New Computationalism Ron Chrisley COGS Department of Informatics University of Sussex School of Humanities and Information, University of Skövde October.

New Computationalism

Ron ChrisleyCOGSDepartment of InformaticsUniversity of Sussex

School of Humanities and Information, University of SkövdeOctober 19th, 2006

QuickTime™ and aTIFF (Uncompressed) decompressor

are needed to see this picture.

QuickTime™ and aTIFF (LZW) decompressor

are needed to see this picture.QuickTime™ and a

TIFF (Uncompressed) decompressorare needed to see this picture.

Overview

Will discuss four related claims/ideas:1. "Transparent" defense of

computationalism2. Falsity of the Church-Turing thesis3. Falsity of pan-computationalism4. Even if computationalism is false,

strong AI is possible

Transparent computationalism

• The claim that cognition is computation can be construed opaquely or transparently

• Opaque construal: The mind is best understood in terms of the concepts from current (or past!) computational theory

• Transparent construal: The mind is best understood in terms of whatever concepts, it turns out, best explain what computers do

• Many critiques of computationalism succeed only on the opaque construal

• Thus, transparent computationalism is not threatened

The transparent strategy

• For each critique, present:– A current (opaque) view of

computation– The critique based on that view– An alternative view of computation

that avoids the criticism– Independent motivation for that view

of computation

Critique 1: Dynamics

• Opaque view: Discrete steps in an algorithm essential to computation

• van Gelder:– Cognition isn't discrete, but

fundamentally dynamical– Therefore, cognition isn't computation

Dynamical computation

• Alternative view: Generalise notion of an effective procedure to include any physically realisable and exploitable process, even dynamical ones

• Independent motivation:Real-time computational control of an airplane wing

Critique 2: Externalism

• Opaque view: Computational properties are syntactic and local

• Fodor:– Psychological properties are semantic

and relational/external/non-local– Therefore, there can't be a

computational psychology

Externalist computation

• Alternative view: Even computational explanations are external/relational (cf Peacocke's "Content, computation and externalism", 1994)

• Independent motivation: Embedded computational systems

Critique 3: The Chinese Room

• Opaque view:1. All essential computational properties are

formal2. Non-formal properties of a computation are

mere implementation detail

• Searle:– Formal properties are insufficient for mind– Therefore, there can't be a computational

psychology

Grounded computation

• Alternative view:1. Having a semantics is crucial to

computation 2. Some properties that current formal theory

takes to be irrelevant play a constitutive role in determining computational state

• Independent motivation:1. Not every process is a computation2. Real-time computational control of an

airplane wing

The Church-Turing thesis

• An example of an explicit acknowledgment of the distinction and relation between informal and formal (theoretical and pre- theoretical) notions

• Diagonal arguments (Gödel, Lucas, Penrose) do not show what they purport to: falsity of Strong or even weak AI

The Church-Turing thesis

• Diagonal arguments highlight a special case of a general property:– For any set of things that can answer

questions, one can construct a question that no member of that set can answer, even though some things outside the set can.

• Implies, e.g., that odd-numbered TMs can compute functions that even-numbered TMs cannot

• And that TMs can compute functions we cannot

Universality

• One might think this violates Turing's famous result, that there exist universal machines

• But no conflict, since Turing's universality result is about simulation, not computation

Against pan-computationalism

• Putnam's sense: Everything instantiates every computation– fails because of the causal aspect of

causation (cf, e.g., Chalmers 1994, Chrisley 1994)

• More plausible sense: Everything has some computational desciption– Yes, but still too broad: IBM vs BMW– Suggests that we need to do more work to

capture real computation: Semantics

Computation and mind

• Traditionally, two ways computation is relevant to understanding or replicating mind:

1. Weak AI: Computational simulation of mind

2. Strong AI: Cognition is computation

Strong AI without Computationalism

• Even if cognition is not computation, does not imply falsity of strong AI– Not because of pan-computationalism– Third way: computation as the ultimate

plastic– Computation is a convenient way to

configure a system's causal/dynamical profile

– In between identity and mere simulation

Strong AI without Computationalism

• E.g. Suppose life is crucial for mind; and (e.g.) Boden is right that life is non-functional – Does not imply that one cannot

program a system to be alive – Falsity of (even transparent)

computationalism does not imply Strong AI is impossible

Thank you!

Video, audio and PowerPoint files of this talk and others can be found at:http://e-asterisk.blogspot.com

Comments welcome: [email protected]