Wiki

Wiki: Thermodynamic Grounding Basics

Arrow-of-time evidence is informative, but it is not direct physical dissipation by default

Mind Uploading Research Project

Public Page Updated: 2026-04-02 Technical / natural science only

How to use this page

Read this first to avoid getting lost

This page explains why thermodynamic language appears on Mind-Upload, what current neural-data papers actually compute, and where the interpretation ceiling still sits.

  • Landauer lower bounds, tissue energy budgets, neural irreversibility metrics, model-based entropy-flow estimates, and implementation-side power accounting are different layers and are not read as one measurement.
  • The main scientific risk is silent translation: state-space construction, partial observation, reverse-transition sparsity, and physiology-side grounding all change what a thermodynamic number can mean.
  • Arrow-of-time results remain auxiliary on this site unless the route card names coarse-graining, hidden-degree risk, estimator family, support coverage, and quantity type.
  • Energetic or metabolic language now requires a separate physiology-side route such as calibrated CMRO2, FDG-PET, or 31P-MRS rather than being inferred from signal asymmetry alone.
  • A clean null or shuffle control is not the same thing as adequate reverse-transition support.
  • Operational promotion also needs a stability / nuisance audit and, when energetic language is used, an explicit same-session physiology bridge rather than a loose multimodal pairing.
  • Implementation-side FLOPs, wall-power, and communication cost stay separate from observation-side nonequilibrium metrics.
Helpful next pages
Verification FAQ Perspective
Best for
Readers who want a rigorous way to read thermodynamic, irreversibility, or entropy-production language without overpromoting it
Reading time
12-18 minutes
Accuracy note
This page treats thermodynamic indicators as auxiliary scientific evidence. It does not treat them as sufficient conditions for identity, consciousness, or WBE success.

Relatively clear at this stage

What we know now

  • The biological brain is a continuously driven physical system rather than a static stored object.
  • Current neural irreversibility papers do report informative nonequilibrium signatures, but they do not all compute the same quantity.
  • State-space construction, coarse graining, and dynamical assumptions are part of the claim rather than implementation footnotes.
  • Partial observation can hide dissipative cycles and induce memory, so weak observable currents do not by themselves imply near-equilibrium.
  • Physiology-side grounding is a separate requirement from observation-side arrow-of-time analysis.

Still unresolved beyond this point

What we still do not know

  • There is still no field-standard route from neural irreversibility metrics to a stable whole-brain entropy-production estimate.
  • It remains unresolved which thermodynamic indicators, if any, will prove decision-relevant for stronger WBE claims.
  • Cross-estimator convergence and physiology-side calibration for energetic interpretations remain open problems.

Learn the basics

Check the basics in the wiki

What the wiki is for

The wiki is a learning aid. For the project's official current synthesis, success criteria, and operating rules, always return to the public pages.

Shortest conclusion

On this site, thermodynamic language is useful only after it is typed correctly. A paper can show temporal asymmetry, a coarse-grained nonequilibrium lower bound, or a model-conditioned entropy-flow estimate without yet measuring microscopic physical dissipation, brain-wide metabolic cost, or a WBE acceptance criterion.

Three Overreads To Stop Early

  • Arrow of time is not direct heat dissipation: observation-side irreversibility and physiology-side energetic cost are different routes.
  • A clean surrogate is not enough: reverse-transition support and sparse-data handling are separate burdens.
  • One thermodynamic word is not one measurement object: lower bounds, asymmetry scores, graph indices, and model-based entropy-flow estimates answer different questions.
How this page fits the site

This is the background page for the Verification: thermodynamic indicators rule and the irreversibility route card. The operational policy lives in Verification; this page explains why that policy is strict.

Scope

This page stays on the technology and natural-science side only. It does not use thermodynamic language to settle identity, consciousness, law, or ethics. The narrower question is: what did the paper actually compute, what assumptions were required, and what ceiling still remains?

2026-03-30 correction: the real danger is silent translation

The main weakness of the older page was not that it mentioned the wrong literature. The weakness was that it still let readers move too quickly across four translations: from signal asymmetry to nonequilibrium inference, from coarse-grained inference to energetic interpretation, from observed trajectory to thermodynamic closure, and from auxiliary physics signal to WBE relevance. The current primary literature does not support those jumps unless each one is disclosed separately.

2026-04-02 correction: estimator meaning is not the same as operational stability

The next weakness was subtler. Even after separating estimator family, hidden-degree risk, and physiology-side grounding, the page still let one more shortcut survive: it was still too easy to read a mathematically interpretable metric as if it were already operationally stable or bridge-ready. The current primary literature does not support that shortcut either. Poudel et al. (2024) showed that small motion can materially alter visibility-graph structure and that only a low-motion subset reached moderate-to-high test-retest reliability for selected graph metrics. Metzen et al. (2024) showed that variability and complexity measures in BOLD fMRI have markedly different reliability profiles, with some functional-connectivity complexity measures remaining in the unacceptable-to-moderate range. Omidvarnia et al. (2021) showed reproducible multiscale-entropy structure for resting-state fMRI, but that result is specific to that estimator family and acquisition setting rather than a blanket license for all irreversibility metrics. Chen et al. (2025) then showed with simultaneous EEG-PET-MRI that temporal coupling across electrophysiology, hemodynamics, and metabolism can be strong while the spatial organization and state trajectories remain distinct rather than interchangeable. Therefore, on this site, stability / nuisance sensitivity, cross-estimator concordance, and physiology-bridge quality are now treated as separate reporting burdens rather than as footnotes under the estimator label.

Why thermodynamics appears on Mind-Upload at all

Mind-Upload does not treat WBE as a static data-storage problem alone. The brain is a continuously driven physical system whose information processing is maintained under non-equilibrium conditions. That is why thermodynamic language appears here. But the site's use is deliberately narrow: thermodynamic indicators are treated as auxiliary constraints on ongoing physical process, not as a shortcut to identity, consciousness, or final success conditions.

Five layers that must stay separate

Layer Representative literature What it can support What it still cannot support
Landauer lower bound Bérut et al. (2012) A lower bound for logically irreversible operations such as bit erasure. Actual brain power draw, actual emulator wall-power, or a WBE pass condition.
Tissue-level energy budget Attwell & Laughlin (2001) A descriptive decomposition of signaling-related energy use in biological tissue. A universal KPI for digital emulation or a thermodynamic success threshold.
Observation-side irreversibility metric Lynn et al. (2021); Deco et al. (2022); de la Fuente et al. (2023); Nartallo-Kaluarachchi et al. (2025) Time asymmetry or broken detailed balance under a declared signal route, state-space construction, and estimator family. Direct microscopic dissipation, direct metabolic cost, or implementation-side efficiency.
Model-conditioned entropy-flow estimate Ishihara & Shimazaki (2025) A time-varying entropy-flow estimate inside a declared state-space kinetic-Ising model. Whole-brain direct EPR measurement without model burden.
Implementation-side cost accounting Project-specific engineering audit Wall-power, FLOPs, communication cost, cooling burden, and hardware throughput of the implementation itself. Observation-side neural irreversibility by itself, unless the bridge is disclosed explicitly.

Three translation gaps the site now blocks

Translation gap What the primary literature shows Safe reading on this site
Estimator label -> thermodynamic meaning Lynn et al. (2021) estimated entropy production only after coarse-graining BOLD dynamics into clustered macrostates and showed sensitivity to the number of clusters. de la Fuente et al. (2023) showed that reversibility detection depends on retained principal components, feature family, and classifier complexity. Ishihara & Shimazaki (2025) added a nonstationary state-space kinetic Ising model precisely because steady-state assumptions fail for neural activity. Teza & Stella (2020) and Cocconi et al. (2022) showed that coarse graining can preserve or rescale entropy production depending on process class and scale. An estimator family name alone is too coarse. State-space construction, timescale, and dynamical assumptions belong in the claim.
Signal irreversibility -> metabolic or energetic interpretation Epp et al. (2025) found that about 40% of voxels with significant task-evoked BOLD changes showed oxygen-metabolism changes in the opposite direction. Observation-side asymmetry therefore does not by itself determine energetic demand. If the text uses words such as energy, metabolism, or dissipation cost, it must name a physiology-side route such as calibrated CMRO2, FDG-PET, or 31P-MRS, or else abstain.
Observed trajectory -> thermodynamic closure Martínez et al. (2019) showed that waiting-time asymmetry can reveal hidden dissipation even when observable current vanishes. Hartich & Godec (2024) showed that this can fail when coarse-graining and time reversal do not commute, and Martínez et al. (2024) limited the earlier claim to local-in-time coarse grainings and, where needed, semi-Markov constructions. Blom et al. (2024) showed that coarse lumping can hide dissipative cycles and induce memory. Baiesi et al. (2024) showed that sparse or unobserved reverse transitions can make direct estimation fail. Low current or a small irreversibility estimate is not read as near-equilibrium unless hidden-degree risk, memory order, and reverse-transition support are disclosed explicitly.

The thermodynamic claim ladder on this site

The safe way to read thermodynamic papers is not to ask whether they are important or unimportant. The better question is which rung they actually reached.

Claim rung Minimum evidence required What it still does not license
Rung 1: signal-side arrow of time A declared observation route plus a reproducible forward-versus-reversed asymmetry or inversion-detection result. Direct entropy production, direct metabolic cost, or implementation-side efficiency.
Rung 2: coarse-grained nonequilibrium lower bound A transition-based or equivalent lower-bound estimate with explicit state-space construction, coarse-graining policy, and finite-data uncertainty. Microscopic dissipation or thermodynamic closure of the full underlying process.
Rung 3: model-conditioned entropy-flow estimate An explicit dynamical model, its assumptions, parameter-identifiability limits, and controls separating coupling-related effects from firing-rate or sampling artifacts. Model-free direct EPR measurement or a unique mechanistic explanation.
Rung 4: physiology-grounded energetic interpretation A separate physiology-side route linking the observed signal-side effect to metabolism, transport, or energetics under a declared bridge. A universal thermodynamic success criterion or identity-relevant sameness.
Rung 5: implementation-side cost accounting Separate measurement of hardware power, wall-clock energy, FLOPs, communication cost, and cooling burden for the implementation itself. Equivalence to the brain's nonequilibrium statistics unless the bridge is argued and tested separately.

Irreversibility is not one estimator family

The same thermodynamic vocabulary still hides materially different mathematical objects.

Estimator family Representative literature What is actually computed Safe ceiling on this site
Coarse-grained transition-flux lower bound Lynn et al. (2021) Entropy-production lower bound from clustered BOLD-state transitions. Broken detailed balance in a declared coarse-grained macrostate system, not direct microscopic heat dissipation.
Time-shifted asymmetry / inversion family Deco et al. (2022); de la Fuente et al. (2023) Forward-versus-reversed asymmetry from time-shifted correlations or inversion classification. Temporal-asymmetry signature under the stated preprocessing and features, not direct EPR.
Directed visibility-graph irreversibility Nartallo-Kaluarachchi et al. (2025) Graph divergence between in/out degree structure of directed visibility graphs built from MEG-derived dynamics. Ordering of nonequilibrium organization across interaction scales, not direct causal wiring or microscopic dissipation.
State-space kinetic Ising entropy flow Ishihara & Shimazaki (2025) Time-varying entropy flow in a nonstationary, sparsely active spike-ensemble model. A model-conditioned entropy-flow estimate in recorded ensembles, not whole-brain direct EPR measurement.

Irreversibility route card

The purpose of the route card is simple: thermodynamic language alone does not tell the reader what quantity was computed, how fragile it is, or how far the interpretation may safely rise.

Route-card field What must be disclosed What misreading it blocks
Signal route and state definition Modality, sensor / source / spike level, preprocessing summary, source reconstruction or parcelization if used, and task / state segmentation rule. It blocks fMRI, ECoG, MEG, EEG, and spike-train results from sounding like the same observation route.
Coarse-graining geometry and timescale Parcel count, clustering rule, retained principal components, source model, window length, temporal bin width, sampling rate, and frequency range. It blocks the same paper title from sounding like the same quantity after state-space construction changes.
Observed-state closure / hidden-degree risk Whether hidden states, hidden cycles, latent variables, or nonlocal coarse-graining could still carry dissipation that is invisible in the reported trajectory. It blocks weak observable current from being misread as near-equilibrium.
Dynamical assumptions and memory order Whether the process is treated as Markov, semi-Markov, or another memory-bearing system, and whether coarse-graining and time reversal commute under the adopted construction. It blocks a Markov-style estimator from being overread when the observed process acquires memory under coarse graining.
Estimator family and quantity type State explicitly whether the result is a lower bound, asymmetry score, graph divergence, model-based entropy-flow component, or fuller entropy-production estimate. It blocks asymmetry scores or lower bounds from being misread as direct microscopic dissipation.
Cross-estimator concordance State whether the qualitative conclusion survives more than one reasonable estimator family, feature construction, or state-space definition, or whether it remains explicitly estimator-specific. It blocks one metric family from sounding like a universal thermodynamic structure of the brain.
Null / surrogate control State what shuffle, time reversal, label randomization, or surrogate was used and what failure it was designed to catch. It blocks a clean null test from sounding like a full thermodynamic validation.
Stability / nuisance sensitivity Report motion sensitivity, denoising / preprocessing sensitivity, split-half or test-retest reliability, session interval, and whether the result is single-site or protocol-scoped. It blocks one clean run from sounding like a stable operational signal.
Reverse-transition support / finite-data handling Transition-count support, whether reverse transitions were observed for the reported state pairs, and how zero or rare counts were handled. It blocks a clean surrogate test from being mistaken for adequate support coverage.
Physiology-side grounding If energetic or metabolic language is used, state whether a direct route such as calibrated CMRO2, FDG-PET, or 31P-MRS was actually measured, or explicitly state that it was not. It blocks signal-side irreversibility from being silently upgraded to metabolic cost or physical dissipation.
Physiology-bridge quality If a physiology-side route is invoked, state whether it was same-session, same-state-window, spatially coregistered, lag-aware, and whether agreement or disagreement across modalities was quantified explicitly. It blocks “paired modalities exist” from sounding like energetic grounding has already been solved.
Cost isolation Report hardware power, wall-clock energy, FLOPs, communication cost, and cooling burden separately from the neural irreversibility metric. It blocks implementation-side engineering cost from being merged with observation-side nonequilibrium evidence.
Abstention boundary Declare what the analysis does not identify, such as microscopic dissipation, direct metabolic cost, direct causal wiring, or identity-relevant sameness. It blocks auxiliary analysis from being promoted to a thermodynamic gate for WBE.

If this card is missing, this site keeps the result at exploratory auxiliary evidence. It is not promoted here to a common thermodynamic gate, direct microscopic dissipation, direct metabolic-cost readout, or WBE-relevant identity evidence.

What would actually strengthen the claim

What to add Why it matters What it still does not prove
Robustness across reasonable coarse-graining choices Shows that the sign or ordering of the result is not a one-pipeline artifact of parcelization, clustering, PCA rank, or temporal binning. It still does not turn a lower bound or asymmetry score into direct physical dissipation.
Explicit model-sensitivity audit Shows what changes when stationarity, pairwise couplings, conditional independence, source-model assumptions, or memory order are altered. It still does not identify a unique neural mechanism.
Cross-estimator concordance audit Shows whether the sign, ordering, or state separation survives more than one reasonable estimator family instead of depending on one mathematical construction. It still does not make different estimators equivalent or erase hidden-state risk.
Observed-state closure / hidden-degree audit Shows whether hidden cycles, latent variables, or nonlocal coarse graining could still carry the dissipation that the observed path misses. It still does not recover total microscopic entropy production by itself.
Within-modality stability / nuisance audit Shows whether the effect survives motion handling, denoising choices, split-half or test-retest checks, and declared protocol changes. It still does not convert an observation-side metric into a universal gate or a cross-site standard.
Reverse-transition support / sparse-data disclosure Shows whether the relevant forward and reverse transitions were actually seen often enough for the chosen estimator. It still does not remove hidden-variable bias by itself.
Physiology-side calibration when energetic language is used Separates observation-side arrow-of-time analysis from actual metabolism-side measurement. It still does not establish identity, consciousness, or a universal thermodynamic acceptance criterion.
Same-session physiology bridge with disagreement handling Shows whether electrophysiology, hemodynamics, and metabolism were aligned in the same acquisition window and whether partial agreement or spatial dissociation was reported rather than hidden. It still does not prove that a neural irreversibility metric is a direct energetic readout.

Nine questions when reading thermodynamic claims

  1. Are we looking at a lower bound, an asymmetry score, a graph index, or a model-based entropy-flow estimate?
  2. How was the state space built? Check parcelization, clustering, PCA rank, source model, temporal bins, and frequency range.
  3. What hidden-state or memory risk remains? Ask whether the observed trajectory is plausibly Markov, semi-Markov, or under-closed.
  4. Did the paper separate null controls from support coverage? A clean shuffle does not guarantee adequate reverse-transition counts.
  5. Does the qualitative result survive reasonable estimator changes? If not, read it as estimator-specific rather than as a general thermodynamic fact.
  6. Does it survive motion, denoising, and scan-rescan checks?
  7. If the paper says "energy" or "metabolism," where is the physiology-side route, and was it aligned in the same session / state window?
  8. Are logical and physical costs still separated? Do not merge FLOPs, wall-power, and signal-side nonequilibrium into one scalar.
  9. What does the metric explicitly abstain from claiming? If that line is missing, the ceiling should be read conservatively.

References

  1. Bérut, A., Arakelyan, A., Petrosyan, A., et al. (2012). Experimental verification of Landauer’s principle linking information and thermodynamics. Nature, 483, 187-189. doi:10.1038/nature10872
  2. Attwell, D., & Laughlin, S. B. (2001). An energy budget for signaling in the grey matter of the brain. Journal of Cerebral Blood Flow & Metabolism, 21(10), 1133-1145. doi:10.1097/00004647-200110000-00001
  3. Seifert, U. (2012). Stochastic thermodynamics, fluctuation theorems and molecular machines. Reports on Progress in Physics, 75(12), 126001. doi:10.1088/0034-4885/75/12/126001
  4. Lynn, C. W., Cornblath, E. J., Papadopoulos, L., Bertolero, M. A., Bassett, D. S., & Daniels, K. E. (2021). Broken detailed balance and entropy production in the human brain. Proceedings of the National Academy of Sciences, 118(47), e2109889118. doi:10.1073/pnas.2109889118
  5. Martínez, I. A., Bisker, G., Horowitz, J. M., & Parrondo, J. M. R. (2019). Inferring broken detailed balance in the absence of observable currents. Nature Communications, 10, 3542. doi:10.1038/s41467-019-11051-w
  6. Hartich, D., & Godec, A. (2024). Comment on “Inferring broken detailed balance in the absence of observable currents”. Nature Communications. doi:10.1038/s41467-024-52602-0
  7. Martínez, I. A., Bisker, G., Horowitz, J. M., & Parrondo, J. M. R. (2024). Reply to: Comment on “Inferring broken detailed balance in the absence of observable currents”. Nature Communications. doi:10.1038/s41467-024-52603-z
  8. Deco, G., Sanz Perl, Y., Bocaccio, H., Tagliazucchi, E., & Kringelbach, M. L. (2022). The INSIDEOUT framework provides precise signatures of the balance of intrinsic and extrinsic dynamics in brain states. Communications Biology, 5, 572. doi:10.1038/s42003-022-03505-7
  9. de la Fuente, L. A., et al. (2023). Temporal irreversibility of neural dynamics as a signature of consciousness. Cerebral Cortex, 33(5), 1856-1865. doi:10.1093/cercor/bhac177
  10. Nartallo-Kaluarachchi, R., et al. (2025). Multilevel irreversibility reveals higher-order organization of nonequilibrium interactions in human brain dynamics. Proceedings of the National Academy of Sciences, 122(10), e2408791122. doi:10.1073/pnas.2408791122
  11. Ishihara, K., & Shimazaki, H. (2025). State-space kinetic Ising model reveals task-dependent entropy flow in sparsely active nonequilibrium neuronal dynamics. Nature Communications, 16, 10852. doi:10.1038/s41467-025-66669-w
  12. Blom, K., Song, K., Vouga, E., Godec, A., & Makarov, D. E. (2024). Milestoning estimators of dissipation in systems observed at a coarse resolution. Proceedings of the National Academy of Sciences, 121(17), e2318333121. doi:10.1073/pnas.2318333121
  13. Teza, G., & Stella, A. L. (2020). Exact coarse graining preserves entropy production out of equilibrium. Physical Review Letters, 125(11), 110601. doi:10.1103/PhysRevLett.125.110601
  14. Cocconi, L., Salbreux, G., & Pruessner, G. (2022). Scaling of entropy production under coarse graining in active disordered media. Physical Review E, 105(4), L042601. doi:10.1103/PhysRevE.105.L042601
  15. Baiesi, M., Nishiyama, T., & Falasco, G. (2024). Effective estimation of entropy production with lacking data. Communications Physics, 7, 264. doi:10.1038/s42005-024-01742-2
  16. Epp, S. M., Castrillón, G., Yuan, B., et al. (2025). BOLD signal changes can oppose oxygen metabolism across the human cortex. Nature Neuroscience. doi:10.1038/s41593-025-02132-9
  17. Poudel, G. R., Egan, G. F., & McIntosh, A. R. (2024). Network representation of fMRI data using visibility graphs: The impact of motion and test-retest reliability. Neuroinformatics, 22, 265-284. doi:10.1007/s12021-024-09652-y
  18. Metzen, D., Fellner, M.-C., Labrenz, F., & Waschke, L. (2024). Reliability of variability and complexity measures for task and task-free BOLD fMRI. Human Brain Mapping, 45(10), e26778. doi:10.1002/hbm.26778
  19. Omidvarnia, A., Pedersen, M., Walz, J. M., et al. (2021). Temporal complexity of fMRI is reproducible and correlates with higher order cognition. NeuroImage, 230, 117760. doi:10.1016/j.neuroimage.2021.117760
  20. Chen, J. E., Lewis, L. D., Coursey, S. E., et al. (2025). Simultaneous EEG-PET-MRI identifies temporally coupled and spatially structured brain dynamics across wakefulness and NREM sleep. Nature Communications, 16, 8887. doi:10.1038/s41467-025-64414-x