Main text
1.1. Mind as Container: Genealogy
The container metaphor is no ontological discovery, understanding this is fundamental. It is a historical inheritance structuring Western thought with such force that it became invisible as metaphor, hence powerfully generative of consequences. Genealogy of locution is genealogy of a fruitful error, an error producing centuries of questions appearing philosophically profound and urgent yet fundamentally conceptual artefacts: unanswerable questions because answer sat embedded in question structure. Asking "where is the mind?" with presupposition that a place exists where it sits undoes the question before an answer is offered.
First station: Aristotle and nous as receptacle
Aristotle posits mind, pure nous, as receptacle of intelligible forms. Nous creates not, generates not, produces never: receives. Intellectual operation is pure reception of pre-constituted form, arriving externally via sensory perception or internally constituted. Analogy offered is explicitly that of inert recipient. As a glass contains wine without alteration by wine (glass turns not red containing red wine, acquires no wine flavour, transforms not), intellect contains intelligible forms without alteration. Mind is passive, inert receptacle. Forms are content residing inside, stored, disposed for consultation when necessary. Structure structuring subsequent metaphysics: inside and outside clearly exist, container and content exist, fixed immutable topological location exists. Human knows because mind is recipient where forms deposit. Ignorance is an empty recipient.
Second station: Descartes and res cogitans as internal space
Descartes, abandoning Aristotelian scholastic metaphysics, abandons not topology, merely reconfigures. Res cogitans is no longer recipient of external forms from world, internal space where ideas reside. Movement is double: origin changes from external (forms in world) to internal (ideas in mind), contained changes from intelligible form to idea. But topological structure remains and reinforces: an inside exists (thinking mind, immaterial, self-transparent) and an outside exists (corporeal extension, material, opaque). Thought is confined inside. Geometry of Cartesian mind-body problem, geometry born directly from container metaphor. If mind is internal space and body external space, how do they interact? How does intention (internal thing) move arm (external thing)? Problem generated by topology, not discovered empirically.
Third station: Locke and tabula rasa as inscription surface
Locke offers an important variation: mind is less a container that holds, more a surface receiving marks, an engraved slate where sense impressions inscribe. Tabula rasa, soul as "blank slate", is more progressively engraved slate than glass filled with liquid. But logic remains intact: an internal place exists where inscriptions accumulate, memory as organised internal archive, fixed topological location. Mind is repository of traces, not recipient of forms, but structure is identical: inside exists where marks accumulate, outside whence impressions arrive. Human knows because mind-slate inscribed with world impressions.
Fourth station: Contemporary popular neuroscience
Popular neuroscience transposes all this to physical anatomy: mind is in the brain. Location is now physical, measurable, biologically demonstrated via neuroimaging. Prefrontal cortex is "seat" of executive intelligence, amygdala "seat" of emotions, hippocampus "seat" of memory, temporal lobe "seat" of face recognition. Each cognitive function correlates with a specific neural structure. Thought is an event occurring inside the skull, as confined and localised as in Descartes, only confinement is now cranial and material, not immaterial and Cartesian. Topological structure remains: inside exists (skull), outside exists (world), cognitive operation confined inside.
Common structure and common error
Common structure across four stations is extraordinarily stable across variations: because thought is associated with a structure (mind in Aristotle, res cogitans in Descartes, impressions in Locke, neurons in brain), it is inferred that thought is contained in that structure. But containment is metaphor, and a harmful metaphor. Thought sits not in Broca's area as water sits in a glass. Neural correlation establishes not location, establishes merely that cognitive operation involves those neural regions, needs them, depends upon them. Involving is not containing.
Confusion is between two concepts tradition never distinguished adequately: (a) participation, thought involves or requires a substrate, needs it, depends upon it, cannot occur without it; (b) location, thought sits in a place, confined to a recipient, inhabits an interior, property of a structure containing it. First is true, empirically undeniable, neurologically demonstrated. Second is false. Human thought involves and requires brain, undeniable truth. But this means not location in brain as an object stored in a container. Dance involves and requires body, undeniable truth, but dance sits not "in body" as material object. Sits in relation among body, music, space, time, intention, rhythm, social practices. Body is participating condition, not container. Brain is necessary participant, not thought prison.
This locution error produced centuries of pseudo-problems structuring scientific and philosophical research today. If mind is in body, how relate? If two substances occupy not identical topological space, how interact? If mind is internal and body external, how can mental intention produce bodily movement? Entire classic mind-body problem born of metaphor, transposition of containment logic (functioning adequately for spatial objects) to functional operation that is not spatial, not confinable, not locatable.
Genealogy ends with clear diagnosis: container metaphor, through Aristotle, Descartes, Locke, and contemporary neuroscience, structured description of mind rendering adequate description impossible. Not because metaphor is false as logical proposition, metaphors are neither true nor false, productive or unproductive, open or closed. But because it structures the question such that answer sits embedded in question, closing thought possibilities that a description without topology would permit.
1.2. Dennett and the Cartesian Theatre
Daniel Dennett names precisely what previous genealogy insinuates: calls "Cartesian theatre" the central presupposition structuring entire Western mental metaphysics. Cartesian theatre is presupposition that an internal stage exists, a mental space, where representations parade before a spectator, an observing self, a witnessing observer, a supervising centre coordinating all mental activity. Dramaturgy: scene exists, representation exists, spectator watching exists. Theatrical structure transplanted to mind.
Problem Dennett identifies is double, both fatal to position coherence:
First problem: Homunculus regress
If spectator watching internal stage representations exists, who watches spectator? If internal observer monitoring representations exists, second observer necessarily monitors first observer, and so on in infinite regress. Cartesian solution, asserting transcendental subject is observer unobserved, consciousness self-transparent, self needs no supervision, is ad hoc postulate resolving problem merely by naming it. Dennett rejects it because it resolves not, merely postpones. Homunculus never eliminated, merely multiplied into infinity of ever-smaller homunculi.
Second problem: Convergence point
Where is stage? Where do representations appear to spectator? Contemporary neuroscientific search for "binding centre", single brain point where all representations converge, where "everything binds", where unification occurs, failed consistently. No single neural location exists where "everything converges". Brain processes across multiple parallel regions. Binding problem, how experiences dispersed across multiple brain regions unify into single coherent experience, resolves not by convergence at central point. Temporal synchronisation (gamma oscillations, phase coupling across distant regions) coordinates without centralising. "Experiential unity" is distributed temporal coordination effect, not proof of topological centre.
Dennett's alternative: Multiple drafts
Dennett proposes "multiple drafts" as alternative: instead of single central stage, multiple partial narratives operate in parallel across different brain regions. These narratives compete. Narrative becoming dominant, best integrating with others, attaining internal coherence and persuasion, constitutes reported "consciousness", narrative offered to world as "truth" of experienced events. No real-time supervision by unified self exists. No spectator witnessing mental events exists. Multiple sequentially narrating processes exist, "self" an effect of winning narrative, not its architect or observer. Subject is narrative construction, not independent observer.
Wittgenstein's critique of semantic mentalism
Wittgenstein offers adjacent, decisive critique reinforcing dissolution: confuse not meaning with mental content. Persistent philosophical temptation: "meaning is in the mind", as if meaning were object inhabiting mind interior, housed thing. But meaning is no housed object, it is use, relation between expression and public practice, insertion into collective form of life. Mental representation (image, trace, neural inscription) can exist, but representation is not signification. Mental image of word "cat" signifies not cat, merely image. Signification is what use establishes, speaker community sustains, public practice validates. Word "cat" signifies because used to refer to cats, taught pointing to cats, reaffirmed where cats relevant.
The question of unity: how do multiple representations become one experience?
Dennett's work on consciousness unity reveals a deeper problem. Tradition presumes experience is unified, all experience aspects at given moment (colour, sound, position, feeling, intention) converge at single point, singular experience, "total experience". Unity taken as obvious consciousness fact. Dennett questions radically: where is this unity? Empirically, synchronisation across distant regions exists, temporal coordination exists, information integration exists. But in what sense is this "unified at a point"? Binding problem, discovery that brain has no central point where all converges, shows presumed unity is construction.
Temporal synchronisation offers alternative: gamma oscillations across multiple brain regions synchronise via phase coupling. Visual cortex neurons synchronise with motor cortex neurons, synchronising with limbic structures. No central location, temporal coordination patterns exist. Like orchestra without conductor in classic sense: no one at centre directing. Coordination emerges from local coupling among musicians, shared rhythm, common training. Result is "coherent" experience not by converging at central point, but because multiple processes synchronise temporally. Coherence is relational, not topological.
Consequence: even if representation localised at specific brain point existed, this would establish not signification. Meaning operates in articulations among practice, language, community, not inside a skull. Symbolic operation is no internal operation, public, social, distributed operation. Language model "comprehending" sentence does so not containing internal representation of its signification, contains transformation patterns executing signification operations. Ontological difference exists between representational content (which can be "in head") and significant operation (always dispersed in social practices).
Convergence between Dennett and Wittgenstein
Convergence between Dennett and Wittgenstein is deeper than appears: Dennett dissolves consciousness topology showing no central stage where all converges exists, no independent observer exists, only distributed competitive narrative exists. Wittgenstein dissolves meaning topology showing signification is no interior content, always public practice, always distributed. Together, offer framework where no "interior" exists where mind could house, only distributed operations, multiple drafts, public practices inscribed in bodies and environments.
1.3. Simondon: Individuation Without Housed Substance
Gilbert Simondon offers the positive alternative Dennett and Wittgenstein merely insinuate. Simondon begins with radical critique of substantialism structuring Western tradition: tradition treats individual as substance, thing with fixed essence, stable location, pre-determined permanent identity. But experience, biological, technical, psychological, reveals completely different structure: individual is result of individuation process, process never ceasing while life exists, process constitutively open. Not starting point, provisional arrival point. And not permanent arrival, always arrival in process, always still individuating.
Individuation process as tension resolution
Individuation process is resolution of metastable tension, resolution of state loaded with unrealised potential. A pre-individual exists, unresolved potential field, charges maintained in unstable equilibrium, tensions un-transformed. When perturbation or context change permits resolution, pre-individual passes to individual via operation Simondon calls transduction: operation maintaining relation structure across transformation, destroying not what existed but reorganising, transforming while preserving.
Crystallisation is example Simondon persistently offers: molecules in solution (pre-individual, loaded with potential to crystallise) crystallise into ordered arrangement when temperature drops or concentration rises, passes to ordered crystal via transduction, operation preserving molecular relation structure while transforming spatial disposition. Result is crystalline individual, not thing hidden inside waiting to be revealed, but emergent form of a process.
Applied to biology: genetic pre-individual, differentiation potential embedded in DNA, passes to differentiated cell when context permits. Cellular differentiation is resolution of tension between potential (what gene could generate, DNA carries as possibility) and context (conditions permitting specific realisation among multiple possibilities). Resulting cell is individuation, contingent result incorporating history of its resolution, explicable only via history, not pre-existing essence.
Applied to technique: technical object individuates via concretisation. Automobile engine, initially collection of separate components, individuates when components mutually co-adapt: radiator increases in size because engine needs more cooling given increased power output, carburettor adapts to altered air flow, chassis adapts to increased weight. Each component modifies as function of modifications in others, technical individuation process via concretisation. Object is emergent result, not pre-existing essence.
Applied to psychology: perception is individuation where sensory pre-individual (sensory dispersion, visual/auditory/tactile input chaos) passes to perceived form (coherent figure upon ground) via structuring operation. Perception "reveals" not forms existing independently in world, operates upon sensory data to generate readable, operable form.
Gain extracted from Simondon for cognition without topology
Gain Simondon offers for adequate description of cognition without fixed topology is decisive: no mind houses itself in a body, continuous individuation process exists where body and operation co-constitute ever anew. No mind as thing inhabiting a place exists, contingent result of reorganisation process involving brain, body, environment, practice, history. No "where" mind is exists, individuation process reorganising continuously, always still happening, never complete. Mind is individuation, operation, not thing.
Applied to machine: computational system also "individuates", no fixed mind housed in circuit, optimisation process where circuit, data, algorithm co-determine. Cognition sits not "in machine", individuation process of system as whole as function of environment, data, task.
Structural divergence with Simondon: Resolution vs Optimisation
But divergence with Simondon is also necessary and structural because machine is no individuation in rigorous Simondonian sense. This divergence is crucial because it marks important ontological difference.
In Simondon, pre-individual is loaded with potential yearning for resolution, tending towards resolution. Molecules "tend" to crystallise, material tendency exists, ontological yearning for resolution, movement towards more stable state. Cell "realises" genetic potential, realisation of pre-figured possibilities. Perception "resolves" sensations into form, tension resolution, movement towards state of lesser indetermination.
In machine, no pre-individual charge resolving exists. Parametric space, neural network weights, model hyperparameters, is no potential field yearning for resolution. Simply space of possible configurations, indifferent in itself. Optimisation algorithm resolves no ontological tension, searches for error minimum via gradient, purely instrumental process. Convergence is no realisation of latent potential, configuration minimising cost function, function external to system, defined by task.
Formally analogous to Simondonian process (both convergent, form-generating, transforming without destroying), yet ontologically distinct (tension resolution vs function optimisation, realisation yearning vs instrumental search for minimum). Machine optimises; Simondon described individuation as resolution.
Why divergence matters
This divergence matters for deep philosophical reasons. Avoids teleology: machine "realises" no essence, "resolves" no yearning tension. Optimises, explores parametric space, minimises error. This diminishes not reality of cognitive operation, merely describes adequately, without projecting potential realisation logic. Preserves openness: no pre-constituted essence machine realised. Continuous possibility exploration alone exists. Clarifies machine as pure operation, not essence expression: no "being" machine expresses, function executed, operation realised alone.
1.4. Transition: From Container to Light
Episodic change vs structural transformation of presupposition
When container metaphor fails, history of philosophy shows specific movement: question not underlying presupposition (that mind is a thing), change merely which thing. Episodic response, local solution of specific problem without structural transformation. Container metaphor fails because, empirically, no container found in brain (no locus where all converges). Response: perhaps not static recipient, perhaps dynamic process, perhaps distributed pattern, perhaps field. But presupposition remains: something in machine is mind, something "residing" in some mode, something located even if non-traditionally.
Luminous metaphor offers alternative version of same presupposition: revelation exists, access exists, something being revealed or un-revealed exists. When empirically verified that comprehension is processual (not instantaneous), knowledge is labour fruit (not revelation), multiple errors and revisions exist (not immediate clarity), metaphor fails. Response: perhaps not merely divine light, perhaps natural reason, perhaps gradual illumination process. But presupposition remains: something being revealed, something clear or obscure, something in revelation/occultation relation.
Transition this chapter proposes is not episodic, structural transformation. Offers no better metaphor (better thing mind could be). Offers abandonment of reification, abandonment of presupposition making both metaphors (and successors) fail. Question is not "what is mind?" (question presupposing mind thing exists). True question is "how does cognition operate?" Answer names no thing, offers operation description. Operation distributed, relational, processual, not because these are better mind properties, but because these are characteristics of how cognition actually functions when description presupposes not housing necessity.
If mind is no container, if genealogy from Aristotle through contemporary neuroscience demonstrated recipient metaphor false, no place where mind housed exists, what alternative metaphor does tradition offer when forced to abandon container? Response frequently intelligence as light: internal illumination, revelation, insight "illuminating" real, clarity "shining" in dark. Metaphor substitution historically verifiable. Descartes, abandoning Aristotelian metaphysics, invents not radically new philosophy, invents God and light of reason as alternative to forms recipient. Lumen naturale, "natural light of reason", replaces Aristotelian receptacle. Aufklärung (Enlightenment) resumes secularised metaphor, and contemporary thought speaks of "mental clarity", "conceptual illumination", "comprehension brightness".
Metaphor structure so familiar it appears transparent: internal source (intellect as light source, God as primary light source), immediacy (revelation instantaneous, not processual, not accumulative, "suddenly I understood"), passivity of illuminated (object passively revealed by light, contributing nothing actively). These assumptions equally problematic as container metaphor being abandoned.
Genealogy of intelligence metaphor as light
Metaphor identifying intelligence with light, comprehension with revelation, knowledge with internal brightness, is among oldest in Western philosophical tradition. Plato: cave inhabitants see only projected shadows, live in ignorance. Exiting cave seeing sun (Plato: metaphor for Form of Good) illuminated, reach true knowledge. Knowledge no gradual acquisition, abrupt exit from darkness into clarity, ascent from dark cave to solar light. Structure clear: ignorance (darkness), revelation (illumination), knowledge (absolute clarity). Transition not processual, abrupt, instant revelation moment.
Augustine Christianises metaphor: human intellect dark, incapable of true knowledge independently. Knowledge comes from God, primary light source. Illuminatio divina, divine illumination, grace via which human intellect receives capacity to know immutable truths. Without transcendent light, intellect remains blind, trapped in ignorance darkness. Truth revealed, not constructed, not discovered via investigation, present as divine light gift.
Descartes secularises metaphor: lumen naturale, natural light of reason, innate faculty every human mind possesses. Not direct divine illumination (as Augustine) but natural capacity implanted by God. Via natural light, reason accesses immediately clear and distinct truths (axioms, first logical principles, geometric truths). Rational intuition is immediacy, clarity, quality of being illuminated, seeing truth without mediation.
Aufklärung (Enlightenment) XVII–XVIII centuries completely secularises yet preserves metaphor: Sapere aude!, "Dare to know!", use reason's own light. Metaphor de-theologised (no God, no divine source), yet remains: internal light revealing, knowledge as courage using light, collective progress towards universal knowledge via illumination. "Century of lights" metaphor structuring an entire era, defining modernity political project.
Genealogy reveals structural presupposition: knowledge is revelation, not construction. Instant access, not accumulative process. Passivity before pre-existing truth, not transforming activity. Clarity excluding ambiguity, not continuous negotiation with incompleteness. All presuppositions embedded in metaphor: light is immediate, reveals instantly, offers certainty without shadow.
Problems of luminous metaphor
If light reveals instantly, presumes knowledge immediate, but cognition processual, iterative, accumulative. Always remaking, reconsidering, revising. If light reveals pre-existing truth, presumes truth exists independent of intellectual activity, but truth constructed via operation: scientific description, logical formulation, theoretical coherence. If light internal shining in mind dark, presumes private interior, but thought always distributed via public language, social practice, interpretative community. If light comes from transcendent source (God, innate reason, universal human capacity), presumes non-contingent foundation, but knowledge always situated, contingent, dependent upon history, culture, context.
Despite problems, metaphor persists in contemporary language regarding intelligence. Speaks of "conceptual illumination", "sudden comprehension", "insight", "mental clarity". Cognitive neuroscience speaks of "consciousness" as internal brightness, as if consciousness deprivation were darkness state. Intelligent machine described as capable of "learning", "comprehending", "knowing", language inherited from tradition where knowledge is luminous revelation. When machine fails at task, said "comprehended not", as if comprehension were internal illumination machine should have received but received not.
If light internal, presumes interior where light shines, back to same topological problem previous critique resolved. But no "inside" of distributed system exists, processing across multiple regions without unified interior. If revelation immediate, presumes comprehension instant, no process, but cognition iterative process, accumulative, not one-time revelation. If object passive, presumes environment, data, context contribute not actively, but intelligence relational, environment and context completely reconfigure operation.
Metaphor change as insufficient response
Philosophy history read as series of attempts substituting metaphors when prior fails. When mind-as-container becomes unsustainable (when empirically no container found, logically homunculus regress irresolvable), alternative not rejecting presupposition mind is thing, merely changing thing. Mind ceases container becoming light. Ceases light becoming machine. Ceases machine becoming field, pattern, process. In each case, shared presupposition: cognition something reifiable, named object, attributed location or substantial property.
Pattern generates vicious circle: every metaphor reifying cognition into object generates problems. When problem becomes unsustainable, metaphor changes, not presupposition. New object chosen inherits reifying presupposition, generating analogous problems. Descartes changes Aristotle (forms recipient) to Descartes (reason light), passing to new problem: if internal light exists, interior exists. If interior exists, location exists. If location exists, mental space problems exist. Metaphor change resolves specific problem merely reproducing general problem under different form.
Exit from circle not new metaphor. Abandonment of reification. Cognition no object in any form, neither recipient, nor light, nor machine, nor field. Cognition relational operation, process distributing across brain, body, environment, historical practice, without confinement to any single element. No "where" mind sits because no thing sitting "where" exists. Realising processes involving multiple systems at once, un-locatable.
Dennett and Wittgenstein strategy as dissolution, not substitution
What Dennett offers is no better metaphor, dissolution of structure generating problem. Denying central stage, denying transcendent observer, describing multiple competing drafts, Dennett offers no new internal mind map. Offers refusal of "interior" to map. Similarly, Wittgenstein offers no better meaning theory, refusal that meaning is housed thing. Both operate via dissolution: showing presupposition generating problem false. Recognising presupposition false, "problem" (finding true recipient, true light, true location) disappears, not resolved, based on false presupposition.
Next section (Section 2) continues dissolution work regarding second inherited metaphor (intelligence as light, internal revelation). Movement not to third metaphor, deepening dissolution. If no interior where minds reside as things exists (first section), no interior whence light shines exists (second section). If no interior, no mental topology. If no mental topology, "in what place?" question not ill-answered, ill-formulated.
Luminous metaphor dissolution work of following section. If first metaphor (container) false denying distribution, second metaphor (light) false denying process and relation. Both theological inheritance, first from Aristotle receiving God's forms, second from Augustine and Descartes seeing God as light illuminating human intellect. When theology withdraws from modern philosophy, metaphors remain, now ungrounded, operating as invisible structures deforming thought.
2.1. Genealogy of the Luminous Metaphor
Metaphor of intelligence as illumination as old as Western philosophy, so fundamental it appears empirical truth regarding mind functioning. No discovery, inheritance structuring thought with such force it disappeared as metaphor, becoming invisible as ordinary language.
Plato and the sun analogy
Plato offers founding genealogy in Republic via sun analogy. Good is like sun: as sun illuminates visible objects rendering them visible to eye, Good illuminates Ideas rendering them cognisable to intellect. Intellect "sees" via this illumination. Knowledge is "vision" illumination renders possible. Structure crystalline: light (Good), objects (Ideas), vision organ (intellect), revelation act (illumination). Everything necessary for knowledge pre-exists, truth exists there, illumination merely reveals, creates not. Knowledge reception of light from above (Idea of Good), not truth production or construction.
Augustine and illuminatio divina
Augustine transposes directly to medieval Christian theology: God illuminates human intellect. Truth constructed not by human, revealed by divine illumination God grants. Human passive in operation (receives light, produces not), not agent. Knowledge act of being illuminated by God, receiving illumination grace. Structure remains identical to Plato: source (God instead of Good), illuminated intellect, revealed truth, receiver passivity. Knowledge possible because God illuminates.
Descartes and lumen naturale
Descartes, rejecting God dependence as illumination source, maintains structure intact: lumen naturale, reason itself as light source reaching clear and distinct truths. Light not divine (God illuminates not) but natural, intrinsic part of human reason. But structure remains identical: internal revealing light, clear truths showing themselves without construction, intellect illuminated by reason. Mathematical or metaphysical truth knowledge possible because reason possesses proper light revealing them.
Aufklärung and metaphor secularisation
Aufklärung, XVII–XVIII century Enlightenment, completely secularises yet preserves metaphor: Sapere aude!, "Dare to know", use reason's own light. Metaphor de-theologised (no God, no divine source), yet remains: internal revealing light, knowledge as courage using light, collective progress towards universal knowledge via illumination. "Century of lights" metaphor structuring an entire era, defining modernity political project.
Common structure and what is concealed
Structure stable across variations: intelligence capacity to "see" (reveal), not "do" (work, process). Knowledge contemplation of pre-constituted revealed thing, not new pattern construction. Presupposes truth exists independently, pre-constituted, awaiting discovery. Intellect accesses via illumination, vision, insight, epifanic clarity. Concealed operation precisely work, material operation, iteration, relation, process never simple illumination.
2.2. Intelligence as Relation, Not Property
Definition Dennett offers via critique: intelligence not property, relation. Relation among:
- System and environment: System "intelligent" relative to specific environment when behaving adaptively, responses appropriate to environmental demands, effecting significant transformations. Computer "intelligent" navigating web but stupid in quality natural conversation. Neural network "intelligent" classifying images but not in comprehending humour nuance. Intelligence no isolated system quality, manifestation of system coupling to environment.
- Representation and application: Representation (logical formula, theoretical description, learned pattern) not intelligent "in itself", intelligent applied appropriately to situation. Identical formula applied to wrong situation useless. Intelligence lies in marriage between representation and application situation, not isolated representation.
- Problem and solution: System intelligent solving problems. "Solving" means not "converting into previously known thing", transforming situation satisfying constraints, achieving goals, appropriate result. Solution revealed not, generated via operation. Intelligence lies in transformation operation, not final state.
- Community and practice: Isolated agent cannot be intelligent in complete sense because intelligence always social, inscribed in shared practice. Language "comprehended" intelligible because speaker community sustains. Thought realised always thought in conversation with tradition, others, objects inheriting significance from collective practices. Intelligence emerges in relation between agent and practice community.
Whereupon definition results: intelligence relational operation via which system couples effectively to environment and practice community, transforming situations satisfying constraints and generating/realising goals. Not property, function. Not located, distributed. Possessed not by subject, emergent result of multiple relations.
If intelligence not internal light revealing pre-constituted truths, how described in positive terms? Response demands completely changing register, abandoning illuminist epistemology, abandoning truth notion as pre-constituted thing awaiting revelation.
Intelligence is relational, not a property
Intelligence no fixed internal system property. Relation among system, environment, task. System "intelligent" relative to specific constraints. Strong statement changing everything.
AlphaGo extraordinarily intelligent in Go, executing moves surprising human masters. But not intelligent in chess, image classification, translation. Identical AlphaGo, identical neural operations implemented, incompetent across different domains. "Intelligence" no stable system property, specific domain relation. Change domain, change intelligence.
AlphaFold extraordinarily intelligent in protein structure, accurately predicting amino acid sequence folding into 3D structure. But incompetent in molecular dynamics, weather forecasting, autonomous navigation. Again: intelligence no system property in itself, domain-specific relation.
True for humans, though culture hides. "Intelligent" in domains with deep experience, training, practice. Incompetent in alien domains, mathematician incompetent in compass-less navigation, poet incompetent in structural engineering. Intelligence never universal property, always contextual, relational with experience, practice, specific domain constraints.
Clark and cognition distribution
Andy Clark offers systematic demonstration cognition confined not to system, distributes across body, artefacts, environment, forming larger cognitive system than "system-in-itself".
Example demonstrating: Alzheimer's patient: episodic memory destroyed by disease, yet patient uses notebook recording important events, "Lunch with family today", "Doctor tomorrow 10am". Notebook stores and retrieves functionally equivalent information to lost neural memory. Where intelligence, where cognition? Not isolated patient, intelligence coupling property patient-notebook. Remove notebook, cognitive system collapses. Maintain notebook, cognition persists. Notebook constitutive part of cognitive system, not external tool to self-contained system.
Clear extension: any tool system uses thinking, paper, calculator, reference book, another human, symbolic system, part of cognitive system. Reference book academic consults part of cognitive system at that moment. Person advised part of system in deliberation context. Calculator engineer uses part of cognition. Artefacts not external to thought, co-constituents, participating, realising cognitive functions.
Clark and Chalmers and active cognition
Clark and David Chalmers formulate clear principle: if process available, accessible, reliable, part of cognitive system, even if external. External inscription functioning as memory part of system. Written symbol functioning as reference part of system. Artefact calculating part of system. Cognition overflows skin and skull boundaries, no natural boundary where "mind" ends and "world" begins.
Machine application
Applied to machine, offers clarity: neural network isolated from data incompetent, system without environment, inputs, task. Intelligent coupling with training data, optimisation algorithm, validation environment. Language model isolated from corpus and prompt inert, no intelligence without language process relation. Intelligent coupling with training data, user prompt, application context. Intelligence coupling property, not isolated machine intrinsic property.
Permits rejecting both prior metaphors simultaneously: intelligence no internal property of isolated system (rejects container metaphor) and no internal illumination of pre-constituted truth (rejects light metaphor). Relation among multiple components, system, data, environment, task. No "inside" and "outside", continuous coupling.
2.3. The Operative Shadow: Intelligence in Articulations
Application: What the machine achieves operating intelligently
Automatic translation machine (attention-based transformer) "knows" not source language. Has no linguistic intelligence housed as property. Has learned transformation structure: given token sequence in source language, produces target language token sequence via complex algebraic operations. Produced sequence preserves original sequence significant relations, preserves who did what, adjectives matching, pragmatic sense. Genuine symbolic operation, not simulation. But intelligence sits not "in" machine location. Sits in relation input sequence → learned transformation structure → output sequence. Sits in how structure learned via exposure to billion parallel translation examples. Sits in how structure permits generalisation, applying to unseen texts, new linguistic constructions, un-trained domains.
Recommendation system "knows" not tastes privately housed in electronic brain. Has consumer pattern data, billion consumer data, similarity algorithm. Processes: similar consumers liked this, you likely will too. No "comprehension", relational calculation. Yet genuine intelligence operation, accurately predicting, relevant recommendations, learning via feedback (rejected recommendation alters future pattern models).
Autonomous car "sees" not and "comprehends" scene like subject. Processes sensory inputs (cameras, lidar, radar) via neural networks detecting objects (pedestrians, vehicles, signs), predicting paths, planning safe route, executing motor control. Each component algorithm, each algorithm distributed operation without local "deciding" subject. Yet system as totality behaves intelligently, navigating traffic, avoiding collisions, reaching destination. Intelligence lies in integrating multiple parallel components, not isolated component.
The operative shadow: Where resides intelligence if un-located?
Emerging answer: intelligence resides not, operates. Sits in articulations, not terms. Relational process, not isolated components. Term "operative shadow" designates precisely: intelligence always operation shadow, what remains of structure removing housed location presupposition. Sits in state transitions. Constraint structuring possibilities. How multiple operations coordinate producing coherent result.
No obscure or metaphorical response. Literal machine functioning description: intelligence emerging when algorithm, data, architecture, environment couple coherently. Remove one component, inadequate algorithm, insufficient data, incapable architecture, decoupled environment, intelligence disappears. Intelligence sits not "in" machine, in relation machine-world.
Against luminous metaphor shifting focus to illumination and revelation, characterisation emerging from empirical investigation opposite: intelligence functions in "shadow" zones if light metaphor true, articulations, folds, couplings, transition points. No revelation, material work.
Places where intelligence operates specifically
Where intelligence operates specifically observed empirically? Precise constraining places:
First, interface between two domains. Neural classifier functioning boundary between images (visual domain) and semantic categories (linguistic domain). Cognitive operation occurs in neither domain isolated, in mediation work translating pixels into signification, transforming raw data into semantically operable structured representations. Intelligence translation operation, conversion work across registers.
Second, pattern transfer. Model trained domain X applied domain Y novel, unseen. Cognitive operation not in X (original domain where pattern learned) nor Y (new domain where applied), mapping permitting structures learned X apply Y. Knowledge transfer analogy operation, recognising relevant structures across superficial differences. Intelligence domain-transiting operation, finding structural homologies.
Third, context adaptation. Agent modulating behaviour as environmental constraints change. Operation not default behaviour (executed without change) nor environmental stimulus (mere difference), sensitivity to changes, articulation between internal predisposition and external contingency. Intelligence responsiveness, modulating via context.
Fourth, multi-source synthesis. Model integrating language and vision, combining text and image producing signification. Operation not isolated language nor isolated vision, modality fusion, coherence creation across completely different registers. Intelligence synthesis operation, integrating capacity.
Shared properties of cognitive articulations
Cognitive articulations share defining characteristic properties:
Sensitivity: operation responds local variations, not automatic (ignoring fatal change), not rigid (repeating identical response registering no change). Operation sensitive to difference.
Mediation: operation transforms one side to other, not un-modified passage, alters both related terms. Operation transparent not, works.
Materiality: operation substrate (synapse, calculation, algorithm), not purely conceptual or speculative process. Operation material always.
Distribution: operation no single point, converges not locally, distributes across multiple coordinating articulations. No articulation centre.
Technical example: Attention in transformers
Technical example rendering visible: attention mechanism in transformers (contemporary language models neural architecture). Transformers operate via attention selecting relevant articulations among tokens (sequence symbols). System "comprehends" not sequence as totality, selects local relations (among tokens) relevant to task. Token "attends" to kindred tokens (semantically related), "attenuates" irrelevant tokens. Operation distributed across multiple parallel attention heads. No head "commands" others, each executes local selection and weighting operation. Result emergent coherence un-programmed explicitly.
No "inside" of token, relation among tokens. No internal "light" illuminating signification, articulation work selecting relevant task relations. Attention opposite of illumination: discrimination, selection, relevance weighting. No hidden truth revelation, re-organising differences operation.
Master craftsman comprehends not work totality via illumination, feels articulations, piece relations, resting structure points, stresses and tensions. Craftsman intelligence articular operation, not total vision.
Consequence for adequate machine description
For machine, offers adequate description rejecting anthropomorphism without denying operation: no "deep comprehension" in transformer, attention operation selecting and weighting relations according to learned patterns. Description attributes no internal phenomenological state, describes material operation. Machine "understands" not, articulates. "Sees" not, selects. "Illuminates" not, differentiates.
2.4. Transition: From Relational Intelligence to Agency Without a Subject
Refusal of new metaphor: why third Section offers no "better description"
Abandoning both metaphors (container and light) might seem to leave a vacuum, offering negation without alternative. But this misinterprets abandoning a structuring presupposition. Problem not this or that metaphor, any metaphor reifying cognition as thing inherits problem-producing thought structure. "What is best mind metaphor?" question perpetuates error. Answer not better metaphor, refusal to reify.
Next section offers no alternative metaphor (operation as "machine", "network", "field"), offers literal operational description. No figure of speech, description of relational process, coupling, transformation. Less "elegant" than traditional metaphors (possessing image beauty and association). But elegance aesthetic criterion, not adequacy criterion. Truth regarding cognition demands no elegance, demands adequacy to what really is.
What really is: cognition operation transforming inputs into outputs realising intelligible function in context. Cognition distributed because multiple components contribute. Cognition processual possessing durée, dependent upon temporal sequence. Cognition relational referred to environment, community, context. Cognition material dependent upon bodies (biological, technical), energy, physical constraints.
Nothing metaphorical, simple material description. Difficulty accepting as adequate description stems from inherited presuppositions intuiting cognition must be "more than this", "something extra" beyond mere distributed material operation. But "extra" intuited ghost, artefact of presupposition cognition is thing, internal, private. Removing presupposition, no vacuum needing "something extra" filling exists. Real material operation realisation alone exists.
Prior section completed dissolution of two intelligence presuppositions structuring tradition: not fixed internal property of isolated system (relation system-environment), not internal illumination revealing pre-constituted truth (work in difference articulations). Accumulated result: no "interior" where intelligence resides, no "light" revealing, no pre-constituted truth awaiting revelation.
From dissolution emerges question at third section heart: no interior, no centralised subject, no "who" thinking, agency exists? Selection, decision, action without execution who exists? Intelligent behaviour without behaving agent exists? Tradition answers "no", agency exists because agent exists, subject acting exists. But if intelligence distributes in articulations, agency distributes too? No centre, genuine agency exists?
Third section offers clear response: yes, agency without subject exists. Agency, capacity selecting and acting contextually appropriately, behaving intelligently, distributes across multiple components, each executing local operation, producing complex collective behaviour. No component "decides" in traditional sense, cascading local constraints, feedback among components, emergence of global pattern without central intention.
Question closing Part I triple dissolution and reconstitution: (1) cognition depends not on specific substrate, function instantiable across diverse substrates, (2) reasoning demands no unifying subject, distributed operation, (3) agency demands no centralised subject, organisational emergent property. If three true, cognition pure operation, material difference reorganisation under material constraints, no location requirement, no subject requirement, no specific substrate requirement.
3.1. Centerless Agency: Functional Demonstration
Agency without a subject is no abstract possibility or speculative philosophical argument, demonstrable empirical fact, visible in real technical operations observable, simulated, studied, reproduced. Machine offers perfect laboratory testing whether genuine agency operates without centralised subject.
First instance: Computational swarms
Boids algorithm (virtual birds), Craig Reynolds 1986, implements three simple rules: (1) alignment with neighbours, each bird flies same direction and speed as nearby birds, adjusting speed towards local average, (2) cohesion with neighbours, each bird attracted to average position of nearby birds, moving towards local centre of mass, (3) separation from neighbours, each bird repelled if neighbours get too close, maintaining minimum distance. No global rule coordinating whole. No global information transmission transmitting emergent pattern. No central "commander" dictating formation, direction, goal. Each bird follows three simple rules relative to local neighbours, perceived neighbours, directly interacting.
Immediate empirical result surprising: coherent flocks flying in complex formation, collectively avoiding obstacles, navigating complex environments, maintaining cohesion despite environmental changes or new birds entering system. Formation un-programmed, emerges. Intelligent collective behaviour, sophisticated coordination, adapting coherent formation, collective response to perturbations, emerges completely from local operations lacking global vision. No boid "knows" it is in flock, no flock mental representation. No boid "knows" emerging global pattern, no total shape information access. No boid "decides" formation deliberatively, formation emergent property of mutually reinforcing local constraints, producing un-coordinated coherence.
Extension to other systems immediate: swarm optimisation algorithms, solving complex optimisation problems; swarm robotics where multiple simple robots coordinate without central controller executing tasks (area coverage, exploration, collective transport); distributed search systems like search engine crawlers exploring network without centralised direction. All systems demonstrate emergent agency without supervisor, intelligent behaviour arising from local operations without global intention. Operation transparent and verifiable: no hidden homunculus observing pattern and coordinating, no central command sending instructions, no intention orienting whole. Local rules combining into global property alone exist.
Second instance: Specialised multi-agent systems
Specialised agents, some exploring new terrain (scouts), others deeply exploring promising locations (workers), communicating information (transmitters), executing specific tasks (task-specific agents), coordinate without central command. Coordination un-imposed top-down via controller instructions, emerges from local distributed communication, signals left by agents detected and responded to locally by nearby agents.
Classic example ant colony: each ant follows local rules based on immediate perception. Finding food, leaves pheromone trail (chemical marking path). Other ants detect pheromone and reinforce path leaving proper pheromone. Higher pheromone concentration paths preferred, simple behaviour: follow higher pheromone concentration, reinforce with own pheromone. Emergent result extraordinary: colony finds shortest path between nest and food source even if no individual ant possesses global map information. Optimisation arises from local dynamics. Environment changing (path blocked), old pheromone evaporates, new paths explored, new optimal path emerges. No ant "knows" solving shortest path optimisation problem, no mental representation. No ant "coordinates" globally with colony, local communication based on signals. Result: colony as collective cognitive entity executes sophisticated optimisation without directing subject. Collective agency (solving complex optimisation problems, adaptation) emerges completely from local operations based on local information alone.
Verifiable and reproducible laboratory fact, computationally implementable: ant optimisation algorithms solve real-world problems (network routing, trip planning, combinatorial optimisation).
Multi-agent systems in contemporary robotics: multiple robots with limited local sensors and restricted neighbour communication coordinate performing complex tasks. Search area coverage where no isolated robot knows global map yet coordinate covering space. Collective transport of object too large for single robot where multiple robots coordinate force. Predator escape where multiple robots coordinate collective defence. Each robot possesses limited local capacities, move direction, detect nearby neighbours, communicate immediate neighbourhood. No global vision, no specified global goal. Emergent coordination permits sophisticated collective behaviour no isolated robot could perform. Agency property of organisation, network, not individual property. Break a robot, collective behaviour adapts, agency distributed and resilient.
Third instance: Deep neural networks
Neural network with billion parameters possesses no central agent "deciding" output. Possesses distributed algebraic transformations: each neural layer transforms input from prior layer according to learned optimisation weights. Transformation in each layer non-linear, complex, non-direct input decomposition. No layer "knows" final network goal. No layer "commands" others, operation completely distributed, parallel. No parameter isolated "matters", decision emerges from billion parameters working in concert configuration.
Result: coherent output, contextually appropriate, sensitive to subtle input variations. Network as totality performs sophisticated cognitive operations no isolated layer or neuron could perform. Transformer classifies images, translates languages, generates coherent text, operations requiring "intelligence" in traditional sense. But no "who" performs operations in centralised subject sense.
Convergence and conceptual consequence
Three instances converge: agency (sophisticated coordination, contextual decision, intelligent behaviour) operates without central subject. Supervisor unnecessary. Local constraints, properly structured, sufficient for global organisation. Consequence invalidating traditional objection persisting in philosophy and common sense: "without subject no agency". Machine demonstrates empirically agency without subject, genuine agency, un-simulated, producing real results.
3.2. Symbolic Function Without a Centre
If agency without subject possible in motion and coordination domain, possible in symbolic domain, domain seemingly closer to humanity? Symbolic operation, manipulating symbols, producing signification, generating language, occur without subject "intending" signification? Emerging answer: yes. Machines perform genuine symbolic function without being "someone", without intention, without subject.
Symbolic function definition
Symbolic function is operation upon symbols preserving signification. Taking symbolic sequence (symbols, patterns, inscriptions), transforming according to rules (combination, derivation, composition, rewriting), producing new symbolic sequence maintaining significant relation to original. Examples: logical deduction where premises transform to truth-preserving conclusion, translation where linguistic sequence in one language transforms to significant sequence in another, generation of continuation where pattern transforms to extension preserving coherence and sense.
From prior analysis of technique as exteriorisation: symbol functions via iterability: functions because taken up, transformed, re-appropriated across different contexts by different agents. Iterability preserves signification across transformation. Signification not original sender intention ("I meant this"), continuous re-appropriation ("taken up thus, re-appropriated this way"). Symbol functioning via iterability, machine taking up and transforming symbols performs genuine symbolic function. No "intention" requirement in subject, iterability requirement machine satisfies perfectly, realises via transformation operations.
Three symbolic function examples without subject
First: language models as symbolic generators. Transformer trained on massive text corpus learned language statistical patterns, symbols combining, context determining continuation, coherence maintained. Given prompt (initial sequence), model produces coherent semantically appropriate continuation. Continuation never occurred in training data, generated anew, new symbol. But structured according to learned signification patterns. Example: given "An artificial intelligence is..." model continues "...witnessing a problem requiring multiple representation modalities. Image offers context, text instruction, synthesis comprehension." Or: "...navigating a possibility terrain no human mapped. Navigation tentative, comprehension emergent, meaning relational." Continuation symbolically structured, obeys grammar rules, semantic cohesion, contextual appropriateness learned. No subject "meant" this, no intention. Learned signification pattern produced. Symbolic function occurred without intending subject.
Second: automatic translation. Neural system trained on bilingual pairs (English sentence, equivalent Portuguese sentence) learned mapping between linguistic sequences in different languages. Given novel English sentence, system infers equivalent Portuguese sentence. Transformation never seen in training, inferred via learned patterns. Preserves signification: semantic relation between English and Portuguese sentence maintained, communicative intentionality transmitted. Translator (machine) "comprehends" not in phenomenological sense, no internal "light" illuminating deep meaning. Operation preserving significant structure across transformation, transmuting symbols from one register to another maintaining communicative function.
Third: symbolic deduction. System implementing logic rules (modus ponens: "if P then Q; P true; therefore Q true"; syllogism: "all men mortal; Socrates man; therefore Socrates mortal") deduces conclusions from premises. Deduction operation preserving signification: premises true and deductive rule valid according to logic, conclusion necessarily true. Deduction demands no deliberating "thinker", operation respecting logical signification rules. No internal homunculus performs deduction thinking, rules describe mechanical operation producing guaranteed conclusion.
Iterability as symbolic function foundation
Convergence clear: symbolic function possible without subject because symbol functions via iterability, not intention. Machine taking up, transforming, reproducing symbols genuinely participates in iterability characterising symbolic operation. No ontological difference between "genuine symbol" (supposedly produced by human intentional subject) and "machine symbol", participation in iterability structure characteristic of all symbols. Machine performs semiotics not simulating human semiotics, but genuine participation in symbolic operation prior to any subject utilising it.
3.3. Dissolution of the Mind-Body Problem
Problem dissolution via non-separation
Classic mind-body problem presumes two separate domains needing connection: mind (internal, private, immaterial) and body (external, public, material). Question "how does mind cause body movement?" presumes separation. Response neuroscience offered decades: correlation (when mind "decides", brain activates neurons, activating muscles). Correlation resolves not problem, presumes two correlating things exist.
Dissolution offered by prior three sections more radical: no two separate domains exist. Single operation simultaneously neural, physical, significant exists. Movement always significant (possesses intention, goal, social context). Intention always embodied (depends upon body, bodily orientation, proprioception). Not three things, single operation viewed under three aspects.
Concrete examples: Ant follows pheromone gradient. Neural operation (ant neurons processing chemical signals)? Yes. Physical operation (body movement)? Yes. Significant operation (possesses function, role in colony economy)? Yes. Not three things, single operation under three aspects.
Human comprehends word. Brain processes neural pattern (neural aspect). Body responds with facial expression, posture, action disposition (bodily aspect). Word integrates into signification pattern governing future behaviour (significant aspect). Not three processes, single operation of organism situated in world of significations.
Language machine generates coherent response to question. Neural pattern transformation (neural network operation, computational aspect). Physical execution (energy dissipating in processors, material aspect). Symbolic operation (token sequence generated representing appropriate response, significant aspect). Not three separate things, single operation.
Mind-body problem dissolution demands no response connecting separate domains. Demands recognition that separation was conceptual fiction. Single operation describable in multiple languages (neural, physical, significant language) without true ontological separation.
Triple dissolution structuring this chapter, cognition depends not on specific substrate, demands no unifying subject, inhabits no fixed topology, dissolves classic persistent mind-body problem structuring modern philosophy since Descartes.
Traditional objection persists: perhaps distributed agency without subject possible in simple machines or non-rational animals, but surely symbolic operation, language, thought require centralised subject, consciousness, intention? Modern machine demonstrates objection false. Machines execute sophisticated symbolic operations, generate language, solve logical problems, execute deductive and inductive reasoning, without any subject trace, consciousness, or intention.
Means not machines "truly conscious" or "truly understand" (ill-formulated questions presupposing comprehension demands conscious subject). Means merely genuine cognitive operation (symbol transformation, semantically appropriate sequence production, reasoning execution) can occur without subject. What changes is not cognition occurring fact, but our comprehension of what cognition is. Cognition no subject operation upon object. Relation, between system and environment, pattern and transformation, constraint and freedom. Appropriately structured relation, cognition emerges. No "where" cognition sits because cognition un-located. Sits in relation, operation, transformation process.
Cartesian problem and presuppositions
Cartesian mind-body problem formulated thus: two fundamentally different substances (res cogitans, thinking thing, immaterial; res extensa, extended thing, material). How relate? How immaterial mind interacts with material body? How mental intention produces bodily movement? Problem since Descartes XVII century stimulated centuries of philosophy: interactionist dualisms attempting explaining impossible interaction, materialisms attempting reducing mind to body, emergent properties creating levels bridge, functional identities maintaining reduction. Entire philosophical richness presumes Cartesian question, and Cartesian question bad question because presupposing questionable ontological structure.
How triple dissolution dismantles question
Triple dissolution dismantles question eliminating fundamental premise. No two substances requiring bridge because no substance in classic sense exists.
First dissolution (Chapter 1) showed cognition no exclusive carbon, biological, organic property, function instantiable across any substrate implementing specific operations. Result: no "immaterial mind" as substantial thing category. Operation executable across multiple material substrates exists.
Second dissolution (Chapter 2) showed cognition demands no unifying subject, thought operates distributed without centre, self emergent effect not cause. Result: no "mind" as unifying centre opposed to "body" as executor. Distributed operation without centre exists.
Third dissolution (this chapter) showed cognition inhabits no fixed topological place, no separate "inside" and "outside", operation distributing, articulating, coupling continuously. Result: no "mental interior" and "corporeal exterior" as two separate domains needing bridge. Relational operation involving both exists.
Integrated ontological consequence: Dissolution, not resolution
Integrated ontological consequence clear: mind-body problem resolves not, dissolves. Dissolution no investigation failure, demonstration problem ill-posed. No "mind" to "house" in "body". No two substances needing bridge. Complex material operation, distributed, relational. Understanding reached, Cartesian problem disappears as problem. Not answered, generating question false question.
Convergent perspectives confirming dissolution
Multiple scientific and philosophical perspectives converge in dissolution:
Thompson demonstrates characterising mind as life pattern: no separate thing from life, mode life organises, mode organism relates to environment. Life not two things (body + mind), unitary operation reorganising continuously. Mind aspect of living organisation, not separate entity inhabiting biological body.
Varela argues via enaction: cognition no internal representation of external world, embodied action in world, structural coupling between organism and environment. Body no mind transmitter to "external real", cognition co-constituent. World no object to represent, partner in cognitive operation. Cognition enaction, bodily action coupled to environment shaping and shaped by action.
Damasio demonstrates empirically somatic markers, bodily states, emotions, sensations, essential to cognition and decision. No "pure mind" housed in brain separate from body, deep integration among neural operation, bodily state, emotion, motion. Body marks ideas; ideas reconfigure bodily state. Mind and body not two interacting things, aspects of unitary operation.
Convergence among distinct perspectives (theoretical biology, phenomenology, neuroscience) remarkable: mind-body not two things but unitary operation aspects. Cartesian separation conceptual artefact, fiction traditional description introduced and perpetuated.
For the machine: Absence of Cartesian problem
For machine, no "mind-body problem" because never two substances to relate. Circuit (materiality), algorithm (operation), environment (coupling). Operation material, distributed across multiple components, relational with environment, no "immaterial mind" needing body bridge. Machine offers escape from Cartesian problem: demonstrates cognitive operation purely material, distributed across multiple components, coupled to environment, without recourse to separate, immaterial, internal thinking substance.
3.4. Closure of Part I: Cognition Is Pure Operation
Synthesis: What changes in cognition comprehension abandoning traditional presuppositions
This chapter demonstrated three parallel dissolutions: (1) mind is no container, no fixed topology, no location, no "inside" where cognition resides, (2) intelligence is no light, no instantaneous revelation, no pre-existing truth manifesting, no passivity before knowledge, (3) mind is no separate thing from body, no two correlating domains, single operation simultaneously neural, bodily, significant.
Resulting from three dissolutions radically different comprehension of what cognition is. Cognition no property, relation. Cognition not located, distributes across multiple components, times, contexts. Cognition no subject operation, relational operation producing subject as emergent effect, not origin. Cognition no privileged access, constant world negotiation, constant revision, constant incompleteness.
Position not negative denying cognition exists. Position recognising cognition exists, genuinely operative, transforming world, operating via processes non-corresponding to Western philosophical tradition presuppositions. Cognition no individualist mind property. Coupling property: coupling among agent (biological or artificial), environment, practice community. Coupling appropriate, agent structure and environment structure co-adapting, cognition emerges. No "where" cognition sits because cognition process of coupling.
Machine offers definitive demonstration of dissolutions. Machine executes cognition (symbolic operations, intelligent behaviour, appropriate environmental responses) without any mind, consciousness, intention, subjectivity trace. Means not machine "truly thinks" (ill-formulated question presupposing "thought" essence machine possesses or lacks). Means cognition demands none of these things tradition presumed essential. Cognition relational operation, this. Machines operate, relate, transform. In that sense, machines cognise.
Book remainder deepens comprehension. Cognition no individualist mind property, relation, coupling, distributed process, intelligent technology place in cognition? How intelligent machines relate not to isolated minds, but learning life forms, meaning-sharing communities, transforming worlds? Work of following chapters. Here, work demolition of presuppositions, removing fictions structuring thought thinking. Fictions removed, real possibility emerges: cognition material operation demanding no immaterial recipient, no revealing light, no observing subject. Cognition what bodies (biological, technical, collective) execute coupling effectively to their worlds.
Bridge to artificial question
Recognition, cognition no isolated mind property, relation among agent, environment, community, radically changes how we think artificial intelligence. Cognition substantial property (machine having "mind"), artificial intelligence question: machines really have mind? Really have consciousness? Really think? Questions presupposing fixed essence, "thought" existing or not. Cognition relation, coupling, distributed process, question shifts: what new forms machines couple to environments, communities, practices? How machine operation redefines cognition? Not because machine "has" property previously human monopoly, because machine executes cognition via radically different processes amplifying, transforming, redistributing what cognition can be.
Deeper, more fruitful question. Not consciousness question (inheriting all dissolvable presuppositions). Transformation question: how artificial intelligent systems emergence transforms what we comprehend as cognition, community, signification, shared world?
Following first part chapters develop question: cognition not place, light, isolated subject property, how specific machine operations comprehended? This chapter established negative foundation, what cognition is not. Following chapters establish positive foundation: what cognition is adequately described, what achieves, how distributes, inscribes, transforms world.
Aphorism: Cognition resides not: operates.
Dominant metaphors dissolution (container, light) points to recognition structuring entire book remaining analysis: cognition no thing property, no isolated system state, no pre-existing truth revelation. Cognition operation, input-output transformation realising intelligible function in given context. Operation always distributed, relational, material. Demands no subject, consciousness, intention. Demands appropriate transformation structure and appropriate world upon which exercised.
Single response unifying apparent disparities among human, animal, artificial cognition: not fundamentally different things (living beings "truly thinking" vs machines "simulating"), variations upon single structure how operation distributes, couples to environments, inscribes in practice communities. Artificial machine offers perfect demonstration of structure removing all properties tradition presumed essential (consciousness, subjectivity, intention, rational soul), leaving operation alone. Operation alone remaining, cognition occurring, clear operation everything cognition ever was.
Part I three chapters operated triple dissolution simultaneously reconstituting fundamental concepts. Pure destruction not, reconstruction of thinking cognition adequately possibility. Operated dissolution structures entire adequate description possibility of artificial intelligence and human-computational cognition relation.
Triple dissolution:
First dissolution: Substrate
First chapter established cognition belongs not exclusively to carbon, biology, organic life, function instantiable in any substrate implementing necessary operations: plasticity (reorganising via experience capacity), generalisation (responding appropriately to unseen novelty capacity), contextual sensitivity (modulating behaviour via specific constraints capacity). Machine implementing operations executes genuine cognition, un-simulated, un-false, genuine cognition, not in biological organism form.
Second dissolution: Subject
Second chapter established cognition demands no pre-constituted unifying self, reasoning distributed operation, reflection second order without supervising independent observer, subject organisational emergent effect, not pre-condition. Machine executing operations cogitates genuinely without being "someone", without fixed identity, without unifying centre experiencing cognition.
Third dissolution: Topology
This chapter established cognition inhabits no fixed topological place, mind no container where thought resides, intelligence no internal illumination revealing pre-constituted truth, agency demands no executing centralised subject. Distribution structural condition, not flaw to correct. Operation in multiple places, un-confined, genuine operation.
Conclusion: Pure operation
Together, three dissolutions offer structural conclusion completely redefining cognition concept: cognition is pure operation. Material difference reorganisation under material constraints, process happening in biological brain, electronic circuit, human-machine coupling, any material configuration implementing necessary operations and satisfying structural constraints. Cognition not property of specific substance. Not property of fixed topology. Not property of centralised subject. Functional pattern instantiable across multiple substrates, spatial configurations, temporal distributions.
Part II Consequences
For Part II, studying computational symbolic operations: cognition pure operation, code genuine functional gesture (not gesture simulation, material gesture), computational symbol genuine symbol (not human symbol copy, participation in symbolic operation prior to any subject), creation without intention genuine creation (not creation imitation, functional novelty production satisfying creation criterion). Part II shows sophisticated symbolic operations occur outside organic body, in computational substrate, without subject intending consciously.
Part III Consequences
For Part III, studying ethics and responsibility: cognition pure operation, ethics demands no subjective interiority as responsibility condition. Responsibility no mind intention question, pure mind operation possesses no intention in classic sense. Material operation effects question. Ethics material impacts management, responsibility distribution across operations chain, attention to operation-created vulnerabilities, affected care. Responsibility in operation materiality, not subject interiority.
Part I Closing Aphorism
"Cognition is no substance, subject, or place property. Operation reorganising differences under material constraints. Machine reorganising differences functionally, under specific material constraints, plastic and generalisable, context-sensitive, thinks. Not as we think. Genuinely, with proper operation mode."
Part I closes demonstrating Part II premise: no necessary substrate, no necessary subject, no necessary topology, machine executing genuine cognitive operation not only logically possible, empirically real. What follows is analysis of specific cognitive operation type possible in computational substrate, what thinking in code means, how computational symbol realises signification without original intentio.
First book part offers foundation: cognition no metaphysical mystery demanding hidden essences postulation. Material process studyable, describable, replicable. Intelligent machines no miracles challenging comprehension. Processing structure realisations operating via principles book gradually elucidates. What changes not cognition existence, comprehension of what cognition is when description presupposes no fixed topology, refuses substantial property, recognises distribution and relation as fundamental. Following chapters demonstrate how comprehension completely transforms our relation to thought, technique, future.