Pith. sign in

REVIEW 4 major objections 4 minor 34 references

Quantum AGI: Ontological Foundations

T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A quantum-native AGI, whose memory and beliefs are unknown quantum states, cannot copy itself, cannot hold context-free knowledge, and cannot introspect without changing itself.

desk verdict A useful taxonomy and correct-but-narrow corollaries; the categorical framing oversells a modeling choice. read the letter →

arxiv 2506.13134 v1 pith:PR2ERIJJ submitted 2025-06-16 quant-ph cs.AI

Elija Perrier , Michael Timothy Bennett This is my paper · ORCID
classification quant-phcs.AI
keywords artificialgeneralintelligencequantumfoundationsKochen-Speckertheoremcontextualityno-cloningBell'sontologyAIXI
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to show that an artificial general intelligence whose memory, beliefs, and program are stored in quantum states is not just a faster version of a classical AI but a different kind of agent, one constrained by quantum foundations. It extends three results of quantum foundations to AGI agents: Bell's theorem, the Kochen-Specker theorem, and the no-cloning theorem. The resulting corollaries say that a quantum agent's knowledge cannot be assigned context-free true/false values, its entangled components can be non-locally correlated, its unknown internal states cannot be copied, and identical components are in principle indistinguishable. The paper also argues that self-observation is destructive for such an agent: measuring a quantum state to inspect it changes the state, so introspection threatens identity. If this is right, proposals for fully quantum-native AGI must be designed around these constraints rather than treated as classical AGI on faster hardware.

What carries the argument

The carrying object is the agent's internal state as a density operator $\rho_A \in \mathcal{D}(\mathcal{H}_A)$, with interactions classified by channel type: classical-to-classical (CTC), classical-to-quantum (CTQ), quantum-to-classical (QTC), and quantum-to-quantum (QTQ). The classical baseline is AIXI, a universal Bayesian agent whose registers are CTC maps and freely copyable; the QAGI baseline is a quantum register updated by quantum channels. Three theorems do the logical work: the Kochen-Specker theorem rules out non-contextual hidden variables for $\dim \mathcal{H} \ge 3$; Bell inequalities bound the correlations of local hidden-variable models of entangled components; and the no-cloning theorem forbids a universal operation that copies an arbitrary unknown state. The identity argument is carried by a channel asymmetry: a classical copy-observation map is injective and has a left inverse, whereas a QTC measurement channel $\Phi_M(\rho) = \sum_k M_k \rho M_k^\dagger \otimes |k\rangle\langle k|$ is non-injective and admits no CPTP left inverse, so the pre-measurement state cannot be reconstructed.

What would settle it

Implement a quantum agent whose entire program is an unknown quantum state and instruct it to read itself; if it returns a complete description of the program while leaving the state unchanged, the paper's copying and introspection limits are contradicted. A more direct test would be any demonstration of a universal operation that perfectly clones an arbitrary unknown qubit, or a non-contextual assignment of truth values to all measurements on a three-level system.

Watch

Extended reading notes

Core claim

The paper's central claim is that the transition from classical to quantum computational substrates is an ontological shift, not merely an instrumental speed-up, and that this shift produces formal constraints on a quantum-native AGI (QS-QAGI) whose internal state is a density operator $\rho_A$ (the mathematical object describing a quantum state). Corollary 1 states that for any internal component with Hilbert space dimension at least three, the Kochen-Specker theorem forbids assigning context-independent classical truth values to all propositions about its state. Corollary 2 states that entangled agent-environment or agent-component states can violate local realism, so distributed QAGI information need not be attributable to local parts. Corollary 3 states that the no-cloning theorem prevents perfect copying of an arbitrary unknown internal state, blocking classical-style recursion, self-replication, and memory backup for quantum beliefs. Corollary 4 states that identical quantum components are indistinguishable and cannot carry persistent classical labels. The identity consequence is that a classical copy-observation channel is injective and has a left inverse, while any nontrivial quantum measurement channel is non-injective and has no completely positive trace-preserving (CPTP) left inverse, so the act of self-observation irreversibly changes the agent.

Load-bearing premise

The argument rests on assuming that a quantum agent's program, beliefs, and self-model are stored as an unknown coherent quantum state, so that inspecting or copying itself must act on that state; if the agent instead kept a classical description of its own code, the stated copying and introspection limits would not apply.

Editorial extensions

If this is right

  • A QAGI cannot run classical-style recursion or self-modification by copying its own code: any attempt to copy an arbitrary unknown internal state is forbidden, so self-reference must be implemented differently, if at all.
  • A QAGI's knowledge base cannot be a context-free list of facts; propositions about its own state become definite only relative to a measurement context.
  • Learning about a quantum environment requires ensembles of identically prepared copies, because a single unknown state cannot be cloned and full characterization demands quantum tomography.
  • Entangled QAGI components share information that cannot be assigned to any single local component, so notions of agent boundary and distributed memory must be revised.
  • Self-inspection that would confirm identity is destructive: the measurement channel has no left inverse, so a quantum agent cannot introspect without changing itself.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The constraints bind only for the fully quantum-native QS-QAGI; a hybrid agent that keeps a classical description of its own code would sidestep the copying and introspection limits, a route the paper's own taxonomy allows.
  • If the corollaries are right, fault-tolerant quantum error correction will not dissolve them: the no-cloning and contextuality constraints apply to logical quantum information, not just to noisy physical qubits.
  • A testable consequence for quantum reinforcement learning is that a policy stored entirely as a quantum state will require an ensemble of copies whose size grows with the information to be extracted, so sample complexity will differ in kind from a classical agent's.
  • The identity result gives a formal sense in which quantum self-knowledge is not observation but transformation, which may require rethinking what it means for an agent to know itself.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper proposes an information-theoretic taxonomy of classical versus quantum AGI, formalizes a 'quantum-native' agent (QS-QAGI) as a register whose internal state is a density operator evolving under quantum channels, and derives four corollaries that apply the Kochen-Specker theorem, Bell's theorem, the no-cloning theorem, and particle indistinguishability to AGI components. It also argues that a QAGI's self-observation is irreversible and therefore problematizes diachronic identity. The advertised contribution is a set of formal constraints showing that quantum AGI differs from classical AGI in kind, not merely in speed.

Significance. If the scope of the claims were properly delimited, the paper would be a useful conceptual contribution: it makes explicit which quantum-information principles constrain an agent whose memory, beliefs, and program are carried by arbitrary unknown quantum states, and its channel-based taxonomy (CS-CAGI, CS-QAGI, QS-CAGI, QS-QAGI) is a clarifying organizing device. The corollaries are, as far as they go, correct applications of standard results, with conditions such as dim(H) ≥ 3 and the symmetric-subspace assumption stated. The paper also honestly acknowledges that practical consequences depend on implementation, decoherence, and error correction. The main weakness is that the categorical conclusions are derived from a specific modeling choice—that the agent's program and self-model are arbitrary unknown quantum states with no classical copy—and the paper does not establish that a quantum-native AGI must abandon a classical control layer. The 'in kind' claim therefore holds only for a constrained subclass, not for all QAGI implementations.

major comments (4)
  1. [§3.1, 'Recursion & Self-Reference'; Corollary 3] The assertion that 'a QAGI's internal program or model ξ_Q is encoded in an unknown quantum state ρ_A cannot be copied for recursive calls or direct self-inspection' is a modeling assumption, not a theorem. The model in §3 explicitly stores percepts o_t and rewards r_t in classical registers, and the paper's own taxonomy includes QS-CAGI and hybrid architectures. In the standard quantum circuit model, the control program is a classical bit string that can be freely copied and used in recursion, while no-cloning constrains only unknown quantum data registers. Corollary 3 and the recursion conclusion therefore apply only to a QAGI whose program and belief content are entirely unknown quantum states with no classical shadow, not to all QAGI implementations. This scope gap is load-bearing because the paper's headline claim that fully quantum AGI differs 'in kind' from classical AGI rests on it.
  2. [§4, 'Identity Consequences'] The argument that self-observation irreversibly changes a QAGI's identity is mathematically sound only under the stipulated identification of identity with recoverability of the density operator from the measurement record. The non-injectivity of Φ_M and the absence of a CPTP left inverse are correct facts, but the conclusion does not follow for an agent that maintains a classical description of the relevant identity-bearing content, or that uses a quantum non-demolition measurement whose back-action is negligible. The paper has not shown that a quantum-native agent cannot keep such a classical description; indeed, the interaction loop in §3 already uses classical percepts and classical action selection. As stated, the identity conclusion is a consequence of the model, not a general impossibility result.
  3. [§4, Corollaries 1–4 and the external appendix [24]] The corollaries are presented as 'formal theoretical results,' but the actual derivations and all detailed working are deferred to an external Zenodo appendix (reference [24]) that is not included in the preprint. The corollaries themselves appear to be direct applications of standard theorems—Kochen-Specker, Bell, no-cloning, and permutation symmetry—so the mathematical content is likely correct; however, the main text should either include the short derivations or state more explicitly that the corollaries are restatements with AGI terminology. In particular, the Identity Consequences section contains the paper's most original formal claim (the injectivity/left-inverse comparison), and that argument should be fully present in the main text or in an included appendix rather than only in an external repository.
  4. [§1 and §4, framing of the contribution] The statement that the paper 'extend[s] three cornerstone results of quantum foundations to AGI agents' overstates what is proved: Corollaries 1–4 restate the standard theorems with an AGI interpretation, and no new mathematical theorem about AGI is established. The contribution is the interpretive transfer and the taxonomy, which is valuable, but the paper should be explicit that the constraints apply under the stated modeling assumptions and do not preclude a QAGI with a classical control layer. Without that qualification, the abstract and introduction promise more than the formal content delivers.
minor comments (4)
  1. [§2, item (vi)] The list of classical ontological assumptions labels the assumption as 'contextuality,' but the sentence describes non-contextuality ('the outcome of measurements is independent of other properties measured alongside it'); the label should be 'non-contextuality' for consistency with the subsequent Kochen-Specker discussion.
  2. [Figures 1 and 2] The captions use CTC, CTQ, QTC, and QTQ before these abbreviations are defined in the text; a short definition in the captions or an earlier mention would improve readability.
  3. [Corollary 2, statement and implications] The phrase 'instantaneous irrespective of spatial separation of the components' could be misread as a claim of superluminal signaling. The implication paragraph correctly notes that no faster-than-light signaling follows, but the corollary statement itself should be reworded to say that the correlations violate local realism, not that they are instantaneous in a causal sense.
  4. [References and appendix] The sentence 'Appendices are available in via [24]' contains a grammatical error, and reference [24] is an external Zenodo link rather than a versioned appendix included with the paper; this should be fixed and, ideally, the appendix should be submitted as part of the manuscript.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the corollaries are direct applications of established external theorems to an explicitly stated toy QAGI model, and the identity discussion is a straightforward consequence of CPTP non-injectivity rather than a self-referential derivation.

full rationale

The paper's central claims are conditional applications of well-known quantum foundations results — Bell's theorem, the Kochen-Specker theorem, and the no-cloning theorem — to an explicitly formalized toy model of a quantum-native agent. Corollaries 1–4 do not redefine their inputs: each restates a standard external theorem and then draws an implication for a QAGI component, with the QAGI register/channel model specified in Section 3. The recursion and self-reference discussion in Section 3.1 relies on the stated modeling assumption that the agent's internal program or model ξ_Q is encoded in an unknown quantum state ρ_A; this is an assumption, not a circular derivation, and the paper explicitly frames the model as a toy model. The Identity Consequences section defines a classical copy-observation channel as injective and a QTC measurement channel as a non-injective CPTP map, then correctly concludes that the non-injective map has no CPTP left inverse; this is a mathematical consequence of the stated definitions, not an equivalence in which the conclusion is assumed. The only self-citation is reference [24], a Zenodo appendix by one author containing detailed working; the main results themselves are standard theorems and the short derivations in the main text are self-contained. The paper's hedging in the conclusion — that practical consequences depend on implementation, decoherence, and error correction — further indicates that the corollaries are not being presented as empirically forced predictions. No step reduces by construction to its own input, and no fitted parameter is renamed as a prediction. The main weakness, a possible overgeneralization from the toy model to all QAGI, is a scope concern rather than a circularity concern.

Assumptions & free parameters 0 free parameters · 7 assumptions · 0 invented entities

The paper's constraints on QAGI rest on accepted quantum theorems (Kochen-Specker, Bell, no-cloning) plus the paper's own model of a fully quantum agent. The model is the main burden: if a QAGI can keep a classical description of its own state, several corollaries would not bind as stated. No parameters are fitted to data, and no new physical entities are introduced.

assumptions (7)
  • domain assumption The postulates of quantum mechanics: stochastic measurement, Born rule, unitary evolution, entanglement.
    Invoked throughout Sections 2 and 3 as the background ontology for QAGI; these are standard, empirically supported postulates, not derived in this paper.
  • domain assumption Kochen-Specker theorem: no non-contextual hidden-variable assignment exists for Hilbert space dimension at least 3.
    Used for Corollary 1; accepted theorem from cited reference [19].
  • domain assumption Bell's theorem: entangled quantum states can violate local realism.
    Used for Corollary 2; accepted result from cited references [2,4,5,10,16].
  • domain assumption No-cloning theorem: an arbitrary unknown quantum state cannot be perfectly copied.
    Used for Corollary 3; established result from reference [33].
  • ad hoc to paper The QAGI model: an agent is a quantum register with a density-operator state, and actions are unitary channels or instruments.
    Defined in Section 3; this is the paper's own toy model, not an established formalism for AGI.
  • ad hoc to paper A QAGI's program and beliefs are arbitrary unknown quantum states with no classical copy available.
    Assumed in Section 3.1 (Recursion and Self-Reference, Adaptivity and Self-Modification); hybrid models listed by the paper would evade this assumption.
  • domain assumption AIXI is the canonical classical AGI and relies on the listed classical ontological assumptions.
    Section 3.1 uses AIXI as the classical baseline; this is a fair characterization of Hutter's model but is an interpretive claim.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Quantum AGI: Ontological Foundations." pith.science (2026). https://pith.science/paper/PR2ERIJJ

@misc{pith2026250613134,
  author       = {Pith},
  title        = {Pith review of: Quantum AGI: Ontological Foundations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PR2ERIJJ}},
  note         = {Machine review of arXiv:2506.13134}
}
read the original abstract

We examine the implications of quantum foundations for AGI, focusing on how seminal results such as Bell's theorems (non-locality), the Kochen-Specker theorem (contextuality) and no-cloning theorem problematise practical implementation of AGI in quantum settings. We introduce a novel information-theoretic taxonomy distinguishing between classical AGI and quantum AGI and show how quantum mechanics affects fundamental features of agency. We show how quantum ontology may change AGI capabilities, both via affording computational advantages and via imposing novel constraints.

Figures

Figures reproduced from arXiv: 2506.13134 by the authors.

Figure 1
Figure 1. Classical agent (CAGI) in￾teracting via CTC, CTQ or QTC maps with classical EC or quantum EQ environments QAGI EC EQ QTC CTQ QTQ QTQ CTQ QTC [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Quantum agent (QAGI) interacting via QTC, CTQ or QTQ maps. Quantum Classical Taxonomies. To struc￾ture our analysis, we introduce a QAGI classifica￾tion taxonomy reflecting the typical demarcation in quantum information sciences between physi￾cal substrates and logical superstrates. Our tax￾onomy comprises four domains: Classical AGI im￾plemented on classical hardware (CS-CAGI); CS￾QAGI (simulating QIP-based AGI on … view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 18 canonical work pages

  1. [24]

    Perrier, E.: Paper Appendices - QAGI - Technical Appendices (2025),https:// zenodo.org/records/15664973

    work page arXiv 2025

  2. [1]

    Cambridge University Press (2013)

    Aaronson, S.: Quantum computing since Democritus. Cambridge University Press (2013)

    2013

  3. [2]

    Physical Review Letters 49(2), 91–94 (1982)

    Aspect, A., Grangier, P., Roger, G.: Experimental realization of einstein-podolsky- rosen-bohm gedankenexperiment: A new violation of bell’s inequalities. Physical Review Letters 49(2), 91–94 (1982)

    work page 1982

  4. [3]

    Physics Physique Fizika1, 195–200 (1964)

    Bell, J.S.: On the Einstein Podolsky Rosen Paradox. Physics Physique Fizika1, 195–200 (1964)

    1964

  5. [4]

    Epistemological Letters9(CERN-TH-2053), 11–24 (1975)

    Bell, J.S.: The theory of local beables. Epistemological Letters9(CERN-TH-2053), 11–24 (1975)

    work page 1975

  6. [5]

    Cambridge University Press, Cambridge, 2nd edn

    Bell, J.: Speakable and Unspeakable in Quantum Mechanics. Cambridge University Press, Cambridge, 2nd edn. (2004)

    2004

  7. [6]

    Nature549(7671), 195–202 (2017)

    Biamonte, J., Wittek, P., Pancotti, N., Rebentrost, P., Wiebe, N., Lloyd, S.: Quan- tum machine learning. Nature549(7671), 195–202 (2017)

    2017

  8. [7]

    Contemporary Mathematics305, 53–74 (2002)

    Brassard, G., Høyer, P., Mosca, M., Tapp, A.: Quantum amplitude amplification and estimation. Contemporary Mathematics305, 53–74 (2002)

    work page 2002

Show all 34 references
  1. [8]

    Nature Communications 13 (2022)

    Caro, M.C., Huang, H.Y., Cerezo, M., Sharma, K., Sornborger, A., Cincio, L., Coles, P.J.: Generalization in quantum machine learning from few training data. Nature Communications 13 (2022)

    2022

  2. [9]

    Nature Computational Science (2022)

    Cerezo, M., Verdon, G., Huang, H.Y., Cincio, L., Coles, P.J.: Challenges and op- portunities in quantum machine learning. Nature Computational Science (2022)

    2022

  3. [10]

    Physical Review Letters23(15), 880–884 (1969)

    Clauser, J.F., Horne, M.A., Shimony, A., Holt, R.A.: Proposed experiment to test local hidden-variable theories. Physical Review Letters23(15), 880–884 (1969)

    1969

  4. [11]

    Fowler, A.G., Mariantoni, M., Martinis, J.M., Cleland, A.N.: Surface codes: To- wards practical large-scale quantum computation. Phys. Rev. A86(3), 032324 (Sep 2012)

    2012

  5. [12]

    Physical review letters 100(16), 160501 (2008)

    Giovannetti, V., Lloyd, S., Maccone, L.: Quantum random access memory. Physical review letters 100(16), 160501 (2008)

    2008

  6. [13]

    Journal of Artificial General Intelligence5(1), 1 (2014)

    Goertzel, B.: Artificial general intelligence: concept, state of the art, and future prospects. Journal of Artificial General Intelligence5(1), 1 (2014)

    2014

  7. [14]

    Quantum information science and its contributions to mathematics, Proceedings of Symposia in Applied Mathematics63, 13–58 (2010)

    Gottesman, D.: An Introduction to Quantum Error Correction and Fault-Tolerant Quantum Computation. Quantum information science and its contributions to mathematics, Proceedings of Symposia in Applied Mathematics63, 13–58 (2010)

    2010

  8. [15]

    Decision Support Systems 46, 318–332 (2008)

    Guo, H., Zhang, J., Koehler, G.: A survey of quantum games. Decision Support Systems 46, 318–332 (2008)

    2008

  9. [16]

    Nature526(7575), 682–686 (2015)

    Hensen, B., Bernien, H., Dréau, A.E., Reiserer, A., Kalb, N., Blok, M.S., Ruiten- berg, J., Vermeulen, R.F.L., Schouten, R.N., Abellán, C., Amaya, W., Pruneri, V., Mitchell, M.W., Markham, M., Twitchen, D.J., Elkouss, D., Wehner, S., Taminiau, T.H., Hanson, R.: Loophole-free b...

    2015

  10. [17]

    Springer Science & Business Media (2004)

    Hutter, M.: Universal artificial intelligence: Sequential decisions based on algorith- mic probability. Springer Science & Business Media (2004)

    2004

  11. [18]

    CRC Press (2024)

    Hutter, M., Quarel, D., Catt, E.: An Introduction to Universal Artificial Intelli- gence. CRC Press (2024)

    2024

  12. [19]

    Journal of Mathematics and Mechanics17(1), 59–87 (1967)

    Kochen, S., Specker, E.P.: The problem of hidden variables in quantum mechanics. Journal of Mathematics and Mechanics17(1), 59–87 (1967)

    1967

  13. [20]

    Minds and Machines pp

    Legg, S., Hutter, M.: Universal intelligence: A definition of machine intelligence. Minds and Machines pp. 391–444 (2007) 12 E. Perrier et al

    2007

  14. [21]

    Physical Review D22(2), 356 (1980)

    Mermin, N.D.: Quantum mechanics vs local realism near the classical limit: A bell inequality for spin s. Physical Review D22(2), 356 (1980)

    1980

  15. [22]

    Cambridge University Press, 10th anniversary edn

    Nielsen, M.A., Chuang, I.L.: Quantum Computation and Quantum Information. Cambridge University Press, 10th anniversary edn. (2010)

    2010

  16. [23]

    Özkural, E.: What is it like to be a brain simulation? In: International Conference on Artificial General Intelligence. pp. 232–241. Springer (2012)

    2012

  17. [25]

    arXiv:2106.10522 (2021)

    Preskill, J.: Quantum computing 40 years later. arXiv:2106.10522 (2021)

    2021 arXiv

  18. [26]

    Preskill, J.: Beyond nisq: The megaquop machine (2025)

    2025

  19. [27]

    Nature Reviews Physics pp

    Proctor, T., Young, K., Baczewski, A.D., Blume-Kohout, R.: Benchmarking quan- tum computers. Nature Reviews Physics pp. 1–14 (2025)

    2025

  20. [28]

    Pearson, 4th edn

    Russell, S., Norvig, P.: Artificial Intelligence: A Modern Approach. Pearson, 4th edn. (2020)

    2020

  21. [29]

    arXiv preprint cs/0309048 (2003)

    Schmidhuber, J.: Gödel machines: Self-referential optimal universal problem solvers. arXiv preprint cs/0309048 (2003)

    2003 arXiv

  22. [30]

    Springer (2021)

    Schuld, M., Petruccione, F.: Machine Learning with Quantum Computers. Springer (2021)

    2021

  23. [31]

    part i & ii

    Solomonoff, R.J.: A formal theory of inductive inference. part i & ii. Information and Control 7(1–2), 1–22, 224–254 (1964)

    1964

  24. [32]

    Cambridge University Press (2018)

    Watrous, J.: The Theory of Quantum Information. Cambridge University Press (2018)

    2018

  25. [33]

    Nature299, 802–803 (Oct 1982)

    Wootters, W.K., Zurek, W.H.: A single quantum cannot be cloned. Nature299, 802–803 (Oct 1982)

    1982

  26. [34]

    Physica Scripta1998(T76), 186 (1998)

    Zurek, W.H.: Decoherence, chaos, quantum-classical correspondence, and the al- gorithmic arrow of time. Physica Scripta1998(T76), 186 (1998)

    1998

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.