Cyc and LLMs: Summary and Discussion
Key Ideas
- Cyc is symbolic rather than neuromorphic. Combining it with an LLM is a
form of neuro-symbolic AI.
- Cyc represents explicit knowledge in CycL and organizes that knowledge
with an ontology and contextual microtheories.
- Its inference machinery answers queries by combining facts, rules,
inheritance relations, contexts, and specialized reasoning methods.
- An LLM can translate language, propose candidate meanings, and explain
results; Cyc can validate representations and supply auditable inference.
- The approach offers precision, context, and explainability, but must pay
the costs of knowledge engineering, symbol grounding, and logical search.
Sources and Further Reading
- Cycorp: Cyc Platform
- Cyc Technology Overview
- Cycorp Glossary (CycL, assertions, microtheories, and inference)
- Douglas B. Lenat, Mayank Prakash, and Mary Shepherd,
“CYC: Using Common Sense Knowledge to Overcome Brittleness and Knowledge Acquisition Bottlenecks”, 1985.
- Douglas B. Lenat and Gary Marcus,
“Getting from Generative AI to Trustworthy AI: What LLMs Might Learn from Cyc”, 2023.
- Abhishek Sharma and Keith M. Goolsbey,
“Simulation-Based Approach to Efficient Commonsense Reasoning in Very Large Knowledge Bases”, 2019.