TLDR Knowledge Map Current Thoughts

Arihant Gadgade
Date Published: 2/26/2025

KM is a project to have a map/graph of all human knowledge expressed in a visual format such that one can easily see the connections between subjects and more importantly so that we can have consensus driven ontologies for all of human knowledge.

However this project may not be the best utilization of my time as LLMs can do some of the functionality of KM. They have a representation of concepts in a high dimensional space. So I can search the LLMs representation space around the concept, to find related concepts; prerequisite topics, specific instances, etc. To a high degree this process emulates how humans would naturally be able to search thru the local area of the Knowledge Map I want to create.

The LLM method isn’t perfect because first, the metagraph isn’t accessible from the LLM method. Second, the visual/graphical representation is very useful for humans (this maybe can be mitigated by LLMs like ChatGPT using Python to create representations of topics in a graphical manner, imageGen capabilities, etc). And last, the voting system allows for the best ontologies to rise to the top. But AI is getting better, so these issues may resolve themselves to a degree.

However, I can always come back to this project.