Technologies are invented for a reason. Usually that reason has more to do with the inventor's personal situation than with the technology's ultimate use. As a result, technologies evolve after they leave the laboratory. As time passes, many innovative practitioners attempt to use the same basic technology to do all sorts of things that were never envisioned or intended. Most of these efforts fail to have much business or social impact -- a few do work out. Occasionally, someone may even find an incredibly successful application for the new technology, one that satisfies some fundamental human need. It is these few successful innovations that "survive" and go on to change our world. Technologies evolve toward human needs.
Expert systems technology was originally invented in the AI laboratories in an attempt to apply the state-space search, knowledge representation, and inference techniques developed in early research to some "real-world problems." The hope of the inventors was to demonstrate, especially to those always-fickle funding agencies, that AI was possible and practical and that thinking about thinking machines was scientifically sound. They succeeded beyond their wildest dreams. Also, they failed.
They succeeded in showing that AI had practical value and that machines could "think through" some very hard problems. But, as inventors often do, they failed to find or create a market demand for the technology sufficient to make their start-up companies successful. With the help of venture capitalists, inventors can be down-right dangerous to the unwary customer or investor: their technology-driven vision can be disguised to look like a real market-driven business.
However, the new technology, if it survives, takes its own course. Once it is out of the inventor's laboratory and in the hands of thousands of people with their own problems to solve, necessity, as they say, is the mother in this script. Most of these inventive engineers and programmers also succeed and fail. With expert systems, they succeeded in building applications that did, indeed, solve real business problems. Ultimately, however, most failed to make the case for expert systems as an essential software technology in their companies. Most fail, but it only takes one lucky application that fits, one simple, elegant solution to a universal problem, to move a technology forward.
All programs contain knowledge. Expert systems is a software technology for more conveniently embedding into programs certain kinds of knowledge (taxonomies, rules, constraints, descriptions of past situations, etc.). The technology consists of software tools and associated methodologies for acquiring and coding the knowledge, for automatically using the right knowledge in the right way at the right time, and for keeping the knowledge up to date.
These "knowledge-based" software tools were originally invented to build a certain class of programs in AI laboratories. In fact, they were invented to avoid laborious work, so that MYCIN would be easier to build than DENDRAL had been. The researchers at Stanford built themselves some tools for editing, using, and debugging hundreds of heuristic rules, so that they didn't have to code the rules directly into LISP and then debug that code as more and more knowledge was acquired (as had been the case with DENDRAL). In the AI labs, it was the problem-solving behavior of programs that was important -- building "expert" systems. The nascent technology for conveniently embedding the knowledge into the programs was just a means to an end.
Of course, building intelligent problem-solving programs is not the only thing that these new software tools and methodologies could be used for. The world has, so far, shown limited interest in purchasing or building intelligent programs. We don't need more intelligent programs. We need better communication. And this very same technology, now called knowledge-based systems (KBS), can be used as part of an unparalleled communications technology. Each major innovation in the history of communication (language, writing, printing, telegraphy, etc.) satisfied a fundamental need to share our experience more effectively. The network of computers that is now stringing the globe is the newest innovation in communication. KBS technology will allow our knowledge to be entered and accessed asynchronously in this new medium.
The series of subsequent inventions that enabled expert systems technology to evolve in this new direction were contributed, for the most part, by programmers in MIS and AppDev groups who were constrained by two things: they had only a week or two to learn "expert systems" and they had to build low-cost systems that had immediately recognizable payback. Under these circumstances, I think you will agree, most classical expert systems applications were ruled out.
Many of these young men and women failed at their over-constrained task. Some succeeded by building on-line policy advisors, regulatory compliance systems, procedures manuals, and even help desks. These systems didn't do much "problem solving" and were considered trivial by AI researchers. However, they are examples of automation of a large class of business activities whose purpose is the distribution of knowledge around the organization. In most corporations, these activities are characterized by expensive and unread documentation, unusable on-line information systems, ineffective training programs, and tiresome meetings. "Passive" information resources like manuals and documentation are hard to use because the they leave to the reader all responsibility for finding the relevant information, which, once it is found, is often incomplete and out-of-date because printed manuals are so hard to maintain.
KBS technology lets knowledge distribution departments shift to the computer system some of the burden of helping people access the relevant information at the time that they need it . What is beautiful, for example, about a modern customer service "help desk" with a case-based problem-resolution system is that everybody contributes to a repository of their collective experience. This database, instead of being manually searched for relevant information when a new situation comes up, actively asks the user about his or her situation until it has determined if anyone has "seen anything like this before." This is a kind of groupware for sharing experience -- in the parlance of automating corporate training, it's called "knowledge-based performance support."
Active information systems, which take an active part in helping us understand our situation and identify what information might be helpful, are the next wave of business automation. We can build them with current KBS tools (although the shells will evolve to meet the specific requirements of this class of applications, which stress intelligent information retrieval over problem solving). Parallel events in the evolution of computing, such as networks, client/server architectures, multimedia and distributed computing, are also essential evolutionary steps.
We are building a new medium for sharing experience. The help desk is to the emerging global digital network what payroll was to business data processing -- a universal application that fits the new technologies like a glove. These systems will change the way business is done. They will transform computers from simple data storage and sorting devices into strategic tools for helping people work smarter, together. No business process re-engineering project should be uninformed about the possibility of automating knowledge distribution. Active information systems will share the workload in the organization's endless efforts to keep people informed and up-to-date. New kinds of business organizations and, eventually, political organizations, will be invented. Technologies evolve toward human needs.
