QNFO Papers
PapersAsk the corpusIdeasArchiveAboutWork with me

Quick read · plain language

Why Perfect Models Are Impossible

A plain-language version of “Epistemic Ineffability in Finite Subsystems: An Information-Theoretic Bound on Coarse‑Grained Reconstructions”

3 min readNo jargon, no mathematicsPhysics

The short version

Any tool that tries to copy a system must lose some information. This paper proves a rule about how much is lost and shows that even very detailed copies keep almost nothing.

The problem

Scientists and engineers want to build perfect copies of real things. Doctors want computer models of patients. Engineers want digital copies of wind turbines. Safety researchers want to control artificial minds. But every copy uses finite tools: a fixed number of sensors, a fixed number of settings, a fixed number of rules. The paper asks: what is the smallest possible loss when the real thing keeps all its information?

The paper calls these tools "handles." A handle is like sorting all possible states of a system into a fixed number of boxes. You can tell which box the system is in. You cannot tell its exact state.

The rule and its proof

The paper proves a bound. A bound is a hard limit that cannot be broken. The bound says: the lost information is at least the total information minus the information from the boxes.

The proof uses two facts. First, the full system has some total information. Second, the boxes can give at most some amount of information. The lost information is what remains. Any uneven sorting of states into boxes gives even less information, so the bound still holds.

The paper also looks at using many handles one after another. If each handle gives a small amount of information and the handles do not overlap, the total captured information adds up simply.

Two examples

The paper gives two examples with made-up numbers.

First, a brain-like system with about a hundred bits of total information and about a million boxes. The lost information is about a hundred bits. The copy keeps only about one part in six. The loss is about five parts in six.

Second, a large system with a thousand bits of total information and five handles, each giving ten bits. The total captured is fifty bits. That is one part in twenty of the whole. You would need a hundred such handles to capture everything.

The paper stresses that these numbers come from the rule alone. No real data was used.

What this means

The paper connects the rule to several fields. Digital copies of physical structures cannot be perfect. Medical models cannot capture all patient facts. Safety changes to artificial minds cannot control all hidden states. The rule gives a hard ceiling for what any finite tool can do.

The paper also proposes a small test. Someone could build a simple reversible system with about a thousand states. They could apply five different sorting rules. They could measure how much information each rule captures. If the total is about fifty bits, the rule holds. If it is much more, the rule fails.

What the paper does not show

The paper lists several gaps. Real systems have noise. The rule assumes no noise. Real sortings of states are uneven. The rule assumes even sorting. Real systems can receive outside information. The rule assumes a closed system. Real handles used one after another may overlap. The rule assumes they do not. If any of these assumptions fail, the actual loss could be larger, not smaller.

The paper does not know how overlapping handles change the total. It does not know whether open systems could break the rule. It does not know what changes for unusual state spaces. The work is purely theoretical and has not been reviewed by other scientists.

Words used here

bound
a hard limit that a quantity cannot go below
handle
a finite tool that sorts system states into a fixed number of boxes
information-conserving
a process where nothing is lost, like a perfect shuffle
partition
a way of dividing all possible states into separate groups
reversible
a process that can run backward without loss

Learn more in plain language

This plain-language version was written by an AI model from the text of the paper and checked against plain-language rules: everyday words, no mathematics, no unexplained names. It leaves details out and can still be wrong; the paper is the source of record. To report a problem, open the paper, select the passage and choose Report.

For specialists: the technical digest, quoted from the paper

In the paper's words

Finite observers embedded in an information‑conserving whole cannot externalize their full epistemic state. Existing digital‑twin, multiscale, and uncertainty‑estimation studies illustrate this limitation but do not quantify the inevitable information loss. We formalize epistemic “handles’’ as finite‑resolution partitions of a system’s state space and derive an incommensurability bound that any such partition must satisfy under unitary or ultrametric dynamics.

The abstract, verbatim, and its opening sentence is the claim. Every line below is quoted from the text.

Numbers it reports

  1. Applying the bound to a brain‑scale subsystem ($N=128$ bits, $M=2^{20}$) yields $\Delta=108$ bits ($\approx$ 84 % loss).

    from Abstract

  2. If the measured mutual information remains $\approx$ 50 bits, the result supports the bound; a significantly higher value would falsify it.

    from 7.3 Minimal Experiment to De‑risk / Falsify

  3. A second illustration with $I_{\text{total}}=1000$ bits, $r=10$ bits per partition, and five successive lenses gives a captured information of $50$ bits (5 % of the whole).

    from Abstract

  4. Example 2 (large system with lenses) – Total information $I_{\text{total}}=1000$ bits, five $10$‑bit lenses capture $I_{\text{cum}}=50$ bits (5 % of the whole), maximal lenses before full capture $n_{\max}=100$.

    from 5. Results

  5. Dual numerical illustrations – Section 5 computes the bound for (a) $N=128$ bits, $M=2^{20}$ cells, and (b) $I_{\text{total}}=1000$ bits with five $10$‑bit lenses, obtaining unrecoverable gaps of $108$ bits and captured information of $50$ bits respectively.

    from 1. Introduction

  6. Parameter Value Source $I_{\text{total}}$ $1000$ bits Assumption (Section 4) $r$ (bits per lens) $10$ bits Assumption (Section 4) $n$ (number of lenses) $5$ Assumption (Section 4) $I_{\text{cum}}(5)$ $5 \times 10 = 50$ bits Multiplication Captured fraction $50/1000 = 0.05 = 5\%$ Division $n_{\max}$ (lenses before full capture) $\left\lfloor \frac{1000}{10}\right\rfloor = 100$ Floor division Step‑by‑step arithmetic Multiply lenses by resolution: $5 \times 10 = 50$ bits.

    from 4.6 Numerical Example 2 (Large System with Sequential Lenses)

How it is done

  1. We formalize epistemic “handles’’ as finite‑resolution partitions of a system’s state space and derive an incommensurability bound that any such partition must satisfy under unitary or ultrametric dynamics.

    from Abstract

  2. If a provable theorem showed that any finite epistemic representation must discard a quantifiable amount of information, designers could treat the resulting loss as an unavoidable approximation rather than a technical flaw.

    from 1. Introduction

  3. The method treats the microscopic lattice as a coarse‑graining of the full microstructure, directly mirroring our notion of epistemic partitions.

    from 2. Background and Related Work

  4. We consider a closed physical subsystem $S$ embedded in a larger information‑conserving whole $W$.

    from 3.1 System Model

What would falsify it

  1. The theorem predicts that no finite collection of partitions can reconstruct the exact microstate of a closed, unitary system.

    from 6.2 Falsification Criterion

  2. An experiment that (i) prepares a reversible dynamical system, (ii) applies a finite set of coarse‑grainings, and (iii) reconstructs the exact microstate from the coarse observations would falsify the bound.

    from 6.2 Falsification Criterion

  3. If the measured mutual information remains $\approx$ 50 bits, the result supports the bound; a significantly higher value would falsify it.

    from 7.3 Minimal Experiment to De‑risk / Falsify

Where the claim stops

  1. Assumption Potential Sensitivity Perfect unitarity (no decoherence) Real systems experience noise; decoherence could increase $\Delta$ beyond the bound.

    from 6.1 Limitations

  2. Closed system External information influx violates the unitary assumption and could alter the bound.

    from 6.1 Limitations

  3. If any assumption fails, the bound remains a lower bound on the gap; the actual unrecoverable information could be greater.

    from 6.1 Limitations

Where to read

  1. Abstract180 words · 1 min
  2. 1. Introduction324 words · 1 min
  3. 2. Background and Related Work386 words · 2 min
  4. 3. Methods327 words · 1 min
  5. 4. Analysis349 words · 2 min
  6. 5. Results78 words · 1 min
  7. 6. Discussion177 words · 1 min
  8. 7. Practical Implications and Engineering Pathways380 words · 2 min
  9. 8. Conclusion113 words · 1 min
  10. Provenance and AI Involvement225 words · 1 min

Read from the paper itself: 32 sections, 2,661 words. Nothing in this digest is written by a model; a line that is not a sentence of the paper is not shown. Have a specific question? Ask this paper.

The full text · Ask the whole corpus