Computationalism Classical view sau Symbolic approach Computationalismul C

Computationalism (“Classical view” sau “Symbolic approach”)

• Computationalismul (C) (Teoria Computationala a Mintii, TCM) • Putnam (‘ 61), Pylyshyn (’ 73), Fodor (‘ 75), Fodor si Pylyshyn (‘ 88) • Newell, Simon, Jerry Fodor, Pinker, Harman, Carruthers, Gary Marcus, Gleitman, Brian Mc. Laughlin, Rey • RTM is restricted to explicit beliefs. (Lower) or “Reprezentari (Rs) amodale”.

• Rs = Structuri complexe simbolice cu semantica si sintaxa combinatoriala • Computarile = Reguli → Manipularea de simboluri proprii. (Fodor, Pylyshyn 1988) • Gandirea= Rs semantic legate! (Fodor) • Gandirea, Mintea = Rs + Computari

• Simboluri + Reguli = LOT (Language Of Thought) • C → Atitudinilor propozitionale = Stari intentionale = Reprezinta ceva + Inteles (credinte, intentii, dorinte, cunoastere) vs. Qualia + experienta fenom-ala (Aydede ‘ 04)

“The LOT hypothesis” • Structura cauzala a intentionalitatii = computationala • Computarile opereaza cu Rs (complexe = propozitii) = Proprietati semantice + Structura recursiva sintactica • Intelesul “propozitiilor” (proprozitii “mentalese”) = Continutul starilor intentionale (Tim Crane · Brian P. Mc. Laughlin 2009)

Formalizarea-computarea, sintaxasemantica • Respingerea intuitilor Kantiene (Bolzano, Gauss, Peano, Frege si Hilbert) → Formalizarea geometriei • Reguli → Adevarul enunturilor • Logica formala: “Un anumit tip de proprietate semantica … fara a se baza pe cineva care sa aprecieze intelesul sirului de simboluri implicate. ” (Clark)

• “Intr-un anumit sens, in aria unde au loc operatiile, intelesul nu exista deloc!” (Clark) • Simbolurile (“tokens”) - manipulate prin reguli asupra propriet. fizice sau sintactice • Newell & Simon (physical symbol system) • Haugeland “Daca ai grija de sintaxa, semantica are grija de ea insasi. ” • Nu partea fizica ci computarea (reguli si simboluri pt sistem)

Masina Turing - Banda infinita - Stocheaza informatie prin simboluri - Instrument – Citeste, scrie simbolurile miscandu-se un patratel inainte/inapoi - Procesor central – memorie + intoarcere la simbolul/starea citit precendent → Actiunea urmatoare a unei masini finite (Clark 2001)

• Masina Turing rezolva probleme bineformate Etapa 2 Computerul digital (von Neumann, ’ 40 s) = “Sisteme automate formale” (Haugeland) sau “motoare sintactice” (Dennett) → Mintea = Sistem computational formal implementat in creier. “Cognitia este doar un nivel-de-program. ” (Clark)

• Rolul causal al R → Corespondenta sistematica a sistemului cu situatia reprezentata de sistem • O stare/proprietate este cea care individueaza in termeni de rol cauzal (Ex. Motorul) • Stari – Rolul cauzal implica conexiuni cauzale cu alte stari, stimuli, comportament (Lower) • Rs-Adecvate causal/sistematic cu continut

Problema = Relatia sintaxa-semantica-cauzalitate • Nu cauzalitate-sintaxa ci semantica • Judecatile implica relatiile sintactice intre constituienti + intelesul lor • Formalizarea→Legatura semantic-sintaxa • Computarea→Legatura sintaxa-cauzalitat

• Formalizarea: Cum proprietatile semantice ale simbolurilor (incodate in reguli de derivare bazate sintactic) → Inferente care au semantica (sensitive doar la sintaxa!) • Masina Turing a computarii leaga sintaxa de cauzare = Mecanism capabil sa evalueze functii formalizabile. (Horst 2005)

• Computerul manipuleaza simboluri fara inteles. Instrument = “motor semantic” (Haugeland) • Omul: Sintaxa, semantica, cauzalitatea - in creier • Analogia Creier-minte si hardware-software

• Procesele cognitive mentale: constiente si computationale • Mintea umana manipuleaza simboluri + acces la intelesul lor • E valabila analogia?

• Newell si Simon: “Sistemului fizic de simbol” (SFS) “Un SFS are mijloace necesare si suficiente pentru actiuni inteligente. ” → Distinctia minte (high-level) - creier (lowlevel) = Niveluri descriere/analiza • High-level (manipuleaza simboluri= computarea) trebuie sa fie “semantically transparent systems” (Clark)

SFS - 3 proprietati 1) Folosirea codului de simboluri = Cunoastere termen lung (memorie) 2) Inteligenta in “problem-solving” prin cercetare = Generarea/modificarea structurilor de simbol → Solutia 3) Inteligenta = Regasirea de informatie stocata in memorie si folosirea ei in cautare

Searle - Camera Chinezeasca • O persoana - camera inchisa. Nu stie limba chineza; mesajele in limba chineza Instructiuni (engleza) – Manipulare caracterele chinezesti → Raspunsuri • Concluzia: Persoana “intelege” chineza dar nu intelege cuvintele! • Relatia sintaxa-semantica! • Computerul - Manipuleaza simboluri, nu le intelege

Searle impotriva C (1980, 84) 1. Un computer are program formal (sintactic). 2. Sintaxa nu e suficienta pentru semantica. 3. Mintea are continut mental semantic. → 4. Programele NU explica mintea! • “Strong AI”: Masinile inteleg, au stari mentale. • “Weak AI”: Computer = Instrument.

LOT: Compozitionalitate, sistematicitate si productivitate 1) Compozitionalitatea • Continutul Rs complexe= Constituienti + Relatii • Semantica e compositionala • “The meaning of a sentence … depend on the meaning of those words together with its structure. ” (Sterlny)

• Conceptele = Mentalese + Computari → LOT (structura logica!) (Limbajul – comunicare; LOT – gandire) 2) Sistematicitatea O persoana care intelege “John loves Mary”, fara invatare → “Mary loves John. ” 3) Productivitatea Reguli sintactictice recursive + lexical de Rs finit → Numar indefinit de ganduri

Fodor – LOT 2 (2008) • RTM necesita “referentialism pur” • Referinta = “Only primitive mind-world semantic property” • LOT 1 failed … to recognize centrality of compositionality in constraining theories about the semantics of Rs • Content of thought - Determined by its structure + content of its constituent concepts • Semantics of Mentalese must be referential. LOT 1 missed the point about referentialism because it misses point about compositionality

Compozitionalitatea necesita: • Referentialism (continutul starilor intentionale) • Atomism (individuatia conceptelor) • Structura constituienta a gandurilor • “… (at least some) mental processes are computations on Rs. ” • Asociationalism (conexionism) vs. C “… associative theories - don’t require thoughts to have syntactic structure. Not right - association doesn’t preserve sense/reference/truth; thinking preserves all three. ” (63)

• Association = Contiguity-sensitive relation • Localitatea - intrinseca proceselor m. : (1) Computarea – proces local intrinsec “computation ‘looks at’ a R - ‘see’/operate –identity/arrangement of its constituents. ” (2) Structura constituenta = Proprietate locala a Rs (3) Localitatea–constangeri pt modele mintii (4) Consideratii empirice – Aceste constrangeri nu pot fi indeplinite. (107)

• Procesele computationale – Definite de structurile sintactice a Rs • Structura sintactica=Str. constituienta Rs • Structura constit. - Relatia parte-intreg → Procesele computa-le = Locale (108) • “Cel putin anumite procese nu sunt computari. ” (112)!! • “Cu cat un proces mental nu e local, cu atat il intelegem mai putin. ” (124)

Fodor - Modularitatea • Module mentale, relativ isolate, realizeaza functii diferite (“domain-specific” + “inferentially encapsulated”) • Informatia - Incapsulata in modul (Ex. Visual illusions vs. Churchland) • “Systems as phonetic feature detectors, color perception, shape analyzers, 3 D relation” = “Highly specialized computational mechanisms” (Fodor, 1983)

• “Informatia - Procesata de procesor central • Directia informatiei – “bottom-up” (automatizata) nu “top-down” (controlata) → Impenetrabilitatea perceptiei de catre cognitie (Pylyshyn) vs. Churchland “laden theory” (Kant)

Critici • Fodor- atomistic (“content of a primitive concept is determined solely by it relationships to something in the world”) vs. Quine- holistic • Retea conexionista: Simbolul - inteles in context • O’Reilly si Munakata (conectionisti) “Computanal Neuroscience”: “Symbolic approach” succes, dar nu retelele (2000) • Rs in cortextul prefrontal – slab inteles relativ la multe alte regiuni • Daca e corect → Modele simbolice pt. “high-level cognition”

• John Anderson (lider “symbol approach”): “… lack of agreement of what symbols are --- No consensus on role symbols into mind; how -- coordinated with brain” • “Debates --- No agreement – what symbols mean → A waste of time”! (2007)

Relatia software-hardware (Searle) 1. Este creierul un computer digital? (Cognitivism) Nu 2. Este mintea un program de computer? (Strong AI) Nu – “Camera Chinezeasca” 3. Pot fi operatiile creierului simulate de un computer digital? (Weak AI) Da “The Rediscovery of the Mind” (1992)

Searle- Probleme pt. C 1. Sintaxa nu e intrinseca fizicii “The ascription of syntactical properties is relative to an agent/observer who treats certain phenomena as syntactical. ” (Searle) Folosirea de 0 si 1 reflecta intelegerea notiunii de computare, algoritm si program.

• Simbolurile “nu numesc trasaturi fizice intrinsici ale sistemului. Starile comp-ale nu sunt descoperite inauntru fizicului, ele sunt atribuite fizicului. ” (Searle) → Sintaxa - Nu e intrinseca fizicii! (Cam. Chinezeasca: “Semantica nu e intrinseca sintaxei. ”)

2. The homunculus fallacy is endemic to cognitivism • “Without a homunculus that stands outside the recursive decomposition, we do not even have a syntax to operate with. ”

3. Sintaxa nu are puteri cauzale • Simbolurile nu au putere cauzala “Ele nu exista decat in ochii celui care lucreaza cu calculatorul. ” • Programul - Nu exista real/ontologic • Dpdv fizic, nu exista separat un “nivel de program”

Alte critici la C • Dreyfus (1972, 1992), Winograd and Flores (1986) – Gandirea umana + comportamentul nu pot fi reduse la reguli explicite = Nu pot fi formalizate/simulate de program computer • Neural networks (conectionism) • Dynamical systems approach + situated cognition • Rs non-logice, non-discursive, nonpropositionale

Dezbateri • Continutul (intelesul) – De cine e dat? • Sunt Rs individuate de inteles? ? ? (“narrow” vs. “wide content”) • Formatul Rs (implicit-explicit, proceduraldeclarativ, etc. )