Worked Example: Cyc as the Logic Component

Question: Does Dana require medical training?

Knowledge:

  1. Dana is a nurse.
    (#$isa #$Dana #$Nurse)

  2. Every nurse is a medical professional.
    (#$genls #$Nurse #$MedicalProfessional)

  3. Every medical professional requires medical training.
    (#$implies
      (#$isa ?PERSON #$MedicalProfessional)
      (#$requiresTraining ?PERSON #$MedicalTraining))

Query:

(#$requiresTraining #$Dana #$MedicalTraining)

Inference Trace

StepReasoningResult
1Dana is an instance of #$Nurse.(#$isa #$Dana #$Nurse)
2#$Nurse is a specialization of #$MedicalProfessional.(#$isa #$Dana #$MedicalProfessional)
3Apply the training rule with ?PERSON/#$Dana.(#$requiresTraining #$Dana #$MedicalTraining)

Answer: Yes.

This is a small example, but it illustrates the central idea. The conclusion was not stored as a separate fact about Dana. It was derived by combining general knowledge with a fact about a particular individual.

Where an LLM Helps

An LLM can recognize that the user's phrase “needs medical training” corresponds to the predicate #$requiresTraining, identify Dana, form the query, and turn the proof trace back into fluent English. Cyc, rather than the LLM, checks whether the conclusion follows from the selected axioms.