← back

Wisdom-Driven Knowledge Augmented Generation at Scale - Chin Keong Lam, Patho AI

3.2K views · Aug 22, 2025 · 18:43 min · Watch on YouTube ↗
Takeaway

For expert-advisor AI systems, model 'wisdom' as a feedback loop over knowledge/experience/insight/situation in a knowledge graph rather than relying on flat RAG retrieval.

Summary

  • Patho.AI (NSF SBIR funded, started for drug discovery) builds 'Knowledge Augmented Generation' (KAG) — knowledge-graph-driven AI advisors, not just RAG retrievers
  • Wisdom graph nodes: wisdom (core) → decision making → situation, with feedback loops from knowledge, experience and insight into wisdom
  • For a competitive-analysis client, wisdom = strategy generator, knowledge = market data, experience = past campaigns, insight = industry DBs, situation = current product performance
  • Prototyped in n8n with AI Agent nodes orchestrating multiple model providers (OpenAI, Anthropic, on-prem) plus a centralized graph updated by specialist sub-agents (e.g. sentiment from social media)
  • Plan to graduate from n8n to LangChain or lighter-weight stacks; emphasizes that final decision quality depends on graph taxonomy more than the LLM choice
knowledge-graphskagn8n
Original description
The main thesis of the video is that by using a Wisdom-Driven Knowledge Graph, we can significantly enhance the quantitative analysis capabilities of Knowledge-Augmented Generation (KAG) systems. This allows for the creation of smarter AI systems that can not only retrieve information but also understand, reason, and provide expert-level advice. The talk argues that this approach surpasses traditional Retrieval-Augmented Generation (RAG) systems, which primarily rely on unstructured vector search.

00:00 Introduction to Patho AI and KAG
01:09 Defining Knowledge and Knowledge Graphs
01:56 KAG vs. RAG
02:37 The Wisdom-Decision Making-Situation Diagram
06:26 Practical Application: Competitive Analysis Chatbot
08:37 Implementation with N8n and Multi-Agent System
11:37 Why Use Knowledge Graphs over RAG
14:01 Challenges with Vector RAG and Numerical Reasoning
15:34 Building KAG Systems and Hybrid Models
16:45 Graph Extraction and Benchmarks
17:42 Conclusion and Resources

Recorded at the AI Engineer World's Fair in San Francisco. Stay up to date on our upcoming events and content by joining our newsletter here: https://www.ai.engineer/newsletter