# Neuro-Symbolic AI **Domain:** Artificial Intelligence / Knowledge Representation / Machine Learning **Doc Type:** Canonical Architecture Node **Maturity:** Developed **Related:** [[wiki/GOFAI (Good Old-Fashioned AI)|GOFAI (Good Old-Fashioned AI)]], [[wiki/Machine Learning|Machine Learning]], [[wiki/Knowledge Representation|Knowledge Representation]], [[wiki/Word Embeddings|Word Embeddings]], [[wiki/Symbolic Language Engine|Symbolic Language Engine]] --- ## Definition **Neuro-Symbolic AI combines learned neural representations with explicit symbols, rules, constraints or knowledge structures.** The goal is to join adaptability and graded generalization with compositional structure, auditability and controlled reasoning. ## Symbolic Language Engine Bridge The [[wiki/Symbolic Language Engine|Symbolic Language Engine]] already separated phonological, semantic, taxonomic, associative and formal layers. Neural pronunciation models, embeddings and local language models could extend coverage and candidate generation without replacing those structures. The useful hybrid is not a neural system wearing a symbolic interface. Each layer should retain operational authority where it is strongest: learned models for expansion and flexible realization; symbolic indexes for explicit relation, constraint checking and provenance. ## Governance Context Hybrid systems can be more legible only if they expose which layer produced a result. Otherwise symbolic rules may provide a false appearance of explanation around opaque model behavior. ## Key Insight **The point of a hybrid is not to average two paradigms. It is to preserve their different epistemic strengths and failure modes.** ## See Also [[wiki/GOFAI (Good Old-Fashioned AI)|GOFAI (Good Old-Fashioned AI)]], [[wiki/Machine Learning|Machine Learning]], [[wiki/Knowledge Representation|Knowledge Representation]], [[wiki/Word Embeddings|Word Embeddings]], [[wiki/Provenance-Sensitive Multiplicity|Provenance-Sensitive Multiplicity]], [[wiki/Semantic-First Constrained Generation|Semantic-First Constrained Generation]] ## Sources / Provenance - Primary project account: [[projects/Ten Years Building a Symbolic Language Engine|Ten Years Building a Symbolic Language Engine]]. - Historical work combining connectionist learning and symbolic reasoning. - Contemporary neuro-symbolic AI research.