Universal Taxonomy¶

Jean-Paul Niko · February 2026

\title{Part VI-B: Universal Intelligence Taxonomy [0.3em] \normalsizeExtract from "Intelligence as Geometry"} \author{Jean-Paul Niko}\date{February 2026} \fi

Part VI-B: Universal Intelligence Taxonomy¶

\addcontentsline{toc}{section}{Part VI-B: Universal Intelligence Taxonomy}

Every cognitive system---human, animal, machine, hypothetical---receives an intelligence vector in a universal type space. Part VI profiled machine intelligences within the human type space $T_{\mathrm{human}}$. This part extends the framework to non-human substrates, defines the overlap subspace for cross-species comparison, and derives communication bandwidth bounds from the Conceptual Irreversibility Theorem.

%═══════════════════════════════════════════════════════════════════════════════

The Universal Type Space¶

%═══════════════════════════════════════════════════════════════════════════════

Universal Type Space

\tB\; The universal type space is [ T_{\mathrm{universal}} = T_{\mathrm{human}} \cup T_{\mathrm{non\text{-}human}} ] where $T_{\mathrm{human}} = \{\mathrm{ling}, \mathrm{spat}, \mathrm{soc}, \mathrm{symb}, \mathrm{mnem}, \mathrm{eval}, \mathrm{aud}, \mathrm{kin}\}$ and [ T_{\mathrm{non\text{-}human}} = {\mathrm{echo}, \mathrm{mag}, \mathrm{elec}, \mathrm{chem}, \mathrm{swarm}, \mathrm{therm}, \ldots} ] with the following non-human modalities: [nosep] - $I_{\mathrm{echo}}$: Echolocation---spatial-auditory fusion with no human analog (bats, dolphins). - $I_{\mathrm{mag}}$: Magnetoreception---geomagnetic navigation (migratory birds, sea turtles). - $I_{\mathrm{elec}}$: Electroreception---EM field sensing (sharks, platypus, electric eels). - $I_{\mathrm{chem}}$: Chemoreception beyond human---olfactory and pheromonal intelligence at sensitivities $10^4$--$10^6$ times human (dogs, ants). - $I_{\mathrm{swarm}}$: Collective/swarm intelligence---emergent computation from simple agents where no individual possesses the capability (ant colonies, bee hives, fish schools). - $I_{\mathrm{therm}}$: Thermal sensing---infrared imaging (pit vipers).

The ellipsis is deliberate: $T_{\mathrm{universal}}$ is open to extension as new modalities are identified. The universal intelligence vector is $\bI \in [0,\infty)^n$ where $n = |T_{\mathrm{universal}}| \geq 14$.

Remark

Humans have $I_{\mathrm{echo}} = I_{\mathrm{mag}} = I_{\mathrm{elec}} = I_{\mathrm{therm}} = 0$ and $I_{\mathrm{chem}} \approx 0.1$ (vestigial olfaction). A bottlenose dolphin has $I_{\mathrm{symb}} \approx 0$ but $I_{\mathrm{echo}} \approx 3.0$---superhuman in a type we lack entirely. An ant colony has $I_{\mathrm{swarm}} \approx 2.0$ while individual ants have $I_t \approx 0$ for all human types. The universal type space makes these comparisons precise rather than anecdotal.

%═══════════════════════════════════════════════════════════════════════════════

Three-Space Grounding: Species-Relative Basis¶

The three-space ontology (Part XIII) provides a crucial insight: the eight-dimensional IAG type space $\mathcal{T} = \{I_L, I_A, I_S, I_P, I_I, I_M, I_E, I_K\}$ is human-specific---it reflects the dimensions activated by the human sensory apparatus and neural substrate as projection channels from $\QS$ to $\PS$.

Different species activate different projection channels. Echolocation (bats, dolphins) may constitute a genuinely different dimension, not reducible to spatial or auditory intelligence as humans experience them. Electroreception (sharks, platypus), magnetoreception (birds, sea turtles), infrared sensing (pit vipers), and polarized-light vision (mantis shrimp) each opens a distinct $\QS \to \PS$ channel unavailable to humans.

Universal Intelligence Space

\tB\; The universal intelligence space across all species is: [ \mathcal{I}{\mathrm{univ}} = \bigcup_s ] where }\; s} \mathcal{I$\mathcal{I}_s$ is the intelligence space of species $s$. The human space $\mathcal{I}_{\mathrm{human}} = \mathbb{R}^{n(e)}_{\geq 0}$ is a subspace of $\mathcal{I}_{\mathrm{univ}}$, which may have $\dim(\mathcal{I}_{\mathrm{univ}}) \gg 8$. The shared space $\mathcal{I}_{\mathrm{shared}}^{s_1, s_2} = \mathcal{I}_{s_1} \cap \mathcal{I}_{s_2}$ between two species is generically impoverished---the overlap of two species' projection channels is smaller than either individual space.

This grounds the animal profiles below: each species' intelligence vector is expressed in its own basis, and cross-species comparison requires basis alignment.

Animal Intelligence Profiles¶

%═══════════════════════════════════════════════════════════════════════════════

The following profiles are the author's informed estimates in cogs, where $1.0$ = human baseline in that type. Calibration via comparative cognition literature is an open research program. The table shows 10 of the 14+ universal types; remaining types ($\mathrm{mag}$, $\mathrm{elec}$, $\mathrm{swarm}$, $\mathrm{therm}$) are zero for most species shown.

[Table — see PDF]

\caption{Intelligence profiles across substrates (selected types). The honeybee colony additionally has $I_{\mathrm{swarm}} \approx 2.0$ (omitted for space). Machine agents have $I_{\mathrm{aud}} = I_{\mathrm{kin}} = 0$ absent embodiment.}

\end{table}

Calibration Notes

Honeybee colony $I_{\mathrm{soc}}$ is set to $0.4$ (not zero): waggle dance, division of labor, and swarm decision-making represent sophisticated social cognition [Seeley2010]. Octopus $I_{\mathrm{soc}} = 0.15$ reflects limited but documented social learning [GodfreySmith2016]. These profiles are calibratable: systematic comparative cognition experiments following the framework in de Waal [deWaal2016] could estimate each entry to within $\pm 0.2$ cog precision.

%═══════════════════════════════════════════════════════════════════════════════

The Overlap Subspace¶

%═══════════════════════════════════════════════════════════════════════════════

Overlap Subspace

\tA\; For two cognitive systems $S_1, S_2$ with intelligence vectors in $T_{\mathrm{universal}}$, the overlap subspace is \begin{keyeq} [ T_{\mathrm{overlap}}(S_1, S_2) = {t \in T_{\mathrm{universal}} : I_{S_1, t} > \theta \;\text{and}\; I_{S_2, t} > \theta} ] \end{keyeq} where $\theta = 0.1$ cog is the activation threshold (Definition 4.2). Using $> \theta$ rather than $> 0$ avoids counting vestigial or negligible capabilities as shared modalities.

Cross-Species Overlap

\tA\; From Table ref:tab:animal-profiles: [nosep] - Human--dolphin: $T_{\mathrm{overlap}} = \{\mathrm{ling}, \mathrm{spat}, \mathrm{soc}, \mathrm{mnem}, \mathrm{eval}, \mathrm{aud}, \mathrm{kin}\}$. Dimension: 7. Rich shared basis for communication. - Human--bee colony: $T_{\mathrm{overlap}} = \{\mathrm{spat}, \mathrm{soc}, \mathrm{mnem}, \mathrm{eval}, \mathrm{aud}, \mathrm{kin}\}$. Dimension: 6. But the overlap strengths are low---the shared types have small $\min(I_{S_1,t}, I_{S_2,t})$ values. - Human--octopus: $T_{\mathrm{overlap}} = \{\mathrm{spat}, \mathrm{soc}, \mathrm{mnem}, \mathrm{eval}, \mathrm{kin}, \mathrm{chem}\}$. Dimension: 6. Notably includes $\mathrm{eval}$: octopuses solve novel problems [GodfreySmith2016]. - Human--Claude Opus: $T_{\mathrm{overlap}} = \{\mathrm{ling}, \mathrm{spat}, \mathrm{soc}, \mathrm{symb}, \mathrm{mnem}, \mathrm{eval}\}$. Dimension: 6. Missing: $\mathrm{aud}$, $\mathrm{kin}$ (no embodiment).

%═══════════════════════════════════════════════════════════════════════════════

Communication Bandwidth Bounds¶

%═══════════════════════════════════════════════════════════════════════════════

\[\begin{modelingprinciple}[Communication Bandwidth] \tB\; The effective communication bandwidth between two cognitive systems $S_1, S_2$ is bounded by the dimension and strength of the overlap subspace: \begin{keyeq} \[ \mathrm{BW}(S_1, S_2) \;\leq\; \sum_{t \in T_{\mathrm{overlap}}} \min(I_{S_1,t},\; I_{S_2,t}). \] \end{keyeq} Communication can only transmit content that both systems can encode and decode. Types outside the overlap are invisible to at least one party. Within the overlap, the weaker system's capability in each type limits the fidelity of transmission. \end{modelingprinciple}\]

Translation Loss in Cross-Species Communication

\tB\; Even within the overlap subspace, cross-species communication is subject to the Conceptual Irreversibility Theorem (Part V). The evolutionary filter mismatch ($F_{\mathrm{genetic}}^{S_1} \neq F_{\mathrm{genetic}}^{S_2}$) means the conceptual topoi of $S_1$ and $S_2$ have different subobject classifiers in the shared types. The round-trip [ S_1\text{-encoding} \;\to\; \text{shared channel} \;\to\; S_2\text{-decoding} \;\to\; \text{shared channel} \;\to\; S_1\text{-decoding} ] is necessarily lossy by the CIT, with loss proportional to the Heyting gap between the species' conceptual topoi restricted to the overlap types.

\[\begin{interpretation} Nagel's question "What is it like to be a bat?" [Nagel1974] receives a precise partial answer. The human--bat overlap subspace is $\{\mathrm{spat}, \mathrm{eval}, \mathrm{mnem}, \mathrm{kin}\}$---we share spatial, evaluative, mnemonic, and kinesthetic intelligence. But the bat has $I_{\mathrm{echo}} \approx 3.0$, a type we lack entirely. The *echolocative* aspect of bat experience is not merely difficult to imagine---it lies outside our overlap subspace, and therefore outside the bandwidth of any possible inter-species translation. The question is hard for mathematical reasons: zero overlap in a type means zero bandwidth in that dimension. \end{interpretation}\]

%═══════════════════════════════════════════════════════════════════════════════

The Effective Attention Simplex¶

%═══════════════════════════════════════════════════════════════════════════════

Species-Specific Simplex

\tA\; Every cognitive system lives on the universal attention simplex $\Delta^{|T_{\mathrm{universal}}|-1}$, but structurally occupies a face: [ \Delta_S = {\alpha \in \Delta^{|T_{\mathrm{universal}}|-1} : \alpha_t = 0 \;\text{whenever}\; I_{S,t} = 0} \;\cong\; \Delta^{|T_{\mathrm{active}}(S)|-1}. ]

Example

\tA\; The human effective simplex is $\Delta_{\mathrm{human}} \cong \Delta^7$ (8 active types, with $\alpha_{\mathrm{echo}} = \alpha_{\mathrm{mag}} = \alpha_{\mathrm{elec}} = \alpha_{\mathrm{therm}} = 0$). The dolphin effective simplex is also $\cong \Delta^7$ but spans a different set of 8 types: it includes $\mathrm{echo}$ but excludes $\mathrm{symb}$. The ant colony lives on $\Delta^3$ (4 active types: $\mathrm{spat}$, $\mathrm{kin}$, $\mathrm{chem}$, $\mathrm{swarm}$).

Cross-species comparison embeds both simplices into the universal simplex. The shared face---the overlap subspace---is the arena where mutual comprehension is geometrically possible.

%═══════════════════════════════════════════════════════════════════════════════

Evolutionary Filters¶

%═══════════════════════════════════════════════════════════════════════════════

\[\begin{modelingprinciple}[Evolutionary Filter Cascade] \tB\; Every biological agent's effective intelligence is its raw capacity passed through a cascade of evolutionary and developmental filters (Part I): \[ \bI_{\mathrm{effective}} = F_{\mathrm{genetic}} \circ F_{\mathrm{developmental}} \circ F_{\mathrm{cultural}} \circ F_{\mathrm{state}}(\bI_{\mathrm{raw}}). \] For non-human animals, $F_{\mathrm{cultural}}$ may be absent or minimal (though cultural transmission is documented in cetaceans, primates, and corvids). The dominant shaping force is $F_{\mathrm{genetic}}$: millions of years of selection pressure have tuned which types are amplified and which are suppressed. \end{modelingprinciple}\]

Primate-Specific Filters

\tB\; The human intelligence vector carries evolutionary priors: $I_{\mathrm{soc}}$ is disproportionately strong (hypersocial primates; Dunbar's social brain hypothesis [Dunbar1998]), $I_{\mathrm{ling}}$ is uniquely developed (recursive syntax), and $I_{\mathrm{kin}}$ is highly refined (precision grip, bipedal balance, throwing accuracy). These are not arbitrary---they reflect $\sim$6 million years of hominid selection pressure [Tomasello2014, Lieberman2013].

Dolphin Echolocation Filter

\tB\; The dolphin's $F_{\mathrm{genetic}}$ amplifies $I_{\mathrm{echo}}$ and $I_{\mathrm{aud}}$ while suppressing $I_{\mathrm{symb}}$ (no evolutionary pressure for symbolic manipulation in an aquatic environment without tool use). The result: a cognitive system that "sees" with sound at resolutions no human visual system achieves in murky water, but cannot perform even basic symbolic abstraction. The filter is not a deficit---it is an optimization for a different ecological niche.

%═══════════════════════════════════════════════════════════════════════════════

Cross-Substrate IdeaRank¶

%═══════════════════════════════════════════════════════════════════════════════

Cross-Substrate Comprehension

\tB\; Given an idea $x$ with requirement vector $R(x) \in [0,1]^{|T_{\mathrm{universal}}|}$, a cognitive system $S$ comprehends $x$ if and only if [ I_{S,t} \geq R_t(x) \quad\text{for all } t \in T_{\mathrm{universal}} \text{ with } R_t(x) > 0. ] The set of ideas comprehensible to species $S$ is bounded by the species' intelligence profile restricted to its active types.

Translation Loss for Ideas

\tB\; Transmitting an idea $x$ from system $S_1$ to $S_2$ incurs loss whenever $R(x)$ has nonzero components outside $T_{\mathrm{overlap}}(S_1, S_2)$, or whenever the conceptual topoi of $S_1$ and $S_2$ assign different truth values to the idea's content within shared types (CIT). Perfect idea transmission across substrates is impossible unless $S_1$ and $S_2$ share identical type spaces, identical intelligence profiles, and identical subobject classifiers---conditions never met across distinct biological species.

%═══════════════════════════════════════════════════════════════════════════════

Open Problems¶

%═══════════════════════════════════════════════════════════════════════════════

[leftmargin=2em] - Empirical calibration. Systematic estimation of animal intelligence profiles using standardized cross-species testing batteries (following the framework of de Waal [deWaal2016] and Griffin [Griffin1992]). - Swarm intelligence formalization. $I_{\mathrm{swarm}}$ is a property of the colony, not the individual. How does emergence theory (Part XII) apply when the "assembly" is a superorganism? - Unknown unknowns. $T_{\mathrm{universal}}$ may be incomplete. Are there cognitive modalities we have not identified because no known species uses them, or because we cannot detect them from outside? - Fossil cognition. Can the framework be applied retroactively to extinct species? Endocast analysis and paleoneurology provide partial data for $\bI$ estimation.

\ifx\STANDALONE\undefined \else