Fusing Knowledge Graphs and Large Language Models

Large Language Models, KnowledgeGraphs, Hallucinations, LLMs, ChatGPT
Large Language Models have a major weakness when it comes to facts and their reliability. Knowledge graphs are great for managing facts. These two technologies complement each other and have the potential to address some of the weaknesses that are preventing businesses from adopting LLMs. How to properly fuse LLMs and knowledge graphs is an active area of research.
This webinar was originally streamed Thursday, August 10th on LinkedIn. During the live stream, there was a major audio problem. This video has been edited to address those issues. As a result of my editing, there are a few sudden transitions, but overall it's coherent.
If you are interested in this topic and wish to connect. Here is a link to my LinkedIn profile:
/ rudy-agovic-phd-a35b9b9
Link to GitHub repository with relevant references:
github.com/RMa...

[@LokeshSharma-hc6vg]

Excellent content Rudy! I have never commented on any video, but here - Please do post more content :-)

[@kushalneo]

17:49 LLMs as Knowledge Base 23:26 Knowledge Graphs 25:30 Use of Ontology in AI 27:31 Use of Knowledge Graph in Business 29:25 Model Journeys 32:26 Learning Over Graph 35:02 Lowest Hanging Fruit: Based On Practical Use case 50:03 Overview of Approaches 1:11:53 Knowledge Graphs enhanced by LLMs.

[@InPennyStock]

Would love an update on this video based on today’s developments and technologies!

[@andydataguy]

Hope you make more long vids like this! Keep sharing 😁 there are very few people talking about KG's

[@angstrom1058]

Post more, please, Rudy :)

[@josephmacdonald1255]

Excellent presentation which clearly delineates the need and benefits of integrating different AI tools to meet business and personal objectives,

[@InPennyStock]

Incredible!

[@mulderbm]

Excellent thanks for sharing

[@PatrickDunca]

I imagine you can keep adding more loops into this to form better guard rails. While I know nothing about this yet really, it seems like an answer generated by the LLM could be fed into a/another BERT LLM. The relations generated by this BERT could then be compared to the original knowledge graph again for consistency. Any inconsistencies could be fed back to a GPT model to form a new prompt. This prompt would ask for another iteration on the originally generated answer to fix the inconsistencies found. Very interesting video. Thank you!

[@va940]

Awesome content, one of the best videos

[@alimahmoudmansour9681]

Thank you. It is a very useful information

[@HelenJackson-pq4nm]

Great overview, thanks for posting