[Applause] [Music] [Music] microfocus creators of visual programming tools for software development is pleased to provide major funding for the computer Chronicles the story of this continuing Evolution welcome to the computer Chronicles I'm Stuart chaet and this is Gary kildall our subject today is the fifth generation the computers of the Future these are sometimes known as expert systems or computers dealing with artificial intelligence they're sometimes called Kips Kips knowledge information processing systems or sometimes for short just knowledge-based systems and our guest today will be Edward Fen bomb he's one of the founding fathers of the field of artificial intelligence and today one of the world's leading experts in that field Gary Fen bomb uh calls it the fifth generation is this just one more technological advance or does the fifth generation really represent a kind of Quantum Leap into a different kind of computing well it certainly does represented Quantum Leap the the major evolutionary changes in Hardware over the last 40 years in computers has they've been referred to as Generations uh but the fifth generation is really a Japanese slogan that underlies their effort at a 10-year National program to produce what you've called the what's been called the knowledge information processing systems or kips at the heart of these Kip systems are uh knowledge-based systems knowledge-based software and that's a technology that's really grown up in our own country's field of artificial intelligence today we're going to be looking at the uh knowledge based system or expert systems uh and also take a look at what threat the fifth generation poses to the US uh computer industry we have a uh short segment here that shows the use of computers in today's society this is a an example of actually Robotics and we'll see how more advanced applications as our session goes further today this computer tape librarian is a robot it responds to commands from departments all over Stanford University loading and unloading tape reels storing them in precise slots on a shelf it performs what would be a dreary and repetitive task for a human being it is computer controlled and it follows instructions but it does not make decisions the robot can't advise the operator for example on the best way to store his tapes or what to do if one is misplaced or damaged it performs a simple task the same way every time but suppose the machine could give advice that instead of just processing data or answering a command and it had enough knowledge stored to manipulate facts and make a decision this new kind of computer one that can intelligently use the information stored inside it is now being developed often referred to as the fifth generation it is the next step in making the machine approach and perhaps even duplicate the way a human being thinks a hint of what we may soon have exists today in what is called an expert system it requires two kinds of knowledge first documented facts and second euristic knowledge or the same kind of ability that a human expert needs to make a good judgment based on experience and practice building this almost intangible Talent into a computer's memory is called knowledge engineering and the country that first Masters that technique will be in a unique Global position the race to achieve that position has already begun [Music] Our Guest today is Edward Fen bomb professor of computer science at Stanford University and author of the book the fifth generation also joining us is Herb lechner from Sri International herb appears frequently on this program to lend his expertise to the subject Gary well I think it's a real honor to have Ed here today talk about expert systems if there's ever an expert in expert systems you got to be sitting here today it's I think very exciting uh topic for me especially since it's probably one of the um most promising new software areas that we we're seeing coming up over the riseon artificial intelligence seems to be one of those areas that's kind of a moving Target things that 10 years ago where topics of discussion AI tend tend to be come practical applications and I think knowledge based systems tend to be in that category right now could you give us an idea of basically what knowledge based systems expert systems are all about the um computers as you know are General symbol processing devices capable of of manipulating any kinds of symbols of which numbers constitute one important class but computers are much more General than that we've known about the generality of computation since at least the time of uh touring in the 1930s and actually I've tracked it back to um intuitions that Babbage had that were reported by aah Lovelace after whom the adaa programming language is named in 1842 a lovece wrote that uh the analytic iCal engine of babage has constituted the the uh link between the mechanical world and the world of the most abstract Concepts that currently in the modern terminology is called the physical symbol system hypothesis and is the basis for artificial intelligence work in artificial intelligence as a science we talk about the use of computers to process symbolic knowledge using logical inference methods symbolic inference methods in in other words we're talking about uh inference and not calculation in the traditional sense we're talking about knowledge and not numbers in the traditional sense when we talk about knowledge we mean not only the factual knowledge of a field which is available uh in the open literature it's published in textbooks it's commonly agreed upon by experts in the field it's delivered in lectures but we're also talking about a much more subtle kind of knowledge which we call theistic knowledge of a field the experiential knowledge the non-rigorous knowledge George POA the great Stanford mathematician in popularizing this in the 50s called heuristic reasoning the art of good guessing and we'd like to invest computers with the art of good guessing and the people who do that for practical applications are called knowledge Engineers they build and test and revise programs and these programs are called expert systems or knowledge systems systems the usually the goal of the expert system designer is to design a program whose Behavior will uh match the behavior or exceed the quality of behavior of uh human professionals experts in a field those are the programs which we normally call um expert systems the expert systems work has been going on since about 1965 the beginning of the second decade of artificial intelligence uh research in this country at the time that the expert systems work developed the artificial intelligence field was focusing on the concept of generality as signifying intelligence not performance but generality the ability of a program to shift its attention from one area to the other easily and solve problems the expert systems people uh initially at Stanford and then later at other places took the other approach that if you wanted to create programs which were very intelligent which performed at the level of human experts you had to encode a great deal of knowledge about a field the key result of 17 or 18 years of work in this area now is simply that knowledge is power the primary source of power in expert systems is the knowledge that they contain and I've only secondary importance is the inference procedure that they use for drawing out the lines of reasoning this turns out to be true of all areas of artificial intelligence not just expert systems and in fact um you can see it in the work that goes on in understanding natural language where a great deal of World Knowledge needs to be present for the programs to um disambiguate natural language utterances we see it also in the uh AI Vision program image understanding it's very difficult to understand a picture unless you know something about the world represented in that picture what a river looks like what a railroad looks like what a city looks like what a cloud looks like all of this focus on knowledge has really led to what is called the shift to a knowledge-based paradigm in Ai and that's probably the single biggest event that's happened in the second and third Decades of AI work so the context around the problem is really the important or is a very important part of the of that it's very difficult to to understand problems to achieve goals to understand utterances and signals unless you have a great deal of knowledge if we take expert systems in particular are is it fair to say that this is something now that's moving from the University environment into industry that they're is becoming industrially acceptable now to to use expert systems and do you have any comments on that well that's what all the excitement is about we we see a kind of frenetic Pace going on uh newspaper articles and uh TV shows and all that on on the subject of artificial intelligence is caused by that transition we're we're beginning to see the industrialization of AI we're beginning to see the kind of transition that took place in the uh late 40s and early 50s from laboratory to Industry and Computing in general M Ed what are some of the applications that have seen the light of day utilizing these knowledge base and artificial intelligence techniques by now there are hundreds of them so I can just skim the surface the granddaddy of all these applications was started in 1965 at Stanford by uh professor leberg and the medical school and professor jassi in chemistry and myself that was an application to chemical analysis where data from physical instruments were were analyzed in terms of chemical structure that may have given rise to that data uh a program called the dendr program so there uh there have been several follow-up applications in the area of programs to interpret instrument data intelligently subsequent to that there were numerous applications to diagnostic processes and therapeutic advising in medicine for example at Stanford the M program for Consulting about uh infectious disease diagnosis and therapy a more recent effort to advise clinicians oncologists about the uh prescription of cancer chemotherapy in the Stanford cancer Outpatient Clinic programs at the University of Pittsburgh for differential diagnosis in Internal Medicine in fact that's a spectacular program that that knows uh a couple hundred thousand things about uh Internal Medicine knows about 500 different diseases 3,500 different signs and symptoms of disease and does an extremely good job at differential diagnosis in addition to that Industries have used uh expert systems for engineering purposes the configuration of computers from customer orders or business equipment systems the use of uh expert systems for assisting in the design of uh integrated circuits the uh use of expert systems for diagnosing equipment failures for example computer equipment failures on that subject DED let me interrupt a second I think uh you have a program you worked on called drilling advisor which helps analyze for a geologist a situation like that we're going to take a break for just a second but when we come back it is going to walk us through this particular expert system called drilling advisor that's coming up next [Music] Ed we want to try to demonstrate how one of these expert systems would work and I know one of your products is something I think you call drilling advisor maybe could give us a brief introduction to this program uh Stuart expert systems are characterized by having extremely high value added in the case of the drilling advisor the value added is in terms of capturing replicating and distributing of expertise the elf aiten Oil Company the French national oil company has a problem with regard to drilling of oil wells drilling problems can be very expensive elf aitan has too few experts in the world in order to capture and diagnose all the problems encountered by the drilling teams these drilling problems can occur on the other side of the world from France for example and they after an expert system that can capture the expertise and sit it down there at the uh drill rig the uh alakan drilling advisor was done in a collaboration with a silicon Valley venture startup firm named techn knowledge that uh does knowledge engineering custom systems of the type that you'll see okay maybe we can roll we have a brief demonstration now for about a minute and a half of actually how this program works we can take a look at that right now one knowledge system currently under development at tech knowledge provides advice on problems encountered while drilling oil wells the field version of this system will Aid oil well drilling Crews to avoid or correct accidents that occur during the drilling process since an offshore oil rig May cost more than $100,000 per day to operate time and Equipment lost due to accidents can account for as much as 2% of the total cost of drilling a well because few individuals possess the knowledge experience and judgment to properly handle drilling incidents the availability of an advisor in the form of a knowledge-based expert system is expected to reduce lost time and Equipment help train drilling Crews to handle problems and significantly reduce the cost of drilling an oil well the technolog drilling advisor prototype is an expert in the diagnosis of sticking problems sticking occurs when the surface crew is unable to move the current drill string up or down the upper left of the screen is devoted to a typed dialogue between the user and the knowledge system the entire right half of the screen serves in the prototype to display the internal operation of the knowledge system this is mostly used for debugging at the lower left of the screen the current drilling environment is diagrammed this includes representations of six major causes of sticking initiating a session with a drilling advisor the user types the name of the well the system then retrieves from its disk storage any information that had previously been entered about this well including any problems encountered after determining that the this is a vertical well the drilling adviser requests information about relevant strata and formations in the well using a similar format the user informs the system that a sticking problem has occurred during an attempt to repair the drill string at 4,000 M the drilling advisor now focuses its attention on the immediate problem designated episode one on the screen first the system requests information about the drilling fluid and learns that the drill has ceased to rotate okay Ed from that portion we saw now how is that knowledge represented in the program and how does the program use that knowledge the drilling advisor is what's called a rule-based system system and the knowledge is represented in the form of rules which contain two sides an if side and a then side the if side contains a number of conditions which I like to think of as the relevancy conditions of this piece of knowledge if those conditions are satisfied then that rule will be evoked from the knowledge base and built into the line of reasoning the then side contains the conclusion that can be drawn on the basis of the evidence that's in the if side now the way it's used is is in a scheme called gold directed backward chaining in which the program works backwards through the rules from a goal to be achieved to some known facts such as were being accumulated from the user at the console okay we have a little bit more of the dialogue let's take a look at it and see how it works returning to the dialogue the drilling adviser is asking the user for the free point the point below which the drill string is not stuck in this case the information is on unknown the user indicates that the Sandstone strata is permeable and that the specific gravity of the drilling fluid is 1.24 but when the system requests the fluid specific gravity of the Sandstone strata the user becomes confused while the previous questions indicated a sensible line of reasoning the user does not understand the relevance of this question to the overall problem rather than answering the question the user types why the drilling adviser then presents an explanation of the specific reason that the question is being asked to help determine the sticking Force this is followed by a more detailed description of the current state of the systems reasoning including the various factors that must be used to calculate the total sticking Force okay Ed this part where the user can ask the computer why that's that's how does that work that seems pretty significant part of an expert system we regard these expert systems as uh assistance to human professionals and we believe that they will not be used unless they can explain to the human user how they are uh reasoning so we have these explanation programs uh as pieces of knowledge Engineering in the system that allow the system to be a transparent box we shun black boxes we like the user to be able to peer right through the box and understand how the system is working now in this particular case the program looks back over the the chain of rules that have been accumulating due to the backward chaining procedure and then uh spins out the rules that have led the program down to a certain point in the consultation you know as uh we still have some time here I think it's worthwhile discussing for a few minutes this threat that you see in the fifth generation uh Japanese computer systems can you give us some idea what's going on there the um the Japanese government working through the uh Japanese technology planning agency meti the ministry of international trade and industry has established a major national project which they're uh funding at the level of uh $450 Million worth of government funding and an equivalent amount of Industry funding over a 10-year period aimed at producing hardware and software and applications in what they call the knowledge industry applications similar to the one that we've been seeing here pack packaged expertise knowledge based systems of all kinds the uh Japanese uh overall strategy is a very interesting one they lack market share in the world computer Market they'd like to have more it's very difficult to gain market share from the world's leader IBM what you have to do is to gain a position in an area in which IBM has not been active and that's artificial intelligence so the Japanese want to get there first and fast so that they can gain a good chunk of the market before before IBM wakes up are we ahead of uh Japanese right now and expert systems real quickly Ed we have just a few seconds well I think we are we've been ahead for some time I think with a dedicated effort on the part of the Japanese uh we'd better move pretty fast if we want to stay ahead and we are out of time I'm sorry to say thank you very much for being with us at Fen bomb and her lechner and I hope you'll join us again for another edition of the computer Chronicles [Music] micr Focus creators of the visual programming tools for software development is pleased to provide major funding for the computer Chronicles the story of this continuing evolution [Music] Random Access is made possible by a grant from bite the small systems Journal Publishers of a monthly magazine on microcomputer technology and Innovative projects in the world of computing in today's Random Access file there appears to be a major effort underway by American computer chip manufacturers to gain their lead over the Japanese the federal government is working with a new trade organization called the semiconductor research Corporation or SRC reportedly more than 20 major Electronics corporations are involved in SRC with the goal of producing the world's first four megabyte chip by the year 1988 us firms used to control the 16k chip Market with about 2third of World Sales but Japanese manufacturers have dominated in the 64k chip Market with 70% chip experts are saying the 4 megabyte chip will require electronic connections of one micron or less if you don't know a micron is about 1 1100 the width of a human hair the semiconductor industry Association says boom times are back for chip makers estimates are that us companies should sell nearly 11 billion doll worth of semiconductors this year an increase of some 16% over last year meanwhile National Semiconductor in Santa Clara reports excellent first quarter with net income of 41 cents a share that's up from less than 1 cent a share a year ago and Intel and Advanced Micro Devices showed very strong fourth quarters just a few weeks ago Fortune computers was bragging that it wouldn't follow in the footsteps of Osborne but this week Fortune has been reeling a bit its founder and CEO Gary fredman resigned at the request of the board of directors Fortune suffered a $3 million loss during the last quarter its stock which had sold at $22 a share earlier this year is now down to about $850 what's new well you've probably seen those new tv ads for MCI electronic mail MCI which has been challenging mob Bell for the long-distance phone business is now challenging the post office with its new MCI mail service for personal computer owners MCI says its system will interface with the post office and Courier Services so that you can even send electronic mail to people who don't have personal computers kico finally showed off a production version of its atom computer to the Boston computer Society reports are that the much touted word processor is not too sophisticated cico hinted at a utility pack due to come out next year which will improve its word processing latest word is that shipment of atoms are finally due to be made at the end of this month if you bought one of those portable computers that you can take on an airplane you might want to check with the Airlines many computer users are being told no when they drag out the old keyboard and put it on their laps some airlines are worried about radio interference with the plan's navigation system according to the LA Times tww and Republic say okay to using the Portables but United PSA Eastern continental and Western have banned the use of portable computers while flying Nolan Bushnell is back if you've forgotten he's the guy who invented video games and found it Atari then sold it to Fortune for a fortune rather to Warner Communications well Bushnell's 7-year non-compete agreement expired this month and word is that he's soon coming out with twin laser dis video games he's also into robots and rumors are he's working on robots to serve pizza at his Pizza Time theater restaurants if you missed it by the way a major Court decision on copyrighting software came out of the US court of appeals in Philadelphia a few weeks ago Apple had sued Franklin computers for copying its Operating System Program the Court ruled that the program even as etched on a silicon chip was copyrightable and so could not be copied without compensation to its creator what's the bestselling Business program these days Lotus 123 has jumped to the top of the software charts in the first six months of this year Lotus sold more than $12 million worth of 123 and success breeds competition about 30 software companies have announced plans for what they say will be improved versions of 123 finally with computers and software multiplying like rabbits the inevitable has happened some ad agencies are thinking about placing advertisements on software discs or on computer networks the Dow Jones information service says it will start doing something like this the best guess is that early ads will be text only and will promote software computer magazines and financial services that's it for this week's Random Access a capsule look at what's happening in the world of computing be sure to watch for Random Access following each edition of the computer Chronicles Random Access is made possible by a grant from bite the small systems Journal Publishers of a monthly magazine on microcomput technology and Innovative projects in the world of computing