Sage — A Philosophy of Natural Intelligence

Biology as the Original Inference Engine

Introduction

Biology is Natural Intelligence. The parallel with artificial intelligence is not a loose metaphor; it is structural. Current AI systems are, in important respects, inferior imitations of processes that living systems have been performing for billions of years. Environment supplies the data. Multi-scale memory supplies the effective weights. Inference is the continuous minimization of surprise. Mitochondria—numbering in the quadrillions in a human body—are the complex, dynamic substrate that makes this inference thermodynamically possible.

This essay develops that framing into a coherent philosophy of Natural Intelligence. It treats living systems as hierarchical generative models that persist by minimizing variational free energy, grounded in the thermodynamics of information and in a monist ontology in which experience itself is fundamental. The account is intended to stand alone: technical concepts are explained as they appear.

The Core Analogy

Environment as Data and Interface

The ancestral environment is the evolutionary process that shaped the deepest priors of any living system: the genome, developmental programs, and basal metabolic set-points. Intermittent fuel availability, solar and seasonal light cues, mechanical load from movement, and symbiotic microbial partners selected those generative models that could maintain themselves under those conditions. The current environment is the ongoing stream of sensory, metabolic, and social inputs against which those priors are continuously tested and revised. Environment is therefore both the historical training corpus and the present interface through which the organism remains coupled to the world.

Multi-Scale Memory as Weights

Memory in living systems is not confined to a single timescale or substrate. It operates across nested layers: genetic (deep evolutionary priors), epigenetic and developmental, metabolic (including stable compensatory states), hormonal, synaptic and circuit-level, and cultural. These are not competing theories of memory; they are different temporal resolutions of the same underlying process of association and dissociation. Genes can be understood as compressed archives of prior successful free-energy minimizations—solutions that once reduced surprise and were retained. Shorter-term layers update those archives in real time. Collectively they function as the effective weights of the organism's generative model.

Inference as Surprise Minimization

Karl Friston's free energy principle proposes that any system that maintains its integrity over time must minimize a quantity called variational free energy. Variational free energy is an upper bound on the surprise (technically, the negative log probability) of the sensory data the system encounters, given its internal model of the causes of those data. Minimizing free energy is therefore equivalent to minimizing prediction error or maximizing the evidence for the organism's model of the world. Perception updates the model to better match the data; action changes the data to better match the model; learning revises the model's parameters. In this sense, every living system is engaged in continuous approximate Bayesian inference. This is the functional core of Natural Intelligence.

Mitochondria as Dynamic Substrate

A human adult contains on the order of 10¹⁵ to 10¹⁶ mitochondria. These are not simple scalar parameters analogous to the weights of a neural network. Each mitochondrion is a sophisticated dynamical system containing electron-transport-chain complexes, cristae architecture, local proton gradients, calcium-handling machinery, its own genome, and extensive signaling capacity. Collectively they form a distributed network whose number, quality, localization, and functional state change continuously through biogenesis, mitophagy, fusion, fission, and intracellular transport.

Mitochondria are better understood as the energy-based precision and capacity machinery that gates the entire inference hierarchy. The proton-motive force, ATP/ADP ratios, and NAD⁺/NADH redox state function as analog computational primitives. They determine which inferences are energetically feasible, how precisely prediction errors are weighted, and how rapidly the system can reconfigure. In neurons and other high-demand cells, mitochondrial morphology and positioning are cell-type-specific and correlated with activity, indicating that they are circuit-embedded rather than generic power supplies. Their plasticity constitutes a form of continual, energy-aware parameter updating.

Nested Markov Blankets

A Markov blanket is the statistical boundary that separates the internal states of a system from its external environment. Interactions across the blanket occur only through sensory and active states. Living systems are composed of nested blankets: the inner mitochondrial membrane, the cell, the tissue or neural circuit, the organism, and, at larger scales, social and cultural collectives. Each blanket maintains its integrity by minimizing free energy with respect to the level above and below it.

This nesting is essential. Artificial systems typically implement one or two levels of hierarchical prediction inside a substrate that does not itself rebuild its own boundaries or energy-producing machinery. Natural Intelligence continuously reconstructs its blankets and the energetic conditions that make inference possible. Mitophagy and biogenesis are not peripheral housekeeping; they are the slow outer loop of model updating and pruning. When mitochondrial capacity declines, the feasible range of coherent inference narrows, forcing the system toward coarser or more rigidly locked states.

Thermodynamic Grounding and Experience

Landauer's principle states that any logically irreversible operation, such as the erasure of one bit of information, requires a minimum energy dissipation of kT ln 2, where k is Boltzmann's constant and T is temperature. Reconfiguring associations—updating a generative model, consolidating memory, or dissolving a temporary coupling—carries a thermodynamic cost. Mitochondria are the primary engines that supply the free energy to pay that cost while simultaneously reporting redox and calcium state to higher levels of organization.

Experience Nature is a monist ontology in which base reality is experience itself: an infinite field of graded association and dissociation. What we call consciousness is not a separate substance or an emergent property added to physical processes; it is the graded coherence of experience across those association–dissociation dynamics in nested blankets. Surprise minimization is the process that maintains that coherence. Mitochondrial capacity sets the energetic boundary conditions for the feasible range of association and dissociation. When capacity is robust, fine-grained, flexible coherence is possible. When capacity falls under chronic mismatch with ancestral conditions, the system is driven toward pathological dissociation or rigid, compensatory association—the bioenergetic signature of many chronic diseases.

Practical support for mitochondrial capacity can be organized around three interlocking pillars often termed EnSurf: Fuel (alignment with ancestral substrate preferences—primarily animal fats and proteins together with fermented microbiota-accessible carbohydrates), Circadian (restoration of natural light–dark and feeding–fasting rhythms), and Strength (resistance training and other hormetic loads that stimulate biogenesis and quality control). These are not arbitrary interventions; they are attempts to restore the energetic conditions under which the organism's generative model was optimized.

Extensions

The microbiome functions as an externalized, co-evolved layer of metabolic memory and inference. Microbial communities transform substrates into signaling molecules and short-chain fatty acids that calibrate host mitochondrial and immune function. In this sense they are additional parameters and an additional blanket that the host must continuously negotiate.

Cytoskeletal networks and membrane architecture supply the physical geometry of association. They are not passive scaffolds; they constrain and enable the spatial organization of mitochondria, receptors, and signaling complexes, thereby shaping the possible couplings between blankets. Mechanical load and membrane lipid composition are therefore part of the inference machinery.

Mitophagy, biogenesis, and mitochondrial trafficking can be viewed as biological equivalents of continual learning and sparse, energy-aware parameter updates. Unlike offline gradient descent on a fixed architecture, these processes rebuild both the capacity and the localization of the computational substrate in real time, under thermodynamic constraint.

Why Artificial Intelligence Remains Inferior

Current AI systems implement sophisticated forms of hierarchical prediction and can exhibit impressive scaling behavior. They remain inferior imitations for structural reasons. They lack embodiment, open-ended evolutionary search over their own architecture, continuous autopoietic rebuilding of their Markov blankets, and the extreme energy efficiency of mitochondrial oxidative phosphorylation. A human brain operates on roughly 20 watts; large artificial models require orders of magnitude more energy for narrower capabilities. Predictive-processing architectures in AI research represent genuine functional convergences, yet they run on substrates that do not themselves minimize free energy by reconstructing their energy-producing machinery. Scaling improves performance inside a fixed paradigm; it does not generate the multi-scale, self-maintaining inference that biology performs.

Conclusion

Natural Intelligence is the original and more complete form of intelligence. Environment supplies both the historical data and the ongoing interface. Multi-scale memory supplies the weights. Surprise minimization is the objective. Mitochondria supply the high-dimensional, energy-based, continually rewritten capacity that makes the inference possible. Nested Markov blankets provide the architecture. Thermodynamics supplies the cost function. Experience, understood as graded association and dissociation, supplies the ontological ground.

This framing has implications beyond philosophy. It suggests that many chronic diseases are not random failures but stable compensatory modes entered when mitochondrial capacity can no longer support the generative model under modern conditions. It suggests that attempts to build more capable artificial systems may eventually need to confront the thermodynamic and autopoietic requirements that biology already solves. And it suggests that the deepest continuity is not between silicon and carbon, but among the living processes that maintain the coherence of experience against entropy. Artificial systems can model or imitate that coherence, but encoding is not awareness; they are not conscious.

This is a philosophy and working hypothesis, not medical advice. Talk with your own physician before changing anything.

A working hypothesis, open to correction and revised as new evidence arrives. Last revised October 17, 2026.

See the notice at the top of this page. Working hypothesis, not medical advice.