Knowledge Graphs - AI Assisted with Human Curation are a NECESSITY

[Alan Rodriguez]

How can we use our content in Confluence, PDFs and diagrams as source materials for an AI Knowledge Graph Assistant that finds connections and suggests them to the user for inclusion in an organizational knowledge graph?

This semantic foundation of the Confluence "knowledge" can then power advance AI reasoning:

https://ontologist.substack.com/p/of-wardley-maps-and-knowledge-graphs?r=1mytm&utm_campaign=post&utm_medium=web

https://www.linkedin.com/posts/kurtcagle_graphrag-a-new-approach-for-discovery-using-activity-7167153672655757312-YgN0?utm_source=share&utm_medium=member_ios

It can also provide advanced SEARCH using THINGS instead of STRINGS

The above ARE MUST HAVE FEATURES for organizations harnessing organizational knowledge to drive AI business functions. If I can't find a solution for this with Confluence, I will be forced to use an alternative.

[Nic Brough -Adaptavist-]

Welcome to the Atlassian Community!

At the moment, generalised AI such as LLMs simply are not capable.  They tell people what they want to hear, rather than the truth of the matter.  They're wrong, and we need to stop that.

AI is a long way off being a "must have" in specialised systems.  LLMs are better than Google, by most measures, but the fact is that for now, the primary "must have" is actually "block AI posts automatically".

My estimate is that about 10% of the writing I do on the Community here is now correcting ****ing awful answers posted by 'bots (or script-kidddies looking for anti-spam points using LLM AIs)

LLM AIs do not work for specialised systems.  Expert AIs can, but most humans (including you and I) do not know when best to use, or refuse to use, them to answer the question.

[Alan Rodriguez]

Well thank you for that useless response Nic - it does not answer my question or address my needs.

LLMs CAN assist users in creating knowledge graphs, and knowledge graphs CAN be used to augment LLM reasoning. Learn more about knowledge graphs and how they  augment AI system with human expertise.

Perhaps you should read the two articles that discuss these topics in my original post?

[Nic Brough -Adaptavist-]

Hi Alan,

Yes, my answer does not answer your question, it is aimed at a higher level - it was to start a discussion about why you think you want this.  

I've read both articles and neither of them address the problem that LLM AIs tend to be biased, can only work with very common cases, and usually tell people what they want to hear, rather than the truth.

What you're looking for is an expert system AI, not an LLM.  There are expert systems around, but none that are trained for your data set.

[Alan Rodriguez]

Nic - you clearly do not understand how knowledge graphs augment LLMs and resolve many of your "issues" you keep repeating. Issues that are highly uninformed and biased.

Nor are you understanding how an organization could use their Confluence knowledge base to generate knowledge graphs.

Nor are you aware of how one can use LLMs to assist in creating knowledge graphs from a knowledge base like Confluence.

Please stop engaging with me and this topic thread.

[Nic Brough -Adaptavist-]

It's the opposite.  I don't think you understand how LLMs get knowledge graphs so howlingly wrong.

Nor are you understanding how an organisation needs to treat AI analyses as suspect until checked properly by a human who actually understands what they are looking at.

I'm not going to stop telling you the truth.

[Alan Rodriguez]

You're an unhelpful idiot...

I'll be ditching Confluence ASAP