Artificial Intelligence in the Personal Computer Environment, Today and Tomorrow

ARTIFICIAL INTELLIGENCE IN THE PERSONAL COMPUTER ENVIRONMENT,

TODAY AND TOMORROW

Tom J. Schwartz

Electronic Engineering Times 1470 Wildrose Way Mt. View, CA 94043

AI AND THE PERSONAL COMPUTER, TODAY

Today many run time expert systems are being delivered on PC's and dedicated workstation's. One of the most ambitious is the GE Cats-1 expert system for assisting in diese! locomotive repair . This expert system is deployed on a dedicated Workstation and contains over 1200 production rules. Cats-1 is an expert tutor. When used in conjunction with a video disk player it can show diagrams and training film sequences in conjunction with the consultation process. PUFF an expert system, written in EMYCIN is used to diagnose obstructive airway diseases. PUFF which originally consisted of 55 production rules, has been rewritten in Basic and is now being commercially provided on a system which incorporates diagnostic hardware and uses an Apple II as a delivery vehicle. Applied Expert Systems has already deployed an expert system on the IBM PC-XT, for use in the financial services industry. The use of PC delivery vehicles is easily predicted because of it's !arge installed base and declining price. Workstations will continue to be popular because their power reduces expert system development time.

Development of expert systems on dedicated workstations has proceeded from the introduction of the LM 2 by Lisp Machines Inc. and continues unabated on workstations from a wide range of vendors. The evolution of expert systems on PC's is now gaining momentum. There are now at least a dozen implementations of Lisp and PROLOG for the IBM PC. ExperTelligence has recently introduced a Lisp and a Smalltalk for the Macintosh with an OPS5 soon to be released. There are already two versions of OPS5 available for the IBM PC as well as a version of Smalltalk. The wide availability of the classic AI languages on the PC will lead to numerous expert systems being developed and deployed on the PC.

ABSTRACT

It has been almest thirty years since the Dartmouth Conference on artificial intelligence. Since then, AI has emerged from a laboratory curiosity to a blossoming technology with world wide strategic implications. As this technology proliferates, the PC will become a major delivery vehicle for expert systems. The PC is already being used to deploy small expert systems and it's power is bound to increase over the coming years. This panel will examine the history of the PC in AI, the current state of development and attempt to en vision future developments in the marriage of these technologies.

PERSONAL COMPUTER POWER

1981 was a watershed year for computers according to market research from INPUT Inc ·· It was the first year that installed end user computing power (PC's and individual workstations) equaled that of mainframe computers. It is estimated that in 1986 PC computing power alone will equal that of all other computers. This pattern will continue as PC's become increasingly more powerful and the boundary between PC's and workstations becomes forever blurred.

VISIONS OF THE PERSONAL COMPUTER

One of the first people to champion the concept of the personal computer was Dr. Alan Kay. Dr. Kay's werk on the Alte and the Dynabook along with others at the Xerox Pale Alte Research Center in the 1970's was the basis of many of the developments we see today. These developments include: bit-mapped graphics, flat screens, personal w6rkstations and networking. Dr. Kay's Dynabook actually predicted the portable, flat personal computer. The Dynabook currently has many commercial realizations including machines by Apple, HP, Data General, Osborne, Morrow, Radio Shack, Grid, Gonvergent Technology and NEC. These machines, while not achieving the full performance envisioned by Dr. Kay, will surely be improved and fulfill his design goals.

Many expert system shells, production rule languages and induction extraction tools have already been developed for the PC (table 1). These include PC implementations of EMYCIN and other English like production rule systems. Systems which learn by example, typically called inductive extraction systems are made available by three manufactures. RuleMaster by Radian is one of these systems and is unique because it combines inductive extraction and a production rule system in one product. A few of these tools also have hooks to other programs to further increase their power. The commercial success of these systems is demonstrated by SRI's Series PC which has been used to develop and field an expert system for copier diagnosis and repair.

AI AND THE PERSONAL COMPUTER, TOMORROW

Many of these PC-based systems are suitable for the development of small expert systems (under 200 rules). These small systems are now being called "technician systems" and will find wide use in business and technical training. Induction extraction tools will enable users of spread sheets to build such technician systems and deploy them with little or no help from knowledge engineers. These systems will be used as learning curve accelerators and to disseminate routine expertise to less experienced members of an organization. They can also be used to enhance homogeneity in the decision making process thereby making delegation more effective.

The deployment of dedicated expert systems on hand-held calculator-like devices using application-specific integrated circuits will make expert systems as pervasive as the microprocessor in the world of tomorrow. Low cost expert systems will aid us in many of our daily activities such as: commuter route selection, business and investment decisions making, human interaction, personal health care and even wagering. Automated knowledge acquisition will allow every expert to become his own knowledge engineer. Even the breaking of the common sense barrier can be foreseen.

ACKNOWLEDGEMENTS

The author would like to thank the panelists Dr. Alan Kay, Stan Curtis and Robert Laddaga. Thanks to Amos Oshrin for the many fruitful discussions on AI and to all the software producers in table 1 for their cooperation.

The following systems run on IBM PC's