Kim (2023): Free Energy and Inference in Living Systems
What the paper establishes
Chang Sub Kim's 2023 paper compares two formulations of the free energy principle applied to living systems. Thermodynamic free energy principle (TFEP) treats organisms as non-equilibrium physical systems. Information free energy principle (IFEP) treats them as systems performing Bayesian inference.
The core result states that adaptive maintenance in organisms exceeds what thermodynamic laws and TFEP alone can describe. Brain-inspired IFEP supplies a more promising route.
Kim hybridizes free energy minimization with Bayesian inference. This yields Bayesian mechanics that govern latent neural dynamics of active inference.
Exact primary work and passages
The primary work is Chang Sub Kim, "Free energy and inference in living systems," Interface Focus 13, no. 3 (2023): 20220041. DOI: 10.1098/rsfs.2022.0041.
Key passages:
"Organisms are non-equilibrium, stationary systems self-organized via spontaneous symmetry breaking and undergoing metabolic cycles." (Abstract)
"We employed a simple model for non-stationary sensory influx and illustrated the development of optimal trajectories in the neural phase space: we numerically observed that the brain undergoes a dynamic transition from a resting state to the stationary attractor, which corresponds to the online inference of the environmental causes in continuous time." (Results section)
"In conclusion, organisms’ adaptive sustentation cannot be described within thermodynamic laws and the ensuing TFEP, for which the brain-inspired IFEP provides a promising avenue." (Conclusion)
"To establish an integrated framework of the operational principle of life, two rationales of FE minimization and Bayesian inference were hybridized, and the BM directing the brain’s latent dynamics of active inference was derived." (Conclusion)
"Consequently, the brain’s perception and motor inference in higher organisms were revealed to operate effectively as Schrödinger’s mechanical machine." (Conclusion)
Convergence patterns evidenced
The work touches scale invariance through stationary attractors that persist across time.
It touches memory and inference through online Bayesian updating that builds internal models.
It touches mind emergence by showing higher organisms reduce free energy via active inference rather than pure thermodynamics.
These map to the Ladder sequence from difference and flow to structure, memory, and mind.
Distance from the full OIP/GRAIN synthesis
The paper remains at the level of formal mechanics for inference. It does not address the Mirror Layer in which the reader sits inside the system under study.
It supports the information side of the synthesis but does not claim the universe possesses an intrinsic grain that reliably produces branching, spirals, or flow networks.
Honest limits and disconfirming edges
No empirical human or animal data appear. All results rest on a simple numerical model of sensory influx.
The claim that TFEP is insufficient rests on conceptual argument rather than direct experimental disproof.
The paper does not test whether IFEP scales to non-neural living systems such as single cells or ecosystems.
Reductionist objections remain open: observed inference-like behavior may reduce to underlying thermodynamic constraints not yet modeled.
Atomic claims
- Claim c1: Organisms maintain non-equilibrium stationary states through symmetry breaking. Tier: mechanistic. Source: Kim 2023 abstract.
- Claim c2: TFEP alone cannot account for adaptive sustenance in organisms. Tier: mechanistic. Source: Kim 2023 conclusion.
- Claim c3: IFEP derived from brain dynamics supplies a viable alternative framework. Tier: mechanistic. Source: Kim 2023 conclusion.
- Claim c4: Neural dynamics transition to stationary attractors that perform continuous-time inference. Tier: mechanistic. Source: Kim 2023 results.
- Claim c5: Perception and action in higher organisms function as a mechanical inference machine. Tier: mechanistic. Source: Kim 2023 conclusion.
Related routes
See /a/oip-the-ladder for the sequence from difference to mind. See /a/oip-the-mirror-layer for the reader-inside-system requirement. See /a/oip-principles for the object-invocation requirements that align with inference loops.
Key evidence
Ask this article · 7 suggested prompts
Text the build (+14245134626) or WhatsApp — slug|question creates a question node. Paste evidence with ingest slug|q:NODE_ID|your paste.