3 The physical interpretation of the Church-Turing Thesis
- 格式:pdf
- 大小:266.28 KB
- 文档页数:43
Hypercomputationandthe
PhysicalChurch-TuringThesis
PaoloCotogno
ABSTRACT
AversionoftheChurch-TuringThesisstatesthateveryeffectivelyrealizablephysicalsystemcanbedefinedbyTuringMachines(‘ThesisP’);inthisformulationtheThesisappearsanempirical,morethanalogico-mathematical,proposition.WereviewthemainapproachestocomputationbeyondTuringdefinability(‘hypercomputation’):supertask,non-well-founded,analog,quantum,andretrocausalcomputation.Thesemodelsdependoninfinitecomputation,explicitlyorimplicitly,andappearphysicallyimplausible;moreover,evenifinfinitecomputationwererealizable,theHaltingProblemwouldnotbeaffected.Therefore,ThesisPisnotessentiallydifferentfromthestandardChurch-TuringThesis.
1Introduction2Computabilityandincomputability3ThephysicalinterpretationoftheChurch-TuringThesis4Supertasksandinfinitecomputation5Computationonnon-well-foundeddomains6Analogcomputation7Quantumcomputation8Retrocausalcomputation9Conclusions
1Introduction
TheChurch-TuringThesisidentifiestheeffectivelycomputablefunctions
withtherecursivefunctions,orequivalentlywiththefunctionscomputable
byTuringMachines;itisusualtoconsidertheThesisasafundamentallawof
mathematics,truethoughnotdemonstrable.Someauthors(e.g.,Gandy
[1980],Copeland[1997])stressthatwhatTuringactuallydidwastoanalyze
computationasaprocessperformedwithpaperandpencilbyahumanclerk
(a‘computor’,inSoare[1996]).Therefore,theChurch-TuringThesisshould
beneatlydistinguishedfromtheclaimthatwhateverphysicalsystemcan
computejustrecursivefunctions(PhysicalChurch’sThesis,or‘ThesisP’).Brit.J.Phil.Sci.54(2003),181–223,axg009
&BritishSocietyforthePhilosophyofScience2003TheinterestingaspectisthatThesisPcouldbefalsewithoutaffectingthe
standardversion:thiswouldbethecaseifsomephysicalsystemswere
capableofcomputingfunctionsthatarenotcomputablebyTuring
Machines,inprinciple.Recentdevelopmentsofnon-conventionalcomputa-
tiontheoriesseemtosupportthisidea.
Inthispaperwediscusstheprincipalapproachestocomputationbeyond
Turingdefinability,orhypercomputation;1ourfocusisonthegeneraldecision
problem,whilecomplexityistouchedononlyasasidetopic.Weshallargue
thatalthoughsomeapproachesare,atleastpotentially,veryefficient,no
modelofhypercomputationcandelivereffectiveresults.Sincethereisno
unifiedtheorythatexplainseverythingandisacceptedbyeveryone,some
thinkthatthepresentlimitationsmaybeovercome,possiblyinthenear
future;weshallsee,however,thatevenifhypercomputationwererealizableit
wouldnotcomputerecursivelyunsolvableobjectsinprinciple,andtherefore
wouldnotfalsifytheChurch-TuringThesis.
Section2recallsthenotionsofcomputableandincomputablefunction,
withsomeemphasisonthelatter.Section3presentsthepureformofthe
Church-TuringThesisanditsinterpretationasaphysicalhypothesis;weshall
reviewsomepossiblecounterexamplesandgivesomecommentsonphysical
incomputability.Section4presentsthenotionofasupertask,aprocess
consistingofanactualinfinityofsteps,executedinfinitetime;thisisakey
point,sincethetheoriesofhypercomputationrest,directlyorindirectly,upon
someformofinfinitecomputation.Weshallseethatsupertasksaresubject
bothtocriticismofaphysicalcharacterandtoobjectionsofalogicalnature.
Section5dealswithcomputationonnon-well-foundeddomains,atechnique
usedinsystemssciencetomodelinteractingandself-referencingobjects;we
shallseethatitshypotheticalhypercomputingpowerdependsessentiallyon
supertasksandnon-effectiveoracles.Section6dealswithanalogcomputa-
tion,oftenheldtobeintrinsicallymorepowerfulthanitsdigitalcounterpart:
weshallseethatanaloghypercomputationisaformofnon-effectiveoracle
computation,possiblydependingontheimplausibleassumptionofinfinite
precision.Section7isdedicatedtoquantumcomputation,themost
promisingapproachtonon-conventionalcomputingatthepresent;we
shallrecallthemainapproachestoquantumhypercomputation,andweshall
seethattheydependeitheroninfiniteprecisionoronsupertaskcomputation.
Section8dealswiththeretrocausalformofquantumcomputation;weshall
seethateventhisspeculativecasecanproducenoviolationsoftheChurch-
TuringThesis.Section9presentssomeconcludingremarks.182PaoloCotogno
1Insomecontexts,‘hypercomputation’refersjusttoparticularlypowerfuldevices,suchasmassivelyparallelcomputers.HereweusethetermintheambitioussensepopularizedbyCopelandandProudfoot([1999]),i.e.computationofnon-recursivefunctions,‘beyondtheTuringlimit’,asSiegelmann([1995])hasit.2Computabilityandincomputability
2.1AfunctionfðxÞ’yis(effectively)computablewhenthereissomemethod
toproduceexplicitlythevalueyandcheckthatyisthecorrectvalue;
computationsmustproducearesult,possiblytheundefinedvalue?,withina
finitelapseofpropertimeafterthebeginningoftheprocess—timebeing
measuredbyelementaryoperationsteps.
Afunctionfisincomputablewhenthereisnomethodtocomputeitsvalues;
itisessentialtodistinguishthecaseswherethereisnosuchmethodasa