Artificial intelligence with Elixir the Rete (...)
summary
Implements the Rete algorithm in Elixir/BEAM to compile rules into alpha/beta networks, using graph libs (libgruff/libgraph), protocols and map-reduce state to avoid mutation; ships a tiny DSL (Sanskrit) and an engine called narrowbridge that runs a dog-sickness expert system (rules → working memory → forward-chaining activations). Claims production-feasible performance: hundreds–thousands of rules are tiny, a cited worst case ~100k rules ≈ 300MB, and the implementation is functional, interactive, open-source and working but still early-stage.
AI-written summary · about
chapters (12)
- 0:00Introduction to AI rules
- 1:03Expert system components
- 2:42Pattern matching in Rete
- 4:20Implementing facts and rules
- 5:57Developing a custom DSL
- 7:16Alpha network compilation
- 8:47Beta network connections
- 10:15Activating the rule engine
- 12:50Elixir benefits and protocols
- 16:03Using the Narrow Bridge lib
- 18:01Limitations and use cases
- 19:17Q&A and technical challenges
related talks
25mApplied Machine Learning in Elixir
01h 00mThe Promise and Limitations of AI
42mReasonable AI
41mHow to Elm-ify Your ML
43mAI Systems: Foundations for Artificial Minds
01h 30mTrustworthy AI & Open Source
20mClojure Tools for Symbolic Artificial Intelligence
47mLivebook & Elixir: Where AI, Web & Concurrency Meet