Knowledge graphs and large language models in cognitive architectures

Ontology Summit 2024 Fall Series with Anatoly Levenchuk, Arun Majumdar and John Sowa - YouTube

Slides: https://bit.ly/3sljmXt

The talk will transfer the style of the definitions for knowledge graphs (KG), their architectures and their systems, taken from Ontology Summit Communiqué 2020, for large language models (LLM). The talk proposes a framework for cognitive architectures based on the use of LLM and KG for joint use in the Popperian evolution of knowledge during 4E (embodied, extended, embedded, enacted) cognition. In this framework ontologies are understood as answers to the question “what is in the world” and can be found in representations with a spectrum of formality/rigor. An example is given of the use of ontology engineering training in management, where upper-level ontologies are given to students in the form of informal course texts (expected to result from fine-tuned LLMs in the “wet” neural networks in the heads of human students) and lower-level ontologies that are more formal and developed as data schemas for databases as well as knowledge graphs. Types (taken from a student’s LLM in their head) in the upper ontology govern types of the middle ontology and the data that resides in corporate DBMS systems.