# Super Creature 1.1 — Scientific Grounding The apex creature's "consciousness" is a set of **explicit, measurable, deterministic mechanisms**, not a claim of sentience. This document is the honest citation trail behind the V89 _consciousness-metrics layer_ (and the V84–V88 quantum-substrate ports it sits on). Every reference below was retrieved from a live source and verified for title, venue, year, and identifier; preprints, vendor language, and contested claims are flagged. **The rule for this project: the docs say _"inspired by / models"_, never _"implements consciousness"_ or _"is conscious."_** Compiled 2026-06-16 (research scout pass). --- ## PRIME neural observatory: measurement audit The World I, Cognition II and Quantum III tabs contain **27 measured cards**, nine per tab. They inspect **PRIME / ORACLE-Σ, Archon index 0**, not all five minds or a population average. `src/ui/super-neural-data.ts` maps each card to copied `SuperMindSnapshot` fields; the UI displays source paths, units, methods, limitations and exact numeric values and exports the complete snapshot, all 27 card datasets and bounded observed history as JSON. The observer does not create intelligence: these are measurements of computational mechanisms, not certification of life or subjective experience. The audit found and replaced clock-modulated bar heights, mislabeled latent-as-sensory displays, a hard-coded pipeline activation, an invented four-way thought tree, proxy-as-drive displays, fabricated pairwise entanglement edges, and repaint-driven history. Current marks have no wall-clock modulation. The five-depth search matrix contains the actual evaluated candidates and winners; echo mode honestly reports its smaller executed search. Signed inputs and scores remain signed. | Tab | Card | Actual data and boundary | | --------- | --------------------- | ------------------------------------------------------------------------------------------------------------- | | World | Original percept | Eleven World-provided input fields, before top-down encoding; phase is a model input. | | World | Encoded senses | Eighteen actual cortex inputs, including internal context and phase encoding; not eighteen physical sensors. | | World | Latent history | Sixteen current workspace components at observed cognitive ticks; no initial fake zeros. | | World | Latent vs imagined | Actual present latent and selected imagined vector; imagined states are internal candidates. | | World | Candidate novelty | Actual evaluated candidate novelty values; not novelty inferred from animation. | | World | Organ outputs | Thirty organs' raw output pairs; attention and aggregation are recorded separately. | | World | Latent RMS | Root-mean-square activation; not physical energy or intelligence. | | World | Scalar memory | Actual bounded scalar memory ring, plotted oldest to newest; not autobiographical recall. | | World | Internal prediction | Reasoned versus recursively predicted vectors; not verified future-world accuracy. | | Cognition | Stage outputs | RMS of actual stage output arrays, with explicit dimensions; no execution pulse or fixed activation. | | Cognition | Workspace | Actual capacity/occupancy/access/pressure values; mechanism telemetry, not conscious access evidence. | | Cognition | Affect | Modeled valence, arousal and dominance; no measured feelings. | | Cognition | Candidate search | Executed search scores and depth winners; no invented recursive tree. | | Cognition | Active inference | Actual discrete posterior plus free energy/surprise diagnostics; explicitly scoped model variables. | | Cognition | Activation homeostat | Branching estimate, gain, proximity and susceptibility; not proof of whole-system criticality. | | Cognition | Plan scores | Seven final scores used by the actual argmax and the selected plan; scores are not probabilities. | | Cognition | Self model | Self-model reliability and body-prediction diagnostics; no awareness percentage. | | Cognition | Ising recall | Actual spin lattice and associative-memory diagnostics; simulated spins. | | Quantum | Control inputs | Ten model drive/aspect values, not FTL, absolute-zero or physical quantum capabilities. | | Quantum | Basis probabilities | All 64 Born probabilities of the six-qubit simulated register; phase is separately recorded. | | Quantum | Metric | Finite-difference circuit metric, determinant, trace and Berry component; not scalar curvature. | | Quantum | Adapter sample stream | Actual marginal probabilities and recorded bits from the seeded classical adapter; no physical entropy claim. | | Quantum | Pair information | Exact pairwise quantum mutual information from the full statevector; diagonal = single-qubit entropy. | | Quantum | Local coherence | Actual single-qubit probabilities and transverse Bloch coherence; these are distinct quantities. | | Quantum | Entropy history | Computational-basis Shannon entropy and mean local coherence; not global von Neumann entropy. | | Quantum | Born sample history | Observed non-destructive basis samples; all six bits retained and gaps disclosed. | | Quantum | Bloch vectors | Actual reduced-state X/Y/Z vectors under a fixed display projection. | Histories contain at most 96 **observed completed cognitive ticks**. World delivers snapshots on its existing 18-frame / 36-mobile-frame cadence; this is sampled telemetry, not every computation and not a guaranteed refresh rate. Hidden tabs collect the same observations. Duplicate snapshots, repaints, paused frames and tab switches append nothing. Tick rollback or source change clears history. The status strip separately reports running/suspended/frozen, tick, sample age and delivery age; after three seconds without a new sample while running it explicitly reports staleness. Missing and non-finite values are not converted into zero. Exact source values remain available below each card. Observation previously perturbed two substrates: qualia was recomputed while reading a snapshot, and quantum geometry probes overwrote the live register. Both reads are now non-mutating. Quantum geometry uses a dedicated scratch register and copied input; its metric describes the **pre-amplification circuit family**, while probability/Bloch diagnostics describe the actual **post-amplification register**. A stencil spanning the discrete Hadamard switch is flagged. The controlled-RY decomposition is corrected to use two CNOTs; this changes prior numerical trajectories. Metric `curvature`/`scalar` names remain compatibility aliases for determinant/trace only. The balanced-cut linear-entropy statistic omits unbalanced cuts: the tested `GHZ5 ⊗ |0⟩` counterexample remains positive despite a separable cut. Participation ratio can be maximal for constant equal activations without any causal interaction. Neither statistic is IIT Phi. IV · BRAIN presents a **sampled, partial module graph** in two layouts. Its 16 groups cover selected stages and state stores of `SuperMind.think()` (senses, cortex, 30 organs, imagitron, thought search, predictor, reasoner, self-model, affect, quantum net, qubit register, reservoir, spin instinct, memory, meta, plan). Each group names its exact snapshot source. Associated core TinyMLP parameter allocations sum to 10,081; that sum is neither a neuron count nor the whole system's stored state. Dedicated `observatory.brain` captures preserve earlier stage values before recurrence and later updates alter them. A final workspace or transformed quantum vector cannot substitute for an earlier network input. Link metadata distinguishes captured payloads, selected operands and unavailable intermediates. A structural relationship with unavailable values has no measured strength; the graph is not a complete execution trace or a causal intervention result. COMPOSITE draws a module flow map and value strips; MEGA 4D maps those same groups to the 16 tesseract vertices and draws each available value as a dot. Both layouts are fixed. A new tick changes the tick label but cannot manufacture rotation or signal motion. Synthetic EEG bands, sine-driven spikes, XOR "axons", traveling pulses and orbiting satellites are absent. Module hue identifies a group; value marks encode sign and normalized magnitude under the displayed method. Positions are layout, dots are model variables rather than neurons, and links are code relationships rather than axons. Neither view measures EEG, subjective experience or intelligence. Missing/nonfinite values remain unavailable, and exact numerical text agrees with the graph exported in JSON. The shared 96-sample history and source/running/suspended/frozen status apply to both views. Relevant CQM-25 axes: `instrumentation_reproducibility`, `visual_embodied_scale`, `substrate_pluralism`, and `causal_integration`. This audit improves evidence fidelity without changing canonical capability scores. See the [measurement contract](./MODULE-CONTRACTS-2026-06-26.md) for controls and falsifiers and the [audit log](./AUDIT-LOG.md) for validation provenance. ## What the code actually implements vs. what the science says | Real research thread | What Super Creature 1.1 actually computes | Honesty caveat | | ----------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------ | | **Global Workspace / GNW** (Baars, Dehaene) — "ignition" | `Consciousness.ignition` — a winner-take-all over the 7 plan-coalitions; when the winner crosses an access threshold _and_ dominates the runner-up it "broadcasts", gating memory consolidation (`super-mind.ts`). | A toy model of the _signature_ (ignition + broadcast), not the cortical phenomenon. The 2025 Cogitate test pressured the offset-ignition prediction. | | **Integrated Information Theory** (Tononi) — Φ | `Consciousness.phi` — the participation/coherence ratio of 8 named module activations (1 ⇒ parts act as one, 1/M ⇒ independent). | Explicitly a **tractable surrogate**. True Φ is super-exponential _and_ non-unique; we never claim to compute Φ. | | **Predictive processing / active inference** (Friston) | `ActiveInference` (active-inference.ts) — a discrete free-energy core: a Bayesian belief over 8 latent situations minimising variational free energy F, then plan selection by **expected** free energy G (epistemic curiosity + pragmatic goal-seeking). Plus the predictor→`surprise` loop. | A faithful discrete-AIF implementation; the Free Energy Principle is a unifying _theory_, not a settled empirical fact. | | **Quantum cognition** (Busemeyer, Trueblood) | The 6-qubit register + Eshkol Born-sampled "thought collapse" + the QGT/Fubini–Study readout (V75–V88). | "Quantum" here is an **algebra on amplitudes/probabilities** in a deterministic sim — nothing about quantum neurons. | | **Higher-Order Theory / metacognition** (Rosenthal; Lau & Rosenthal; Fleming & Daw) | `Metacognition` (metacognition.ts) — a second-order **confidence** in the decision (decision margin + Φ + belief certainty + calm), spent as cognitive control: low confidence ⇒ EXPLORE to resolve uncertainty, high ⇒ commit/exploit. | A computational confidence/control model — NOT a claim of subjective higher-order awareness. | | **Organoid intelligence / "wet computing"** | `Reservoir` (reservoir.ts) — a 64-node **echo-state network** (the algorithm physical/organoid reservoirs use): a fading nonlinear echo of recent latents giving temporal memory + a novelty signal that sharpens curiosity. | We implement the reservoir-computing **algorithm**, NOT wetware — there is no biological substrate. Organoid reservoirs are the real-world inspiration only. | --- ## 1. Scientific theories of consciousness — the IIT vs GNWT adversarial collaboration (Cogitate) - The flagship result is published and **challenges key tenets of both theories — asymmetrically, not a win for either side** ("double-bind" is our shorthand, not the paper's term; GNWT's central prefrontal-ignition prediction was challenged more severely): the preregistered adversarial test (n = 256; fMRI + MEG + iEEG) aligned with _some_ predictions of both IIT and GNWT while **challenging key tenets of both** — IIT contradicted by a lack of sustained posterior synchronization, GNWT by a lack of prefrontal "ignition" at stimulus offset (Ferrante et al., 2025, _Nature_). - The collaboration was **designed up front to falsify**, with both camps co-signing predictions before data collection (Melloni et al., 2023, _PLOS ONE_). - **Adversarial note:** the GNWT proponents dispute the framing, arguing the offset-ignition prediction tested was not core to the theory (Naccache, Sergent & Dehaene, 2025, _Neuroscience of Consciousness_). Cite the _outcome_, not a verdict. ## 2. Integrated Information (Φ) — tractable proxies and intractability - True Φ is not just expensive, it is **under-determined**: Hanson & Walker (2023) show Φ in IIT 3.0 is _non-unique_ — systems exist for which Φ = 0 and Φ > 0 are simultaneously valid. - The proxies used in practice (decoder-based Φ\*, geometric Φ_G, stochastic interaction) are openly acknowledged approximations to a "computationally intractable" measure (Nilsen, Arena & Storm, 2024). - A 2024 peer-reviewed alternative sidesteps Φ with a tractable **integration–segregation difference** (ISD), classifying conscious states at ~93% accuracy — a _correlate_, not an identity claim (Jang, Mashour & Hudetz, 2024, _Nature Communications_). **This is the family our `phi` proxy belongs to.** - **Caveat:** integrated-information measures sometimes _increase_ under anesthesia, contrary to IIT — they are state-discriminators, not validated "amount of consciousness" meters (Nilsen et al., 2024). ## 3. Global Workspace / "ignition" - Ignition remains the named neural signature of conscious access in the Dehaene line, but the strongest 2025 test (Cogitate) **failed to find the predicted offset ignition** (Ferrante et al., 2025). - The GNWT camp's rebuttal reasserts the framework and defends ignition as still viable (Naccache et al., 2025). - Convergence work argues GNWT is compatible with the complexity-based PCI historically tied to the _rival_ theory (Farisco & Changeux, 2023). ## 4. Organoid intelligence / "wet computing" / biocomputing - **Brainoware** — a human brain organoid as a physical reservoir for speech-recognition / nonlinear tasks; a real, peer-reviewed proof-of-concept with modest accuracy (Cai, Ao, Tian et al., 2023, _Nature Electronics_). - **DishBrain** — in-vitro neurons "learn Pong"; the word _sentience_ is the authors' **operational redefinition** (responsiveness), not evidence of consciousness (Kagan et al., 2022, _Neuron_). Underpins Cortical Labs' 2025 commercial CL1. - **FinalSpark Neuroplatform** — 16 organoids / ~160k neurons accessible over the cloud; the "1,000,000× less energy" figure is a **vendor projection**, not a measured benchmark (Jordan et al., 2024). - The **Organoid Intelligence** roadmap is an aspirational position paper, not a result (Smirnova et al., 2023, _Frontiers in Science_). ## 5. Active inference / predictive processing & quantum cognition - The **Free Energy Principle** is mature and widely applied but is a unifying _theory_; whether it adds mechanistic content beyond Bayesian inference is debated (Friston, 2010; Piekarski, 2023). - A 2024 bridge approximates the otherwise-intractable exact free-energy computation with diffusion models — an engineering demo, not neuroscience evidence (Durand, Joffily & Khamassi, 2024). - **Quantum cognition** is a legitimate mathematical-modeling program, **not** a claim the brain is quantum; order effects fall out of non-commuting projection operators (Huang, Epping & Trueblood, 2025; Busemeyer & Bruza, 2024). --- ## 6. World models & predictive maps — the successor representation - The **successor representation** (SR) reformulates value as expected discounted future state-occupancy, M = Σ γᵗ P(sₜ | s₀) (Dayan, 1993): a predictive cache that supports fast, flexible re-planning when rewards shift. The landmark neuroscience claim is that the **hippocampus encodes exactly this predictive map** — place/grid fields as discounted future occupancy, not mere present location (Stachenfeld, Botvinick & Gershman, 2017, _Nature Neuroscience_); human reinforcement learning shows SR-consistent behavioural and neural signatures (Momennejad et al., 2017, _Nature Human Behaviour_). - **What Super Creature 1.1 computes:** `src/sim/successor-representation.ts` — a deterministic SR matrix over the 7 plan-archetypes, learned online by the TD rule M(s,·) ← M(s,·) + α[φ(s) + γM(s²,·) − M(s,·)] toward the closed form (I − γT)⁻¹, used each beat as a model-based **look-ahead** that biases plan selection toward behaviour whose learned successor-states carry high value. - **Honesty caveat:** a tabular SR over 7 discrete plan-states — a faithful implementation of the SR _algorithm_, not a claim about biological place cells. It is "model-based planning over the creature's own behavioural dynamics," not a world model of the external scene. ## 7. Empowerment — information-theoretic agency / intrinsic motivation - **Empowerment** is a reward-free, agent-centric measure of control: the channel capacity I(Aⁿ; S² | s) between an agent's action sequence and its resulting future state — the maximum information the agent can inject into its own future (Klyubin, Polani & Nehaniv, 2005; Salge, Glackin & Polani, 2014). It is a distinct intrinsic motivation from novelty/surprise and from the active-inference _epistemic_ term: an empowered agent seeks states of high **optionality** (where its choices matter), not merely surprising or uncertainty-reducing ones. It remains a live frontier — Tiomkin, Nemenman, Polani & Tishby (2023) give the dynamical-systems treatment; Levy, Allievi & Konidaris (2024) estimate empowerment from a _latent_ predictive model with no simulator (exactly this setting); Lidayan et al. (2025) find entropy/empowerment both **positively track** human open-world exploration (information-gain does _not_), playing complementary roles — not a blanket "empowerment beats curiosity" result. - **What Super Creature 1.1 computes:** `src/sim/empowerment.ts` — each beat it bins the 16-D world-model latent into M = 64 coarse cells by the signs of a frozen random-hyperplane LSH, maintains an online exponential-forgetting estimate of the K × M channel q(cell | plan) from its committed-plan → next-cell transitions, and runs the **Blahut–Arimoto** algorithm (Blahut, 1972; Arimoto, 1972) for a fixed 12 iterations to read the channel capacity C in nats; empowerment = C / ln K ∈ [0, 1]. The result feeds a curiosity lift toward controllable regions and a bounded vote for the plan whose action-row most steers the future. Blahut–Arimoto's iterate provably raises the capacity lower bound monotonically toward the true capacity — a closed-form property the module unit-tests, alongside the two known limits (a deterministic, fully-controllable channel ⇒ empowerment 1; a degenerate channel ⇒ 0). - **Honesty caveat:** a discrete, online empowerment estimate over 7 plan-archetypes and 64 LSH cells — a faithful implementation of the empowerment _objective_ and the Blahut–Arimoto capacity _algorithm_, not a claim of true n-step empowerment over a learned world model. It measures "how much this beat's plan-choice shapes the creature's own next coarse state," not control over the external scene. ## 8. Vector Symbolic Architectures / holographic memory — compositional binding - **Vector Symbolic Architectures** (VSA) / **Holographic Reduced Representations** (HRR) encode symbolic structure in distributed high-dimensional vectors via three algebraic operations — BIND (tie a role to a filler), BUNDLE (superpose a set), PERMUTE (order) — plus a cleanup memory that snaps noisy vectors back to clean atoms (Plate, 1995; Kanerva, 2009; Gayler, 2003). Because random high-D vectors are near-orthogonal, many bindings superpose in ONE vector and are recovered by unbinding — a candidate bridge between connectionist substrate and symbolic structure, and an active neuromorphic-computing frontier (Kleyko, Rachkovskij, Osipov & Rahimi, 2022, _ACM Computing Surveys_). - **What Super Creature 1.1 computes:** `src/sim/holographic-memory.ts` — the MAP-VSA model on bipolar {−1,+1}⁵¹² hypervectors: binding is the element-wise product (exactly self-inverse, a⊙a = 1⃗), bundling is the majority sign, cleanup is nearest-atom cosine. Each beat the mind encodes its situation into a context hypervector (a sign-bundle of feature atoms), binds it with the committed plan, and folds (context ⊙ plan) into a decaying holographic trace; to recall it unbinds the trace by the current context and cleans up against the 7 plan atoms — an analogical prior ("in situations like this, I chose ·") that biases the next plan. - **Honesty caveat:** a faithful implementation of the VSA/HRR _algebra_ over 7 plan-atoms and an 8-feature context — a real holographic associative memory, not a claim of human-level compositional reasoning. It recalls "the plan I bound to contexts like this," not arbitrary symbolic inference. ## 9. Quantum reservoir computing — the register as a cognitive reservoir - **Reservoir computing** harnesses a fixed high-dimensional nonlinear dynamical system as a temporal kernel and trains only a linear readout. **Quantum** reservoir computing (Fujii & Nakajima, 2017, _Phys. Rev. Applied_; Mujal et al., 2021, _Adv. Quantum Technol._) makes that reservoir a quantum system: the exponentially large Hilbert space of a few qubits is itself the rich nonlinear feature space, and only a linear readout of its measured observables over time is used/trained. - **What Super Creature 1.2 computes:** `src/sim/quantum-reservoir.ts` — the apex mind's existing 6-qubit statevector register IS the reservoir; the module is the readout, a fixed seeded linear projection of the register's per-qubit Bloch observables (⟨X⟩,⟨Y⟩,⟨Z⟩ × 6) over a leaky temporal trace, plus a quantum-state-velocity (`qFlux`) signal that feeds the curiosity drive (a churning wavefunction ⇒ a restless, exploratory mind). Deterministic, allocation-free, bounded, unit-tested. - **Honesty caveat:** a deterministic 6-qubit _simulation_ used as a reservoir — a faithful implementation of the QRC _algorithm_ on a perfect noiseless register, not a physical quantum processor. The value here is exactness and reproducibility, not a quantum speed-up. ## 10. Open quantum systems — the Lindblad decoherence of deliberation - A real qubit is never closed: coupled to an environment it undergoes **decoherence** (loss of phase coherence, T₂) and **relaxation** (energy decay, T₁). The mathematically complete description is the **Gorini–Kossakowski–Sudarshan–Lindblad** (GKSL) master equation — the most general generator of a completely-positive, trace-preserving quantum dynamical semigroup (Lindblad, 1976; Gorini, Kossakowski & Sudarshan, 1976; Breuer & Petruccione, 2002). For a single qubit it reduces exactly to the optical-Bloch equations on the Bloch vector. - **What Super Creature 1.1 computes:** `src/sim/quantum-deliberation.ts` — a deliberation qubit whose coherent superposition of candidate actions DECOHERES into a committed classical decision. A Rabi drive Ω (from curiosity) sustains the superposition, a detuning Δ (from dominance) leans the preference, and a pure-dephasing channel γ_φ (from arousal — the environmental noise) collapses the off-diagonal coherence; amplitude damping Γ₁ relaxes toward a ground "rest" decision. Coherence √(x²+y²) reads as "still deliberating", decisiveness 1−√ as "decohered into a commitment", and the mind lifts EXPLORE while it stays undecided. Deterministic midpoint integration; the Bloch vector stays inside the unit ball (a valid ρ). - **Honesty caveat:** a faithful implementation of the qubit Lindblad/GKSL _master equation_ as a model of deliberation→commitment — open-system physics, not a claim that decisions ARE quantum-mechanical. A dynamical metaphor with real, conserved (trace-preserving, completely-positive) mathematics underneath. ## 11. Stabilizer formalism — large-scale Clifford simulation (Gottesman–Knill) - The **Gottesman–Knill theorem** says a circuit of only Clifford gates (H, S, CNOT, Pauli) on |0⟩ inputs with computational-basis measurement is **classically simulable in polynomial time** — not by tracking 2ⁿ amplitudes, but the n Pauli operators that STABILISE the state. The **Aaronson–Gottesman tableau** (Aaronson & Gottesman, 2004) makes each gate an O(n) binary row update and measurement O(n²), so GHZ/Bell/graph/stabiliser-code states on dozens-to-hundreds of qubits are exact and cheap — precisely the regime the dense statevector (capped near 8 qubits) can never reach. - **What Super Creature 1.1 computes:** `src/math/clifford-tableau.ts` — the Aaronson–Gottesman tableau reimplemented deterministically (a port of the **Moonlab** simulator's Clifford backend; see THIRD-PARTY-NOTICES.md → Ported primitives), a LARGE-scale stabiliser substrate complementary to the small dense register: O(n) Clifford gates, a seeded O(n²) measurement, and the bipartite **entanglement entropy** read in ebits straight off the stabiliser matrix as a GF(2) rank (Fattal et al., 2004) — intractable to compute on the dense register, cheap here. - **Honesty caveat:** a faithful implementation of an EXACTLY-simulable subclass of quantum mechanics (the Clifford group), not a universal quantum computer — by Gottesman–Knill no quantum speed-up is implied. Its value is exact, reproducible, large-n stabiliser dynamics + entanglement structure. ## Honest framing (cite without overclaiming) 1. **Established / peer-reviewed empirical results:** the Cogitate 2025 _Nature_ adversarial test, Brainoware, DishBrain, the FinalSpark Neuroplatform — these happened and are published; describe what was _measured_, not what it "proves." 2. **Established theoretical frameworks (not facts):** IIT, GNWT, the Free Energy Principle, quantum- probability cognition are coherent, productive _theories_ — none is confirmed as "the" account of mind. 3. **Genuinely contested:** IIT faces a "pseudoscience" open letter (2023, PsyArXiv — itself controversial) and a non-uniqueness proof of Φ; treat Φ as mathematically problematic, not a working meter. 4. **Hype-prone terms handled with care:** "sentience" (operational redefinition), "biocomputer" (narrow + early), the "1,000,000×" energy figure (vendor projection). 5. **Aspirational, not delivered:** the Organoid Intelligence roadmap and any "conscious organoid" framing. 6. **For this art project this is ideal source material** — real and genuinely unsettled — _provided the docs say "inspired by / models" rather than "implements" or "is conscious."_ --- ## References - Aaronson, S., & Gottesman, D. (2004). Improved simulation of stabilizer circuits. _Physical Review A, 70_(5), 052328. https://doi.org/10.1103/PhysRevA.70.052328 - Arimoto, S. (1972). An algorithm for computing the capacity of arbitrary discrete memoryless channels. _IEEE Transactions on Information Theory, 18_(1), 14–20. https://doi.org/10.1109/TIT.1972.1054753 - Blahut, R. E. (1972). Computation of channel capacity and rate-distortion functions. _IEEE Transactions on Information Theory, 18_(4), 460–473. https://doi.org/10.1109/TIT.1972.1054855 - Breuer, H.-P., & Petruccione, F. (2002). _The Theory of Open Quantum Systems_. Oxford University Press. - Busemeyer, J. R., & Bruza, P. D. (2024). _Quantum Models of Cognition and Decision_ (2nd ed.). Cambridge University Press. - Cai, H., Ao, Z., Tian, C., et al. (2023). Brain organoid reservoir computing for artificial intelligence. _Nature Electronics, 6_, 1032–1039. https://doi.org/10.1038/s41928-023-01069-w - Dayan, P. (1993). Improving generalization for temporal difference learning: The successor representation. _Neural Computation, 5_(4), 613–624. https://doi.org/10.1162/neco.1993.5.4.613 - Durand, E. D., Joffily, M., & Khamassi, M. (2024). A diffusion model-based approach to active inference. _IEEE FMLDS_, 75–80. https://doi.org/10.1109/fmlds63805.2024.00024 - Farisco, M., & Changeux, J.-P. (2023). About the compatibility between the perturbational complexity index and the global neuronal workspace theory. _Neuroscience of Consciousness, 2023_(1). https://doi.org/10.1093/nc/niad016 - Fattal, D., Cubitt, T. S., Yamamoto, Y., Bravyi, S., & Chuang, I. L. (2004). Entanglement in the stabilizer formalism. _arXiv:quant-ph/0406168._ https://arxiv.org/abs/quant-ph/0406168 - Ferrante, O., Górska, U., Henin, S., et al. (Cogitate Consortium). (2025). Adversarial testing of global neuronal workspace and integrated information theories of consciousness. _Nature, 642_(8066), 133–142. https://doi.org/10.1038/s41586-025-08888-1 - Friston, K. (2010). The free-energy principle: a unified brain theory? _Nature Reviews Neuroscience, 11_, 127–138. https://doi.org/10.1038/nrn2787 - Fujii, K., & Nakajima, K. (2017). Harnessing disordered-ensemble quantum dynamics for machine learning. _Physical Review Applied, 8_(2), 024030. https://doi.org/10.1103/PhysRevApplied.8.024030 - Gayler, R. W. (2003). Vector symbolic architectures answer Jackendoff's challenges for cognitive neuroscience. In _Proceedings of the ICCS/ASCS Joint International Conference on Cognitive Science_, 133–138. https://arxiv.org/abs/cs/0412059 - Gorini, V., Kossakowski, A., & Sudarshan, E. C. G. (1976). Completely positive dynamical semigroups of N-level systems. _Journal of Mathematical Physics, 17_(5), 821–825. https://doi.org/10.1063/1.522979 - Hanson, J. R., & Walker, S. I. (2023). On the non-uniqueness problem in integrated information theory. _Neuroscience of Consciousness, 2023_(1). https://doi.org/10.1093/nc/niad014 - Huang, J.-Q., Epping, G. P., & Trueblood, J. S. (2025). An overview of the quantum cognition research program. _Psychonomic Bulletin & Review, 32_(6), 2507–2556. https://doi.org/10.3758/s13423-025-02675-9 - Jang, H., Mashour, G. A., & Hudetz, A. G. (2024). Measuring the dynamic balance of integration and segregation underlying consciousness, anesthesia, and sleep. _Nature Communications, 15_(1). https://doi.org/10.1038/s41467-024-53299-x - Jordan, F. D., et al. (2024). Open and remotely accessible Neuroplatform for research in wetware computing. _Frontiers in Artificial Intelligence, 7_, 1376042. https://doi.org/10.3389/frai.2024.1376042 - Kagan, B. J., et al. (2022). In vitro neurons learn and exhibit sentience when embodied in a simulated game-world. _Neuron, 110_(23). https://doi.org/10.1016/j.neuron.2022.09.001 - Kanerva, P. (2009). Hyperdimensional computing: An introduction to computing in distributed representation with high-dimensional random vectors. _Cognitive Computation, 1_(2), 139–159. https://doi.org/10.1007/s12559-009-9009-8 - Kleyko, D., Rachkovskij, D. A., Osipov, E., & Rahimi, A. (2022). A survey on hyperdimensional computing aka vector symbolic architectures, Part I: Models and data transformations. _ACM Computing Surveys, 55_(6), 1–40. https://doi.org/10.1145/3538531 - Klyubin, A. S., Polani, D., & Nehaniv, C. L. (2005). All else being equal be empowered. In _Advances in Artificial Life (ECAL 2005)_, LNCS 3630, 744–753. https://doi.org/10.1007/11553090_75 - Levy, A., Allievi, A., & Konidaris, G. (2024). Latent-predictive empowerment: Measuring empowerment without a simulator. _arXiv:2410.11155._ https://arxiv.org/abs/2410.11155 - Lidayan, A., Du, Y., Kosoy, E., Rufova, M., Abbeel, P., & Gopnik, A. (2025). Intrinsically-motivated humans and agents in open-world exploration. _arXiv:2503.23631._ https://arxiv.org/abs/2503.23631 - Lindblad, G. (1976). On the generators of quantum dynamical semigroups. _Communications in Mathematical Physics, 48_(2), 119–130. https://doi.org/10.1007/BF01608499 - Melloni, L., Mudrik, L., Pitts, M., et al. (2023). An adversarial collaboration protocol for testing contrasting predictions of GNW and IIT. _PLOS ONE, 18_. https://doi.org/10.1371/journal.pone.0268577 - Momennejad, I., Russek, E. M., Cheong, J. H., Botvinick, M. M., Daw, N. D., & Gershman, S. J. (2017). The successor representation in human reinforcement learning. _Nature Human Behaviour, 1_(9), 680–692. https://doi.org/10.1038/s41562-017-0180-8 - Mujal, P., Martínez-Peña, R., Nokkala, J., García-Beni, J., Giorgi, G. L., Soriano, M. C., & Zueco, D. (2021). Opportunities in quantum reservoir computing and extreme learning machines. _Advanced Quantum Technologies, 4_(8), 2100027. https://doi.org/10.1002/qute.202100027 - Naccache, L., Sergent, C., & Dehaene, S. (2025). GNW theoretical framework and the adversarial testing of GNW and IIT. _Neuroscience of Consciousness, 2025_. https://doi.org/10.1093/nc/niaf037 - Nilsen, A. S., Arena, A., & Storm, J. F. (2024). Exploring effects of anesthesia on complexity, differentiation, and integrated information in rat EEG. _Neuroscience of Consciousness, 2024_(1). https://doi.org/10.1093/nc/niae021 - Piekarski, M. (2023). Incorporating (variational) free energy models into mechanisms. _Synthese, 202_(2). https://doi.org/10.1007/s11229-023-04292-2 - Plate, T. A. (1995). Holographic reduced representations. _IEEE Transactions on Neural Networks, 6_(3), 623–641. https://doi.org/10.1109/72.377968 - Salge, C., Glackin, C., & Polani, D. (2014). Empowerment — An introduction. In M. Prokopenko (Ed.), _Guided Self-Organization: Inception_ (pp. 67–114). Springer. https://doi.org/10.1007/978-3-642-53734-9_4 - Smirnova, L., Caffo, B., Gracias, D. H., et al. (2023). Organoid intelligence (OI): the new frontier in biocomputing and intelligence-in-a-dish. _Frontiers in Science, 1_, 1017235. https://doi.org/10.3389/fsci.2023.1017235 - Stachenfeld, K. L., Botvinick, M. M., & Gershman, S. J. (2017). The hippocampus as a predictive map. _Nature Neuroscience, 20_(11), 1643–1653. https://doi.org/10.1038/nn.4650 - Tiomkin, S., Nemenman, I., Polani, D., & Tishby, N. (2023). Intrinsic motivation in dynamical control systems. _arXiv:2301.00005._ https://arxiv.org/abs/2301.00005 _Supporting / lower-confidence (verified to exist; secondary): the IIT "pseudoscience" open letter (124 signatories, Sept 2023, PsyArXiv — a contested preprint); the Ferrante et al. (2023) Cogitate bioRxiv preprint (10.1101/2023.06.23.546249); Cortical Labs CL1 (2025) — a product announcement, not a paper. The Busemeyer & Bruza 2nd-edition and the FinalSpark DOI were confirmed via publisher/NCBI listings rather than a direct index hit._