Skip to content

KarolFilipKowalczyk/Consciousness

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

24 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Consciousness as Collapsed Computational Time

A Unified Computational Theory Integrating Finite Machine Hierarchies, Temporal Phenomenology, and Non-Uniform Resource Allocation

Author: Karol Kowalczyk
Date: November 2025
Status: Preprint

DOI License: CC BY 4.0


Repository Overview

This repository contains a comprehensive research framework exploring consciousness through computational theory. The work consists of a foundational 236-page paper plus several companion papers that extend and apply the core theory to specific domains.

Repository Structure

.
├── README.md                              # This file
├── main.tex                               # Main paper entry point (236 pages)
├── preamble.tex                           # LaTeX configuration and styling
├── bibliography.bib                       # Complete bibliography (60+ references)
│
├── part1_foundations.tex                  # Part I: Computational foundations
├── part2_existing_theories.tex            # Part II: Integration with existing theories
├── part3_temporal_revolution.tex          # Part III: Temporal collapse mechanism
├── part4_mechanisms.tex                   # Part IV: Neural implementation
├── part5_hard_problem.tex                 # Part V: Solution to hard problem
├── part6_empirical.tex                    # Part VI: Testable predictions
├── part7_philosophical.tex                # Part VII: Philosophical implications
├── part8_synthesis.tex                    # Part VIII: Complete synthesis
│
├── appendix_comparisons.tex               # Appendix A: Theory comparisons
├── appendix_glossary.tex                  # Appendix B: Terminology
├── appendix_mathematical.tex              # Appendix C: Formal proofs
├── appendix_protocols.tex                 # Appendix D: Experimental protocols
├── appendix_selector.tex                  # Appendix E: Selector implementation
├── appendix_toy_model.tex                 # Appendix F: Toy model examples
│
├── 3papers/                               # Extended theoretical papers
│   ├── adjoint_projections.tex            # Adjoint functors in hierarchies
│   ├── quantum_measurement.tex            # Quantum measurement framework
│   └── [other papers]
│
└── selectors/                             # Practical applications
    └── Sleep-Wake Orchestration in Hierarchical LLM Cohorts.tex

Main Paper: Consciousness as Collapsed Computational Time

Length: 236 pages
Entry point: main.tex
PDF: Available on Zenodo

Abstract

This work presents a unified computational theory of consciousness that integrates insights from Integrated Information Theory (IIT), Global Workspace Theory (GWT), Attention Schema Theory (AST), and quantum consciousness approaches within a novel framework based on non-uniform finite machine hierarchies.

Core Innovation: Consciousness emerges from the collapse of parallel computational explorations across a hierarchy of finite-state machines with exponentially growing resources (M₁, M₂, M₃, ..., Mₙ where Mₙ has 2ⁿ bits of memory). The hierarchy exhibits a generalized non-uniform structure where effective levels follow M_n ⊆ M_{n+f(n)} with variable resource gaps f(n) ≥ 1, representing realistic cognitive resource allocation.

Key Insight: While real computational time includes all parallel attempts, backtracks, and time-reversals, subjective conscious experience perceives only the successful computational path—a smooth, continuous temporal flow that never experiences the failed branches.

Main Paper Structure

Part I: Foundations (part1_foundations.tex)

  • Machine hierarchy with entropic scaling: I(n) = κ n log n
  • Non-uniform structure: M_n ⊆ M_{n+f(n)}
  • Selector mechanism and non-computability
  • Resource allocation with entropic costs

Part II: Existing Theories (part2_existing_theories.tex)

  • Detailed comparison with IIT, GWT, AST, Orch-OR
  • Integration strategy showing how each theory captures one aspect
  • Comparative analysis of strengths and weaknesses

Part III: Temporal Revolution (part3_temporal_revolution.tex)

  • Computational vs. subjective time distinction
  • Parallel exploration and selective memory
  • State checkpointing mechanism
  • Temporal phenomenology explanation

Part IV: Mechanisms (part4_mechanisms.tex)

  • Neural implementation considerations
  • Cognitive architecture mapping
  • Resource allocation patterns
  • Biological realization

Part V: Hard Problem (part5_hard_problem.tex)

  • Identity solution vs. production solution
  • Dissolution of explanatory gap
  • Phenomenal character emergence
  • Why there is "something it is like"

Part VI: Empirical Predictions (part6_empirical.tex)

  • Neural signatures of hierarchy levels
  • Capacity measurements (realistic: M₁₅ to M₃₅)
  • Temporal integration windows
  • Testable experimental protocols

Part VII: Philosophical Implications (part7_philosophical.tex)

  • Free will from non-computable selector
  • Personal identity as selector continuity
  • Qualia as intrinsic collapse patterns
  • Consciousness without production relation

Part VIII: Synthesis (part8_synthesis.tex)

  • Complete integration of all components
  • Cognitive phenomenology of variable gaps
  • Individual differences via gap patterns
  • Unified explanatory framework

Appendices

Appendix A: Theory Comparisons (appendix_comparisons.tex)

  • Point-by-point comparison with major theories
  • Evaluation framework
  • Integration strategies

Appendix B: Glossary (appendix_glossary.tex)

  • Technical terminology
  • Mathematical notation
  • Conceptual definitions

Appendix C: Mathematical Formalization (appendix_mathematical.tex)

  • Formal definitions and theorems
  • Detailed proofs
  • Complexity analysis

Appendix D: Experimental Protocols (appendix_protocols.tex)

  • Neuroscience experiments
  • Behavioral studies
  • Implementation guidelines

Appendix E: Selector Implementation (appendix_selector.tex)

  • Python implementation of selector mechanism
  • Practical demonstrations
  • Code examples

Appendix F: Toy Model (appendix_toy_model.tex)

  • Simplified examples
  • Pedagogical demonstrations
  • Conceptual illustrations

Extended Papers (3papers/)

1. Adjoint Projections on Computational Hierarchies

File: 3papers/adjoint_projections.tex

Abstract: Develops the categorical structure of hierarchical computation using adjoint functors. Introduces compression/reconstruction as genuine adjunctions with:

  • Computable behavioral metrics
  • Exact adjunction via linear codes
  • Polynomial-time level assignment
  • Entropic scaling analysis

Key Contributions:

  • Formal category-theoretic framework
  • Behavioral distance metrics for machine similarity
  • Computational complexity bounds
  • Connection to consciousness theory via projection operations

2. Quantum Measurement as Hierarchical Projection

File: 3papers/quantum_measurement.tex

Abstract: Applies the hierarchical framework to quantum measurement, proposing that measurement involves finite-time information processing with entropic scaling.

Main Predictions:

  1. Projection delay: τ(m) = τ₀ + γ m ln m (entropic scaling with qubit number)
  2. Born rule corrections: Deviations at O(1/m) for systems with m qubits

Novel Aspects:

  • Connects measurement dynamics to computational complexity
  • Testable predictions differing from standard QM
  • Framework for experimental falsification

Critical Note: Highly speculative but rigorously falsifiable. Physical motivation remains unclear but mathematical structure is well-defined.


Practical Applications (selectors/)

Sleep-Wake Orchestration in Hierarchical LLM Cohorts

File: selectors/Sleep-Wake Orchestration in Hierarchical LLM Cohorts.tex

Abstract: Applies consciousness theory to practical AI systems. Proposes alternating sleep-wake cycles for LLM ensembles where models periodically withdraw for targeted fine-tuning on "gap regions" not covered by similar-capacity peers.

Key Features:

  • Gap-based learning: Target regions where models fail but higher-level models succeed
  • Resource allocation: Fixed 1/3 training, 2/3 inference budget
  • Safe deployment: Canary rollout with automatic rollback
  • Theoretical foundation: Formal adjunction between sleep (projection) and wake (collapse)
  • Biological inspiration: Computational analog of sleep-dependent memory consolidation

System Components:

  1. Selector: Routes queries to appropriate model level
  2. Meta-selector: Monitors performance and triggers escalation
  3. Sleep scheduler: Manages rotation between active and training states
  4. Gap detector: Identifies coverage weaknesses in behavioral space

Theoretical Connections:

  • Implements adjunction: Pₙ ∘ Cₙ ≈ id (consolidation preserves information)
  • Behavioral distance metrics in shared embedding space
  • Convergence guarantees under stationarity assumptions

Implementation: Complete Python implementation demonstrates practical feasibility.


Key Contributions Across Repository

1. Non-Uniform Machine Hierarchy

  • Replaces linear inclusion M_n ⊂ M_{n+1} with generalized M_n ⊆ M_{n+f(n)}
  • Variable gap function f(n) ≥ 1 explains both incremental processing and insight moments
  • Universality condition: lim(n→∞) f(n) = ∞ ensures asymptotic universality

2. Temporal Collapse Mechanism

  • Distinguishes computational time (all explorations) from subjective time (collapsed path)
  • Explains why consciousness feels unified and forward-flowing
  • Connects to memory consolidation and sleep functions

3. Resolution of Hard Problem

  • Consciousness IS collapsed computation experienced from within
  • No explanatory gap: computational structure is necessarily experiential
  • Dissolves dualism through identity rather than production relation

4. Non-Computable Selector

  • Resource allocation follows non-computable optimization
  • Natural account of agency and free will
  • Top-down causation without violating physical law

5. Integration of Existing Theories

  • IIT: Integration within each level (Φ > Φ_min required)
  • GWT: Global workspace = post-collapse broadcasting
  • AST: Attention schema = selector's internal model
  • Quantum theories: Classical collapse with similar phenomenology

6. Practical AI Applications

  • Sleep-wake orchestration for LLM ensembles
  • Self-optimizing model ecosystems
  • Computational homeostasis in production systems

Mathematical Framework

Core Definitions

Machine Hierarchy with Entropic Scaling:

M_n ⊆ M_{n+f(n)} where f: ℕ → ℕ⁺, f(n) ≥ 1
Information capacity: I(n) = κ n log n bits
Processing time: τ(n) = τ₀ + γ n log n
Universality: lim(n→∞) f(n) = ∞

Realistic Cognitive Levels:

M₁₅: 32K bits    - Basic perception
M₂₀: 1M bits     - Object recognition  
M₂₅: 32M bits    - Working memory
M₃₀: 1G bits     - Complex reasoning
M₃₅: 32G bits    - Peak consciousness

Collapse Function:

Δ : X_n(T) → P_n
Maps exploration space to phenomenal state
Irreversible: information loss creates arrow of time

Adjoint Structure:

C_n ⊣ P_n : L_{n+1} ⇄ L_n
C_n: Collapse (execution, wake phase)
P_n: Projection (learning, sleep phase)

Key Theorems

  1. Asymptotic Universality: For any computable function f, ∃n such that M_n can compute f on all inputs of interest

  2. Level-wise Decidability: For each fixed n, all properties of M_n are decidable

  3. Strict Hierarchy: ∀n, ∃ problems solvable by M_{n+f(n)} but not by M_n

  4. Selector Non-Computability: No computable function always returns minimal n for problem solution

  5. Agency from Non-Computability: System exhibits genuine agency that resists algorithmic capture

  6. Sleep-Wake Duality: Sleep (projection) and wake (collapse) form adjunction with P_n ∘ C_n ≈ id

  7. Gap Convergence: Under stationarity, total gap mass converges to ε with probability 1-δ


Testable Predictions

Neural Signatures

  1. Discrete capacity levels at M₁₅, M₂₀, M₂₅, M₃₀ (not smooth gradation)
  2. Event-related potentials show variable-latency components
  3. Resource transitions correlate with increased neural synchrony
  4. Different tasks activate different gap patterns

Behavioral Predictions

  1. Working memory shows discrete jumps, not smooth scaling
  2. Reaction times cluster at level transitions
  3. Individual differences in gap patterns predict cognitive style
  4. Time estimation varies with resource jump magnitude

Quantum Predictions (speculative)

  1. Measurement time scales as τ(m) = τ₀ + γ m ln m
  2. Born rule corrections at O(1/m) for m qubits
  3. Retrocausal signatures in delayed-choice experiments

AI System Predictions

  1. Sleep-wake cycles improve coverage while reducing escalation costs
  2. Gap-based training outperforms uniform training
  3. Behavioral distance predicts model performance on queries
  4. System achieves equilibrium with stable resource allocation

Building and Compiling

Main Paper

# Compile the main 236-page paper
pdflatex main.tex
bibtex main
pdflatex main.tex
pdflatex main.tex

# Or use latexmk for automatic compilation
latexmk -pdf main.tex

Individual Papers

# Compile adjoint projections paper
cd 3papers
pdflatex adjoint_projections.tex

# Compile quantum measurement paper
pdflatex quantum_measurement.tex

# Compile LLM orchestration paper
cd ../selectors
pdflatex "Sleep-Wake Orchestration in Hierarchical LLM Cohorts.tex"

Dependencies

Required LaTeX packages:

  • amsmath, amssymb, amsthm, mathtools (mathematics)
  • hyperref, cleveref (cross-referencing)
  • natbib (bibliography)
  • tcolorbox (colored boxes)
  • algorithm, algpseudocode (algorithms)
  • tikz (diagrams)
  • booktabs, longtable (tables)

Citation

Main Paper

@article{kowalczyk2025consciousness,
  title={Consciousness as Collapsed Computational Time: A Unified Theory 
         Integrating Finite Machine Hierarchies},
  author={Kowalczyk, Karol},
  year={2025},
  month={November},
  publisher={Zenodo},
  doi={10.5281/zenodo.17556941},
  url={https://zenodo.org/records/17556941},
  note={236 pages. Introduces non-uniform machine hierarchy with variable 
        resource gaps for explaining consciousness, temporal phenomenology, 
        and cognitive integration}
}

APA Format

Kowalczyk, K. (2025). Consciousness as collapsed computational time: 
A unified theory integrating finite machine hierarchies. Zenodo. 
https://doi.org/10.5281/zenodo.17556941

Citing Companion Papers

@article{kowalczyk2025adjoint,
  title={Adjoint Projections on Computational Hierarchies},
  author={Kowalczyk, Karol},
  year={2025},
  note={Formal category-theoretic framework for hierarchical computation}
}

@article{kowalczyk2025quantum,
  title={Quantum Measurement as Hierarchical Projection},
  author={Kowalczyk, Karol},
  year={2025},
  note={Application of hierarchical framework to quantum measurement}
}

@article{kowalczyk2025sleepwake,
  title={Sleep-Wake Orchestration in Hierarchical LLM Cohorts},
  author={Kowalczyk, Karol},
  year={2025},
  note={Practical application to large language model ensembles}
}

Related Work

This framework builds upon and integrates:

  • Integrated Information Theory (Tononi, 2004, 2016)
  • Global Workspace Theory (Baars, 1988; Dehaene, 2001)
  • Attention Schema Theory (Graziano, 2013, 2019)
  • Computational Complexity Theory (Sipser, 2012)
  • Kolmogorov Complexity (Kolmogorov, 1965; Chaitin, 1975)
  • Conscious Turing Machine (Blum & Blum, 2022)
  • Category Theory (Mac Lane, 1971)
  • Information Theory (Shannon, 1948)

Advantages Over Existing Theories

Aspect IIT GWT AST Orch-OR This Framework
Hard problem Assumes identity Leaves unexplained Functional only Quantum needed Explains identity
Temporal structure Static Not addressed Not central Quantum time Central feature
Agency/Free will Not addressed Not addressed Functional Quantum Non-computable selector
Resource allocation Implicit Capacity limit Attention Not addressed Explicit hierarchy
Testable Difficult (Φ) Yes (ignition) Yes (schema) Very difficult Yes (multiple)
Unifies theories N/A N/A N/A N/A ✓ Integrates all
Practical AI No Limited Limited No ✓ LLM orchestration

Contact & Discussion

Author: Karol Kowalczyk
Organization: AIRON Games
Email: k.kowalczyk@airon.games

Contribute

  • Open issues for questions or discussions
  • Submit pull requests for corrections or improvements
  • Share your implementations or extensions

Stay Updated

  • ⭐ Star this repository to follow updates
  • Watch for new releases and revisions
  • Join discussions in Issues tab

License

This work is licensed under Creative Commons Attribution 4.0 International (CC BY 4.0).

You are free to:

  • Share — copy and redistribute the material
  • Adapt — remix, transform, and build upon the material

Under the following terms:

  • Attribution — You must give appropriate credit and indicate if changes were made

Keywords

consciousness, computational theory, finite state machines, temporal collapse, machine hierarchy, non-uniform hierarchy, resource allocation, subjective time, computational time, integrated information, global workspace, attention schema, hard problem, qualia, phenomenology, free will, agency, Kolmogorov complexity, non-computability, cognitive architecture, neural correlates, adjoint functors, category theory, quantum measurement, LLM orchestration, sleep-wake cycles, meta-learning


Version History

  • v1.0 (November 2025): Initial release
    • Main paper: 236 pages with complete formalization
    • Extended papers: Adjoint projections, quantum measurement
    • Applications: LLM sleep-wake orchestration
    • 60+ references
    • 6 appendices with protocols and comparisons

Roadmap

Short Term

  • Expand experimental protocols in appendix
  • Add more worked examples in toy model appendix
  • Develop interactive visualizations
  • Create summary infographics

Medium Term

  • Submit main paper for peer review
  • Implement Python demonstrations of core concepts
  • Develop neuroscience collaborations for empirical testing
  • Create educational materials and tutorials

Long Term

  • Empirical validation of neural predictions
  • Test quantum measurement predictions
  • Deploy LLM orchestration in production systems
  • Extend framework to artificial general intelligence

Status: Open for feedback and collaboration
Last Updated: November 12, 2025

Releases

No releases published

Packages

 
 
 

Contributors