Post
382
I wrote about the Event System Model (ESM), a symbolic AI approach that replaces token prediction with a simulate → validate → commit/repair cycle against a "lawbook" of invariants.
Key ideas:
No gradients: Learns by promoting successful repair sequences into new programs.
Explainable by construction: Produces a hash-chained ledger of every state change.
Runtime corrigibility: Fixes violations on the fly under a defined resource budget.
It’s an attempt to build a more deterministic and auditable AI for systems where trust is non-negotiable. The full architecture and theory are in my latest blog post. Would love to hear your thoughts.
Link: https://huggingface.co/blog/dimentox/esm
#AI #SymbolicAI #Programming #SysArch #ExplainableAI
Key ideas:
No gradients: Learns by promoting successful repair sequences into new programs.
Explainable by construction: Produces a hash-chained ledger of every state change.
Runtime corrigibility: Fixes violations on the fly under a defined resource budget.
It’s an attempt to build a more deterministic and auditable AI for systems where trust is non-negotiable. The full architecture and theory are in my latest blog post. Would love to hear your thoughts.
Link: https://huggingface.co/blog/dimentox/esm
#AI #SymbolicAI #Programming #SysArch #ExplainableAI