What is an Expert System

(Last updated 11 March 2007)

 

The term Artificial Intelligence (AI) was first used at a conference at Dartmouth College in 1958. Since that time, AI has come to mean many things; robotics, neural networks, rulebased systems and knowledgebased systems. KBSC deals mostly with rulebased systems in the business world where they are sometimes referred to as Business Rule Management Systems, BRMS.

A BRMS is normally one or more sets of if-then-else rules managed by a non-monotonic inference engine that is Turing Complete. The Universal Turing Machine as well as the Church-Turing Thesis is also discussed on Wikipedia.  An even more enlightening look at Church-Turing is kept at Stanford and points out many of the “mis-quotes” that have been attributed to their thesis in the past in many different disciplines.  Alan Turing, one of the leading mathematicians of his day, greatly contributed to breaking the German Nazi Enigma cryptology machine during WW II.  Dr. Charles Forgy, inventor of the Rete Algorithm, Rete 2 Algorithm and the Rete III extensions to Rete 2 Algorithm, had this to say about Turing, Church and G
ödel:

 

The Church-Turing thesis basically says that all computing machines beyond a certain (surprisingly low) level of complexity are equally powerful.  That is, anything that can be computed by one computing machine can be computed by all computing machines.  The Turing Machine is one example; Church’s lambda calculus is another.  To show that a new model is equivalent to the others only requires that one show the new model can simulate one of the old models.  Regarding production systems, it is trivially easy to write rules to simulate a Turing Machine.

 

There is another result, the Gödel Incompleteness Theorem that is often mentioned in relation to intelligence.  It says that every formal system of a certain power (basically powerful enough to model itself) is either incomplete or inconsistent.  Some people claim that this is evidence that machines can never be as intelligent as people.  After all, people can create new axioms to create ever more powerful logical systems when they need to.  The flaw in this argument, in my opinion, is that the argument focuses on the “incomplete” part of the theorem while ignoring the “inconsistent” part.  If someone can show me a person who has not a single inconsistency in his world model, then I might grant that people are not subject to the incompleteness theorem.”

 

Normally, the rules are gathered in an intelligent manner, called Knowledge Extraction or Knowledge Acquisition, by a Knowledge Engineer (KE) and implemented by a Rule Engineer (RE) into the engine along with any other computer programs necessary for the rules to function.  In a true BRMS the rule architecture may be designed by the KE but the rules could be entered by a Business Analyst.

While non-monotonic is a term used to describe the actions of an inference engine wherein the if-then-else clauses are examined over and over, a more proper definition was offered by Dr. Charles Forgy et al: “A logic system is monotonic if the truth of a proposition does not change when new information (axioms) are added to the system. In contrast, a logic is non-monotonic if the truth of a proposition may change when new information (axioms) is added to or old information is deleted from the system.” 

 

There is a rather good link on First Order Propositional Logic (FOPL) at CMU.  As you can see, I offer very little in the way of theoretical studies but rather I am an implementer (KE) as well as maintaining a small research lab where we gather tools, analyze those tools, write articles on those tools and run benchmarks.  Some benchmarks may be found at the following links:

 

 

Expert System Links

 

 

we specialize in Intelligent Business Rule Management Systems (iBRMS) - the same thing that used to be called rulebased systems by the AI companies.