4.1 KiB
| description | type | domain | created | confidence | source |
|---|---|---|---|---|---|
| Free energy is an upper bound on surprise and its long-term average equals entropy -- by minimizing free energy organisms keep themselves in a small set of viable states far from thermodynamic dissolution | claim | critical-systems | 2026-02-16 | likely | Friston 2010, Nature Reviews Neuroscience; Friston et al 2006, Journal of Physiology Paris |
biological systems minimize free energy to maintain their states and resist entropic decay
The defining characteristic of biological systems is that they maintain their form and states in the face of a constantly changing environment. Mathematically, this means the probability distribution over an organism's physiological and sensory states must have low entropy -- there is a high probability of being in a small number of states and a low probability of being elsewhere. A fish out of water is in a surprising state both emotionally and mathematically. A fish that frequently forsook water would have high entropy and would soon cease to be a fish.
Free energy is an information-theoretic quantity that provides an upper bound on surprise (the negative log-probability of a sensory state). Since organisms cannot directly evaluate surprise -- they would need to integrate over all possible causes of their sensations -- they instead minimize free energy, which they can compute because it depends only on their sensory states and an internal probabilistic representation (a recognition density) of what caused those sensations. This representation is encoded by the organism's internal states: neuronal activity, synaptic weights, and connection strengths. Minimizing free energy therefore implicitly minimizes surprise, which over the long term minimizes entropy and keeps the organism within its viable states.
The mechanism is elegant in its circularity: organisms resist the second law of thermodynamics not by violating it but by actively sampling and modelling their environment in ways that confirm their own continued existence. Since Markov blankets enable complex systems to maintain identity while interacting with environment through nested statistical boundaries, the free energy principle explains what happens at those boundaries -- the internal states behind the blanket encode a generative model of the external world, and both perception and action serve to minimize the discrepancy between model and reality. Since emergence is the fundamental pattern of intelligence from ant colonies to brains to civilizations, free energy minimization provides the mathematical account of why emergent systems persist: they exist precisely because they have found configurations that resist surprise.
Relevant Notes:
- Markov blankets enable complex systems to maintain identity while interacting with environment through nested statistical boundaries -- FEP explains the dynamics at these boundaries: what internal states do and why
- biological organization nests Markov blankets hierarchically from cells to organs to organisms enabling local autonomy with global coherence -- free energy minimization operates at every level of the hierarchy
- emergence is the fundamental pattern of intelligence from ant colonies to brains to civilizations -- FEP provides the formal account of why emergent self-organizing systems persist rather than dissipate
- active inference unifies perception and action as complementary strategies for minimizing prediction error -- the two mechanisms through which free energy minimization is accomplished
- hill climbing gets trapped at local maxima because it can only accept improvements and has no way to see beyond the nearest peak -- free energy minimization IS hill climbing (descending) on a surprise landscape; organisms converge to local minima that keep them alive
- companies and people are greedy algorithms that hill-climb toward local optima and require external perturbation to escape suboptimal equilibria -- FEP establishes that greedy optimization is the mathematical default for all self-maintaining systems, not just a behavioral tendency
Topics: