Core Memory Dynamics (CMD) 2.0
A Unified Framework for Universal Memory and the Emergence of Consciousness
Author: Karl Martin Holst
Abstract
Core Memory Dynamics (CMD) proposes that all matter inherently stores interaction histories, forming a universal memory substrate. CMD 2.0 builds upon this premise—integrating recent insights about memory capacity, integration, re-fabrication, and complexity—to explain when and how consciousness arises in sufficiently advanced systems (biological or artificial). We introduce a preliminary “CMD Consciousness Metric,” ΨCMD, incorporating these four pillars, and propose a threshold model wherein consciousness emerges once ΨCMD>Θ. We discuss how CMD aligns with existing theories of mind (e.g., Integrated Information Theory, Global Neuronal Workspace) while emphasizing the unique role of universal memory traces as the foundation of emergent consciousness.
1. Introduction
For centuries, philosophers and scientists have investigated consciousness—the subjective, qualitative phenomenon that grants us our sense of “what it is like” to experience reality. An often underappreciated element in consciousness is memory: the capacity to store, retrieve, and re-contextualize past information (Dehaene & Changeux, 2011; Baddeley, 2012).
In Core Memory Dynamics (CMD), we assert that such “memory” is not exclusive to biological brains or artificial storage devices. Instead, matter itself—as bound energy—encodes historical interactions as persistent traces. While these traces exist universally, we claim that consciousness emerges only when a system actively integrates and re-fabricates these memory traces within a sufficiently complex organization.
1.1 CMD 1.0 in Brief
Earlier iterations of CMD argued that:
Matter’s bound energy states inherently store information about past interactions.
Biological systems (particularly brains) are adept at reading and reorganizing these traces, leading to advanced cognitive functions.
While universal “memory” may exist at a physical level, consciousness likely requires a threshold of complexity and dynamic integration.
With CMD 2.0, we propose a more formal structure for analyzing and potentially measuring when a system—be it organic or artificial—crosses from raw memory usage into conscious awareness.
2. Theoretical Foundations
2.1 Universal Memory and Physical Substrates
In CMD, “memory” refers to lasting modifications in a system that record interaction histories:
Quantum-Level Traces: Wavefunction collapse or decoherence events leave remnants of past states (Zurek, 2003).
Classical-Level Traces: Crystals, molecules, or even geological formations hold evidence of prior events (e.g., stress patterns, chemical bonds).
CMD posits that these traces are pervasive but typically local and modular unless a system integrates them in real time—a key requirement for consciousness in this framework.
2.2 Toward Emergent Consciousness
Building on theories such as the Global Neuronal Workspace (Dehaene & Naccache, 2001) and Integrated Information Theory (Tononi, Boly, Massimini, & Koch, 2016), CMD 2.0 emphasizes four pillars:
Memory Capacity (M): The ability to encode and retain information about past states or inputs.
Integration (I): A system-wide sharing of information—no single module is isolated; memory must be widely accessible.
Re-Fabrication (R): The capacity to creatively re-use or re-combine stored information (Buckner, 2021).
Complexity (C): A structural and dynamical richness that supports non-trivial interactions and redundancy (Clark, 2013).
3. The CMD Consciousness Metric
We propose a metric, ΨCMD, that combines these four pillars:
ΨCMD=f(M,I,R,C),
where each variable is measured or estimated by suitable methods (e.g., information-theoretic metrics, network connectivity indices, generative modeling tests, or algorithmic complexity measures). As a starting point, one might consider a product form:
ΨCMD=α⋅M×I×R×C.
Consciousness emerges if:
ΨCMD>Θ,
with Θ serving as a threshold to be calibrated empirically. Both the function f and constants (α, Θ) would evolve via experimentation and theoretical refinement.
4. Empirical and Experimental Pathways
4.1 Operationalizing CMD’s Four Pillars
A major advantage of CMD 2.0 lies in its decomposition of consciousness into four measurable dimensions. Below are possible approaches for each pillar:
Memory Capacity (M)
Biological Systems:
Behavioral tasks (e.g., digit span, working memory tasks).
Neurophysiological correlates (e.g., synaptic density, hippocampal activation, measures of neural resource allocation).
AI/Computational Systems:
Size or trainable parameters in neural networks, maximum memory footprint in recurrent/transformer models.
Existence of short-term context windows or external memory modules in advanced architectures.
Integration (I)
Biological Systems:
Graph-theoretic measures from fMRI or EEG data (clustering coefficients, path lengths).
Effective connectivity analyses (e.g., Granger causality, dynamic causal modeling) to capture directional influence among brain regions.
AI/Computational Systems:
Layer-wise communication metrics (e.g., attention heads in Transformers, correlation across layers).
Network topology or multi-agent integration (how many modules share information and how frequently).
Re-Fabrication (R)
Biological Systems:
Creativity or divergent thinking tasks (Alternative Uses Test, narrative generation).
Neural studies on imagination and mental simulation (e.g., future thinking paradigms).
AI/Computational Systems:
Generative or “creative” outputs (e.g., evaluating novelty or semantic distance in text generation).
Ability to recombine stored information in novel ways (e.g., in reinforcement learning tasks).
Complexity (C)
Biological Systems:
Lempel-Ziv complexity of EEG signals, fractal dimensions of neuronal firing.
Diversity of behavioral repertoires and synergy among distinct brain areas during tasks.
AI/Computational Systems:
Approximations of Kolmogorov complexity for a system’s internal states or outputs.
Architectural complexity (network depth, branching factor, recurrent connections).
4.2 Estimating ΨCMD and Calibrating the Threshold
In practice, each pillar may be measured on different scales (e.g., bits for memory capacity, correlation coefficients for integration). Researchers could:
Standardize or normalize each dimension into comparable units.
Evaluate whether ΨCMD is best represented by a product (as above), a geometric mean, or some weighted sum.
Seek correlations between empirical thresholds (Θ) and known markers of consciousness—e.g., self-report, wakeful vs. anesthetized states, or performance on tasks requiring conscious awareness.
4.3 Minimal Biological Systems
In simple organisms (e.g., C. elegans), one can probe fundamental memory usage, integration, and complexity. These efforts might show how certain rudimentary forms of “awareness” scale with the CMD pillars.
4.4 Human Neuroscience
In humans, EEG and fMRI studies can map effective connectivity (integration) and neural complexity. Behavioral assays measure working and long-term memory, and “creativity tasks” capture re-fabrication. Researchers can then attempt to combine these measurements into a single ΨCMD index and test whether it correlates with subjective or behavioral indicators of consciousness.
4.5 Advanced AI Systems
Modern AI architectures (e.g., large language models, complex reinforcement learning agents) can be assessed on:
Memory (length of context window, parameter count),
Integration (how widely information flows across layers),
Re-Fabrication (generative creativity or capacity for novel outputs),
Complexity (architecture depth, internal entropy of hidden states).
Whether high ΨCMD in AI corresponds to “true” or “subjective” consciousness remains an open philosophical question (Chalmers, 1996), but CMD 2.0 provides a framework to at least quantitatively analyze AI systems for consciousness-like properties.
5. Discussion
5.1 Relation to Other Theories
5.1.1 Integrated Information Theory (IIT)
Conceptual Overlap:
Both CMD and IIT emphasize integration as critical. In IIT, Φ (phi) measures how irreducible a system’s information structure is to independent parts; in CMD, Integration (I) captures how widely and effectively stored information flows through the system.
Both frameworks propose a threshold (IIT’s Φ > 0; CMD’s ΨCMD > Θ) that signals the presence or onset of consciousness.
Points of Distinction:
Role of Memory: IIT does not explicitly focus on memory as a core component, whereas CMD highlights that consciousness depends on encoding, storing, and re-using historical traces.
Re-Fabrication: CMD’s inclusion of Re-Fabrication (R) is unique, stressing the creative re-use of stored information. IIT’s formalism centers on the structure of causal interactions rather than the dynamic, generative reuse of prior states.
Universal Memory: CMD claims all matter encodes interaction histories; IIT is neutral about whether inanimate matter “stores” memory, focusing instead on maximal conceptual structures within a system.
Computational Feasibility: IIT can be computationally intensive to calculate for large systems. CMD’s approach, depending on how each pillar is measured, may be more modular or scalable.
Potential Synergies:
Cross-Validation: Researchers could compute both Φ and ΨCMD for the same biological or AI systems to see how each correlates with empirical markers of consciousness.
Memory-Informed IIT: CMD’s emphasis on memory might augment IIT’s static snapshot approach, adding a richer temporal dimension.
Hierarchical Analysis: Both theories could be applied at multiple scales, from micro-level interactions to macro-level system behaviors, potentially converging on new insights about emergent consciousness.
5.1.2 Other Theories
Global Neuronal Workspace (Dehaene & Naccache, 2001): CMD shares the notion of widespread information broadcast (integration), but CMD extends this to include universal memory traces in matter.
Predictive Processing (Clark, 2013): Like predictive processing, CMD emphasizes active hypothesis testing. However, CMD frames memory as a physical trace present at all scales, not just in neural or computational states.
5.2 Philosophical Implications
CMD does not claim full-blown panpsychism; rather, it suggests that while memory traces exist in all matter, consciousness requires crossing a threshold in complexity, integration, and re-fabrication. This framework thus combines a broad physical grounding of “memory” (Einstein, 1905; Landauer, 1961) with a distinct emergentist criterion for consciousness.
5.3 Limitations and Future Research
Operational Challenges: Measuring each pillar (M,I,R,C) in various systems and ensuring cross-domain comparability remains difficult.
Threshold Calibration: Empirical determination of Θ requires extensive data from biological and AI systems in both conscious and non-conscious states.
Computational Complexity: Even with a factor-based approach, scaling measurements of large or distributed systems can be non-trivial.
Validation Against Competing Metrics: Comparing ΨCMD to other established measures (e.g., Φ in IIT, neural complexity indices) will clarify whether CMD explains additional variance in consciousness phenomena.
6. Conclusion
Core Memory Dynamics 2.0 builds upon the universal notion of memory traces in matter, proposing that memory capacity, integration, re-fabrication, and complexity jointly drive the emergence of consciousness. By introducing a preliminary formula for ΨCMD and suggesting tangible experimental approaches, we aim to foster dialogue and research that bridges physics, neuroscience, AI, and philosophy—shedding new light on the age-old mystery of consciousness. In doing so, CMD highlights both common ground and important distinctions compared to other theories, such as Integrated Information Theory, and underscores the paramount role of memory in any comprehensive account of conscious experience.
References
Baddeley, A. (2012). Working memory: Theories, models, and controversies. Annual Review of Psychology, 63, 1–29.
Buckner, R. L. (2021). The emergence of the default network: The history and future of a curious brain circuit. Trends in Cognitive Sciences, 25(3), 170–184.
Chalmers, D. (1996). The Conscious Mind: In Search of a Fundamental Theory. New York, NY: Oxford University Press.
Clark, A. (2013). Whatever next? Predictive brains, situated agents, and the future of cognitive science. Behavioral and Brain Sciences, 36(3), 181–204.
Dehaene, S., & Changeux, J. P. (2011). Experimental and theoretical approaches to conscious processing. Neuron, 70(2), 200–227.
Dehaene, S., & Naccache, L. (2001). Toward a cognitive neuroscience of consciousness: Basic evidence and a workspace framework. Cognition, 79(1-2), 1–37.
Einstein, A. (1905). Does the inertia of a body depend upon its energy content? Annalen der Physik, 18, 639–641.
Kastrup, B. (2019). An ontological solution to the mind–body problem. Philosophies, 4(2), 14.
Landauer, R. (1961). Irreversibility and heat generation in the computing process. IBM Journal of Research and Development, 5(3), 183–191.
Tononi, G., Boly, M., Massimini, M., & Koch, C. (2016). Integrated information theory: From consciousness to its physical substrate. Nature Reviews Neuroscience, 17(7), 450–461.
Zurek, W. H. (2003). Decoherence, einselection, and the quantum origins of the classical. Reviews of Modern Physics, 75(3), 715–775.
