How Graph RAG and rules-based AI can power expert chatbots and other applications

[@csmac3144a]

Interesting experiment but as an enterprise architect I'd perhaps wait for a bit on the "enterprise ready" designation.

[@phenanrithe]

This demo perfectly illustrates why RAG won't improve the situation much, disregarding the fact it has higher requirements. As long as the experience is driven by an LLM that "blurts out" the first answer that comes to mind, without iterating and fact-checking, we can't hope to get any reliability. Here, we saw that 1) it forgot or mixed up recipes when asked to show the instructions and 2) it forgot the dairy requirement in its request. We don't even know if the recipe happened to be correct by chance or not, but we know for sure that 82 isn't the correct answer - maybe try 82, 82, 82?

[@bastabey2652]

twenty years of work in traditional data warehouse and business intelligence.. despite all its shortcoming, what modern LLM does already is amazing.. RAG is a cool tech to leverage the power of LLM.. that hiccup in a live demo is not unusual.. hallucinations can be grounded with source material...

[@bastabey2652]

what is the difference between synonym and ontology? kzbin.info/www/bejne/iZDGhJiHo6hqps0

[@limjuroy7078]

Is Cen Xi Toh from Malaysia? Her accent sounds very Malaysian?