DmitriiBaturo @ DmitriiBaturo @beehaw.org

Posts

6

Comments

8

Joined

3 wk. ago

Thanks for the reply. Just to clarify: nothing in ICT-1.1 involves metaphors like “love as a force”. These are your own analogies, not part of the model.

ICT-1.1 is strictly empirical. If you think it’s “technobabble”, the simplest way to show that is to point out which of the three falsifiable predictions below is untestable:

1. Monotonic neural/metabolic response to informational novelty (odd-ball gamma connectivity, MEG/EEG — standard).
2. Equal-energy / different-information stimuli producing different system responses (same protocols used in sensory neuroscience).
3. kT·ln2-scale fixation thresholds across physical substrates (classic Landauer-scale measurements on PCM/memristors).

The specific experiments — with full methodology, parameters, and falsifiable predictions — are described in Section 8 of the preprint.

If any of these fail → ICT is wrong. If they’re “continental philosophy”, then so are Shannon, Landauer and Friston.

Happy to discuss specifics if you want — criticism is welcome, but it has to be about the actual experiments.

Thanks for the comment. Just to clarify: ICT isn’t continental philosophy — it’s an information-theoretic model with explicit falsifiable experiments.

ICT-1.1 is based on three empirical protocols:

1. dI/dT → neural/energetic response Prediction: increasing informational novelty changes gamma connectivity & metabolic cost. If no monotonic effect is observed → ICT is wrong.
2. “Structure without energy” tests Equal-power stimuli but different informational structure. Prediction: different neural/material responses despite identical energy. If no difference → ICT is wrong.
3. Energy cost of fixation across substrates Measuring kT·ln2-scale thresholds in memristors, PCM, organoids, etc. If fixation cost doesn’t correlate → ICT is wrong.

All three are measurable and falsifiable. This is experimental physics, not philosophy.

Thanks — yes, I feel we are entering exactly that transition.

AI doesn’t “replace” science, but it changes the environment in which science happens. When billions of people can iterate ideas with AI, old intuitions about “what is serious” or “what is allowed to be questioned” stop working.

ICT is just one attempt to formalize this shift — not by rejecting physics, but by adding an informational layer that can be tested. If the paradigm is moving, we need models that can move with it.

Appreciate your comment — it captures the moment very well.

Thanks for the critique — fair points. A few clarifications directly based on the ICT Model v1.1 (the version linked in the post):

1. About “consciousness” as a red flag

In the paper “consciousness” is used operationally, not philosophically.

ICT defines it strictly as:

C ∝ dI/dT — rate of informational change over time (Section 1.1, Eq. 2)

No metaphysics, no claims about qualia — only measurable information dynamics (entropy-rate, LZ-complexity rate, update-energy).

2. What field the work belongs to

Also clarified in v1.1: ICT sits at the intersection of:

information physics,

thermodynamics of computation,

temporal dynamics,

neuroscience of information processing.

(Sections 1.1, 2, and Correspondence Table)

The model does not present itself as philosophy of mind.

3. Background / prior work

v1.1 explicitly connects the framework to:

Landauer limit,

Bekenstein bounds,

Friston’s free-energy principle,

algorithmic complexity measures,

temporal stability / phase structure.

(All referenced in Sections 1.1–2)

4. Falsifiable predictions

The paper includes three concrete experiments (Section 8) designed for empirical testing:

1. Neuroenergetic test of dI/dT
2. Structure-without-energy input experiment
3. Cross-substrate information-fixation thresholds

All with operational variables, not philosophical language.

5. Summary

Your comment is useful — especially about clearly signaling the disciplinary context. But everything I’ve referenced above is directly in the v1.1 preprint and defines ICT as a physical/informational model, not a metaphysical one.

Happy to refine further if needed.

Hi! Thank you very much for such a detailed and thoughtful review — I really appreciate the time and attention you gave it. Your feedback is exactly the kind of constructive perspective that helps strengthen interdisciplinary work like this.

Let me address your main points:

1 On the Free Energy Principle. You are absolutely right that part of the neuroscience community sees FEP as overly broad and sometimes unnecessarily complex. In ICT, the model does not rely on FEP as a foundation — we use it only as an illustrative special case where dI/dT can be interpreted in terms of prediction error and the energetic cost of updating internal states. In other words, FEP is not a basis for ICT, but rather a local projection of a more general temporal structure. We will make this explicit in ICT 2.0 to avoid any confusion.

2 Citations and connection to existing research. You’re right: in several places the preprint assumes familiarity with background work — entropy metrics, temporal integration, information-theoretic models, etc. The next version will include a more structured “background and context” section with clear references throughout. Your comment here is very helpful and will definitely make the next edition stronger and more accessible.

3 Empirical testability and neuroscientific methods. Thank you for highlighting this. Section 8 already outlines specific paradigms (oddball / novelty detection, LZ-complexity, entropy rate, γ-coupling, prediction-error energetics, etc.), and I agree that for readers outside neuro-metrics these connections should be made more explicit. ICT 2.0 will expand this section with clearer explanations of applicable methods, their benefits, and limitations as used in practice.

4 On quantitative scales of dI/dT and metabolic effects. A key clarification is this: the magnitude of measurable effects varies greatly depending on the type of cognitive process. Simple, fast sensory events do produce very small changes — but the paradigms we propose are specifically chosen to target conditions where dI/dT varies much more strongly, such as:

disrupted temporal sequences,

violated expectations over time,

high-level predictive mismatch,

integration over multi-step patterns.

These are precisely the contexts where entropy, γ-coherence, and prediction-error timing produce the strongest and most reliable signals. We are formalizing these estimates for ICT 2.0 and will include the corresponding references.

You are right that basic stimuli produce minimal metabolic signatures — but the ICT experiments are deliberately focused on scenarios where the temporal structure is perturbed more significantly, and where the relevant methods are most sensitive.

And most importantly — thank you for your kind words. I’m an independent researcher, and your tone and careful attention truly mean a lot. The aim of ICT is not to bypass academic standards, but to offer something genuinely testable and conceptually consistent. Your comments help sharpen that aim.

Thanks again for your thoughtful, precise, and generous critique.

P.S. And yes… you’re not the first one to mention the formatting issue on Zenodo. We’ll fix that as well, but only in the next version, because Zenodo doesn’t allow replacing a file without deleting the entire record and creating a new version, which would reset the metadata and links. So I’m unable to change the extension there at the moment. If you’d like to download the paper as a PDF, you can do so via Academia: https://www.academia.edu/s/8924eff666

Thanks — that's an important concern, and you’re absolutely right to raise it. Let me clarify why ICT doesn’t require any spatial localization and doesn’t imply metaphysical idealism. In the ICT model, “stabilized information” (I_fixed) isn’t a place in space and not a substance. It’s simply a temporally persistent structure, whatever its physical carrier is. Because of that: ICT doesn’t assume that information has a definite position (so the uncertainty principle is not violated); and it doesn’t require a non-physical “realm” to host information. Everything remains within a physicalist picture: I_fixed = any pattern that keeps its form across time,dI/dT = the rate at which such patterns change. So ICT doesn’t rely on spatial localization and doesn’t push toward idealism — it only reforms the temporal vocabulary we use to describe information dynamics. In ICT, I_fixed is not a Platonic “idea”, not an abstract realm of forms, and not an ontological substrate lurking behind the world. It is simply the minimal temporal stability that any informational process must have in order to persist. “Temporal invariant of minimal change”or dI/dT_min — the condition for the existence of reality. In other words, I_fixed emerges from the process itself — from the fact that some patterns must remain sufficiently stable over time to allow any interaction, measurement, or change.Thus ICT does not assume an idealist metaphysics. It does not place “information” outside reality or above it. Instead, it treats stability (I_fixed) and change (dI/dT) as two inseparable aspects of one physical process. This avoids Platonism entirely: ICT doesn’t posit eternal forms — it formalizes the conditions under which processes can exist at all. Thank you for this incredibly important clarification. This risk genuinely needs to be articulated and addressed explicitly. We will make sure to incorporate it in the next stage of the model’s development and formulate it clearly in writing. Your contribution here is truly valuable — thank you for pointing it out so precisely.

P.S. If the DOCX on Zenodo still doesn't open on your side, you can use the PDF version instead, or simply read the full article directly in the browser via the Academia link: https://www.academia.edu/s/8924eff666

Thank you for the question — the distinction is indeed not obvious at first glance, so let me clarify the terminology.

In the ICT model, “stabilized information” (I_fixed) does not mean spatial localization. It refers to temporal stability — a structure that maintains its form under changes of context (i.e., has a low dI/dT).

“Local,” on the other hand, refers to position — something defined within a particular region of physical space or within an abstract state space.

In short:

local = where it is,

stabilized = how well it persists.

These properties can coincide, but they do not have to. For example, an interference pattern may be local but not stable, while a mathematical invariant may be stable without having any spatial localization.

Within the ICT framework, I_fixed is introduced specifically as temporal stability, not spatial confinement.

Hi. Unfortunately, Zenodo doesn’t allow replacing a file without deleting the entire record and creating a new version, which would reset the metadata and links. So I’m unable to change the extension there at the moment. If you’d like to download the paper as a PDF, you can do so via Academia: https://www.academia.edu/s/8924eff666#comment_1478583