PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
June 17, 20260 citationsOpen Access

Deep Structural Theory: Formalizing AGI Selfhood and Reconstructing Post‑LLM Systems via R/C/D Dynamics

View Full Paper
SNShintaro Nakamura

Key Points

  • This work aims to formalize intelligence and autonomy through a new framework called Deep Structural Theory (DST).
  • Defines minimal organizational units R (structured reality), C (conceptual structure), and D (dynamic patterns) for AI systems.
  • Classifies the coupling forms of these units into eight structural types (S1-S8).
  • Utilizes topological analysis to investigate limitations of current LLMs and introduces AI Structural Safety Principle.
  • Demonstrated that S5 (cyclic structure) meets stability conditions for self-referential autonomy (Selfhood).
  • Identified structural limitations in LLMs including high energy consumption and hallucination issues.
  • Introduced controls for R/C/D coupling that enhance structural safety without external monitoring.

Abstract

Conventional AI research and psychology have lacked explicit structural variables capable of formally describing intelligence, consciousness, or autonomy. This paper presents Deep Structural Theory (DST), which defines three minimal units of system organization—R (structured reality), C (conceptual structure), and D (dynamic patterns)—and classifies their coupling forms into eight fundamental structural types (S1–S8). Using this framework, we provide a topological analysis of the structural causes underlying the limitations of current LLMs, including their dependence on scaling, high energy consumption, and susceptibility to hallucination. We further demonstrate mathematically that S5 (cyclic structure) satisfies the stability conditions for autonomous self‑reference (Selfhood). Finally, we introduce the AI Structural Safety Principle, which controls R/C/D coupling at the OS level and enables the possibility of structural safety without reliance on external monitoring.Japanese 従来の人工知能研究および心理学には、知能・意識・自律性を形式的に記述するための「明示的な構造変数」が欠如していた。 本論文は、システムを構成する最小単位としてR(構造化現実)・C(概念構造)・D(動的パターン)を定義し、 それらの結合形式を 8 種の基本構造(S1–S8)として分類する「深層構造理論(DST)」を提示する。 本枠組みにより、現行の LLM が抱えるスケーリング依存・高エネルギー消費・ハルシネーションの構造的原因をトポロジー的に分析し、 S5(循環構造)が自律的自己参照(Selfhood)の安定条件を満たすことを数学的に示す。 さらに、R/C/D の結合を OS レベルで制御する「AI 構造安全原則」を導入し、外生的監視に依存しない構造的安全性の可能性を論じる。

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Shintaro Nakamura (2026) studied this question.

synapsesocial.com/papers/6a323bbcd50b63ecad2062adhttps://doi.org/10.5281/zenodo.20694515
Ask AI
Helpful
Bookmark
Share
View Full Paper