Status and Resources

help edit space_dashboard

System Interfaces

  • PLN: MetaMo biases inference search toward contextually appropriate goals
  • SubRep: MetaMo defines what the system cares about; SubRep validates which skills serve those motives. Together they form PRIMUS's goal management layer.
  • ECAN: MetaMo's goal priorities influence attention allocation, directing STI toward goal-relevant atoms
  • TransWeave: MetaMo's motive signals guide transfer decisions across cognitive paradigms

Implementation Anchors (cluster-pilot trilateral classification, Source 2 close 2026-05-06)

The Non-clustered HAA cluster pilot Source 2 reconciliation locked a trilateral classification across the MetaMo-related repositories. Wiki text describing MetaMo implementations should distinguish formal-reference, heuristic-prototype, and predecessor-substrate roles rather than collapsing them as a single implementation:

  • Primary formal reference: iCog-Labs-Dev/MetaMo-Python (Python; HEAD 3ec409b; ~33 commits, primary committer Nahom Senay). Compact reference implementation of the MetaMo categorical skeleton: MotivationalState (\(X = G \times M\)), AppraisalComonad + DecisionMonad ABCs, MetaMoPseudoBimonad composition, lax-distributive-law numeric check, parallel_merge, contractive update law, homeostatic damping, blend-states, and a research_assistant.py interactive demo with OpenPsiAppraisal + MagusDecision + TranslationFunctor. [IMPLEMENTATION-BACKED-CORE] at the skeleton level; [FORMAL-LAWS-PAPER-ONLY] for categorical proofs; no test suite per AGENTS.md:23. See Core Mechanisms and Formalism for file:line citations.
  • Heuristic prototype: glicerico/MetaMo-Prototype (Python; ~24 commits, forked from dagemawinegash/MetaMo-Prototype, primary committer Dagemawi Bekele). Active LangGraph/LLM rewrite with rich goal/modulator dynamics in engine.py. No formal categorical structure; targeted searches at HEAD for "bimonad" / "comonad" / "MAGUS" / "OpenPsi" return zero matches. Useful as a heuristic routing engine; NOT a duplicate of MetaMo-Python and NOT the categorical implementation. [HEURISTIC-PROTOTYPE].
  • Predecessor OpenPsi substrate: iCog-Labs-Dev/hyperon-openpsi (MeTTa/Python; HEAD 3b356c5; ~575 commits). Mature MeTTa OpenPsi implementation with traditional MindAgent + Demand/Modulator patterns and use cases (ping-pong, curious-agent with LLM integration, qwestor web API). NOT the primary MetaMo implementation at HEAD — targeted searches for "MetaMo" / "MAGUS" / "individuation" / "transcendence" return zero matches. Recent OpenPsi paper-alignment work (commit 01936f1, 2026-02-24) is OpenPsi-side, not MetaMo-side. [OPENPSI-PREDECESSOR-SUBSTRATE]; the operational substrate that MetaMo's appraisal comonad abstracts at one level higher.
  • Prolog-port variant: iCog-Labs-Dev/PeTTa-OpenPsi — Prolog-based MeTTa port with A* and Thompson sampling planners; NARS-based trading agent use case. Operational variant of the OpenPsi substrate, separate from the trilateral above.

Current Status

  • Operational: hyperon-openpsi (MeTTa OpenPsi substrate, with use cases); MetaMo-Python (Python reference implementation, demo-grade); MetaMo-Prototype (heuristic engine, separate paradigm); PeTTa-OpenPsi (Prolog port variant)
  • Under development: Formal pseudo-bimonad proofs; MeTTa-native MetaMo integration with AtomSpace and PLN; production Hyperon runtime binding
  • Proposed: Multi-agent coordination via MetaMo (Reciprocal State Simulation beyond identity-matrix demo); verification methods for motivational stability; benchmarking against alternative AI motivation approaches

Open Problems / Research Directions

  • Formalizing and proving stability via contractive updates in the full pseudo-bimonad structure (categorical proofs, not just numeric distortion checks)
  • Embedding MetaMo into Hyperon AtomSpace for motivation-guided PLN inference
  • Multi-agent MetaMo — coordinating motivational state via reciprocal state simulation with non-trivial translation functors
  • Blending dynamics between MetaMo's motive geometries and SubRep's option certificates
  • Temporal goal reasoning — implementing PredictiveImplication-style temporal windows (inherited gap from OpenPsi 2016 review)
  • MetaMo-side test suite (currently absent per AGENTS.md:23 in MetaMo-Python)

Primary Sources

  • Goertzel, B. and Lian, R. (2025). MetaMo: A Robust Motivational Framework for Open-Ended AGI. AGI 2025. RawData.
  • Goertzel, B. and Lian, R. (2025). Embodying Abstract Motivational Principles in Concrete AGI Systems. AGI 2025.
  • Goertzel, B. (2025). Hyperon for AGI⇒ASI Whitepaper, §5.10: MetaMo.
  • Mikeda (2024). MAGUS. (External decision-theory framework cited in AGI 2025 paper-2 as the decision-monad component of the MetaMo bimonad; see Historical Lineage for the boundary statement.)
  • Cai, Z., Goertzel, B., and Geisweiller, N. (2011). OpenPsi. AGI 2011. RawData.
  • See also: BBM Publication Map for the CogPrime→Hyperon correspondence.

Cluster-pilot extraction archive: scripts/archive/non_clustered_haa_pilot/source2_metamo_paper/ (findings_codex.txt + findings_gemini.txt + findings_reconciled_crossmodel.txt). V2-1..V2-7 carry-forwards locked.