The founders of TrustGraph, Daniel Davis and Mark Adams, discuss their journeys with big data, knowledge graphs, and data engineering. Knowledge graphs are hard to learn - no matter what Mark says, and he gives everyone a crash course on them, why querying graphs is tricky, and what makes for reliable data services. The conversation ends with a discussion of what makes for "explainable AI" and the future of AI security. Topics: 0:00:00 Introductions 0:03:25 Mark's background 0:06:23 Are Knowledge Graph's more popular in Europe? 0:08:27 Past data engineering lessons learned 0:17:15 Knowledge Graphs aren't new 0:22:42 Knowledge Graph types and do they matter? 0:27:10 The case for and against Knowledge Graph ontologies 0:39:40 The basics of Knowledge Graph queries 0:45:42 Knowledge about Knowledge Graphs is tribal 0:47:50 Why are Knowledge Graphs all of a sudden relevant with AI? 0:53:45 Some LLMs understand Knowledge Graphs better than others 0:58:30 What is scalable and reliable infrastructure? 1:01:45 What does "production grade" mean? 1:04:45 What is Pub/Sub? 1:09:40 Agentic architectures 1:12:17 Autonomous system operation and reliability 1:16:50 Simplifying complexity 1:19:48 A new paradigm for system control flow 1:23:45 Agentic systems are "black boxes" to the user 1:24:55 Explainability in agentic systems 1:30:05 The human relationship with agentic systems 1:32:00 What does cybersecurity look like for an agentic system? 1:35:30 Prompt injection is the new SQL injection 1:37:00 Explainability and cybersecurity detection 1:39:40 Systems engineering for agentic architectures is just beginning 🔗 TrustGraph Links: ➡️ GitHub: https://github.com/trustgraph-ai/trustgraph ➡️ TrustGraph Config UI: https://config-ui.demo.trustgraph.ai/ ➡️ Website: https://trustgraph.ai/ ➡️ Discord: https://discord.gg/sQMwkRz5GX ➡️ Blog: https://blog.trustgraph.ai ➡️ LinkedIn: https://www.linkedin.com/company/trustgraph/
Daniel Davis of TrustGraph and Kirk Marple from Graphlit discuss the 2024 state of RAG. Whether it's RAG, GraphRAG, or HybridRAG, a lot has changed since the term has become ubiquitous in AI. Where are we, where are we going, and where should be going are all answered in this discussion. 00:00 00:20 Introductions 04:10 The Term "RAG" Itself 06:20 Long Context Windows 08:10 Claude 3.5 Haiku 11:20 LLM Pricing Variance 14:11 What Happened to Claude 3 Opus? 19:03 AI Maturity 23:22 What is AGI? 26:40 Entity Extraction with LLMs 32:18 RDF? Cypher? Something else? 36:36 Why so many new GraphDBs and VectorDBs? 42:23 Reinventing the Wheel 42:48 "You Don't Need LangChain" 44:20 How to Promote Emerging Projects 46:53 "Hype Matters" 49:15 Where is RAG 1 Year from Now 54:09 Should AI Model Itself on Human Cognition? 58:45 The DARPA MUC AI Conferences 🔗 Graphlit Links: ➡️ Website: https://graphlit.com 🔗 Kirk's Links: ➡️ Twitter: https://x.com/kirkmarple 🔗 Daniel's Links: ➡️ Twitter: https://x.com/trustspooky 🔗 TrustGraph Links: ➡️ GitHub: https://github.com/trustgraph-ai/trustgraph ➡️ TrustGraph Config UI: https://config-ui.demo.trustgraph.ai/ ➡️ Website: https://trustgraph.ai/ ➡️ Discord: https://discord.gg/sQMwkRz5GX ➡️ Blog: https://blog.trustgraph.ai ➡️ LinkedIn: https://www.linkedin.com/company/trustgraph/