months or so while working on projects of interest to their employers.
The Awareness part of the Alvey program has proved to be a great success. The Alvey Tapes are a set of videotaped lectures of distinguished experts on artificial intelligence prepared for the Alvey program by the Open University. The 16 one-hour tapes cover subjects like logic programming, dealing with uncertainty, image understanding, machine learners, natural-language processing, and various aspects of expert systems. One of the most popular tapes, designed to counter the oft repeated canard that expert systems are not really used, shows examples of such systems at work in British industry.
Another way of helping British industry to get started on building expert systems has been through the Expert Systems Starter Pack put together by the National Computing Centre. These consist of computer tapes for four different expert-system design tools, together with training guides. These tools provide elementary experience of such techniques as backward- and forward-chaining inference mechanisms and the use of dialog generators.
To spread the knowledge of expert systems widely in British industry and commerce, a set of Community Clubs has been established. A Community Club typically consists of 10 to 30 likeminded firms who join together to study the development of an expert system in an area of particular interest to them. They work with a team from the computer industry or the universities that has had experience in building expert systems, and together they draw up a specification, recruit staff to build the system, construct the system, and then test it. Often one or more of the firms acts as the guinea pig for trying out the system. The Alvey program has sponsored nine such
Community Clubs in fields like financial services (Small Company Health Adviser), insurance, data processing (Help Disk Adviser for data-processing installations), econometric modeling, planning for real-time manufacturing projects, quantity surveying, real-time quality control of processing plants, transport-route planning, and water-industry construction planning. Altogether, some 200 organizations are participating in these clubs and learning how to build and use expert systems for a purely nominal fee. They have greatly helped to spread awareness of the power and limitations of expert systems in the U.K. The initiative is being copied by others, so that knowledge about expert systems has grown very fast in the U.K. It is not accidental that the expert-systems group of the British Computer Society is perhaps the most flourishing such society in the U.K. today. It is probably true to say that there is more true awareness of the implications of expert systems in the U.K. than elsewhere, even if the magnitude of some of the applications in the U.S. is far larger.
No picture of the AI scene in the U.K. today would be complete without mention of the part played by the EEC’s ESPRIT program. ESPRIT is a program very like the Alvey program except that the cooperative teams are drawn from throughout the European Community of 12 nations. In scale it is, for the U.K., about half the size of the Alvey program, which means that in total for Europe it consists of about twice the number of workers. AI is about a fifth of the program, being largely represented in the Advanced Information Processing part of the program. Of course, ESPRIT has the crucial role of bringing the research workers of Europe together. Perhaps for the first time the AI community in the U.K. is coming into active contact with the other research workers in Europe not just to carry out project work but also to plan, evaluate, and implement the program. And in several countries in Europe there are also nationally sponsored programs, so the total AI effort in Europe is far from negligible and its historic problem of fragmentation is being overcome.
Courtesy of Brian Oakley
Brian Oakley, a physicist who started his career working on microwaves and computer applications in what is now the Radar & Signals Research Establishment in England, is the former director of the Alvey Program, a British national cooperative program in information technology formed in response to the Japanese Fifth-Generation Computer Systems Project. He is currently chairman of Logica Cambridge.