- good explanation on fundamental difference between knowledge graphs and llms
- unifying large language models and knowledge graphs: a roadmap | property | knowledge graphs | large language models | |---------------------------|--------------------------|--------------------------| | representation | structured: triples | unstructured: text | | knowledge storage | explicit | implicit | | interpretability | high | low | | generalizability | low | high | | factual accuracy | high | low | | reasoning | symbolic and probabilistic | probabilistic | | knowledge update | easy | hard | | decisiveness | high | low | | unseen facts | no | yes | | incompleteness | high | low |
- TODO create visualization
- insights are the following that knowledge graphs and llms are
- fundamentally different
- must interact in any intelligent system
- key to superintelligence
- go on to dive into
- cybergraph and tru
- as beautiful counterpart probabilistic model for llm
- in a modern soft3 stack