GraphGeeks Podcast Practical Trends in Graphs and AI with Paco Nathan
25 snips
Sep 24, 2025 Join Paco Nathan, an experienced AI and graph consultant, as he shares insights on the latest trends in AI and graph technology. He discusses how companies are shifting from basic code generation to real-world applications like fault detection. Paco elaborates on the rise of hybrid AI, combining neural and symbolic systems for better model efficiency. He also highlights practical workflows with tools like MLflow and DSPy, and touches on the growing interest in probabilistic graph approaches. A must-listen for anyone curious about the future of tech!
AI Snips
Chapters
Transcript
Episode notes
Neurosymbolic Methods Are Returning
- Neurosymbolic and probabilistic approaches are resurging to combine symbolic reasoning with neural models for explainability and efficiency.
- These hybrid methods can constrain LLM outputs with external symbolic systems instead of expensive internal guardrails.
Constrain LLM Outputs With Probabilistic Models
- Use simple probabilistic models to restrict LLM outputs to a domain-limited set when possible.
- This yields faster, more explainable, and cost-effective reasoning than brute-force neural computation.
Graph RAG Improves Recall And Semantics
- Graph RAG blends vector retrieval with symbolic graph traversal to improve recall and semantic relevance.
- Adding graphs helps find multi-hop semantic relations that pure vector RAG often misses.