Machine Learning + Symbolic Reasoning
summary
A Quarkus web app wires an LLM to a Drools rule engine using Quarkus + langchain4j + drools and demonstrates an airline refund flow: the LLM extracts customer and flight data, Drools computes refunds at $2 per minute +10% over 65, expecting 240 minutes → $528, but a local Mistral 7B model hallucinates (injects a comment into JSON) and breaks the demo. A separate pipeline auto-generates rules from data via a Revised OneR algorithm, reducing errors from 28 to 17.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:12Combining ML with symbolic reasoning
- 4:26What is symbolic artificial intelligence
- 7:00LLM and rule engine example
- 12:22Live hybrid AI demo
- 23:40Hybrid AI use cases
- 25:07Rule engines vs ML
- 26:32Hybrid AI approach with Drools and machine learning
- 30:44Automating rule creation
- 32:52Refining rules for accuracy
- 33:13Conclusion
related talks
48mAgentic AI with Quarkus, LangChain4j and vLLM
47mQuarkus and AI Integrating Java and LLM to build intelligent applications
32mScaling Generative AI: Building Production-Ready LLM Applications
01h 01mScaling AI with Java: Building Production-Ready LLM Applications
49mPushing LLMs over the Edge: Exploring the Limits of the Possible
42mBuilding reporting services using Quarkus and AI
02h 37mDeveloping Cloud-Native Java AI applications with DJL and LangChain4J
24mBuilding Real-Time AI Decision Systems with LLMs