A Clojure fusion of symbolic and data driven AI
summary
Claims a REP DSL embedded in Clojure plus a hybrid architecture — symbolic “bones” and data‑driven “flesh” — gives the fastest path to usable chatbots. Concrete stack: Clojure, Postgres for configs, a state‑machine compiler that turns scripts into per‑session state machines, and TensorFlow/Kafka for ML tasks. Uses sentence embeddings and relevance models for pattern matching. Deployed as Juji bots for surveys, recruiting and reception; studies show higher completion and more informative responses than form surveys.
AI-written summary · about
chapters (15)
- 0:00Introduction to AI chatbots
- 2:23Practical AI applications
- 4:45Hybrid AI approach
- 7:23Perception vs intelligence
- 10:45Symbolic and sub-symbolic AI
- 13:27Bringing back symbolic AI
- 15:49Limitations of data-driven AI
- 18:10Integrating AI approaches
- 22:04System architecture
- 25:07Topic-based chatbot language
- 28:11Pattern matching in AI
- 30:42Machine learning integration
- 34:01Rules and similarity scores
- 37:49Conversation flow management
- 40:19Conclusion
related talks
27mReal World AI Integration
44mData Science in Clojure
36mLearning to Talk to Machines with Speech Acts
29mDeep Learning Needs Clojure
48mChatbots in Scala: meet your new robot overlords
59mMaking the most of artificial and human intelligence for data science
28mMachine Learning Live
20mClojure Tools for Symbolic Artificial Intelligence