Showing posts with label Edward Feigenbaum. Show all posts
Showing posts with label Edward Feigenbaum. Show all posts

Artificial Intelligence - What Is The MYCIN Expert System?




MYCIN is an interactive expert system for infectious illness diagnosis and treatment developed by computer scientists Edward Feigenbaum (1936–) and Bruce Buchanan at Stanford University in the 1970s.

MYCIN was Feigenbaum's second expert system (after DENDRAL), but it was the first to be commercially accessible as a standalone software package.

TeKnowledge, the software business cofounded by Feigenbaum and other partners, offered EMYCIN as the most successful expert shell for this purpose by the 1980s.

MYCIN was developed by Feigenbaum's Heuristic Programming Project (HPP) in collaboration with Stanford Medical School's Infectious Diseases Group (IDG).

The expert clinical physician was IDG's Stanley Cohen.

Feigenbaum and Buchanan had read stories of antibiotics being prescribed wrongly owing to misdiagnoses in the early 1970s.

MYCIN was created to assist a human expert in making the best judgment possible.

MYCIN started out as a consultation tool.

MYCIN supplied a diagnosis that included the necessary antibiotics and dose after inputting the results of a patient's blood test, bacterial cultures, and other data.



MYCIN also served as an explanation system.

In simple English, the physician-user may ask MYCIN to expound on a certain inference.

Finally, MYCIN had a knowledge-acquisition software that was used to keep the system's knowledge base up to date.

Feigenbaum and his collaborators introduced two additional features to MYCIN after gaining experience with DENDRAL.

MYCIN's inference engine comes with a rule interpreter to begin with.

This enabled "goal-directed backward chaining" to be used to achieve diagnostic findings (Cendrowska and Bramer 1984, 229).

MYCIN set itself the objective of discovering a useful clinical parameter that matched the patient data submitted at each phase in the procedure.

The inference engine looked for a set of rules that applied to the parameter in question.

MYCIN typically required more information when evaluating the premise of one of the rules in this parameter set.

The system's next subgoal was to get that data.

MYCIN might test out new regulations or ask the physician for further information.

This process was repeated until MYCIN had enough data on numerous factors to make a diagnosis.

The certainty factor was MYCIN's second unique feature.

These elements should not be seen "as conditional probabilities, [though] they are loosely grounded on probability theory," according to William van Melle (then a doctoral student working on MYCIN for his thesis project) (van Melle 1978, 314).

The execution of production rules was assigned a value between –1 and +1 by MYCIN (dependent on how strongly the system felt about their correctness).

MYCIN's diagnosis also included these certainty elements, allowing the physician-user to make their own final decision.

The software package, known as EMYCIN, was released in 1976 and comprised an inference engine, user interface, and short-term memory.

It didn't have any information.

("E" stood for "Empty" at first, then "Essential.") Customers of EMYCIN were required to link their own knowledge base to the system.

Faced with high demand for EMYCIN packages and high interest in MOLGEN (Feigenbaum's third expert system), HPP decided to form IntelliCorp and TeKnowledge, the first two expert system firms.

TeKnowledge was eventually founded by a group of roughly twenty individuals, including all of the previous HPP students who had developed expert systems.

EMYCIN was and continues to be their most popular product.


~ Jai Krishna Ponnappan

Find Jai on Twitter | LinkedIn | Instagram


You may also want to read more about Artificial Intelligence here.



See also: 


Expert Systems; Knowledge Engineering


References & Further Reading:


Cendrowska, J., and M. A. Bramer. 1984. “A Rational Reconstruction of the MYCIN Consultation System.” International Journal of Man-Machine Studies 20 (March): 229–317.

Crevier, Daniel. 1993. AI: Tumultuous History of the Search for Artificial Intelligence. Princeton, NJ: Princeton University Press.

Feigenbaum, Edward. 2000. “Oral History.” Charles Babbage Institute, October 13, 2000. van Melle, William. 1978. “MYCIN: A Knowledge-based Consultation Program for Infectious Disease Diagnosis.” International Journal of Man-Machine Studies 10 (May): 313–22.







Artificial Intelligence - What Is The MOLGEN Expert System?

 



MOLGEN is an expert system that helped molecular scientists and geneticists plan studies between 1975 and 1980.

It was Edward Feigenbaum's Heuristic Programming Project (HPP) at Stanford University's third expert system (after DENDRAL and MYCIN).

MOLGEN, like MYCIN before it, attracted hundreds of users outside of Stanford.

MOLGEN was originally made accessible to artificial intelligence researchers, molecular biologists, and geneticists via time-sharing on the GENET network in the 1980s.

Feigenbaum founded IntelliCorp in the late 1980s to offer a stand-alone software version of MOLGEN.

Scientific advancements in chromosomes and genes sparked an information boom in the early 1970s.

In 1971, Stanford University scientist Paul Berg performed the first gene splicing studies.

Stanford geneticist Stanley Cohen and University of California at San Francisco biochemist Herbert Boyer succeeded in inserting recombinant DNA into an organ ism two years later; the host organism (a bacterium) subsequently spontaneously replicated the foreign rDNA structure in its progeny.

Because of these developments, Stanford molecular researcher Joshua Lederberg told Feigenbaum that now was the right time to construct an expert system in Lederberg's expertise of molecular biology.

(Lederberg and Feigenbaum previously collaborated on DENDRAL, the first expert system.) MOLGEN could accomplish for recombinant DNA research and genetic engineering what DENDRAL had done for mass spectrometry, the two agreed.

Both expert systems were created with developing scientific topics in mind.

This enabled MOL GEN (and DENRAL) to absorb the most up-to-date scientific information and contribute to the advancement of their respective fields.

Mark Stefik and Peter Friedland developed programs for MOLGEN as their thesis project at HPP, and Feigenbaum was the primary investigator.

MOLGEN was supposed to follow a "skeletal blueprint" (Friedland and Iwasaki 1985, 161).

MOLGEN prepared a new experiment in the manner of a human expert, beginning with a design approach that had previously proven effective for a comparable issue.

MOLGEN then made hierarchical, step-by-step changes to the plan.

The algorithm was able to choose the most promising new experiments because to the combination of skeleton blueprints and MOLGEN's enormous knowledge base in molecular biology.

MOLGEN contained 300 lab procedures and strategies, as well as current data on forty genes, phages, plasmids, and nucleic acid structures, by 1980.

Fried reich and Stefik presented MOLGEN with a set of algorithms based on the molecular biology knowledge of Stanford University's Douglas Brutlag, Larry Kedes, John Sninsky, and Rosalind Grymes.

SEQ (for nucleic acid sequence analysis), GA1 (for generating enzyme maps of DNA structures), and SAFE were among them (for selecting enzymes most suit able for gene excision).

Beginning in February 1980, MOLGEN was made available to the molecular biology community outside of Stanford.

Under an account named GENET, the system was linked to SUMEX AIM (Stanford University Medical Experimental Computer for Artificial Intelligence in Medicine).

GENET was able to swiftly locate hundreds of users around the United States.

Academic scholars, experts from commercial giants like Monsanto, and researchers from modest start-ups like Genentech were among the frequent visitors.

The National Institutes of Health (NIH), which was SUMEX AIM's primary supporter, finally concluded that business customers could not have unfettered access to cutting-edge technology produced with public funds.

Instead, the National Institutes of Health encouraged Feigenbaum, Brutlag, Kedes, and Friedland to form IntelliGenetics, a company that caters to business biotech customers.

IntelliGenetics created BIONET with the support of a $5.6 million NIH grant over five years to sell or rent MOLGEN and other GENET applications.

For a $400 yearly charge, 900 labs throughout the globe had access to BIONET by the end of the 1980s.

Companies who did not wish to put their data on BIONET might purchase a software package from IntelliGenetics.

Until the mid-1980s, when IntelliGenetics withdrew its genetics material and maintained solely its underlying Knowledge Engineering Environment, MOLGEN's software did not sell well as a stand-alone product (KEE).

IntelliGenetics' AI division, which marketed the new KEE shell, changed its name to IntelliCorp.

Two more public offerings followed, but growth finally slowed.

MOLGEN's shell's commercial success, according to Feigenbaum, was hampered by its LISP-language; although LISP was chosen by pioneering computer scientists working on mainframe computers, it did not inspire the same level of interest in the corporate minicomputer sector.


~ Jai Krishna Ponnappan

Find Jai on Twitter | LinkedIn | Instagram


You may also want to read more about Artificial Intelligence here.



~ Jai Krishna Ponnappan

Find Jai on Twitter | LinkedIn | Instagram


You may also want to read more about Artificial Intelligence here.



See also: 


DENDRAL; Expert Systems; Knowledge Engineering.


References & Further Reading:


Feigenbaum, Edward. 2000. Oral History. Minneapolis, MN: Charles Babbage Institute.

Friedland, Peter E., and Yumi Iwasaki. 1985. “The Concept and Implementation of  Skeletal Plans.” Journal of Automated Reasoning 1: 161–208.

Friedland, Peter E., and Laurence H. Kedes. 1985. “Discovering the Secrets of DNA.” Communications of the ACM 28 (November): 1164–85.

Lenoir, Timothy. 1998. “Shaping Biomedicine as an Information Science.” In Proceedings of the 1998 Conference on the History and Heritage of Science Information Systems, edited by Mary Ellen Bowden, Trudi Bellardo Hahn, and Robert V. Williams, 27–46. Pittsburgh, PA: Conference on the History and Heritage of Science Information Systems.

Watt, Peggy. 1984. “Biologists Map Genes On-Line.” InfoWorld 6, no. 19 (May 7): 43–45.







What Is Artificial General Intelligence?

Artificial General Intelligence (AGI) is defined as the software representation of generalized human cognitive capacities that enables the ...