Today we travel back to 1972 and run MYCIN, one of the first expert systems, on the Pentium 1.
MYCIN was developed at Stanford University in the early 1970s to diagnose bacterial infections and recommend antibiotics. In 1979, it achieved a 90.9% agreement rate with Stanford experts.
The system was built around roughly 450 production rules, each representing clinical decision criteria from infectious disease experts.
So I got the code, installed CLISP on the Pentium 1 and started the consultation.
There was one small compatibility issue: Peter Norvig's implementation uses two functions not defined in modern CLISP, requires and rest2. A short shim file fixed that and the inference engine came to life.
The version available today is only a fragment of the original MYCIN. The implementation from Peter Norvig's book "Paradigms of AI Programming" contains the inference engine and just 6 rules for identifying organisms. The remaining rules for therapy, dosage and patient-specific recommendations are gone.
The original MYCIN was written in Interlisp for a DEC PDP-10 running WAITS. It was never released as open source. What survives today is essentially the shell of the system, not the knowledge it contained.
MYCIN is a monument to AI history. The way its source code was handled is both a memorial and a warning.
ELIZA (1966) survived because Weizenbaum published the code. CLIPS (1985) survived because NASA released it as open source. MYCIN (1972) did not.
Publicly funded research. Publicly funded code. Never released to the public.
This is why publicly funded research should be open source. Always!
#MYCIN #AIHistory #ExpertSystems #ArtificialIntelligence #AI #Stanford #Lisp #CLISP #NetBSD #Pentium1 #Pentium #Pentium166 #RetroComputing #vintagecomputing #OpenSource #FOSS #OpenScience #PubliclyFundedResearch #ComputerHistory #Science #Research #AcademicChatter
