Citation:Morgan, JA (2018) Yesterday’s tomorrow today: Turing, Searle and the contested significance ofartificial intelligence. In: Realist Responses to Post-Human Society: Ex Machina. Routledge, pp.82-137. ISBN UNSPECIFIED

Link to Leeds Beckett Repository record:https://eprints.leedsbeckett.ac.uk/id/eprint/5275/

Document Version:Book Section (Accepted Version)

The aim of the Leeds Beckett Repository is to provide open access to our research, as required byfunder policies and permitted by publishers and copyright law.

The Leeds Beckett repository holds a wide range of publications, each of which has beenchecked for copyright and the relevant embargo period has been applied by the Research Servicesteam.

We operate on a standard take-down policy. If you are the author or publisher of an outputand you would like it removed from the repository, please contact us and we will investigate on acase-by-case basis.

Each thesis in the repository has been cleared where necessary by the author for third partycopyright. If you would like a thesis to be removed from the repository or believe there is an issuewith copyright, please contact us on openaccess@leedsbeckett.ac.uk and we will investigate on acase-by-case basis.

1

Yesterday’stomorrowtoday:Turing,SearleandthecontestedsignificanceofArtificialIntelligenceChp5pp.82-137inAl-Amoudi,I.andMorgan,J.editors(2018)RealistResponsestoPost-HumanSociety:ExMachina(Volume1)London:RoutledgeJamieMorganIntroductionArtificial intelligence (AI) has become an increasingly important issue in and forsocieties.IthasalsobecomeentangledwithwhatistermedTranshumanism(TH).InthispaperIexplorethewayAIisconceivedandfocusedupon(withsomecommentonTH).TwoconceptualfocusesofAIhaveemergedandtheseinturnhaverootsinandarerelatedtokeyworksthathavedominatedhowAIhasbeenaddressedthroughphilosophy. The key works are Turing (1950) and Searle (1980).1 I explore theargumentsofthesetwo.Theintentisnotapejorative‘backtobasics’,butratheranexploration of sophisticated origins in order to identify how dividing lines andomissionscanbecomeinsomewaysingrainedandinothersinterstitial.Bothinvolveproblemsofontologyandsocialontology,whichinturncreatesproblemsforhowweseektoshapethefuture.FrombothTuringandSearleaweakandstrongfocusinAIhasdevelopedandthishashadavarietyoffurtherconsequences.Theconsequencesarethemselvescomplicatedandinter-connectedandsocannotbesimplystatedorenumeratedbutarisecumulativelyastheargumentproceeds. InthefinalsectionIdraw the whole argument together in terms of the social significance of actualtechnologicalchangesoccurringundertheaegisofAI,anddosofinallywithreferencetotheconceptofrelationalgoods(DonatiandArcher,2015).Twoconceptsofartificialintelligence(AI)ForAIaconvenientplacetostart iswiththemuch-publicisedfirst report fromtheStanfordOneHundredYearStudyonArtificialIntelligence.Theprojectwaslaunchedin2014andasthetitlesuggestsisalong-rangestudyofAI.ItsremitistoexploreboththestateofandconsequencesofAI.Theprojectbringstogetheradesignatedmulti-disciplinarypanelofexpertseveryfiveyearstoprovideanupdateonprogress.AsanauthoritativeexerciseitexpressessometypicalpositionswithinAIresearch.2Thefirstreportnotesthereisa‘lackofaprecise,universallyaccepteddefinition’ofAI(Stone,

1 In stating the problem is not new and in drawing attention to Turing and Searle as longstandingsourcesofkeyideasIdonotmeantosuggestthatthehistoryofAI,computing,robotics,animatronicsand automata beginswith these two.Neologismsmay be new but ideas far older. JohnMcCarthycoinedthetermAIataconferencein1956.Thereare‘robots’intheIliadandactualanimatronicsofvaryingdegreesofsophisticationarescatteredthroughouthistory.SeeAdamRutherford‘RiseoftheRobots:Thehistoryofthingstocome’Radio4broadcastMonday13thFebruary2017:http://www.bbc.co.uk/programmes/b08crvz32Therearerecognizedlimitations.ThefirstreportrestrictsitselftoimpactsforatypicalAmericancitybasedon8domains. It explicitly excludesmilitary and security aspectsofAI andassumes that thematerialpresentedwillberelevanttodifferentdegreesonaglobalbasis.

2

2016:p.12).However,thisisnotquiteright.Therearetwobasicconceptualfocusesfor AI and some confusion about how they relate. How something is conceived isusuallymorecomplexinitsformationthanhowitisdefined.Furthermore,whattheconceptualfocusemphasiseshasconsequencesforhowaconceptoperates.3WhatImeanbythiswillbecomeclearerasweprogress.

Forourpurposes,SearleintroducestherelevantbasicconceptualdistinctionforAIinhis‘Minds,brainsandprograms,’(1980).HisconcernwasthesignificanceofAIforthestudyofthemind.ForSearle,‘weakAI’concernscomputersasatoolforthestudy of the humanmind,whilst ‘strong AI’ assumes or asserts that a sufficientlycomplexcomputerisamind.WewillreturntoSearlelater.Thedistinction,thoughstillpartlyrootedinSearle’swork,hasgeneralisedbeyondafocusonphilosophyofmind.WeakAIfocusesprimarilyonAIasfunctionsandstrongAIasentities.Weakartificialintelligence(AIw)In contemporary usage theweak concept of artificial intelligence (AIw) focuses onfunction and is on closer inspection semantically minimal and tautological. TheStanfordreportadoptsNilsNilsson’swell-knowndefinitionandneatlyencapsulateskeyaspectsofAIw:

Artificial Intelligence is that activitydevoted tomakingmachines intelligentandintelligenceisthatqualitythatenablesanentitytofunctionappropriatelyandwithforesightinitsenvironment[...Accordingtothisview]thedifferencebetweenanarithmeticcalculatorandahumanbrainisnotoneofkind,butofscale, speed,degreeofautonomyandgenerality […There isa] intelligencespectrum[…and]thecharacterizationofintelligenceasaspectrumgrantsnospecialstatustothehumanbrain.Buttodatehumanintelligencehasnomatchinthebiologicalandartificialworldsforsheerversatility,withtheabilitiestoreason, achieve goals, understand and generate language, perceive andrespond to sensory inputs, provemathematical theorems, play challenginggames,synthesizeandsummarizeinformation,createartandmusicandevenwritehistories[…ButforAI]matchinganyhumanability isonlyasufficientcondition,notanecessaryone.(Stone,2016:p.12)

ItshouldbeemphasisedthatAIwisadefaultposition,andoftenaplaceholder.TheStanfordreport,likemanyothers,attemptstoleaveopenexactlywhatAImightbeandbecome.NewreportsaretobecommissionedbyaStandingCommitteefortheStanford project every 5 years. The first reportwas producedby a selected StudyPanel,mainlycomprisedofexpertsinrobotics,programming,dataanalysis,systemstheoryandplanning,andeconomics(drawnfromMicrosoft,MIT,Harvardetc).ThedefaulttoAIwreflectsthelackofagreementandcoherencyacrossthesefields.Thislack,awkwardthoughitseems, isnotemptyforthepurposesofhowsomethingis

3Tobeclear,asanexerciseinanalyticalphilosophy,onecouldformallydistinguishdefinition,conceptandconceptualfocus.InwhatfollowsIdonotelaboratesignificantlyregardingdistinctions.ThemainpointIammakingisthatdefinitionstartsfromasimplestatementaboutintelligenceasakindofdoing,andthattheconceptofAIobservablyemphasiseseitherfunctionorentity,andsoinapracticalsenseofdevelopmentofconcernsisboundupwithfocus,whichinturnhasconsequences.

3

conceived,sinceit invitesreduction,minimalismandtautologytocreateapointofdeparture: the troublingproblemofdefiningAIhasbeenaddressedandsowecanmoveon. This tacitly introducesa ‘be’ intoAIw. Ifone reads theStanford report itclaimsAIwillbewhatAIresearchersdoandexplore.So,AIbecomesanaccidentalidentity:AIisAI.Concomitantly,sincethereisagreatdealofactivityanddevelopmentoccurringundertherubricAI,whatitdoesbecomestheconvenientfocusforwhatitis.Thus,AIwfocusesonfunction,albeitincreasinglycomplexfunction,andwhatcanbe tested and (mainly) observed. Here, function is primarily an expression of‘intelligence’. The main subsidiary distinction becomes one between specific‘intelligence’,definedasthemore-or-lessefficientorappropriatedoingofsomething,andgeneral‘intelligence’,definedasthereplicationofthisfunctionalityinmultiple,differentandnewdomainsofapplication.4 Many AI researchers are aware that the distinction between specific andgeneralintelligenceisbasedonablurringofwhat‘intelligence’itselfis.Forexample,Legg andHutter (2007) collate around 70 available definitions of intelligence, andsubcategorisetheseintothosederivingfrompsychologistsandfromAIresearchers.TheirintentistosynthesiseahumanindependentconceptofintelligenceamenabletoAIresearchers.Theyidentify3primaryfeatures:1)intelligenceisapropertyofanindividualagentasitinteractswithanenvironment,which2)relatestoitsabilitytosucceed with reference to an objective, and 3) depends on agent adaptability toobjectivesandenvironments.‘Intelligence’canthenbemeasuredvia‘achievement’,subjectto1-3.Clearly,sincethepaperisdevelopedinthecontextofAIconcerns,thisdefinitioncanbereadastendingtowardsageneralintelligenceconceptofAI.Still,itremainsheavilyweightedtowardsAIw.Moreover,thederivedconceptofintelligencehasbeenbasedonaparticularselectivesynthesis.Itnotesbutputsasideawholesetofkeyaspectsofthepsychologists’definitionsandbroaderconceptualstatements.Whatseemscommonisselected,whilstwhatmaybeimportantbutonlyhighlightedby some is omitted. Several non-AI definitions highlight the composite andinterpretivebasisof intelligence,andtherangeofaspectsof intelligencestatedbypsychologistsbutomittedinthefinaldefinitionincludes:toplan,tothinkabstractly,toshowgoodsense,practicalsenseandinitiativethroughjudgementandassociatedactivity, to learn factsandskillsandappropriatelyapply themand todemonstrateawarenessoftherelevanceofbehaviour.Theseaspectsaresievedfromthesynthesis,andsotheconceptofintelligenceispre-structuredinLeggandHut’ssynthesisinawaythatencouragesafocusonfunction.AIasafocusofpracticalconcernscanfocusonlittleelsethanfunction,andsoseekingacommondefinitioniseithercircularor

4 There have beenmajor developments in specific ‘intelligence’, inways that address someof thespecific areas of complexity in the Stanford report quoted list of human achievements. The EMIprogramhassuccessfullyimitatedtheworkofBach.Inamorehighprofilecase,theDeepMindprojectatGoogleisresponsiblefor,AlphaGo,whichisnowabletodefeatahumanGoworldchampionusingMonte Carlo simulation and tree search within a system based on multiple self-play. See neuralnetworksmateriallatersectionandalso:https://deepmind.com/research/publications/mastering-game-go-deep-neural-networks-tree-search/Ingeneral,machinelearningusinglargedatasetshaschangedsomeaspectsofhowacomputerplaysgames.However,thisisstillcurrentlyfarfromgeneralintelligence,evenintheAIwsense,sincetheAIsofarareunabletosimplyturntheir‘intelligence’toadifferentsettingorgameetc.Theymustbeginaneweachtime.Thismaychangeofcourse,butthisinitselfwouldnotsatisfycharacteristicsforstrongAI.

4

self-propagating. The intent to develop something human independent becomessimply isolated (rather than abstracted) from important characteristics that maysituateintelligence.Function,isthecommonconcernthatoccursacrossalldefinitionsandsobecomesthecentralaspectofintelligence.

Concomitantly,thefocusonfunctioninAIisapragmaticresponsetopotentialbasedononesetofkeycontemporaryconcerns.Ittakesthepresentandexpressesafuturebasedonanengineerordesigner’sframeofreference.SinceAIisnotcurrentlyconsciousetc.itseemsconvenienttoconcentrateonwhatidentified‘AI’candoandmaydo rather thanwhat a ‘intelligent entity’ is andmaybe.5 There is a tacit andsometimesacknowledgedsemanticslippageinvolvedsinceAIresearcherssometimesrefertoAIandthen‘trueAI’,withthelatterreferringtoafullyrealisedentity.YetAIwstill involves a version of ‘be’. AIw adopts, often inadvertently, an external andbehaviouristicapproachtohowAIisdefinedandsohowAIisconceived.Thisismostreadilyunderstoodintermsofspecific‘intelligence’,butissignificantalsoforgeneral‘intelligence’.Differenceofkindisputasideinfavourofpositionalongaspectrumandthis has consequences: a calculator and a human function differently and aredifferentlycomplex,butareas‘intelligences’notproperlydistinguishedconstitutivelyorqualitatively.ThisambiguityofdistinctionisbasictoAIw.Itis,asweshallsee,rootedin Turing’s approach, though arguably Turing also made strong AI claims thatencourageafocusonfunctionviafunctionalismasconstitutiveforanentity.

One should note that the problem of kind for AIw is not clear-cut. Readsympatheticallyincontext,thetypicalintentofanAIwdefaultistodeferdiscussionofwhat AI may be in a constitutive-qualitative sense, until such time as it becomesgermane.Thisisconceptuallyproblematicandpotentiallydangerous,evenifoneputsasidedoomsdayTerminatorscenarios.ThepreviousquotefromtheStanfordreportstates that matching human abilities is sufficient but not necessary (SbnN) as abenchmark for AI. Given that AI has already been subject to a spectrum thatencompassesfunctionreferencedtoacalculator,SbnNintroducesacomponentthataffectsthecoherenceofthiswayofdefiningAI.Onemightnowinferanythinglessable than a human in its flexibility, diversity and complexity (i.e. AIw generalintelligence)wouldbeinsufficienttobeAI.6TheauthorsseemtomeannomorethanthatthefutureabilitiesofAImayexceedthecomparableabilitiesofhumansandthatgivenabilitiesmaynotberestrictedtothoseofhumans.Thisisstillaboutfunctionandsosaysnothingdirectlyaboutwhat issufficientornecessary to theconstitutionof‘intelligence’ortothehuman,despitethatthehumanisstatedasthebenchmark.

ThereisthusabasicambiguityinAIwregardingwhatitisthatisdifferent,whatthisdifferencederivesfromandwhetheremergenceisameaningfulconcepttoapplytoboththehumanandAI(forcontextseeStephan,2006).7Itisperhapsworthnoting

5Clearlyaprogrammerordesignermustgivethoughttowhatsomethingisinordertomakeit,butthisis inordertomake itdosomethingand isdifferentthanextendedruminationorreflexivefocusonwhatsomethingisasanentityintermsofmattersofstatus,kind,comparativeconstitution,qualitiesandsoforth.6AndyetmaycontainsomethingnecessarytoAI,sothereisthepossibilityofinsufficientnecessaryaswellassufficientnon-necessaryconstitution.Theproblemislogical-semanticandsoinfectsclaimsofsubstanceformulatedwithintheschema.7Theimplicationisnotthatintelligenceemergesfrompriorformssuchascalculatorsalongaspectrum(sincethiswouldbeabsurdinvariousways)butratherthatanyparticularentityonthespectrummaybedescribedinemergentterms.

5

thatDarwinultimatelypositionedhumansandanimalsinthesameway,sothisisnota procedure that is especially controversial when dealing with distinctions. Theproblemitcreatesisthatitseemstoinadvertentlyequateentitiesthataredistantonthespectrum,sincethespectrumimpliesnobreakpointsorthresholdsfordifference,evenifonemightwanttoacknowledgethatathresholdcouldbebroadandblurredratherthannarrowandneat.

InthecaseofAI,thecuriouscorollaryofthisAIwfunctionalapproachisthatitdecentreswhat it istobehumanor,tobe lessprejudiciallyanthropocentric,tobeequivalentintermsofpossibleessentialcharacteristics(sincetheseinvokeissuesofkindandmayberelevantmatterstoaddressintermsofsomeanimals,aliensetc.).This is important because it is built into the conceptual construct, and so cannotreadily be rectified by subsequently acknowledging the importance of the humanwhilstanalysisandargumentproceedsonthebasisofAIw.Aconceptualdisjunctureisliabletolinger.Thisisalsoproblematic,sincethefutureisamatterofhowtechnologyisdesigned,shaped,andusedby,inandforhumanswithinsocieties.Astheauthorsof, forexample, theStanford reportareaware,whatmatters ishowtechnology isshaped,andhowtechnologyinturnwillshapesociety.FewAIscholarssetouttobesimple technological determinists.However, thedisjuncture and focuson functiondoesencourageakindoftacitweightingtowardscharacteristicsofdeterminism:AIiscoming and we have to cope with it. Concomitantly, the disjuncture createsproblematic beginnings for howwe deliberate regarding human futures based onwhatitistobehuman,whichcanaffectinturnhowhumansflourishorsuffer.Thesearequintessentiallyissuesofontology.Strongartificialintelligence(AIs)WhilstAIw focusesonfunction, incontraststrongartificial intelligence (AIs) takesastepbacktoconsiderwhatdirectsfunction.AIsthuslocates‘intelligence’withinanexpandingsetofcharacteristicswhichmaybeassociatedwiththisdirection:purpose,awareness, cognitive unity, consciousness, self-consciousness etc. AIs is mainlyconcerned with the nature of entities. It focuses on the constitution that affectsexternalexpression,andsomediatesandenablesfunction.Moreover,functionisnottheonlyconcern,beingmerelyasubsetoftheconsequencesoftheexistenceofanentity. There are two main subcategories of an AIs conceptual focus. First, onesubcategory focuseson theequivalencebetweenhumanand ‘AI’,and thuson thevalidity of analogical claims. This locates what it means to be intelligent withinphilosophyofmind.Interalia,whenposedasamachine-mind(program)problematicit invites disputes regarding what function alone can reveal or allow one to inferregardingthenatureofmind,organicorotherwise.ThoughconcernedbyAIthemainconcerniswiththehumanandwhatAIdoesordoesnottellusaboutthehuman.Thisfocus follows from Searle’s work. Searle is concerned by the dominance offunctionalism in cognitive science and with the mutual relation between thisfunctionalismandAI.His‘Minds,brainsandprograms,’(1980)raisessignificantissuesthat we will set out later. Second, andmirroring this, another subcategory of AIsincludestheworkofspeculativescience,futuristsandofsciencefictionwritersthatprojectorimaginethepotentialsofahuman(animal/alien)equivalentAIandthenofsuper-AI.ThoughthislattersubcategoryismainlyconcernedwithwhatAIisnotyet

6

(andmayneverbe),discussionintermsofitalreadyhasmaterialconsequenceandsome focus on imminent prospects. This is because AIs encompasses legal andregulatorydiscussionanddevelopment.

OneofthemoreprominentcurrentexamplesofthelattersubcategoryofAIsis provided by the work of the European Parliament Committee on Legal AffairsCommissiononCivilRulesonRobotics.SinceAIsfocusesonintelligenceintermsofwhatitisandwhatitmaybe,itraisesissuesintermsof‘autonomy’.Clearly,mattersofautonomyareimmediatelysignificantforthelegalstatusofAI.OnceanAIbecomesa seat of decision-making it becomes a source of concern regarding its materialconsequences. Here, there is some ambiguity that glosses over the differencebetweenalocus(site)ofdecision-makingandasourceofdecision-making,withtheformertendingtoinformhow‘autonomy’isconceived.Inanycase,sincedecision-making can be programmed, an AIs set of legal concerns need not wait on anydemonstratedextensivelistofallimaginableAIscharacteristics.‘Autonomy’createsalegal issue regarding liability for actions, harms etc. since it introduces a potentialbreak inchainsofcausationwithreferencetoowners,designersandbuilders,andmakesambiguoustheconceptof‘operator’(thinkofadriverlessvehicle,awarehousemobiledeliveryunit,anadaptive-targetingdroneweapon).

Moreover, since AI are also manifestly developing or changing, a legalperspective immediately invites forward thinking regardingwhat anAI is andmaybecome.Thesearealreadygermane inawaythat theyarenotnecessarily for theprogrammer, the data analyst, the systems theorist, the economist etc. Note, theStanfordproject,likeothersofitskind,doesnotsimplyignorealegalperspectivebutthis is not its core concern, and so matters of what constitutes intelligence aremarginalised.ThetypicaldefaulttoAIwisamatterofdominanceratherthanasimpledenialoftheconcernsofAIs.Incontrast,alegalperspectiveonthepresentexpressedintothefuturecannotmarginalizemattersofAIspreciselybecauseofthenatureoflaw.Inlaw,functionbegsquestionsregardingconsequencesandthusresponsibility,whichcannotevadeissuesoftheconstitutionandstatusofentities.ThecurrentdraftreportpreparedfortheEuropeanUnionCivilLawonRoboticsconvenientlyillustratesthis:

[T]hankstotheimpressivetechnologicaladvancesofthelastdecade,notonlyare today’s robotsable toperformactivitieswhichused tobe typicallyandexclusively human, but the development of autonomous and cognitivefeatures -- e.g. the ability to learn from experience and take independentdecisions--hasmadethemmoreandmoresimilartoagentsthatinteractwiththeir environment and are able to alter it significantly; whereas in such acontext,thelegalresponsibilityarisingfromarobot’sharmfulactionbecomesacruciallegalissue[…]themoreautonomousrobotsare,thelesstheycanbeconsideredsimpletoolsinthehandsofotheractors(suchasthemanufacturer,theowner, theuser etc.)whereas this in turnmakes theordinary rulesonliabilityinsufficientandcallsfornewruleswhichfocusonhowamachinecanbeheld--partlyorentirely--responsibleforitsactsoromissions;whereasasaconsequence,itbecomesmoreandmoreurgenttoaddressthefundamentalquestionofwhetherrobotsshouldpossessalegalstatus[…]ultimatelyrobots’autonomy raises the question of their nature in the light of existing legal

7

categories -- or whether a new category should be created, with its ownspecific features and implications as regards the attribution of rights andduties, including liability for damage […The report recommends] creating aspecific legal status for robots, so that at least the most sophisticatedautonomous robots could be established as having the status of electronicpersonswithspecificrightsandobligations,includingthatofmakinggoodanydamagetheymaycause,andapplyingelectronicpersonalitytocaseswhererobots make smart autonomous decisions or otherwise interact with thirdparties.(EP,2016:pp.5&12)

TheEUdraftreportdemonstratesthatanAIsframeofreferenceisalreadywithinthepurviewoforganizations.However, it isprobablymore reasonable to suggest thatmatters of AIw have provoked a partial AIs response.8 The driving force even hereremains for themoment function and consequences. The draft report defers anddelegatesfullerdevelopmentofmattersrelatingtothestatusofanelectronicperson.Forexample,theprofoundquestionofwhetheranAIcan/couldflourishorsufferandwhetherthisisthebasisfortheascriptionofrights.So,whilstAIwdecentreswhatitmeanstobehuman,currenttentativestepstowardsgreaterfocusonAIs,intermsoftheAI,forandfromthepositionoftheAI,arealsolimited.Thisispartlybecauseitisdifficulttoconsiderderivativeissuesofelectronicpersonsforthosepersonsuntilonehas an adequate concept of the person and then some knowledge, based onrealisation,ofanactualelectronicperson--soanAIisAIproblemreturnsinadifferentguise (however, seeCalverley, 2007). In any case, explorationofAIs is a small butimportantaspectofthewholeEUreport.Thispatternisrepeatedintermsofcurrentlegalandregulatorycommitteeinvestigationsinmanycountries.9AIsremainsmainlyamatterforphilosophers,futurists,andsciencefiction.Shifting,adequacyandtheinterstitialproblemUltimately,whenAI is consideredasadiscourse, there isa shiftingbackand forthbetweenAIwasafocusonfunctionandAIsasafocusonentities(whichmayexplicitlybefunctionalist).AgoodexampleofthisisprovidedbyarecentcollectioneditedbyBrockman(2015).Thecollectionbringstogethercontributionsbyleadingscientists,socialtheoristsandphilosophersfirstpublishedasresponsestothe‘questionoftheyear2015’ontheEdgeonlinesciencesalon:whatdoyouthinkaboutmachinesthatthink?10SomecontributionsareoverwhelminglyAIwandsomeAIs.However,giventhe

8GaryLea,visitingresearcherinAIregulation,makesthisclearinhisblogonthewell-knownsite,TheConversation. As already noted AI researchers are aware that the concept of intelligence remainsultimatelyambiguousandthatitmightbepreferabletohavea‘humanindependent’measure;variousattemptshavebeenmadetoaddressthisintermsoffunctionthatincreasinglyrecognizeaspectsofgeneral intelligence along AIw lines, but do so with an AIs set of concerns related to foreseeableregulatory issues: https://theconversation.com/why-we-need-a-legal-definition-of-artificial-intelligence-46796Manyindividualgovernmentsarenowstartingtotakeaninterestalongtheselines.9Forexample,theUKHouseofCommonsScienceandTechnologyCommitteehasproducedaseriesofreports on Robots and Artificial Intelligence. These invite expert evidence, and this includes fromprominent figures concerned with AIs. For example, Alan Winfield, Professor of Robot Ethics atUniversityoftheWestofEngland.10https://www.edge.org/about-edgeorg

8

natureofthequestion,whichinvitesdeliberationregardingthescopefor‘thought’,somecontributionsshiftbetweenthetwo,takinganAIwapproachtocontemporarytechnology and its immediate prospects and an AIs approach tomore speculativepossibilities. For example, the Harvard psychologist and public intellectual, StevenPinkerpraisesthefocusonfunctionasameanstoovercome‘spiritualism’alongAIwlines,butadoptsaspeculativeAIspositiontomakethepointthatfutureAIneednotbe innately aggressive, since this is amasculine trait and not necessary to a non-masculineAI;athoughtlikelyprovokedbyhisownworkongradualprogressinhumancivilization(PinkerinBrockman,2015,pp.5-8).Bycontrast,butstillshifting,theMITNobelprizewinningphysicist,FrankWilczektendstowardsAIw,butdoessointermsof an AIs first subcategory framing: ‘What distinguishes natural from artificialintelligenceisnotwhatitisbutonlyhowit’smade’(WilczekinBrockman,2015:p.121). Following this example, one might conclude that since both AIw and AIsconceptsoccurandaresubjecttodevelopment,and,moreover,itispossibletoshiftbackandforthbetweenthem,thatthereisnoproblemregardinghowAIisconceived.However,twoproblemsarise.First,theexistenceofconceptsdoesnotentailconceptsareadequate.Forexample,varietiesofentityrelatedfunctionalismmaybecritiquedin different ways. Second, legitimating shifting does more than legitimate bothconcepts (AIw and AIs), it tends to put aside how focus affects the way conceptsoperate.TheexistenceoftwobasicfocusesforAI,onewherefunctiondominates,andonewherethenatureofanentitydominates,createsscopeforoneconceptualfocustobemoreinfluentialthananother.Currently,andinmanywaysunderstandably,thatisAIw.So,theexistenceoftwofocusescreatesthepotentialfordominanteffectsfromthedominantconcernsofaconcept,sinceothersaremarginalized.Moreover,issuesmaybemarginalizedinwaysthatdonotappearintheconcernsofthesubordinatedconception, becoming rather interstitial. There may then be a link between theadequacyofeachconceptandtheproblemofbothconcepts.There is thusapriorproblemofontologythatmaybeusedtoappropriatelyexplorethesematters.Iwillreturntothislaterintermsofrelationalgoods.AtthisstageIsimplysuggestthattheexistenceofajuxtapositionofconceptsandfocicreatesgroundsforperpetuationofproblemsofmanykinds.Considerthisintermsofthedifferencebetweenaspectrumviewofintelligenceandthepossibilityofbreakpointsandemergence,andhowthismayrenderwhatitistobehumandecentred.11OnewayinwhichthisisimportantstartstobecomeclearwhenonebeginstothinkaboutTranshumanism(TH).Transhumanismloweranduppercase(th«TH)In an ordinary language sense, lower case transhumanism (th) is a portmanteaublendingof‘transitional’and‘human’,thoughonethatalsoinvokestransformingandtranscending some prior limit on the capacities, abilities or typical observable lifeoutcomesofthehuman.Assuch,thetermisextremelybroad,ifnotamorphous,but

11NoteonemightcategorisesomecomplexitytheoryapproachestoAIasAIsandcomplexitytheoryistypicallydefinedintermsofemergentproperties(thoughwhatthismeansishighlyvariable),soitisimportantnottogivetheimpressionthatthepurposeofdifferentiatingAIwandAIsistocreateasimpledichotomy.Ratheritistoestablishthatdichotomisationisatendencythataffectsnuanceandsubtletyandfocus.

9

itisalsointimatelyboundupwithfunction.Astransitionetc.itinvolveschange,andsincechangestocapacity,abilityandobservablelifeoutcomesarenotnew,thereisnothingintrinsicallynewtothenotionofth.Throughouthistoryhumanshavebeenchanging, augmenting and enhancing their bodies, and also the context in whichbodies are capable of achieving (and suppressing) things -- in relation to persons,roles, agency etc. Consider some of the range of means by which this has beenachieved over time: artefacts (tools, machines, prosthetics, exo-tech etc.),pharmaceuticals, surgical-intervention (transplants, implants, amputation),technologicallybasedservicesthatcreategroundsfororaffecthumanactivityandsoforth.Consider,inadditionto‘augment’and‘enhance’,someofthefurtherlanguagewe apply to changes in terms of these means: facilitate, stimulate, extend,(re)generate,suppress(negate),mutilate,delegate,perfect…And,considersomeofthe historical consequences of related change for the human. Health immediatelyspringstomind:lifeexpectancy,vitality,heights,weights,shapes,mentalstatesetc.However,manyotheraspectsoflivingcanalsobethoughtthroughinsimilarways.For example, our relation to time in termsof how long given activities take,whatactivities are possible, our sense of distance as a timemeasured relation, our lifeorderingthroughclocktimeetc.

Iftranslatedintomattersofsocialontology,theaboveseemsnomorethanaspecificwayofmaking the general point that humans livewithinopen systems inprocess.Humanhistory,thehistoryofthehuman,hasalwaysbeenentangledwithinventionandinnovation.TooluseisasoldandolderthanHomosapiens.However,changes, recognized potentials and speculations have made possible a particulardiscursive response regarding transition and transformation.Modern surgery, theprospects for genetic manipulation, and continual development of informationtechnologiesashardwareand softwarehaveprovoked issues regarding interfaces,melding,mutation,andperhapsevenre-embodimentanddisembodiment.HerethbecomesentangledwithAIinvariouswaysandthishasbeenrecognizedasuppercaseTranshumanism.ProponentsofTranshumanismtendtouseH+orh+toreferto it.However, IwillcontinuetouseTH.WithinTHtherearetransitionalhumansandapotential for a new kind of entity, the ‘posthuman’,whowill live in new kinds ofsocietiesthatwelcomeandcelebratefutureAIasequivalent(anddifferent)entities(see Regis, 1991; O’Connell, 2017). Though it has antecedents, the termTranshumanismbegantobeusedinthe1980s.AWorldTranshumanistAssociation(WTA) was founded in 1998, and there are several different versions of aTranshumanistdeclarationormanifesto.KeyaspectsoftheWTAdeclarationare:

Transhumanists advocate the moral right for those who so wish to usetechnology to extend their mental and physical (including reproductive)capacitiesandtoimprovetheircontrolovertheirownlives.Weseekpersonalgrowthbeyondourcurrentpersonal limitations[…] Itwouldbetragic ifthepotential benefits failed to materialize because of technophobia, andunnecessaryprohibitions[…]Transhumanismadvocatesthewell-beingofallsentience(whetherinartificialintellects,humans,posthumans,ornon-humananimals)andencompassesmanyprinciplesofmodernhumanism.(WTA,2005:p.1)

10

The core emphasis on the benefits of AI and on augmenting and ultimatelytransforming the human (and enabling posthumans) has outlasted the variousincarnationsofTHorganizations(whicharethemselvesinflux).12Notably,allversionsincorporate a moral argument: the right of free expression (where TH and theposthuman are creative and liberating) and a duty or obligation to recognize andaccommodate fully realised AI entities. Many adherents of TH also prioritise a‘proactionaryprinciple’overa‘precautionary’one:therequirementtotranscendriskthroughactivity rather thanavoidingor suppressingchangebecauseof recognizedrisks (such as the loss of identity, quasi-Gattaca coercive/competitive eugenicsocieties,worse-casescenariosofabrupttransitiontofullyrealisedAI--a‘singularity’--withTerminatorconsequencesetc).13Theemphasisonbenefitsandtheweightingtowards a proactionary principle is also associated with an ‘abolitionist’ thesis:biotechnology and social transformation can (and should) eradicate suffering as ahumanexperience(byalteringthecapacitytoexperience).ThisthesisismostcloselyassociatedwithDavidPearce,oneofthefoundersoftheWTA.14

TheexoticaofTHispartofitsappealandhascreatedpubliccuriosity,oftenwithinFutureStudies(seeGidley,2017).Moreover,sincethhascreatedthegroundsforTHasadiscursiveresponseitisunsurprisingthatthesamesetofpossibilitieshasprovoked an interest fromother perspectives. Variousmedia have pursued a ‘thefuture is now’ theme in relation to current forms of th advances in science thatnecessarilyraiseconcernsregardingaTHpositionofadvocacy.Forexample,generaladvancessuchasCRISPRgeneediting,recentadvancesinsyntheticbloodproduction,experimentalimplantstocontrolParkinson’s,aswellasmorespecificpioneeringworksuchasthesuccessful implantationofmicrochiptechnology inaquadriplegicmaleenablinghimtograduallyrecoversomeuseofanarm(Mason,2016).OnemightalsonoteElonMusk’srecentcorporatelaunchofNeuralink,acompanydedicatedtothedevelopmentofaneurallaceforhuman-computerinterfacing.Clearly,aspectsofTHpotentials immediately invite legal and regulatory concern, and so parallel someaspectsofhowAIshasbeenmotivated.Thisisamatterof(inthepoliticalsense)publicinterestandpublicconcern. In theUS, forexample, research fromthePewCentreindicatespopularmisgivingsaboutatransitionfromhelpingtheunhealthytoshapingandinterveninginthelivesofthealreadywell(Funketal,2016).Infocusgroups,themoreparticipantswereinvitedtoconsidertheissues,themoreitbecameevidentthatsocietywasunderpreparedtodealwithanyconsequencesbecauseofalackofpublicdeliberationandinformedawareness(Rainieetal,2016).1512TheWTAisnowHumanityPlus:http://humanityplus.orgSeealsoTranshumanity:http://transhumanity.net13FortheproactionaryargumentseeBostromandOrd,2006.Attheextremethisbecomesareversalofaposited‘magic-in-themeat’position(seecritiqueofSearlelater).ForanaccountoffutureriskfromsuperintelligenceseeBostrom(2016)14Thereisapotentialdangerouselisionintheabolitionistpositionsincethereisadifferencebetweenremovingsufferingfromtheworld--changingtherelationsoftheworld--andremovingthecapacityto suffer from the human; the latter does not mean ills are removed merely one’s capacity toexperiencethem.Thisinscribesanunpleasantpotential:createinhumansbecauseweact‘inhumanly’,perpetuateharmsbecausewearesomethingotherthanhuman.15ThePewfocusgroupsweredesignedtocoversignificantsubcategoriesofUSsocietybuteachwasconstructedwithaninternalsimilarityofmembersinordertoexpeditefreeflowofconversation.IngeneralparticipantsexpressedviewssharplyatoddswithaTHproactionaryapproachandemphasisedtheneed for caution and interventionbeginning froma ‘first donoharm’ principle for thehuman

11

Giventhegrowingattentionabove,variousacademicdisciplines,andnotablyethics, have oriented on th and TH. Again, there is some crossover here with AI,notably AIs. For example, there are now Professors of Robot Ethics, such as AlanWhitfield,aswellasBioethics,suchasRobertSparrow.Ethics’disciplinaryinterestandcrossoveralsoextendstocritiqueordefenceofTH(foraninitialrangeseeSandel,2007;Cabrera,2015;Clarkeetal,2016).THhasalsoattractedcritique fromotheracademicpositions.ThoughTHreferstoposthumans,THcanalsobedifferentiatedtosome degree from a set of humanities and cultural studies-based social theoriescollectivelyreferredtoasPosthumanism.ThesetendtobecriticalofTH,locatingitashastyvalorisationofnoveltyandfantasythatdoesnotpaydueattentiontofeasibilityor to the social theory basis of society.16ManyPosthumanists are still engaged inrethinkingtheoryasanecessaryprecursortoanyemphasisonremakingsocietyorunderstandingthehuman(seeBadmington,2000;Herbrechter,2013;Wolfe,2009;Braidotti2013).ThesourcesforthisrethinkingrangeacrossButler,Deleuze,Derrida,Foucault,LatourandWoolgar,Haraway,Luhmannandmanyothers.

Otheressaysinthiscollectionhavemoretosayabouttheinter-connectionsbetweenTHandPosthumanism.Forourpurposes,itissufficienttonotethatthemainthreadsofPosthumanismtakeadifferentapproachtodecentringthehumanthanweintroducedearlierinthisessay,andthisneedstobedistinguishedheretoavoidanyconfusion. Literary theory, cultural studies, post-structuralism and postmodernismemphasise the entanglement of knowledge and power and tend to associate theEnlightenmentandhumanisttraditionwiththeuncriticalprojectionofuniversalsthatare actually expressions of oppressive andmarginalizing particularities, as well assourcesofdangerouslyposeddiscoursesofscientisticsciencethatfosterharms.Assuch, decentring the human is seen as an important theory move in opposingproblemsofgenderconstructs,ecologicaldestructionandsoforth.Manyrealistshavegreat sympathy with the intent, but are sceptical regarding the ontologicalimplicationsofsubsequenttheoryandcritiques.17ThereissomecrossoverherewithissuesalreadyexploredinthefivevolumeCentreforSocialOntologyworkinggroupMorphogenic Society project (see Morgan, 2016). Realists argue that the newmaterialism, vitalism, actor network theory etc. replace one set of problemswithanotherset(flatontologies,lackofadequatelyexploreddifferentiation,conflationofparticularsasuniversalsinepistemologywithessenceandkindinontologyetc.).

(basedonunintendedconsequences),andwithdueattentionpaidtopreventingtheexacerbationofcurrentinequalitiesbasedonprivilegeenablingthefewtopayforaugmentationsandchangesthatputthemandtheirdescendantsapart(notinghoweverthatsocietyisalreadyunequalandthattechnologymight actually allow for equalisation, depending on how it inhered in society). The responses alsomirroredtheidiosyncraticUSsuspicionofbiggovernmentpreferringsomemorecross-socialmeansofoversightoftechnologicalchange(seeRainieetal2016).16ForanearlyfeasibilitycritiqueseeNordmann,2007.NotethatrealistcritiqueseemstologicallytradeontheirreversibilityofTHtransformations(sosomethingisharmedorlostwithoutfullconsiderationofwhatisharmedorlost).However,reversibilitymaychangesomeoftheforceofargument--sinceexperimentalTHtocreatemutableentitiesmaycarrydifferentforceofargument(andisastapleofspaceoperasci-fiutopia,suchasIainBank’sCulturenovels.17OnemightalsonoteSteveFuller’sHumanity2.0 (2011)here.Fullerraisesmany important issues(especiallythemoralhorizonofthehuman)butdoessointermsofhisusualsocialepistemology.Hesetsouthowdiscoursedisputesandmakesambiguousscienceinsocietybutultimatelyprovidesnodefiniteposition regardingwhat itmeans tobehumanorwhat theprospects forhumanityare (orshouldbe).Thoughthoughtprovoking,theworkexploresevasionevasively(seeMorgan,2013).

12

The initial point we have made in this essay is that an AIw functional approachdecentreswhatitistobehumanor,tobeequivalentintermsofpossibleessentialcharacteristics,andthatanAIsentityfocusmaynoteasilyresolvethis.ClearlymattersofkindandessenceareproblematicforPosthumanismbasedonitstheorysources,and for TH based on a combination of optimistic emphasis on the benefits oftransformationsandaproactionaryprinciple.Again,othershavemoretosayaboutthis.MyconcerniswithproblemsofsocialontologythatmaybecomeingrainedorinterstitialbasedonhowAIhasbeenconceived.Assuch,InowmoveontoTuringandSearletoconsiderthe‘sophisticatedorigins’andimplicationsofAIwandAIs.Turing,AIwandAIs:findingaquestionthatcanbeansweredInhisseminal1950paper‘Computing,machineryandintelligence,’TuringclearsthegroundforadominantAIwfocusonfunction,whilstalsosettingthesceneforAIssub-categoricalconcerns.Turing’spointofdepartureis:canmachinesthink?However,forTuring, the ordinary language sense of this question is too ambiguous and thisimpedesanysatisfactoryanswer.Assuch,what isrequiredisasubstitutequestionthat can inprinciplebeanswered. Specifically, couldamachineprovide responsesindistinguishable from those a human provides, and so pass for human? Thesubstitution takes the guise of a thought experiment in the form of a test, the‘imitationgame’.Turingdescribes thegame initiallyasoneplayedby3people:aninterrogator (C) and a man (A) and woman (B). C is in a separate room andcommunicationisviasomemedium(anotherpartyorbycards,teleprinteretc).Theinterrogator(C)isunawarewhichofthetwoothersisthemanorwoman.ForCtheyaremerelyxandy.ThetaskforCistodecidewhichofxandyisthemanandwhichthe woman. The man (A) is given the task of confounding the interrogator’s (C)attemptstoidentifywhichofxandyisthemanandwoman,andthewoman(B)isgiventhetaskofhelpingtheinterrogator.However,neitherAnorBcansimplystatewhoiswho.

TuringthenproposesthatamachinetakethepartofA,andbymachinehemeansadigitalcomputerofsomepossiblefuturevariety.Thisdigitalcomputerisamachinedesignedtocarryoutoperationsthatcouldbedonebya‘humancomputer’.Ahumancomputer in these terms isone thatwouldbe ‘following fixed rules’andwithout‘authoritytodeviatefromthem’(Turing,1950:p.436).Thedigitalcomputer(whichstandsintheplaceofAI)isthendescribedasanextrapolationofcontemporarytechnology: a technology that follows fixed rules where the technology is acombinationofastoreof information,anexecutiveunit thatcarriesout individualoperationsinacalculation,andacontroltableofinstructions(aprogramcode).Turingdescribes this digital computer as for-all-intents-and-purposes a ‘discrete-statemachine’.Thatis,onethatfollowsrulesandshiftsfromonedefinitestatetoanother,whichunlesserroroccursisultimatelypredictable(inabasic‘Laplacian’sense).Itthushas a clear set of input-outputpathways.However, given that one canprogramadigitalcomputertofulfilanyfunctionthatcanfollowthisprocedureonecandescribedigitalcomputersas‘universalmachines’(Turing,1950:p.441).Turingthensuggeststhatadigitalcomputerwithasufficientlylargestoragecapacityandprocessingspeedcouldinprincipleplaytheimitationgame.

13

Readcarefullytheimitationgameisclearlyatestoffunctionalequivalence.TheinterrogatorCissimplybeingaskedcanyoudistinguishbetweentwoentitiesinrelationtotasksandindicativecharacteristics.Thoughtherangeoftaskscanbewidetheunderlyingstructureofthemfortheoriginalargumentishighlycircumscribed:asetofrulefollowingactions.However,thereismoreinvolvedsinceTuring’sunderlyingargumentcanbedifferentiatedfromtheinitialunderlyingstructureofthegamehesetsout.Thisbecomesclearbasedontheexamplesusedandthewaysomeofthe9counterargumentsorobjectionstothetestareaddressed.

Theunderlyingstructureoftheimitationgamerequiresamachinetofollowaprogramthatenablesittosimulatetheresponsesofahuman.Thehumanpointofreferencediscussedasequivalentintheargumentinthepaperisahumancomputerwho does not deviate from fixed rule following behaviour. However, the humanresponses actually illustrated for the game are wide-ranging and include morenaturalisticresponses.Theydonotfocusonlyonahumanengagedinunequivocallyfixed response answers to given questions. For example, they are not closed-endyes/no issues or simplymatters of a human calculating. Clearly thiswould be toonarrow to satisfy any reasonable form of imitation test. Significantly, then, thestatementofarangerequiresanintuitiveleaporinferencethatadiscrete-stateyet‘universal’machinecanatsomepointinthefutureplaytheimitationgameinawaythatcananswerthepossiblerangeofquestions.Ataminimum, itmustbeabletoaddressTuring’sillustrationsinthepaper.Mostnotablyinsettingthesceneforthegame: ‘C:Will Xplease tellme the lengthofhisorherhair?’ IfA is X: ‘Myhair isshingledandthelongeststrandsareabout9incheslong’(Turing,1950:pp.433-434).And then in the following section when identifying different ‘specimen’ questionforms:‘Q:PleasewritemeasonnetonthesubjectoftheForthBridge.A:Countmeoutonthisone,Inevercouldwritepoetry’(Turing,1950:p.434).ForTuring,passingsuch a test is sufficient to answer the question: can amachine imitate a human?However, there ismore to it than this, sinceTuring is reasonably confident that amachinewilleventuallypassthetestandinintroducingpossibleobjectionshestates:‘Ibelievethatattheendofthecenturytheuseofwordsandgeneraleducatedopinionwillhavealteredsomuchthatonewillbeabletospeakofmachinesthinkingwithoutexpecting to be contradicted’ (Turing, 1950: p. 442).18 He situates this claim as‘conjecture’,buttheimpressionconveyedisimportant.Theclaimismadebasedonthedigitalcomputer,thehumancomputerandthentheintuitiveleap.ForTuring,thecurrent impediment to playing the game is processing capacity and speed.19 Theinference is that it is based on future technology following similar lines and withreferencetothegamethatitwillbereasonabletoclaimthatamachinethinks(despitethatthegameisaboutimitation).

18Analternativeargumenttotheonethatfollowsisthatthe‘useofwords’changes,andsothinkingisdefineddifferently,suchthatadigitalcomputerandhumanthinkwithoutaddressingthepointsImake.This,however,doesnothelpTuring’sposition,sinceitreliesonsemanticincoherenceasasolutiontosubstantive incoherence of argument, and so replaces one problemwith another, where the newproblemfailstorespondtotheoriginalissue:itmerelyrepositionsandevadestheissue.19‘Ibelievethatinaboutfiftyyears’time,itwillbepossibletoprogrammecomputerswithastorage

capacityofabout109,tomakethemplaytheimitationgamesowellthatanaverageinterrogatorwill

not have more than 70 percent chance of making the right identification after five minutes ofquestioning’(Turing,1950:p.442).

14

Now,considerwhatisinvolvedinTuring’sposition.Thepointofdepartureisequivalencebetweenahumancomputerandadigitalcomputer,butthelanguageuseof the former is ambiguous once the argument starts to be extended. There issemanticslippagefromtheclaimthatahumancancomputeandthisfollowsfixedrulesorprocedures(whosearchetypalformiscalculationaccordingtoformulae)totheimplicationthatahumancomputingisequivalenttoahumanthinking.Turing’sargumentclaimsadigitalcomputer following fixedrulescanachieve imitationandalso(asconfidentconjecture)impliesthatthetechnology(presumablywithreferencetothegame)willsettletheissueofwhetherdigitalmachinesthink.Forthis lineofreasoningtobeplausibleitmustrelyontheassumptionthatahumanthinkingandadigitalcomputercomputingareequivalent.ThereisthusalogicalsubstructurethatencouragesbothanAIwfocusforfutureAIresearchers(aconcentrationonfunctionwhere the problem of the entity will take care of itself based on technologicaldevelopment)and(beneaththeconjecturecaveat)anAIsfirstsubcategoryclaimthatcaneasilybecomeaconflation:thehumanmindisequivalenttoacomputer,sothehumanmindisacomputer.Thislatterformalsotradesontheinferencethatahumanmindcanbereducedtoinput-outputprocedures(andsoisalsooverwhelminglyaboutfunction).

AtthisstageitisimportantnottotraduceTuring,butrathertohighlighttheproblems of his lines of reasoning. Turing’s claim implies equivalence but heintroducesfurthercaveatsandconsiderationsinaddressing9objections.Threeareimmediatelyrelevanthere.Turingacknowledgesthatahumannervoussystemisnotadiscrete-statemachinebutarguesthatifamachinecanplaythegameandthusbeindistinguishable fromahuman then this is irrelevant. The implication is thus thatequivalenceisnotidentityofconstitutionbutsimilarityofoutcome.Inthecontextofthe claim about machines thinking, the implication is thus that behavour is thesignificantlocusthatallowstheinferencethatamachinecanthink.Ifonereversesthelineofreasoningthentheimplicationisalsothattheinternaloperationofthehumanin the act of thinking is conducive to theequivalenceof outcomes andequivalentstatus(boththehumanandmachine‘think’).The‘difference’thusseemstomakeno‘difference’.However,Turingalsonotestheobjectionthatthegameisnotatestofconsciousness, since thedigitalmachine isnot required to (ordemonstrate that itactuallycan)knowwhatitisdoing,orfeelemotion.ThisseemstoindicatethatTuringissimplyclaimingthatthetestispurelyamatterofsimulationbasedonthegame.

However, the verypoint of the game is to replace the ambiguousordinarylanguagevarietyofquestion(canamachinethink?)withanoperationallyanswerablequestion(canamachineplaytheimitationgame?).ItisthisthatunderpinsTuring’sclaimthatinthefutureitwillbepermissibletostateadigitalcomputercanthink.Forthis to be so, one must, therefore, invoke the underlying assumption that it ismeaningfultosubstitutethelatterquestionfortheformer,whichinitselfcanonlybemeaningfulasanact ifoneassumesthesubstitutedquestionbearson theoriginalone.Thecontextandpurposeofthegamethuscreate implicationsthatshapethesubsequent caveat: ‘I do not wish to give the impression that I think there is nomystery about consciousness’ (Turing, 1950: p. 447) in terms of the subsequentstatement:‘ButIdonotthinkthesemysteriesnecessarilyneedtobesolvedbeforewecananswerthequestionwithwhichweareconcernedinthispaper’(Ibid).Onemightinferthenthatthe‘mystery’canbeanswered--Turingdoesnotsay.However,

15

he immediatelymoves onto the possible objection that there is no evidence thathumans follow lawsofbehaviourandactualactivity isvariable in termsofspecificconduct.Here,henotesthatgenerallawsofbehaviourmayexistthatconditionthescopeofvariabilityinspecificconduct.Turingseemstobeimplyingherethatgenerallaws inhere in thehumanthinkingandsoequivalence isan interiormatterof rulefollowingbehaviourratherthanmerelyexteriorcircumscribedequivalentoutcomes.There is, therefore, grounds to infer based on Turing’s argument that a humanthinkingandadigitalcomputerthinkingare(orcouldinthefuturebe)doingthesamething.

It remains the case that Turing’s position is sufficiently underdeveloped toallowdifferent interpretations.Asaconsequence,thereisa lineofAIssubcategorywork thatexploresTuring’s imitationgame in termsofwhatheactually intended,includingwhetherheintendedthegametobebehaviouralinitsimplicationsbecauseof its functional focus (forearlyexamplesthatassessthedebatesee,Millar,1973;Lassegue, 1988). This notwithstanding, the point I want to emphasise is that thedirectionofargumentTuringpursuesintermsofkeyobjectionsfollowsapattern.Heaffirmstherelevanceoftheimitationgameasavalidtestanddoessobyorientingonthesignificanceofbehaviour, functionandequivalence.Thisenablesa slide in theargumentsuchthatequivalenceisamatteroffunction,whichissuggestiveofmorethanmerefunction:functionbecomesthesignificantindicatorof‘thinking’.So,whilstthere are manifest tensions in the way Turing reasons, his argument conveys animpressionregarding‘thinking’relevantto(andencouraging)bothAIwandAIsdespitethat the game is about imitation.20 Since it has givenencouragement toboth, theoriginalproblematiccreatedbyTuringisthusmultiplysuggestive,andsoambiguousin its particular implication, despite that it is constructed to enable definitiveconsideration of what Turing considers would otherwise be too amorphous aproblem.ThisreturnsustoTuring’sintent.Turingintendstofindaquestionthatcanbeansweredthatcanstandinforanordinarylanguageapproachto:canmachinesthink? However, one can reasonably ask: 1) does the form of the imitation gameargumentasconstructedbyTuringactuallyprovidegroundsforconcludingthatthegamecanbeplayedeffectivelybyadigitalcomputer?2)isthenewquestionactuallyanappropriatesubstitutefor:canmachinesthink?ExploringtheformerhighlightsthetensionsinTuring’sposition,whichhavecontemporarysignificance,whilstidentifyingthelatterprovidesanentrypointtoSearle’sapproach.Playingtheimitationgame,substitutionasseduction

Thequestion:doestheformoftheimitationgameargumentasconstructedbyTuringactuallyprovidegroundsforconcludingthatthegamecanbeplayedeffectivelybyadigitalcomputer?createsmultiplegroundsfordispute.Theoriginal imitationgame20This impressionisalsocreatedbyTuring’sreplytothe‘solipsist’ implicationsoftheargumentforconsciousnesswherehestatesthatasonnetwritingmachinecapableofdemonstratingopinionwouldbeunlikelytobedescribedasmerely‘signalling’(seeSearleonthefallacyhere).Note:ithasalsobeenpointedoutthatthe9objectionsmainlyrelyonemphasisingthelackofevidencetorefuteTuring’spositioni.e.absenceofproofisnotproofofabsence.However,suchargumentsforthenegativehavethegeneralpropertyasargumentationstructuresthattheycontainnoevidencefortheargument,andasreductioadabsurdumthisallowsanynon-evidentialclaimtohold(particularlythoseforwhichnoevidenceseemslikelytobeeverforthcoming).ItisarguablehowfarthisactuallyappliestoAI.

16

requiresanAandBasxandytoeitherseektofooloraidtheinterrogator.Thisisanopen-endedstrategicproblemofcontext,eventhoughitisstatedastworoleswithamaster directive for each role (in addition to the role of C). There is a significantdifferencebetweenacodifiedresponsetoaspecificquestionwithinthisremitandthestrategicnarrativethatemergestoexpressthatremit.Afixedruleapproachmust,therefore,solvetheproblemofstrategicconversation ratherthanmerelysyntacticand semantic consistency of an individual response. The terms of the game thusconcealtheproblemsofcompleximprovisationandthenaturalisticfeelthatsuchaconversationmustconvey.Turingdoesnotresolvetheseproblems,sincehisactualexamples (quoted previously) focus on individual or single responses rather thanstringsorpathwaysofinteractivedialogue.21Hisfirstexample(hair)seemstoreducequestions and answers to a simple problem of logic where a game is deductiveelimination(e.g.xisnotBbecauseofanswerz).Hissecondexample(thesonnet)issimplyaformofevasion.However,theformerexampleashumanconversationcouldquickly become confounded by non sequitur, confusion and ambiguity ascharacteristicsoftheconversation,andsoadeductiveapproachascodingwouldfindthisdifficultifnotimpossibletocopewith.Codingresponsesthatsoughttorectifytheproblem by putting the conversation back on track would immediately strike aninterrogator as non-naturalistic, creating suspicion likely leading to failure in theseamlesssubstitution(machineforman)aspectoftheimitationgame.Theverynameof the game trades on a conflation of two different purposes: imitation quaman/woman for any player (implicit in the different role remits of A and B) andsubstitutionofmachineinonerole.Inthecaseofthesecondexample(thesonnet),ifcodifiedresponsestosuchquestionsareallintheformofevasions,ratherthaneitherdemonstratedabilityorqualitativeopinion,thenthereisanothercumulativeeffectofsuspicion.Theinterrogatorwillbecomesuspiciousastowhatkindofentitytheyaredealingwith.ThisisadoublehermeneuticproblemoncetheTuringtestbecomesamatterofpublicknowledge(asitwouldovertherestofthecentury).

Both examples highlight that an interrogator is an interlocutor within adialogicalopenprocess.Assuch,itisnotclearlyestablishedthattheTuringtestcanbepassedbasedonthebasicfoundationsoftechnologyasstated,andextrapolatedfrom,by Turing: a discrete-if-universalmachine.Multiplyingdiscrete functions is aconfusionofwhatuniversalimplies,sinceitindicatesuniversalityismerelyadditive.This is a problem that continues to dog contemporary AI technology. There is adifferencebetweenoperativeefficacy ina taskandnavigatingseamlesslybetweentasks (AI researchers typically refer to this in terms of the specific and generalintelligence problematics). For our purposes, there is something analogous inconversationsinceitisthewayoneadaptsandcontributesthatsignalsconstructiveappropriateness.Atrootthereisanotherproblemofemergencehere(toaddtotheissueofspectrumintelligence).

Moreover,carefulconsiderationoftheexamplesandtheproblemofdialogueindicatethatwhatitmeanstopassthetestisalsodisputable.Thepurposeofthetestisclearlystated,butthespecificdesignofthetestandthecontextinwhichitisappliedarenotfullydevelopedbyTuring.Assuch,hedoesnotaddresstheissuethatcodingcouldbedevelopedinaccordancewithhowthetestisspecificallyoperationalised.So,

21Thoughonemightrefertohisobjectionbasedonconsciousnesspointhere.

17

‘passing’thetestcanbecomeamatterofpassingatestinthecontextofgamingthegameorsubstitutingsomeotherversionofthetest.ThisbecameamatterofsomecontroversyinJune2014whenthe‘chatbot’programEugeneGoostmanpersuaded10of 30 judges (33%) at a Royal Society organizedAI event that it could pass forhuman.22The2014RoyalSocietyeventwasnotrecognizablyTuring’simitationgame.Itwasa5minutekeyboardbasedinteractionbetweenjudgesandtheprogram,wherethejudgeswereaskedwhethertheycoulddistinguishaprogramfromahuman.Theeventsetathresholdofsuccessatpersuasionof30%ofjudges.TheEugeneprogramsimulatesa13-year-oldUkrainianboy,andsotheformoftheprogramcreateslimitedexpectations for the range of interactions and builds in anticipations of errors,evasionsandinconsistenciesthatthenbecome‘idiosyncrasies’.Thisisquitedifferentthan simulating a fully operational adultmeeting core normswithin any given yetopen-ended socio-cultural milieu over an extended duration. Passing the RoyalSocietyeventtestthusquicklybecameamatterofwhatkindoftestwaspassedbasedonwhatkindofcoding.Thebest-knownAIeventistheannualLoebnercompetition,which has offered a cash prize of $100,000 since 1991 for a program fullyindistinguishablefromahuman,andasmallerprizeforthebestentryoftheyear.Asof2016,noprogramhadwonthe$100,000.

However,disputeregardingtestdesignisnotitselfadecisiverefutationthatadigitalcomputercanbeprogrammedtosuccessfullyplaytheimitationgame.WhatonecanstateisthatTuringdoesnotestablishthatitispossible.However,againitisimportant not to traduce Turing.His specific developmentof the formof a digitalcomputer is as a discrete-state machine with definite input-output relations.However,indiscussingobjectionsheconsidersthepossibilityofa‘learning’machinecapable of ‘induction’ (not abduction/retroduction). He makes no attempt toarticulatehowthismightbeconstructed,norisitcentraltohisargument,buthedoesatleastintroducethepossibility.Chatbotsandrelated‘AI’technologiesareconstantlydevelopingandtheuseofbigdataanalyticsdrawingonahugepoolofconversationandcommunicationcreates thepossibility thataneffectivedigital computer coulddraw on blocks of similar responses from similar situations, and so simulatenaturalisticlanguagewithalevelofapparentsophisticationthataninterlocutorwoulddeemappropriate.Thispossibility fallsunder theremitof ‘learning’programs.Theimplication is that it is,asTuringclaimed,only (thoughperhapsmainly)processingcapacityandspeed,and time (time forAI to ‘learn’, time forcoding todevelopasproblems are identified and solved) that stand between the digital computer andsuccessfulplayingofthegame.Thereisalsoaconvergenttechnologyargumenthere,since imminent developments such as quantum computing offer the possibility ofsignificant leaps inprocessingcapacityandspeed(ifsothenMoore’s lawdoesnotconfront the impending limit entailed by non-quantum processing). The inferencedrawn would then be that chatbots such as Apple’s Siri and Amazon’s Alexa willbecome or have descendants that are increasingly naturalistic in their interactions(pushing past momentary embarrassments, such as those created by Microsoft’sTay).23

22SeeBBCcoverage:http://www.bbc.co.uk/news/technology-2776208823RecentresearchpublishedinScienceindicatesthatthoughitmaybepossibletomakechatbotsmorepolite,sincelearningisbasedonbigdatasetsinwhichmeaningisembedded,AIfacesadeeperproblemofabsorbingpre-existinghumansocio-culturalbiasbasedonthewaylanguageisassociatedandused:

18

Clearly,technologicalchangeisalteringwhatitmeanstotalkabout‘AI’andthe terminology across the field is likewise changing as new fields propagate. AIresearch has moved on from simple discrete-state input-output concepts andapproaches, and Bayesian or Boolean solutions. There is a heavy emphasis onquantifying‘uncertainty’.Thereisalsoincreasinguseofthelanguageof‘complexity’todescribeAI.24TheStanfordreport,forexample,makesmuchoftheseasthecuttingedgeofthefield.25Atfirstsight,therewouldthusseemtostillbesomecredenceinthewayTuring’sapproachseemstohaveclearedthegroundforadominantAIwfocusonfunction.AIwillbewhatAIresearchersdoandinsofarasAIbecomecapableofpassing theTuring test,whatAI researchersdowillultimately (if later thanTuringanticipated)fulfiltheexpectationthat‘onewillbeabletospeakofmachinesthinkingwithoutexpectingtobecontradicted’(Turing,1950:p.442).However,considerhowthis claim is positioned. Both ‘thinking’ and ‘learning’ are deeply ambiguous andcontestabletermswhenappliedtoanAI,andarguablytheshiftfromdiscrete-stateasadefiniteinput-outputrelationtomorecontingentapproachesisachangeofscopenotofform.

One might argue that it is the extension of terms under ambiguity thatunderpinssomeofthedifferenceinhowargumentispositionedandclaimsarenowmade.Forexample,oneofthemajorinnovationsincurrentAIis‘deeplearning’usingartificialneuralnetworks(ANN).ANNaredescribedassoftwaresimulationsofneuronconnectivity (Economist, 2016).26 That is, they aremultiply layered sets of ‘neuralunits’creatingmultipledividingpointsfordirection,asprocessing,fromsomegiveninputtosomeoutput.Thesophisticationofthesystemoritscapacityfordifferenceandrangeisbasedonthenumberoflayers,the‘depth’,inthestructure.Whatthesystemisdirectedtocanthen(currently)besetupinthreewaysexpressedaslearningmodes: 1) supervised learning (a network system is fed an example dataset thatexemplifies what it is intended to achieve, such as spam identification) 2)unsupervisedlearning(anetworksystemisfedanexampledatasetandissetuptolook for patterns, clusters anomalies in the data, which then become the specific‘Weshowthatstandardmachinelearningcanacquirestereotypedbiasesfromtextualdatathatreflecteveryday human culture... stereotypes and empirical associations, has long been known in corpuslinguistics… since we performed our experiments on off-the-shelf machine learning components[primarilytheGlobalVectorsforWordRepresentation(GloVe)wordembedding],weshowthatculturalstereotypes propagate to artificial intelligence (AI) technologies in widespread use.’ Caliskan et al(2017:p.183).Thisisinadditiontorelatedproblemsthatcommercialchatbotsareoftendesignedwithfemale voices, which are considered, non-threatening and submissive (drawing on and creatinggenderedeffects).24Inrealistcritiquethereareclearontologicalproblemsinvolved:theshiftfromsimpledeterministicinput-outputtomodelled/programmeddefinedrangesofreactionandresponseproducesaproblemofprobabilisticframingthatdoesnottranscenddeterminismbutmerelyresituatesit(andalanguageof quantified uncertainty cannot disguise this). Note also that complexity theory claims as a keycomponent‘emergence’.OnemightcategorisesomecomplexitytheoryasAIswork,butthereisalsoatendencytouseemergencelooselytofocusonfunctionalefficacyasrepeatedoutcomeachievement(ratherthandifferentiateeventsandconsideremergenceasapropertyofanentityasasourceofcausalpowerdistinctfromoutcome).Interalia,complexitysitsawkwardlywithspectrumclaims.25Revolutionshavebeenheraldedbefore,seeChurchlandandChurchland(1990)26 So, there is an immediate issuehere since simulationofneurons is a claim that theoperation isneuron-like,thoughonceonestartstoconsidertheactualstructureofthetechnologyratherthantheclaim made then it becomes clear the statement of neuron-like owes more to metaphor thansubstantiveevidence.

19

output within a broader data-defined remit, such as fraud patterns in insuranceclaims)3)reinforcementlearning(anetworksystemisfedanexampledatasetandrefines its behaviour based on rewards as feedback to achieve goals, creating asimulationof‘dowhatworksbestinsituationx’,suchasplayingandwinningavideogame).27Inallthreecasesthekeyinnovationisthatthenetworkprogressivelyrefinestheweightingbetweenconnections,anditthusfine-tunesthenetworksystem.Themoredatathesystemhastoworkwith,themorelayerstotheneuralnetworkandthemoresimulationsrun,thenthemoreeffectivethesystembecomes,overtimeandinrealtime,subjecttoprocessingcapacityandspeed.

Since2012therehavebeensignificantadvancesinANNAI.28However,ifonedecodesthelanguageoflearningbeingusedthenitisaboutrefiningasystem.Thisistermedtrainingand/orlearningbecauseANNdoesnotdependonprecisecodingofeverypossiblesituationasan‘ifythenx’.Ratherthandefiniteinput-outputrelationsonenowhasdefinedrelationsofinputsandoutputs,butstillafocusedsystemthatisallaboutachievementofsomegoal.ClearlyANNhasscopetobemoreflexiblethanthe coding that Turingwasworkingwith, but as yet a barrier still exists based ontransferring between different specific functions (the problem of ‘generalintelligence’),sincethisstillrequiresreconstructionofthesystem.Moreover,systemsstillrequirespecificationthroughfunctioninordertoexistatall.Ahuman,arguably,isnotreducibletospecificationquafunction(andsoasabeingisunspecifiedinthissense).Thereisno‘Iamfunction’forthehuman,asarestrictiononconstructionandexistence.Moreover,ifoneplacesthepotentialofANNinthecontextoftheTuringtest,andconsidersthecommunicationmilieuthenarguablyany‘learning’AIisusinglanguage,itisnotintheordinarylanguagesense,alanguageuser.

Thefundamentalquestioniscanonebeintelligentorlearningoralanguageuser if one lacks consciousness, self-consciousness or awareness? One can simplyloosentheuseofthetermsbyextension,andtradeonambiguity.However, ifonelacksconsciousnessetctheninwhatsenseisitsemanticallyappropriatetousetermsthatattributeunderstandingtoanentity?Atroot,intelligencerequiresonetomakeintelligible,andsoforth.Inthecaseofchatbots,theycanbemoreorlessnaturalisticandsomoreor lesseffectiveinsimulatingauthenticity,butthisdoesnotinandofitselfchangethestatusofthechatbotintermsoftheimitationgame,unlessfunctionalefficacyconfersthestatusofthinkingorunlesssomethingadditionalhasoccurredthatisnotyetdemonstratedabouttheentity inquestion.Thisreturnsustothecentralproblem created by the very existence of the imitation game. Turing replaces thequestion,canamachinethink,withthequestion:canamachineplaytheimitationgame?Aswehavenoted,thisonlymakessenseifweassumethelatterbearsonthe

27 TheEconomist articlenotes thatDeepMind’sAlphaGo systemuses twodeepneuralnetworks, areinforcementlearningnetworkandarandomsamplingnetwork.Onethrowsuppossiblemovesthattheotherthenplaytests.Itwasthisthatenabledthesystem’smuchpublicisedachievementsinthegameGo.28Ingeneral,‘deeplearning’programsarecapableof‘recognition’(objects,audio,speechetc)andhavemultipleapplications;interaliaonemightalsonotealgorithmicgametheorycreatesdecisionmakingmatriceswhererulescanbeadjusted.

20

former.However,thereplacementinvokesthesecondquestionweidentifiedintheleadintothissection:isthenewquestionactuallyanappropriatesubstitute?29

Based on the argument so far Turing’s replacement seems like a shift thatencouragesproblematicfoci,ratherthanawarrantablesubstitution.An‘answerable’questioncanbeinappropriateoratleastinsufficient,andthiscanbecomeincreasinglyevident as time passes and subsequent work is undertaken. So, onemight argueTuring creates a point of departure that ingrains a bifurcation between focus onfunction(withatleastimplicitproblematicconsequencesforentitycharacteristics-intelligence,thinking,learning)andreactionsthatdrawattentionbacktoentitiesthatreconsider the nature of intelligence, thinking and learning as well as furthercharacteristics.

Atthispointonemightbetemptedtosay:sowhat?IfAIwisafocusonfunctionand AI research cumulatively develops to achieve specific functions then does itmatterwhetheranAIreallythinks,hasintelligenceandlearns?Inatrivialsensetheanswermaywellbeno.However,inamorebasicsensewhetheranAIreallythinksetc really matters because of the many social consequences of AI. Failure toappropriatelyconceiveofthenatureofentitiesistoinviteobfuscationandthisinturnisindicativeofabasicontologicalomission.HowwerefertoAIactsbackonhowweconceiveofthehumanandsohaspossibleconsequencesforhowwevalue,preserve,developandnurturethehuman.Thisrelationisnotamatterofeffectivecause,butofdiscursivecontext,raisingissuesregardingcausationbasedondistinctionsbetweenexercisingapowerandbeinganoperativesourceofinfluence.Moreover,thefocuson function without due consideration to entities creates a kind of residualbehaviouristicpresumption.Thereafter,anAI functionalismcanserveto legitimateTH,insofarasfunctionalismlendsitselftotheinferencethatanAIisthemeasureofthehuman(allowingamovewherenothingsignificantaboutthehumanislostinaTHorposthumanfuturebecauseoftacitequivalenceassumednow,whichinturnleadstocomplacencyregardingwhathappensnowas itaffectsanypossible future).AIwthus feeds TH. Concomitantly, the focus on function can marginalise properconsideration of what is also lost for humans through what is done on behalf ofhumans.Iwillsaymoreaboutthisintermsofrelationalgoods.

AsalastpointhereconsidertheinteraliaeffectoftheTuringimitationgame.Function is highly seductive. It can become its own self-confirming technocraticdiscourse followingAIw rationales,and,asalreadynoted, it can inspiredisciplinaryresponsesalongAIslines.Itis,forexample,easytobecomeseducedbytheminutiaeoftheimitationgame.Wehaveprovidedmorethan3pagesofanalysisregardinghowandifthegamecanbeplayed.Thiscouldeasilybeextended(andhasbeen:Crockett,1994;Millican andClarke, 1996; Saygin et al, 2000).30As JamesMoornotes inhisintroductiontothespecialissueofMindcelebrating50yearssincethepublicationofTuring’spaper,‘Thisarticleisarguablythemostinfluentialandwidelyreadarticleinthephilosophyofartificialintelligence.Indeed,mostofthedebateinthephilosophy

29Note,Turingputsasidethequestionwhatwouldamachinethatcouldthinkthink?Thisalsoturnsout tobe importantonceone shifts to Searle’s critique (see subsequent argument andalsoPinsky1951).30NotelongerworksregardingTuringalsoincludeanalysisofSearlesincethelatterfollowsfromtheformer.

21

ofartificialintelligenceoverthelastfiftyyearsconcernsissuesthatwereraisedanddiscussedbyTuring’(Moor,2000:p.461).

ThisbringsustoSearle.SearleisnotthefirsttorespondtoTuring(seePinsky,1951;Mays,1952),buthesetsinmotionmanyofthepointsIhavealreadyraisedandprovides the archetypal AIs argument that contests the consequences of Turing’sformulation:itsseductivequalities,whichinviteafocusonfunction,aconcernwithminutiaeandaproblematicslidefromequivalencetoconflation,wherethehumanmindisequivalenttoacomputersothehumanmindisacomputer.31Inrespondinginthenegativeregardingthislatterposition,Searleintroducesaparticularemphasisto the framework where simulation is the basis of the substitute question: canmachinesthink?Itisbasedonthisemphasisthatinterstitialproblems(basedonthefociofAIwandAIsandtheemphasisofthelatter)canbeidentified.Respondingtotheissue of simulation creates further grounds for argument and these have resistedagreement,despitethesignificantplausibilityofSearle’scase.Searle:AIs,semanticversussyntaxandthefailureofsuccessfulsimulationTobeclear,TuringisnotSearle’simmediatetargetinhis‘Minds,brainsandprograms’(1980). His point of departure is the mutual influence that the prominence ofcomputerisationhashadforandwithcognitivescience,andhenceproblemsinthephilosophy of mind. He distinguishes the use of computers to study the mind (afunctional tool) from the claim that a mind and a computer are the same (theyfunctioninthesameway).TheformerisSearle’sversionofAIwandthelatterAIs.HisuseisthusnarrowerthanIhavepreviouslysetoutfortheseterms.Afocusonfunctionisnotreallyhisprimaryconcern,atleastinthesenseofthesubsequentfocuson,andconsequencesof,howAIdevelopstofunctionintheworld.Heisratherconcernedwith the problem of functionalism, and initially with behaviourism. His aim is todemonstratethatthemindandacomputer,ascurrentlyconceived,arenotthesame(andsoonecannotclaimthathowacomputerworksexplainshowamindworks).Inso doing he acknowledges that Turing attempts to put aside the problem ofconsciousness,andyettheimitationgameasasimulationremainssubjecttocritique.The point of the critique is to establish that a computer and a human mind aredifferent, even if the superficial consequences can be the same: a successfulsimulationremainsmerelyasuccessfulsimulation,unlessonecandemonstratethattheinnerworkingsofbothmindandcomputer(againstandinginforAI)havesimilarcharacteristics.His focus is thuson theentity rather thanmerely theoutcome.HeclearlysetsoutafirstsubcategoryAIsposition,andtheargumentcanbelocatedasaprimaryrefutationofthesubstitutequestionthatsimulationissupposedtooffer. Searle’s critique takes the form of a thought experiment. The thoughtexperimentreversesTuring’sgame.Searlecreatesahumansimulationofacomputer,ratherthanintroducesacomputerasasimulationintoagametoidentifywhois(whatgenderof)human.Thecritiqueisnowcommonlyreferredtoasthe‘Chineseroom’thought experiment. A person is placed in a locked room, which contains some

31PreSearle,perhapsthemostnotableareBlockandGunderson.Note,BlockprovidesaprototypeChinese Room argument in ‘Troubles with functionalism,’ commonly referred to as the ChineseGym/Nationargument(citizensaregiveninstructionstophoneanotherinanetworkcreatingapatternofcallingthatreplicatesneuron’sfiring:isthecollectiveChinaconsciousandcoulditbeinpain?’)

22

materialwritten inChinese.Theperson in the roomknowsnoChineseandso theChinesecharactersaremeaninglesstoher.Asecondsetofmaterialsistransmittedtotheperson.ThissetconsistsoffurtherChineseandasetofrulesinhernativelanguage(English)thatenableidentificationofthesymbolsbasedonshape.32Thisenablesthepersonto ‘correlate’theformalsymbols inthefirstandsecondsets.AthirdsetofChinesematerialsisthentransmittedwithfurtherinstructionsinhernativelanguage.Thissetofinstructionsenableshertocorrelatesymbolsinthethirdbatchwiththefirst two batches. Significantly, the new instructions are rules that dictate whichsymbolstoreturntooutsidetheroominrelationtothefirsttwosets.ThepersonintheroomisunawarethatthefirsttwosetsofChinesearedesignatedasstories/scriptsandthethird,questions,andthatthethirdsetofinstructionsisessentiallyaprogramfacilitatinganswerstothequestions.

Searle’spointisthatinsofarastheinstructions(program)areadequatelysetoutandfollowed,thepersonisabletotransmit‘answers’thatareadequate,andsoindistinguishable from a native (literate) speaker of Chinese.However, the personknowsnoChineseandhasmerelyengaged in formalsymbolmanipulationwithoutcomprehensionofmeaning.Theprogramissyntaxbutfortheoperatorthereisnosemanticcontent.Theyhaveactedinaccordancewithaprogramalonginput-outputlines, and there is no interpretation-as-translation of the symbols. This is quitedifferent thanwhat a human doeswhen communicating. To emphasise this pointSearleintroducesanadditionalfeaturetotheexperiment.Thepersonisalsorequiredto answer a parallel set of questions in English. As a native speaker, the person’sanswerstothesequestionsarealsoadequate,andsotheappropriatenessofanswersto both the English and Chinese questions are indistinguishable, despite that theformerarecommunicative-interpretiveactswithsemanticsignificanceforthepersonandthelatterarenot.Searle,therefore,concludesthatsuccessfulsimulationisnotatest(isinsufficient)toestablishthatamindandacomputerarethesame.SuccessfulsimulationisstillafailureinstrongAIterms(thereisaconfusionofsimulationandduplication--meaningisnotduplicatedfortheoperator). As with Turing’s imitation game, Searle’s Chinese room has invited manycritiquesandresponses(e.g.Anderson,1987;Harnad,1987;Hauser,1997;PrestonandBishop,2002).Thisbeganwithmorethan25briefresponsesandSearle’srepliesthatappearwiththeoriginalessayinthejournalBehaviouralandBrainSciences.Intonetheyrangefromthehostiletothesympathetic.Moresignificantly,therepliesarefor-all-intents-and-purposesvarietiesofthestandardobjectionsSearlesetsout(followingtheformatdictatedbyTuring)aspartoftheoriginalessay:thesystems,robot,brainsimulator,otherminds,manymansionsandcombinationreplies.Notallare relevant here.What is relevant is that this sets a pattern. Searle and others’responses are unable to decisively refute the objections to the satisfaction ofinterlocutors.Thisisdespitethatthemoreconsideredrepliesconcedethatthereisacasetobeanswered.Theproblemisthatitremainspossibletoplaceaquestionmark

32Note,theBushoustructureofChinesesymbolsisconducivetoshapematchingforpartsofChinesecharactersandsomeEnglish-Chinesedictionariesusethisformattoexpeditefindingthepinyin,thoughitisnotentirelyclearhowthisrelatestotheChineseroomexperimentintheoriginalargument,whichispurelyaboutformalsymbolmatchingratherthanidentifyingmeaningbasedondecompositionofcharactersinamainlybi-syllabiclanguageofthetypeChineseis.

23

againstthetermsofSearle’scritique,andsoofferalternativetermsormerelyconsidertheargumentincomplete.

Forexample,similartoTuring’sgametheChineseroomprovokesa‘canthegame be played?’ response in the form of ‘can the experiment be constructed?’.However,therearelimitstothislineofreasoning,sincetheargumentationschemestatus of Searle’s Chinese room differs from Turing’s game. Clearly, both requireconsistency,andbothinvolveaclaimthatfollowsfromtheinitialconstruct:inTuring’scasetheclaimisthatimplications(orreasonableinferences)followregardingthinkingfromfunction,inSearle’scasethattheydonot.However,itisintrinsictoTuring’scasethataversionofthegamebeactuallyconstructible,whereasitissufficientforSearle’scasethat thethoughtexperimentbeconceivable.Turing’scasehingesonpractical(albeit future) demonstration,whereas Searle’s needonly demonstrate in thoughtthatsuchapracticaldemonstrationisinsufficientforinferencestobemaderegardingthoughtforAI.However,itisherethatdisputearises,andthetermsofthisdisputeunderpincontinuationofversionsofallthestandardobjectionstoSearle’sargument.Specifically,whatdoesSearle’sargumentassume,whatdoesitreasonablyallowonetoinfer,andwhatarethelimitsofanysubstantivecase?Disputeperpetuatedthroughthelimitsofargument:designingasuccessfulfailureandthecontextof‘insufficient’ManyrepliesarguethatSearlehastakenthepositionofapartinawholeandthenmade inferences from the part. The Chinese room argument is in this sense‘reductive’, thoughtheproblem isvariouslyreferredtoasa levelofanalysiserror,categorymistakeetc.Searleintroducesahumanoperatorthatcarriesinformation(anoperatingunitorprocessorquaprogram),butthisissimplyonecomponent.Totakethe‘pointofview’ofacomponentistomissthepossibilitythatitisnotoperativelysignificant in isolation,andsothecharacteristicsdeniedonthebasisofapartmayexistbasedonarelationalwhole.Fromthispointofview,theformalstructureoftheChineseroomargumentdescribedintermsofacomponentisdesignedbySearletobeasuccessfulfailure.Searleissettingupaconstructthatmustfailbecauseofthepositionfromwhichinferencesaremade,ratherthanhegenuinelyestablishesthatanAIcouldnotpassanappropriatelyconceivedanddescribedtest.Thus,intermsofthe systems critique, he has taken a sub-system position, which cannot actuallyaddressitstarget.Thiscritiquethenbecomespartofiterationsofotherobjections:semanticsmaybeapropertyofthesystem,analogoustoamind,and intelligence,awarenessetcmaybeconceivableaspotentialsofcomplexartificialsystems,wherethesesystemsmayemulateneuralpatterns,andifembodied(arobot),andsotactile-as-experiential in the world, could have or develop to be what Searle claims theChineseroomestablishesthatAIcannotdemonstrate(intelligence,internalsemanticsignificance,awarenessetc.).

For Searle, all these replies miss the point. The room orients on a coredifference: formal symbolmanipulation in contrast to comprehensionofmeaning.Therearedifferent terms involvedand thesearenot synonymous (understanding,meaning,intelligence,awareness,consciousness),andsomoremightbesaidabouteach, but this is irrelevant or superfluous to the initial insight of the thoughtexperiment. Expanding from a sub-system to any defined actual system, and so

24

alteringthelevelofanalysis,doesnotin-and-of-itselfchangethestatusoftheclaims.Theentityeitherhasordoesnothavethecapacitytocomprehendmeaning(andbyextensionhasothercharacteristicsassociatedwiththehuman).ForSearle,criticsdonot establish that it has the capacity. So, simulation remains merely successfulsimulation.However, this clearly does not deter critics, since they are still able toreverseSearle’spoint,partlybecauseofthelimitsofwhatcanbeclaimedfromtheChineseroomargument.

The argument only establishes that successful simulation is insufficient toestablishthatacomputerandamindarethesame,andthatanAIcanhavesignificantmind-like characteristics. It does not establish impossibility. As such, responsescontinue to develop along three mutually related lines: (1) what intelligence,understandingetcareismoreambiguous,contingent,contestable,andnuancedthanSearle allows; (2) Searle is missing something in terms of the actual equivalencebetweenhowamindandcomputeroperate(properlydescribedthe/atechnologycanbewhatSearlestatesitisnot,andsocanamind);(3)Searleisaddressingaproblemthattechnologyisprogressivelyovercoming(andsopotentialisbeingmissedbytheway Searle mis-specifies the problem in parts). From this point of view, Searle’sinterventionhasultimatelybecomepartof thecontinuingdiscourse initiallysetbyTuring -- not least because it reprises and so iterates the problem that disputeregardingatestdesignisnotitselfadecisiverefutationthatadigitalcomputercanbeprogrammedtosuccessfullyplayanimitationgame,fromwhichinferencescanthenbemade.WhilstonecanargueTuringdoesnotestablishthatitispossibletoconstructadigitalcomputerforsuchagame,ithasremainedthecasethatperhapsonecouldbeconstructed.Searle,doesmorethananyothertoestablishthatsimulationisnotsufficient for an inference to equivalent characteristics of an entity. However, hisclaimshaveremainedsubjecttodispute,sincetheprimarypoweroftheargumentisbasedoninsufficiency.

Tobeclear,SearledoesnotclaimafutureAIcouldneverhavecharacteristicsweassociatewiththehumanmind(heacknowledgesthistobeanempiricalissue).HisargumentisthatproponentsofAIhaveestablishednothingbeyondformalsymbolmanipulation,andsoclaimsincognitivesciencethatthemindislikeacomputer(andacomputerislikeamind)shouldnothavefoundationalstatus.Ifreadinthisnarrowsense,itseemscuriousthenthatthemajorityofrepliesseektoquestionthebasisofwhat is, in this context, meant only to question what seems an unthinkingpresumptionaboutthenatureofthought.TheproblemisthatSearledoesmorethanquestioninthisway(hiscontext isbroader).AsSearlenotes,thebroader issueforphilosophyofmindiswhywouldoneassertthatamindandacomputerarethesame,and why would one persist and pursue lines of reasoning that first require thisassertion.Thecontext,seemingly,isabasicfallacyofreasoningthathasthendoggedcognitivescience,whichinturnaffectstheAIproblematic:initiallyintheformofanovertbehaviourismandeventuallyinaresidualform(Searle,1980,1985,2002,2010).ForSearle,thisinturnisindicativeofanoddformofdualism,wheremindisseparablebothempirically and conceptually from thebrain (a computer andamindare thesame,andsothebrainiseitherirrelevantinitselforequivalentinitsfunctioningtoacomputer).Theimplication isthatcognitivescienceseemstobemis-specifyingtheproblemofconstitutionandcausalpowersofanorganicbrainintermsoftheproblemofmind(optingforamindislikeacomputerapproach,whichissyntacticinstructure,

25

andfailstoaccountformeaningascontentinrelationtoitsbiochemistry),whilstAIisbenefitingfromthemis-specificationthroughaunifyingfunctionalism(whatbothAIandamindareisdefinedbywhattheydo).

Clearly,forthisbroadersetofissuestomakesenseoneneedstogobeyondtheChineseroomargument.Oneneedsaphilosophyofmindargument.ForSearle,thishasinvolvedanexplicitturntoontology,andintermsofthebroaderaspectsofthedevelopmentofthehumanandoftheproblemofAI,asocialontology.Thelinkhere is not immediately obvious. However, consider that Searle’s argument isunderpinnedbyhisgeneralapproachtointentionality.Intentionalityisthecapacityfor‘aboutness’ofthemind;itscapacitytocreatementalstatesinregardoforwithreferencetostatesofaffairsintheworld(andthisismorethanjust‘Iintendtodo’,includingalsobelief,desireetc.).ForSearle,itismanifestlythecasethatinhumansthis is a consequence of the human brain as organic or biological phenomena.Awareness,consciousness,self-consciousness,understanding,meaning, intelligenceand so forth are likewise biologically caused (or as Searle more often states,constituted)andinvolvecausalpowersintheworld.

ForSearle,akeyfeatureofhumansistheirsociality.Intentionalityallowsforcollectiveintentionality,definedasthecapacityforintentionalitytobewe-directed:this‘we’remainsindividual,anddoesnotnecessarilyrequirejointthinking,butalwaysinvolvesareferencingtowhatotherscanincombinationenableordothataffectsthecapacityforwhatoneisalsodoing,throughtheassumptionthatothersarefollowingsimilar mutual points of reference. This mutuality is grounded in or becomes anorganizing feature of the socialworld built up from ‘status function declarations’,typically in the formof ‘XcountsasY inC’.Thedeclaration imposesa functiononobjectsandpeople thatarenot simplyperformablebyvirtueofphysical structure(they are products of the ascription or recognition of status -- a wall becomes apropertyboundary).Fromthisgeneralform,complexinstitutions(essentiallygroupedrules)develop,andthisisthebasisofaconstructedsociallyrealitywithinwhichonecanreferto institutionalfacts(JohnisaProfessoratBerkeley),wherethewhole isheavilydependenton linguisticrepresentationandhencemeaning.Humansarebyvirtueofbiologycapableofmeaning,anditisthroughvariousrelatedcapacitiesthattheverypossibilityofAIarisesthroughtechnologywithinsocietiesthatalsodependsonmeaning.Searle,ofcourse,mayrefertofunction(statusfunction),butthereisonlyasuperficiallexicalsimilaritybetweenhisconcernsandthoseofthefunctionalismheopposes. For example, status functions are about themeaningful pursuit of livingsociallynotthedeterminedefficacyofcompletingatask(thoughthismaybeagoaloflivingsocially).Concomitantly,toemphasisefunctionalityratherthancausalpowerorconstitutionistomis-specifybeingintermsofdoing,andthisisabasicproblemofhowboththehumanandsocietyareconceived.

ThepointtomakehereisthatthereiscontinuityandcoherencebetweenthedifferentaspectsofSearle’sgeneralargument,andthatthecombinationmakessenseofhisoppositiontofunctionalism.ForSearle,developinghispositiononphilosophyofmindandsocialontology(beginningfirstwithhisworkonspeechacts)hasbeenalife’swork,andfromhispointofviewthatworkaugmentshisspecificclaimssetoutintheChineseroomargument.Forcritics,however,itindicatesthathisclaimsintheChineseroomargumentdonotstand-alone.TheyrequirecommitmentsthatarenotpartoftheChineseroomargument.Searleessentiallycreatesthechallenge:prove

26

thatanAIunderstandswhatitisdoing.ThedefaultintrinsictoSearle’spositionisthatwejustknowthatitdoesnotunderstand(basedoncomprehensionofmeaning).Weknowthisbecausewedesigned/programmedtheentity,andweknowthisisdifferentthan what we know about ourselves. The case of AI seems to be, therefore, aneither/orissuethathingesonevidenceinawaythatisdifferentthanhowweattributein other cases. Those other cases are shared aspects of the human (degrees ofawarenessetc)onemightattributetoanimals,whatweattributetootherhumans(since it is reasonable toassumetheyare likeus),andwhatwemightattribute toaliens(sincetheymaybelikeus).Eachoftheseisnotdesignedbyusandsodoesnotrequiresome ‘forgetting’,which theAIcase, fromSearle’spointofview,seemstorequire.

However,thoughhingingonevidenceSearle’scaseisgroundedontheory;thatis,Searle’scumulativelydevelopedbroaderposition.Moreover,plausibleasthefocusonevidenceseems,itraisestheproblemofwhatexactlywouldsatisfySearleintheformofproof.Itcannotbemerefunctionandsocannotbeablackboxbehaviouristproof,butthistooseemstodisallowasmuchasitdisprovesafunctionalistresponse.Sincethetestwillbeofsomethingartificial, tocritics,at theextreme,anddespitecaveats, Searle seems to have (at least inadvertently) disallowed any practicaldemonstration because he seems tacitly committed to the claim that an artificialentityisbydefinitionsyntheticandsocanonlydemonstratesimulation.Putanotherway, thoughSearle claims thatAI confuses simulationwithduplication (simulatingunderstandingisnotactualunderstanding,soisnotduplicatingit)thereverseisthathe resists the possibility that anything other than a brain can duplicate, and thusrealise given states (and yet a synthetic heart is not a simulation-only). Searleinadvertentlyover-writesfallibilityandfuturecontingencyviacurrent‘forgetting’.

So, for critics, Searle has donemore than he set out to do. This is a basicvulnerabilitythatcriticshavedevelopedtodifferentdegreesandwithmoreor lesssympathyfortheoriginalargument:hispositiontakesaplausibleintuitionregardingdifference, but requires its own assertions regarding what might be the basis ofunderstanding; it thus involves a tacit certainty, which critics can draw furtherinferences from regarding prejudice or ‘magic in the meat’ ‘chauvinism’ -- thesignificantstatusaccordedtothebrainastheseatofgivencharacteristicsconfusessignificancewithspecialoruniqueorspiritualormysterious…33OnecanreadSearleashingingthedifferenceonthenatureofresponse:ahumanrespondstomeaningbecauseofmeaning(andsomeaningfully),anAIrespondstomeaningbecauseofform(andsomechanistically).Thetwoarecausallydifferent.Meaningisnotthereason-as-causefortheresponseoftheAI(sothereisnosemanticjustsyntax).However,forcritics,ifthereiscausationthencausemotivatescontent,so,subjecttoredescription,onecanbegintoclaimthatmeaningisproduced(semanticscanbederivedfromthecausal process of which syntax is a part). The question then becomes, how it isproduced andwhat produces it, re-opening up lines of inquiry and argument thattradeoncomplexsysteminterconnectivityforbothmindandAI,contestingthewaysemantic,intelligenceetc.aredefinedandused,speculatingonthebasisofpossiblewaystoduplicate(ratherthanmerelysimulate),reconstructinganargumentthata

33Magicinthemeat,derivesfromthesciencefictionwriterTerryBissonwhoresituatesthecontextasaconversationbetweentwoentitieswhohavenevercomeacrossorganicmindsbefore.

27

robotcanduplicateaboutnessthroughsensorycapacity(asexperience)andsoderive(afunctionalistfounded)meaningcapacityandsoforth.

Clearly,forSearlethisagainallmissesthepoint.Asfarasweknowthereisnoequivalent mental state in an AI as-is. Subtle violence seems to be perpetratedregardingourcommonunderstandingofunderstanding(thoughthistoobecomesacounter-argument--sincecommonsenseplaysnoroleinmuchofscienceandcan,asconvention,impedeproperunderstandingorinvestigation).However,thepointhereisnotwhetheroneendorsesorconcurswitheachargument,butratherthattheformof Searle’s case continues to provoke responses along these lines. Insufficiency,assumptions, consequences for argumentation, and then dependence on contextmaterial,provide thegrounds for reasoned ifnotnecessarilyalways reasonableorsympatheticresponses.

AndsodebatecontinuestoevolveinandaroundtheChineseroomargumentbasedonmultiplelinesofdevelopmentthatuseitasapointofdeparture,tradeonthelimitsofwhatitcanclaim,andsituatethistovariousconcernsofthecritic.34Theargumenthasunifiedcriticsasapointofconvergencefordisparateargumentratherthanagreement:eliminativematerialistsworkinginneuroscienceandphilosophyofmind (e.g. Churchland and Churchland, 1990), professional philosophers able todeconstruct thecaseandparse itsmany implicationsastheypertaintosemantics,functionalism, alternative views of intentionality etc (e.g. Boden 1988; Chalmers,1992; Pinker 1998; Fodor, 1992; Dennett, 2013), and futurists with agendas thatentangle AI and TH (Kurtzweil, 2000).35 As a point of convergence, then, Searle’sChinese room has not created consensus, but rather a focal point around whichdisagreementcoalesces.

Again,thisisostensiblyoddifonetakestheChineseroomargumentatfacevalue:athoughtexperimentthatplacesaquestionmarkagainstequivalenceofAIandthat reminds cognitive science that the brain is significant for the

34ItisironicperhapsthatIampointingthisoutasanexerciseindoingthesame.35Pinker,forexample,standsontheoppositesidetoSearleascommittedtoavarietyoffunctionalistinformation processing (‘computation’ and ‘program’)mind-AI approach: ‘Themind is a system oforgansofcomputation,designedbynaturalselectiontosolvethekindsofproblemsourancestorsfacedintheirforagingwayoflife,inparticular,understandingandoutmanoeuvringobjects,animals,plants,andotherpeople.Thesummarycanbeunpackedintoseveralclaims.Themindiswhatthebraindoes;specifically, the brain processes information, and thinking is a kind of computation. The mind isorganizedintomodulesormentalorgans,eachwithaspecializeddesignthatmakesitanexpertinonearenaof interactionwiththeworld…Onthisview,psychologyisengineeringinreverse. Inforward-engineering,onedesignsamachinetodosomething;inreverse-engineering,onefiguresoutwhatamachinewasdesignedtodo…Thecomputational theoryofmind is indispensable inaddressingthequestionswe long to answer… the contentofbrain activity lies in thepatternsof connections andpatternsofactivityamongtheneurons.Minutedifferencesinthedetailsoftheconnectionsmaycausesimilar-lookingbrainpatchestoimplementverydifferentprograms.Onlywhentheprogramisrundoesthe coherence become evident… The computational theory of mind is not the same thing as thedespised‘computermetaphor.’Theclaimisnotthatthebrainislikecommerciallyavailablecomputers.Rather, theclaim is thatbrainsandcomputersembody intelligence for someof thesamereasons.’(Pinker, 1998:pp. 21, 25-6). Pinkerpositions Searle’s Chinese roomas an argument that tradesoncommonsensebutnotscience,makesambiguousthelinktobiochemistryofthebrainformind(inawaythatinformationprocessingdoesnot),whilstclaimingthatthebiochemistryofthebrainiscoretoconsciousness,intentionality,awarenessetc.whichheneverproperlydefinesorexplains(Pinker,1997pp:93-96).Note,thoughSearle’sChineseroomhasbecomeakeyissueandsopointofdeparture,itisnotcentraltoallworksthatexploreit.Fodor(1992)isequallyconcernedwithChurchlandonmeaning.

28

experienced/observedcharacteristicsofthehuman(ithascausalpowersbyvirtueofits constitution). However, it is less oddwhen one considers that the strength ofSearle’sargumentisalsoitsweakness:ithasatightargumentforacleardistinction,seems to demand an empirical response, extends to expose the assumptions thathavepreventedananswerhedeemsplausible(themindisacomputer,thoughthistooisambiguousifonemeansonlycomputational),andissituatedtoanalternativetheorizationthatgroundsananswer(hisontologyofhumanintentionalityetc.).Theweakness is that the combination provides formultiple lines of reasonable reply.Concomitantly,Turing’sgameandSearle’sroomarebyfarthemostcitedworksonAI(partlybecauseuseexceedsafocusonAIonly).InApril2017,aGooglescholarsearchon ‘Mind, Brains and Programs’ returned over 59,000 results, and ‘Computing,MachineryandIntelligence’morethan176,000.

Ironically, Searle’s attempt to simplify and focus debate in terms of a coredifference, which implicitly contests the problem set in motion by Turing of aseductivetendencytoproliferatedebatebasedontheexplorationofminutiae,hasproducedamaelstromofminutiae.Thishashadsomepositiveconsequences,sinceithasfosteredanAIsfocusonentities,pushingfunctionalismtobecomemorecarefullyconsideredregardinghowandwhatproducesfunction.Interalia,ithascontributedtoformalrepudiationsofbehaviourism(and,lessconstructively,foratendencyforittobecomeatermofabuseindiscoursewhosereversalisapejorativereferenceto‘mentalism’).However,thepositioningandinfluenceoftheargumenthashadfurtherconsequences.AIwandAIs:resolutelyunresolvedlinesofinquiryWebeganthisessaybydifferentiatingAIwandAIsbasedonfocus.AIwinvolvesafocusonfunctionthattendstosetasideentitystatus.AIsinvolvesafocusonentitiesthatdecomposesintotwosubcategories:afocusontheequivalencebetweenhumanandAI,withafocusonsignificanceforthehuman,primarilywithinphilosophyofmind,andamirroringfocusonthestatus,characteristicsandpotentialofAIentities.Searle’sChineseroommaybeaboutAIbutitisintendedtodemonstratesomethingaboutthehuman.ItisinthissensefirstsubcategoryAIs.However,aswehavesetout,Searle’sargument has not created agreement. It has become the point of departure forreasonabledisagreement.Ifoneworksbackwardsthroughallthematerialwehaveconsidered so far in this paper it should be clear that disagreement is not mereformlessness.Thegroundsofdisagreementareconcernedwithfunction.Moreover,insofarasresponsestoTuringandtoSearlehavecontestedfunctionality,muchofthedebatehasconcernedwaystopreservefunctionalistwaysofthinking(noironyintended) about the problem of AI and of the human (mind). Much of thedevelopmentofargumenthasbeenneo-functionalist,andthatwhichhasnotbeenhasbeenaboutthelimitsofcritique(sufficiencyofargumentviagames,tests,thoughtexperiments)ofwhatisalsofunctionalistbyfocus.DespiteSearle’sintervention,therehasbeennoimmanentcritiquethathasdecisivelyshiftedthetermsofdebate.Thoughitisnotfalsetosaytherehasbeenanontologicalturninphilosophyofscienceandsocialscienceinsomewaysandtosomedegree,onecannotreasonablyclaimthatexplicitontologyorsocialontologyarestandardpointsofdeparturefortheproblem

29

of AI.On the contrary, that Searle’s position is situated to an ontology and socialontologyisusedagainsthisChineseroomargument. Thereis,ofcourse,nothingintrinsicallyillegitimateincritiquethatquestionstherelationsinasituatedargument.However,thereisadangerthatoneconflatescritique of the specifics of the way an argument is situated with a basic generalproblemthatanargumentissituatedatall.ThisisparticularlyimportantinlocatingSearle’s work. All argument-as-claim ultimately involves an ontology, and so thealternativetoSearle’sargumentshouldnotbearefusalofontologybutanexplicitlyaddressedontologicalargument.ReadappropriatelySearle’sChineseroomargumentisnotmerelyaugmentedbyhisontology,butspecificallyrootedinontologicalissues.It begs questions of functionalism regarding the constitution of entities and theircausalpowers.Assuch,muchoftheperpetuationoffunctionalisminspiteofSearle’sargument is a refusal to engage at the level of ontology, whilst pursuing tacitontologicalissues.Thishashadobservableconsequences.36

FunctionhasnormalisedastheoverwhelmingconsiderationinandforAI.Ofcourse,noting thisphrasingcanseemsuperficially ridiculous.HowcouldAInotbeconcerned with function? However, the nature of concern flows from what isconsideredand inwhatways. Focusdoesnotnecessarily create clarity,but ratherpotentiallyadversenormativity.TheveryexistenceofAIwisadefermentofthestatusofentitiesthatpresupposestheineluctabilityofAIastechnology.Thereisan‘AI iscoming andwemust cope’ that decentres the seat of decisionmaking, as thoughhuman choiceswere not dictatingwhether andwhat kinds of AI develop and areadopted. ‘Cope’becomes ‘let’s getonwith it’ as though thebasisof functionwassettled.Thishasa‘meanwhile’orinteraliacontext.AIsmaynothavesettledanything,butthebasisofnon-settlementinvitesafocusonfunction,andtypicallypresumesafunctionalist frameof reference.So,dominantaspectsof theconceptofAIsareatleast associated with the general pervasiveness of AIw. As such, the bifurcationbetweenthesetwothatIpreviouslyreferredtoisnotwithoutmutuality,andthisisimportant, since it is because of mutuality that some issues or foci or ways ofconceivingaremarginalised,inadequatelydevelopedorbecomeinterstitial.

As already noted, if functionalism dominates then the problem of beingbecomesaproblematicofdoing,whichinturncanbecomeaproblemofefficiency.Thehuman‘doing’,astasks,becomesataskmastermasteringoursenseofwhatthehuman is. This is sociological rather than purely philosophical (involving thepositioningandrelativepowerof ideas,ratherthanjustthesubstantivecontentofthoseideas).Afocusonefficiency,forexample,mayabsorbthesocialcontextthatdominatesandexpresses (represses) intrinsicaspectsof thehuman.Efficiency isatechnical term but also a shaped value; it is referenced to the socio-economic

36ThoughSearleisnotconcernedwiththeproblemofAIw,aswehavesetitout,hisownAIsargumentrefutingafunctionalist(behaviourist)varietyofstrongAI,withreferencetophilosophyofmind,doesprovideanargumentforhowasplit infocusmightarisebecausebothsidesofthesplitcanshareaconcernwithfunction.Thisthencanoperateideologicallyandsobeinfluentialfarbeyondanyissueof‘onethingdirectlycausesanother’.Itisnotdifficulttounderstandhowaprogram-centredviewofmindincognitivesciencecanoperatetodecentreaconcernwithwhatisspecificallyhuman.Itisnotdifficulttounderstandhowthiswayofthinkingaboutthinking(andothercharacteristics)canaffectprocessesof change. Clearly, one can recognize basic entanglements herewith TH, specific and general. ThevariousexchangesbetweenSearleandtheTHfuturistKurzweilarethemostobviousmanifestationofthis.

30

conditioningofthesysteminwhichvaluesinhere(principlesofcapitalismthataffectwhathumansdointhenameofefficiency).

ItmaywellbethecasethatethicsareaprofoundlyimportantpartofAIs,buttheyarenotcentral,controllingormostimportantasasourceofconsequenceintheworldbecauseofAIs.Furthermore,ifthedominantAIwdiscoursestartsfrom‘cope’,thenthepowerofethicaldiscoursetoshapethesocialconsequencesofAI(andalsothenTH)issubverted.Itisalsoobfuscatedinsofarasfunctionalismmayinformtheconceptsofentitiestowhichethicalanalysisisapplied.Moreover,considerthewaylanguageusehaschangedasandforAI.Despitecritiqueoftheextensionofmeaningregarding thinking, learning etc, meaning has still been extended, and so hasultimatelybeenappropriated.TuringclaimsandSearleconcedesthatafutureAImaybethinking,intelligentandsoforth,subjecttohowtheseareconceivedandsubjecttosomeempiricaltestordemonstration.Inthemeantime,thelanguageofwhathasnotyetbeenincontestablyachievedhasbeencolonisedandsonormalised.37Afuturethatmayneverbeisalreadyhereintermsofthelanguageweuseregardingwhatartificialentitiesdobutwhichultimatelyexpressesan is (abe): ‘intelligence’ isanacceptedeverydayassociativetermquaAI,accordingtoeverydayreferentialcommunicativeactsbetweenhumansAI do ‘learn’... Socialisation through languageuse is alreadyoccurringaroundfunction.Theworldweliveinisthusdrawingusintoafutureweareconstructing as ineluctable via language that contributes to a ‘we must cope’mentality.Whatfollowsfromfunction:InterstitialproblemsandontologyascritiqueThefirstpointtomakehereisthatontologyitselfhasbecomeinterstitial.ItisnotthetypicalpointofdepartureforAIsand,byvirtueofitsfocus,isnotaprimaryconcernforAIw.Notingthisisbynomeanstodenigratethesophisticationinitsowntermsofwork that has been done. Nor is it to invite unreasonable expectations of whatontology can achieve. Rather it is to note that all arguments-as-claims involve anontology,andclarityhereaffordsclaritytootherandsubsequentissues.Thisisnottosuggestphilosophyisunclear,oratleastsetsouttoclarify,sinceanalyticalphilosophyinparticularisconcernedwithprecisionandclarity.However,suchclarityisepistemicandneednotbeontologicallyposedingeneralorrealisticallyreferencedinparticular.Arguably,thefamiliarfunctionalismarticulatedviatheoryisimposedonrealityratherthanderivedfromit.

Moreover,clarityissituatedtofocus.This,essentially,iswhatthispaperhasargued by exploring the development and consequences of focus. Much of theseductionofTuring’simitationgameputsasideontology,invitingdevelopmentofthegame,whilstseemingtotradeonafunctionalistsetofclaimsthatthenbecomethepointofdisputeintermsofwhatthegamecandemonstrate(viaitssubstitutionofquestions).Ultimately,functionalismandtheequivalencebetweenAIandthehumanareclaimsaboutbeing.Assuch,Searle’sresponseisanontologicallymotivatedreply.TheChineseroomthoughtexperimentconteststheequivalenceofAIandhumaninorder to highlight that the significant characteristics of the human in relation to

37Thereisagreatdeallesstalkofartificialstupidity.

31

meaning cannot be demonstrated for an AI, and yet are a consequence of theconstitutionoftheorganicbrainthatcreatescausalpowerintermsofmind,thoughSearle is wary of this language of distinction between brain and mind due to itshistoricallegacyandconnotations. Of course, many ontologically oriented elaborations are possible, and anontology isnotassertionbut reasonedargument subject toevidence, critiqueandsubsequentcriticallyposeddevelopment.38So,asalreadynoted,Searle’sargumentisnotimmunefromcritique,butcritiqueneednotbetheabandonmentofontology.Forexample,asTonyLawsonnotes,Searleresiststhelanguageofemergence.Accordingto Lawson, the subject matters of Searle’s ontology and social ontology seem torequireaconceptofemergencetoexpresspropertiesforandinentitiesandsystems(Lawson, 2016). In his reply to Lawson Searle rejects this claim, in so far aswhatemergence actually is, is not made clear, and what emergence explains thatconstitutiondoesnotislikewiseunclear(Searle,2016).39However,Searle’sargumentturnsonthenamingofanentity,whichhasanorganization,andthenthedistinctionbetweenwhatfollowsfromtheorganizationintermsofproperties,andwhatisknownandaccountedforintermsofthoseproperties.Thisdistinctionisrecognizablefromearlytwentiethcenturyemergentistphilosophy,butcreatesnewproblemsintermsof the distinctions, if the point of emergence is first-and-foremost to express thepropertiesthatwouldnotoccurwithouttheorganization,andsothepropertiesareirreducible to the properties of the decomposed parts -– structural integrity is acharacteristicofabuilding,consciousnessofthebrain,andtrustofacommunity,insofarasappropriatelyconstitutedandactive/activated.Allrequiretheorganization-as-constitution to bemanifest or to be possible, but not all involve an additionalpropertythatisunexplainedthatwedeemadditionalsinceitissofarunexplained(asitdoes in thecaseof consciousness).40 It is thisadditionalunexplained thatSearle

38So,ontologymayprovideclaritybutitdoesnotguaranteethatmattersaresettled…Itmerelycreatesanadditionalandappropriatecontextofargumentbyrecoveringthetraditionaldomainofmetaphysicsthatmuchofmodernphilosophyeschews.39ForanotherconstructivecritiqueofSearleseeElder-Vass(2012),andforanassessmentofElder-VassseeMorgan(2014)40Thepointatissueseemstobethatadditionalimpliescannotbereducedto,andthemeasureofthisiscannotbeexplainedintermsof;however,theonticcharacteristicoftheorganizationisthecreationof thegroundsofapropertynotwhether in fact theproperty is fullycomprehendedregarding theorganization-as-constitution. If this were so then the very moment we had a full explanation ofconsciousness itwould be fully accounted for in relation to constitution and sowould cease to beemergentbydefinition,andyetremainapropertythatwouldnotexistwithoutconstitution.Thehingethustacitlyseemstobeadifferenceofmeaningandemphasisregardingtherelationofknownandadditionalwhilsttheargumentitselfisexpressedas‘fullyaccountedforbyxandthereforereducibleto x’, where to x becomes confused as a which to x? the parts or the constitution, and then theconstitutionasanxoranewmeta-encompassedconstitutionofthisoriginalx–brainisnotsufficientthereforebrainandnewxisconstitutionismind…etc?Inallcasesconstitutionisthenamedentitythatisthenasourceoftheproperty,butthepropertyisalwaysdescribableasarisingfromtheconstitution.Just becauseone knowsanddesigns ahouse tohave structural integritydoesnotmean structuralintegrity does not ‘emerge’ from the constitution-as-combination (it is a product of…). One can ofcourseargueaboutwhetheronewantstotermthisemergenceorsimplystatethattheconstitutionproduces(wherethelatterstilldoesnotdifferentiateconsciousnessandstructuralintegrityexceptinso far as currently known – an epistemic rather than ontic distinction), and one can argue aboutwhetheronewantstorefertoconsciousnessandartifactsasofthesamekindorcategoryofthismoregeneral category. It is these that Searle seems toactuallybe contestingand Lawsondoesnothelp

32

associates with emergence. However, if social ontology requires a concept ofemergence, it is not quite the one that Searle is rejecting. It is rather emergencebecauseofconstitution-–areferencetoarangeofpropertiesofpropertiesinrelationtoorganization.Whatseemstobeatstakeinitiallyislanguageuseforclarity,butthisultimatelycreatesdifferentwaysofdevelopingsimilarconcerns.

Lawson’ssocialontologyoverlapswithbutisdifferentfromSearle’sapproach.Itdistinguishescommunities(organizedsystemsofrelationsexpressedthroughrightsand obligations for socially positioned humans) and artefacts (non-human thoughsocially constituted and specified objects) as emergent social entities, and makesparticularreferenceto languageasan importantemergentsystem.Onemightalsonote that since for Lawson artefacts are not communities, AI introduces a furtherpotential issue for Lawson’s social ontology, by virtue of what an AI may be. InLawson’s terms an AI could be an artefact that participates in language usingcommunities,thoughwhetheronewoulddescribeAIasanartefactwouldthenbeatissue.

Lawson’s isaconstructivecritiqueofSearle’s socialontology.41ButSearle’suseofAI isalsoinstructivehereregardingtheconsequencesofthesub-categoricalsplitinAIs.InChapterSixofMakingtheSocialWorldSearlemakesthecasethatsocialrealitycreatesdesire-independent reasons foracting.Hedoessobyextending theargumentsetoutintheChineseroomthoughtexperiment.Specifically,hewantstocontestthatreasonstoactonlyariseifonealsohasadesiretosoact(Hume’s‘reasonisaslaveofpassion’).However,forSearle,deonticpowerstranscenddesireorpurelypersonalmotivation. Social reality creates cross-referencedduties, obligations andrequirements. In simplest form, promise keeping demonstrably creates desire-independent reasons for action. In general,weuse institutionswithoutdestroyingthem (extending sometimes to recognizing that reproducing the institutions isnecessaryorimportanttodo--thedutytovote).Weroutinelysuppressinclinationsandmodifybehaviourinwaysthatoverride‘Idon’tfeellikeitrightnow’or‘I’dratherjustdox’(wegotowork,werespectpropertyrightsetc).Insofarasthisisso,deonticpowersareafeatureofinstitutions,andthecombinationaffectshowandaboutwhatreasoningoccursand,therefore,whatwedo.

himself here by emphasizing novelty, which seems to imply cannot be or is not known (which isdifferentbutrelatedtocannotbepredicted–thequalifiersmatterasdomattersaposteriori).Actingbackuponcreatesadisputeintermsoforganizationasconstitution(butevenhereonecanarguethatahouseorartifactbyvirtueof constitutionaffects thedurabilityofandenvironmentwithinof theartifact–structuralintegrityisalsoaconstitutedcausaleffectfordecayifredescribed,whilstmaterialcauseremainsadifferentissue…41Forexample,Lawsonstates: ‘Insteadofviewing individualsasmaterially/practicallypositionedascomponentsofatotality,however,Searle,seeminglyproposesamorementalisticorrepresentationalapproach. According to it individuals rather are merely ‘counted’ as in effect being appropriatelypositioned,withassociatedpositionalpowersorfunctions.Isay‘ineffect’becausepositions,positionalpowers/functions,andpositioningarenotSearle’slanguage.’(Lawson,2016:p.370).ForLawson,thedevelopment of language presupposes practices; social objects do not require contradiction in thesenseanobjectbecomesmaterially two things, rather that for it tobe social involvesaprocessofemergencethatcreatescommunitieswithinwhichorganisation,socialpositioninganduseallowsfortheobjecttobesocial(oneobjectmaybemanythings,butitisneithercontradictionnormeaninglesstorefertothemassocialobjects,anditisinsufficienttoredescribethewholeasanobjectwithastatusfunction - a computermay be a component in a variety of systemswithout beingmore than onecomputeroracomputerandnotacomputer).

33

Crucially,thedeonticpowersofsocialrealityonlymakesenseifoneassumes‘agap’.Thatis,thepossibilityofnotfulfillinganobligationorrecognizingaright.Thisisfreewill,thoughthisisnotaphrasingSearleiscomfortablewith.Rather,hechoosesto emphasise that the constitution of social reality presupposes the possibility ofrefusal,andsoofchoice.Forexample,anobligationisnotanobligationifthereisnopossibilityofdeciding to fail toexecute. Itbecomesmerelystimulus-response,andthiswouldbeunrecognizableashumansocialreality(andthesameappliestopromisekeepingandothervariations).Searleclarifiesthispointbycontrastinghumanactivitywithaprogrammedrobot.TherobotinSearle’sconstructlacksconsciousness-- itsresponseisa‘determinedmechanicalemission’(Searle201:p.136).However,deonticpowers require consciousness and reflexive capacity in order for recognition ofinstitutionstoaffectchoice.‘Imaynotfeellikedoingxbutwilldoitanyway…’,isquitedifferentthandetermination-as-compulsionintheformexpressedbythebehaviourofnon-consciousartificialentities.

The point here is that Searle’s main social ontology use of AI remainscontrastivewith the human, following the insight developed in the Chinese roomthoughtexperimentregardingsymbolmanipulation.Searlerelocateshiscritiqueoffunctionalism, residual behaviourism and cognitive science in order to support hisargumentfortheconstructionofsocialreality:

The notion of a deontic power makes no sense unless you presupposeconsciousness and the gap. Once you regard the creatures as like thecomputationalmodelscommonincognitivescience,then,itseemstome,youcannot have institutional reality in our sense. You might program themachinery to resemble some of the forms of institutional reality, but thesubstancewouldberemoved.(Searle,2010:p.137)

Ibynomeanswishtosuggesttherobotargumentusedhereisill-founded.RatherIwanttosuggestthatitillustratesontologycanbepursuedinavarietyofways,oncesituatedtoorusingtheproblemofAI.Otherelaborationsarepossible.Consider,thatSearle’smainfocusonAIistoemphasisethatthebrainmattersforthecapacitiesofmind we are familiar with. As such, his focus is what AI does not demonstrateregardingwhatmindis.Equally,however,onemightaskwhatwouldbethecausalcapacitiesofaconceivableAImind,sinceitsurelyfollowstheyneednotbethesameas those of a human, if the constitution of the entity is different. This is secondsubcategoryAIs.Itiscontrastiveinamirroringsense.AnAIentitymayhaveequivalentcharacteristics as categorisations to the human (intelligence, understanding,consciousness, self-consciousness… intentionality), but it does not follow theywillhaveallorsimilarcharacteristicswithinthosecategorisationsofcharacteristics.Aswillbecomeclear,thiscanhaveadditionalimportantconsequences.ConstitutionandsecondsubcategoryAIsOne can imagine that differences create basic Nagel (1979) problems ofphenomenologicaldivides.42Considerthathumanmemoryisnoteideticfunctionality42Phenomenologyconcernstheworldasweareratherthanasitis;thefocusisexperiential.Thisisadifferentemphasisto,butnotadenialof,ontology-as-realism.Phenomenologymayargueforbutdoes

34

dictated by processing power and subject to cumulative 'error'. It is the drawingtogetheroffragmentsaccordingtonarrativesandpurposes(projects).Itisinherentlycreative and transitional; the very act of recalling repurposes and overlays in andthroughtime.Weareusedtoframingaproblemofmemoryasoneofunreliabilityofmemoryandwitnessing,butthisisareductivesenseofwhatmemoryis,referencedtoacapacitythevastmajoritydonothave(perfectrecall).Ittellsuslittleabouttheactual significance of memory. Memory is active, and in this sense cannot be'dispassionate' (see McGilchrist, 2009).43 This is different than an objectivity-subjectivity dichotomyor a reality and appearance problematic (see Collier, 2003;Rescher,2011).Activememory isaconstituent inthetemporalityofhumanbeing;ourbeingdevelopsasweexistthroughtime,andaswereconceiveourpastanddirectittothefuture.

Contrast this with AI as archetypally conceived.What kind of being is theproductofeideticfunctionality?Whatislearningandexperiencetoanentityoferrorcorrection?Moreover,considerthedifferenceinthegroundsofbeingthroughtime.Human temporality is finite; not only does it have an expectation of an end, it isexperiencedasanunsteadyprocessofdegenerationandbodilychangeinourselvesandobservedinothers.LeonardCohenexpressesthisas:

Everybodyhasexperienced thedefeatof their lives.Nobodyhasa life thatworkedoutthewaytheywanteditto.Weallbeginastheheroofourowndramasincentrestageandinevitablylifemovesusoutofcentrestage,defeatsthehero,overturnstheplotandthestrategy,andwe’releftonthesidelineswonderingwhywenolongerhaveapart-orwantapart-inthewholedamnthing.Everybody’sexperiencedthis,andwhenit ispresentedtoussweetly,

notentail a radicaldisjuncturebetween reality andappearance.Nagel’s famousbatargumentalsocommentsonTuring:‘Consciousnessiswhatmakesthemindbodyproblemreallyintractable…[andreductionist] discussions of the problem give it little attention or get itwrong.’ (1979: p. 165). Hedifferentiates this from the Turing machine-IBM problem which are ‘successful reduction’ but areunlikelytoshedlightonthemindbodyproblemsince‘wehaveatpresentnoconceptionofwhatanexplanation of the physical nature of a mental phenomenon would be.’ (1979: p. 166) ‘Themostimportantandcharacteristicfeatureofconsciousmentalphenomenaisverypoorlyunderstood.Mostreductionisttheoriesdonoteventrytoexplainit…thefactthatanorganismhasconsciousexperienceatallmeans,basicallythatthereissomethingitisliketobethatorganism…fundamentallyanorganismhas conscious mental states if and only if there is something that it is like to be that organism -somethingitislikefortheorganism.Wemaycallthisthesubjectivecharacterofexperience.Itisnotcapturedbyanyofthefamiliar,recentlydevisedreductiveanalysesofthemental,forallofthemarelogically compatible with its absence. It is not analyzable in terms of any explanatory system offunctional states, or intentional states, since these could be ascribed to robots or automata thatbehavedlikepeoplethoughtheyexperiencednothing.’(1979:p.166)AccordingtoNagel,ifphysicalismasreductiontomaterialstatesistobedefendedthenphenomenologicalfeaturesmustthemselvesbegiven a physical account, but this seems impossible in so far as ‘every subjective phenomenon isessentially connected with a single point of view.’ (1979: p. 167) Nagel’s main point is thatpsychophysicalreductionisamovetowardsgreaterobjectivitybyremovingspecies-specificpointsofviewtowardtheobjectofinvestigation,intermsofgeneraleffectsandpropertiesthatarenotsimplya matter of human senses. However, experience cannot follow this pattern, since moving fromappearancetorealitymakesnosenseasawaytoconceiveexperience(1979:p.174).43Wealsoexperiencemoment-to-momentincoherencies,disjunctures,fragments,flashesofimagesetcbutwedonotexperiencetheseaserrorwesimplyacceptthemaspart-and-parcelofbeingandfromincongruityotherexpressivepotentialsarebuilt:humour,irony,inspirationandsoforth.

35

thefeelingmovesfromhearttoheartandwefeel less isolatedandwefeelpartofthegreathumanchain,whichisreallyinvolvedwiththerecognitionofdefeat.(Ellen,2016)

Brainstatesmaybebio-chemical(involvingneurons,oxytocinetc),butwhatwouldsympathy(recognitionofanother’s-typically-adversesituation)andempathy(theexperienced feeling for another’s - typically- adverse situation) be for an entitywithout the biochemistry, andwithout first person expectation and experience offinitude,degenerationandsuffering?Wherewouldsentimentalityandcompassioncome from and howwould these be exhibited? How and inwhatwayswould anequivalentAIsocializeandbesocializedvis-à-visfellowfeeling?Onecould,ofcourse,seektocodeordesignforsyntheticfeeling(butagainwhatcanthismean?),andinteraliadesignanAIwithanervoussystem,andsoasystemforpleasureandpain.Onemightconstruct thisasacomponentof sensuality,butwhatwouldbe thegeneralquality of emotion in relation to sensuality for such an AI? In a human senses,sensuality,experienceandemotionarenotidentities,andhowonebecomesanotherislittleunderstood(evenifitcanincreasinglybemechanisticallycorrelated).So,whatkindofemotivestatesifanywouldanAIhaveorcouldanAIbegiventhecapacitytohave?

Moreover,considerwhatmaybebasictodifferenceinAI.IamawarethatIhavebrainstates,butawarenessandself-consciousnesshaveonlylimitedaccesstoandcontroloverbrainstates(Icancalmmyselfbutcannotnegatestresscreation--thoughsomeTHadvocatesaspiretothis).ButAIconsciousnessseemstoinvolvecode-awarenessofadifferentordertobrainstateawareness.AnAImaypossesseideticfunctionality,butitmayalsopossesthecapacityto(re)writecode.Asabeingitwouldthusbepotentiallyself-alteringandtheconsequencesofthisforbeingintheworldaredifficulttoconceive.Eideticfunctionalityandfundamentalmutabilityofselfseemsa profound clash, creating a conflicted constitution of self (an inner life andintrospection that confrontspreservationandcontinuity challenges tobeing). Self-cultivationmaysharealanguageframebutnotarealmeaningsensebetweenhumanandAI.44AImaybeitsownproject inquiteadifferentsensethanisthecaseforahuman(unless,ofcourse,oneaspirestoTH).

Consideralsowhatkindofexperienceoftheworldwouldthisself-cultivatingbeingbedirectedthrough?Self-consciousnessandembodiedfirst-personperspectiveare basic to a developed average adult human. An AI could have dissipatedconsciousnessandmultiplepointsofsimultaneousperspectiveandexperience;evenifitwas(partlyorsomeofthetime)embodied.45AndinwhatsensecouldanAIdesire?ItmaybethecaseasSearlesuggeststhatwehavedesire-independentreasonsforacting,buthumanswantthingsandthisissubtlydifferentthanhavinggoals.Societywould be different without desire, since many aspects of how we engage with

44Interalia,AImaybeanefficientcausebutalsomaybeitsownmaterialcause;categorisationofcausewithinatypologymaybecomeblurred.45 It could also have quite different quasi-neurological and hence sensory relations to the world,followingpatternsfoundintheanimalkingdom,thoughnotnecessarilythese.Somespeciesofoctopus(ascephalopods)haveapproximately500millionneurons,butmoreoftheseare intheirarmsthan‘brain’andexperimentindicatessomespecieshavehighcapacityproblemsolvingskillscombinedwitharadicallydifferentsensoryexperienceoftheworld(Godfrey-Smith,2017)

36

organizationsdependasmuchonourdesiresastheydoonourcapacitytooverridethem(theybecometherootofaspirationandsoaredelayedoralteredratherthanrepressed).Marketingwouldmakenosensewithoutdesire,andyetmarketingisacentralcomponent inrealsocio-economicsystems.Onemightgofurtherandnotethataspectsofthe‘gap’asstatedbySearleimplyalsotheintegrationorperhapsmoreaccuratelyinteractionoftheindividualandaspectsofsocietythroughdeontologythatdependontherelationbetweenlifeprojectsofindividualsandtheaspectsofsociety(howevernamed)thatexistasorganizedcomponents.Thisisnottotrivializehumanprojectsas simplyproductsofdesire (by suggesting they lack reflexivity,durabilityetc),butrathertosuggestdesiremaybesomethingdifferentorabsentforanAI.

Desireingeneralisexpressiveinawaythathavingagoalonlyneednotbe.Itisabodily relationof thought (seeDamasio,1994).ThebodyofanAIasaseatofthought could be quite different here and one might extend this thought aboutthoughtandthebodytomanyotherdifferences.Thoughthepurposeofsleepremainsamatterofdispute,recentresearchindicatesoneoftheprocessesthatoccursduringsleepisb-amyloidclearancebasedonconvectiveexchangeofcerebrospinalfluidandinterstitialfluid.Putanotherway,sleep(apotentiallydangerousperiodofrequiredinactivity) involves a shutdownwhere neurotoxic waste products producedwhilstawakearecleanedfromthecentralnervoussystem(Xieetal,2013).However,this‘shutdown’isalsoaperiodofdream-states.OnemayreasonablyaskonwhatbasisanAIwouldshutdownandwhetherthisissleepfromwhichdream-statescouldarise?Ina human, dream-states are important for waking states as sources of inspiration,reflexive change andmany other consequences.Would a self-conscious AI have asubconsciousandanunconscious?

Onecouldgoonlistingpointsofpotentialdifferencebecauseofconstitution.Alternatively,onecouldnotethatmuchoftheaboverequiresonetoassumethatanAIcaninfactthink,anassumptionthattakesforgrantedthetechnologicalcapacitytocreatethought(Turing’saspirationcompleted),butsimultaneouslyquestionsthatthetechnologycanbeassumedtoresolveotherproblemsof (throughduplication)theself(anembodiedthinkingfeelinghuman-like-as-human-similar/emulatingentity).Assuch,thefocusondifferencesplitsdifferencebasedonanon-necessarydivideinwhatisassumed(apositivesolutiontooneaspect,anegativeforanother).However,itisprecisely because assumptions regarding duplication in one aspect do not requireassumedduplicationintheotherthatthepotentialdifferencescanbeexplored.Itiswhy they have been a staple of science fiction and of futurist speculation fordecades.46ItisalsowhysuchspeculationhasalsobeentranslatedascritiqueoftheChinese room thought experiment (most obviously via the robot objection). In TH

46 The speculation sits within a broader universe of philosophical-as-speculative argument. Forexample,thepossibilitythatourrealityisacosmologicalvirtualrealityspace:ifamaterialspeciesinamaterialuniversesurviveslongenoughtoachieveadvancedtechnologyitisreasonabletoassumeitwill alsoproducecomputer technology,and thiswill involveexponential advances in simulation;atsomepointadvancedcomputerscouldrunindistinguishable‘real’simulations.Thisbeingsoonemightthenassumetheratioofcomputersimulationrealitiestomaterialrealitiesfavourscomputerrealities,andthiswouldimplythatthelikelihoodisthatanygivenspeciesthatisself-aware,suchasourselves,livesinsucha‘reality’.Asafurtherstep,anyself-awareentitythatevolveswithinavirtualrealitywilldevelopsciencetointerrogatethatreality.Thecloserinvestigationcomestothefundamentalsofthatrealitythemoreitwillberevealedthatthebasisoftherealityismathematical(astrongPlatonicclaimforthestatusofmathematicsinherentinacoded/designed/syntheticvirtualreality).

37

thesearetheseductionsofaworldweareonthecuspofcreating.Noneofwhichistosuggestexplicitontologyhasbeenthetypicalpointofdeparture.

The immediatepoint,however, isthatsubcategoriesofAIsarealsodividinglines.Searle’sapproachhasbeensignificantasakeypointaroundwhichdisagreementcouldcoalesce,andhasinthissensehelpedtocreateandfosterthesubcategoriesofAIs.ThoughSearle’sChineseroomishelpfulinsettingoutdifference(whatanAIdoesnot demonstrate about a human) it is less helpful in terms of what may be thedifferentconstituentaspectsofanAIonceadividinglinehasbeendrawn.Itisunderelaborated in this way (however, see Preston and Bishop, 2002). Posed in purelyphilosophical terms this may seem unimportant, since Searle’s original argumentregardingwhetherAIisinfactthinkinghasnotsimplydisappeared.Buttheproblemisnotjustphilosophical,itissociologicalregardingtheconsequencesofphilosophy.The very separation into subcategories has become a problem in this sociologicalcontext.Workingacross thesubcategorieshasbecomeachallenge,partlybecausethere is something disempowering about the initial location of the majority of asubcategorywithinsciencefictionandfuturism.Oneisconstantlydealingwithwhatcanseemsimultaneouslyprofound,butalsooverblownandperhapsunserious.Thisisbynomeanstodenigrateinterestingworkthatisdone.Forexample,inphilosophyShanahan(2010)onembodiedAI,orSparrow’s(2004)alternativetoTuring’sgame,whichtakesitsinspirationfromBladeRunner’sVoight-Kampfftesttocreateathoughtexperiment to assess themoral capacities of an artificial entity. But consider thebroader problem of context based on increasing recognition of problems andpotentialsoftheactualtechnologiesthataredevelopedundertheaegisofAI.

Hereonemightnotethe201723AsilomarAIPrinciplesthatareintendedtoguide the future development of AI.47 These heavily emphasize control, benefit,common good, risk assessment and caution.However, the principles are not legalinjunctions,nordotheyrefertointrinsic(orsetbindingextrinsic)limitstotechnology,norcantheypreventalternativeinterest-incentivesthatmaysubverttheprinciples(astate’s concernswith security, surveillanceand superiorityof arms; a corporationsconcernswithcompetitiveadvantageandmarketcapture).48ItremainsthecaseaswefirstnotedearlyintheessaythatthesubjectmattersofsecondsubcategoryAIsaregraduallycomingintothepurviewoforganizations,butdoingsoisasocio-politicalactivity that is affectedby thedominanceof concerns thathave accompanied thedevelopmentofcategorizations.RecallthecaseofEUdeliberationsonAI,theideaofan electronic person was not central, and yet was recognized to be increasinglyimportanttoaddress.Manifestly,basedonthedevelopmentofcategorizations,theproblemsaremulti-faceted:normalisationofAI,theissueof‘cope’,functionalismandfunction, but now also dividing line inertia’s that reduce the urgency or resist thecentralityof concernswith real technologicaldevelopments thatmaybeoccurringundertheaegisofAI.Onemightwanttoconsiderthisalsointermsofotheressays’

47https://futureoflife.org/ai-principles/48Forexample,cyberwarandthenewsecuritisationaredeepproblems.Thereisnomutuallyassureddestructionapproachtocyberwar,andthismakesself-restraintlogicsdifficulttoimplement.Cyberwarisnotjustinformationextractionputtingsecurityoperativesatrisk.Itisalsoanabilitytoattackandparalyzeelectricity systems,hospitals,welfareadministrationorany complexbureaucracy. It is thecapacitytomanufacturefakenewsasinterventionsindemocraticprocesses.Thesewarscanbefoughtinproxy;dumpinginformationontoWiki-leaks…Cyberwarinformationispowerbutnottruth.

38

commentsontendenciestowardsTH.Inanycase,itisnotamatterofblametonotethatSearle’sChineseroomislimitedasaresourceintermsofanyreasonableanalysisofactualAI,butitisimportanttonotethatitsdiscursiverolehasbeenimportant.FromSearletorealproblemsof‘AI’Furthermore,onceonestarts thinkingaboutsociologicalnuanceonealsostarts tothinkaboutthewaythisinheresinsocialontology,andthisbringsusbacktomattersofelaboration.ForSearle,hissocialontologyisaphilosophyforthesocialsciencesnot a philosophy of the social sciences; it is a simplified (rather than simplistic)apparatusthatexpressesthecommonconstituentsfromwhichsocialrealityisbuilt.Theclaimisthatallhumaninstitutionalrealityiscreatedandmaintainedbya‘singlelogico-linguistic operation’: status function declaration, and so has a ‘commonunderlingstructure’ (2010:p.201).Twoproblemsarise.First, thepositionrequiresthat status function declaration be fully descriptive of the structure and,concomitantly,second,thepositionrequiresthatotherandfurthermattersarenotsignificantforboththestructureofsocialrealityandwhatoccursoninandthroughthatstructure,wherethis isdeemedtobeactualconstitutedsocial reality,since ifthey are significant then the structure itself is also not quite the structure-in-operation, which seems like a tension or incompleteness if not a contradiction intermsofwhatstructureisvis-à-viscreationandmaintenance.

OneneedstobecarefulnottotraduceSearlehere.Searle’ssocialontology(likeallofhiswork) isbrilliant.However, italsohas its limitsand likeallworks, itspointsof pressure. Searlehas configuredhis claim regarding social ontology tobeinternally consistent but in so doing he has preconfigured it to be potentiallymisleading.Theclaimisthathumaninstitutionalrealityisfullyaccountedforandsocreatedandmaintainedbyaninstantiatedvarietyofhissocialontology.SinceSearleis the one to define institutions through rule construction, and defines ruleconstruction as a consequence of status function declaration, then it is by logicalconsistencythattheclaimacquirescoherence.Buttheinferenceisthatsocialrealityis fully accounted for, rather than it is an internal system of rule creation andreproductionthatisaccountedfor.Coherencebecomescredence.Socialontologyisaconcernwithsocialbeing,MakingtheSocialWorldandTheConstructionofSocialReality are titles that convey the impression that more than linguistically stated(statable)rulesystemsarebeingaccountedfor,unlessallthatissignificantaresuchrulesystems.

ThoughSearle’sapproachisstrippeddownandelegantitcanalsoreadlikeanattempttoconstructacodeforhowsocietyoperates,wherethesysteminoperationseemstobedependentonsomuchmorethanthecode,andsothecodeisnotfullyexpressive of what society is. This seems slightly ironic if one considers how theChineseroomthoughtexperimentisdirected(codeisinsufficientforcomprehension)rather than how it is formulated. Searle’s structure (his social ontology as status-function declaration) does not internalize error, ambiguity, conflict, contingency,multiplicity, materiality, the constitution of distinct parts and their interactionsthroughaformalaccountofemergence,anditprovideslittlesenseofthedifferentvarietiesoftransactionswithrealityabeingisengagedin(bodily,personally,sociallyetc),whichmayalsoextendtohowapersonwhoisnotsimplyanexpressionofthe

39

sumofsocialconstructsnegotiatesand livesthroughsocialconstructions (andthismust be important if Searle’s ‘gap’ -- free will -- is to be real, and then alsoinstantiated).

Ifstatusfunctiondeclarationsareallthatisrequiredtocreateandmaintainsocialrealitytheneverythingelsebecomessuperfluous,anetcordetails(andmuchof this forSearle isdelegatedas ‘background’).49Theformaloperativepotentialofdirected language use becomes the overwhelmingly significant aspect of society.However,onemightarguethateverythingelsematterstowhatactuallyoccursandsowhatthingsbecomeinandthroughtime.Ifoneistowarranttheclaimthatasocialontology can be the philosophy for social science itmust also be a philosophy ofsociety.50Itmustbesociologicallyoperative.Searleisconfidentthathisapproachis.Status-function declaration as social ontology does more than merely confirm aninternallyconsistentclaimregardingthelogico-linguisticstatementofitselfastheory.Theclaim isalsothat ithasmorethanmerelysomepurchaseonsocial reality.ForSearle,itisthebuildingblocksofinstitutionalreality,andalsoabasisforexplanatoryinvestigation. Here Searle argues that institutions ground institutional facts, andbehavioroperateswithentitiesreferencedtothesethroughrationalityintheformof‘propositionalstructures’thatexpressreasonsforacting(everyactualdeonticpowerhasawhyandbecausepotentialforinvestigation).

However, though Searle’s social ontology allows one to ask and answerimportantquestionsaboutsociety,itisquestionablethatitcanprovideappropriatelydevelopedexplanatoryaccountsofwhathasandisactuallyhappeninginsociety(ithasasomeratherthantherelationtoaccountingforandexplainingsociety).ThisisdespitethatSearleclaimsinreplytoLawsonthat‘Mostimportant,theanalysishastoreallyanalyze.Nothingmustbe leftunexplained’ (Searle,2016:p.402).51ArguablySearle’ssocialontologyproducessociologicallylimitedaccountsofsocialreality:itisaninstitutionalfactthataisPresident.Hehadrightsbanddutiesdwithininstitutionewherexcountedasyinc.Itisbyvirtueofrationalityhthatpersonjdidg,andthiswasaccountedforbydeonticpowerpunderreasonsforactingn.OnecoulddescribeDonald Trump in terms of deontic powers, institutions, institutional facts, andrationalitiesthroughreasonsforacting,butarguablydoingsowouldnotprovideasatisfying explanation (of personhood, life projects, integration into existingpossibilitiesofinstitutionsindecay,changesthroughtimebasedoninteractionsinthe

49ThisisimplicitinLawson’scritiqueofSearleandoccurredtomewhilstreadingthatcritique;notably,‘the sort of totality it is… has a bearing on the sorts of positions and power relations thatwill beinvolved’(2016:p.388).50Consider:actualoperativedeontology isnotabstract logic it isalsothe integrationof institutionalconditionswithpossiblelifeprojectsandultimateconcerns.ThisisasociologicalproblemnotreducibletoconstitutiverulesorinstitutionalfactsalongthelinestypicallystatedbySearle.Itisthefeelingofand for a system. Searle’s logic of reasonmaynot be ill-foundedbut the framingof reason seemsinsufficient to capture the fullness of human lived experience as social reality. As abstraction it isimpersonalanddisembodiedandthisseemstoabstractfromwhatseemextremelyimportantaspectsofthehumanconditioninordertomakeclaimsaboutwhatconditionshumanexistence.51Searle’sresponsetoLawsonalsoseemstotradeonanambiguitybetweenonlyfoundinhumansandtheonlythingimportantfortheconstitutionofsocialreality:status-functiondeclarationseemstobeuniquetohumancivilizationthereforestatus-functiondeclaration isallthat issignificantforhumancivilizationasconstructed.Itisnotclearthatthisshouldfollow,andsoitcannotbetheactualforceofargumentasclaimthatcanbeusedtorefutebyreplythestatementofalternativesthatseemtoshareconstituentswithanimals.

40

context of disintegration, creating unintended consequences etc). Trump is at theleastastatusdysfunction.Associologytheaccountwould(withoutaugmentation)beexsanguinated,aspsychologyitwouldbesilent.52

PerhapsthisseemslikesomethingofadetourfromthefocusonfocusintermsofAIwandAIs.However,recallthatthepointistoassesstheconsequencesofhowargumentshavebeenconstructed,positionedandpursued.Wehavealreadynotedthat the problem of AI is not just philosophical, it is sociological regarding theconsequencesofphilosophy.Onemightnowalsoaddthattheproblemissociologicalregardingtheformofphilosophyandthatincludesontology.Searle’sChineseroomhasplayedamajorroleindevelopingthesub-categorizationofAIs.Hisownapproachto theChinese roomsituated in termsofhisontologyand socialontologyprovidelimitedresourcesaswaystothinkaboutanyrealizedAI(sinceitsconstitutionislikelytobedifferent).Onemightnowaddtothat,thoughthepointiscontestablebasedoncompeting ontologies, the ontology and social ontology within which Searle’sapproachtohisChineseroomargumentissituated(areversaloftheabovephrasing)provide limited resources for addressing the problem of any actual technologicalchangesundertheaegisofAI.ThisisadifferentpointthanthattheexistenceofthesubcategoriesofAIsseemstohavecreateddiscursiveconstraintsonaddressingtheproblemofanyactualtechnologicalchangesundertheaegisofAI(wheretheprimaryconcernisnothowactual‘AI’developstofunctionintheworld).Thepointhereisthatthe social ontologyprovides limited resources for exploring societyasa system inoperation,and‘AI’isanimportantsourceofchangeandchallengewithinthatsystem.

Ifoneistoconsiderprocessesintimethenoneneedsanontologyofprocessand time with a developed and consistent methodology. Archer’s realistmorphogenesisseemsanobviouscandidate(seeArcher,1995,2012).However,ifwearetoconsiderspecifics,thentheconceptofrelationalgoodsseemsausefulwaytobringthisessaytoaclose,sinceitallowsustofollowonfromcontemporarychangeinandaroundactualAIwhilstreturningtotheproblemoffocusonfunctioninamoreimmediateway.Function,AI,substitution,delegationandRelationalGoodsThesensethatAIiscomingandwemustcopeisnowwidespread.However,therearevariousattemptstopositionthisasintrinsicallybeneficial,orliabletobesobasedonthealreadyexistingprocessesthatexistbywhichAIisdevelopingorthroughwhichitsdevelopmentcanbemanaged.Ifwereturntothe100yearsStanfordprojectreportwebegan from, I suggested thisexhibitedanAIwposition,whilst it simultaneouslyacknowledged the roleof lawetc in relation to this,which I suggestedwas also amatter thatevokedAIs issuesofentitiesandshapedthewaysomeof those issueswerestatedandmademoreorlesssignificant.Wearenowinapositiontosaymoreaboutthis.AccordingtotheStanfordreport:

ThemeasureofsuccessforAIapplicationsisthevaluetheycreateforhumanlives…Given the speedwithwhichAI technologies are being realized… the

52 As philosophy the problem of logico-construction as presupposition, where primary statementsdefineandconfinesubsequentdevelopments,alsocreatesaproblemoftranspositionifonewantstoconsiderthewholeasnecessaryratherthansufficientforexplanation.

41

Study Panel recommends that all layers of government acquire technicalexpertise inAI…Facedwith theprofoundchanges thatAI technologies canproduce,pressurefor‘more’and‘tougher’regulationisprobablyinevitable.Misunderstandings about what AI is and is not could fuel opposition totechnologieswiththepotentialtobenefiteveryone.Inappropriateregulatoryactivity would be a tragic mistake. Poorly informed regulation that stiflesinnovation,orrelocatesittootherjurisdictions,wouldbecounterproductive…Inprivacyregulation[weadvocate],broadlegalmandatescoupledwithtoughtransparencyrequirementsandmeaningfulenforcement–ratherthanstrictcontrols… This in turn supports the development of professional tradeassociationsandstandardscommitteesthatspreadbestpractices…(Stoneetal, 2016: p. 10) ‘Policies should be evaluated as to whether they fosterdemocraticvaluesandequitablesharingofAIsbenefits,orconcentratepowerand benefits in the hands of a fortunate few… [Thereafter, AI must beintroduced] inways thatbuild trustandunderstanding,andrespecthumanandcivilrights.’(Stone,2016:p.11)

Now,considerthecontext inwhichthestatementsabovearemadeandalsowhattheyandthereportingeneraldonotstate.Thepage11quotesitsawkwardlywiththepage10, since the latter (takingprivacyasanarchetype)emphasises thatoneshould resist regulation until well informed and suggests the best source of suchinformationisthebestpracticethatemergesfromtradeassociationsandstandardscommittees. It leads to a dominance of self-regulation in market situations bypowerfulprivatepartiestothosesituations.Ingeneral,thisassumesthatinformationand practice are already, or are developing along, lines that are objectively-as-universally beneficial and that this, furthermore, is either normatively beneficialthrough development and discussion by parties or is a situation of normativeneutrality in relation to technology (since more and better information and bestpracticeare intrinsictoprocessesandtheseareassociated).Butthisthenrequiresthat the driving force of change and innovation within the world and under theauthorityoftherequisitebodiesisexpressibleinthesebeneficialwaysandthatnoother considerations can also exist that subvert, shapeor co-optwhat occurs andunderwhatcircumstances.Thisisasophisticatedwaytoexpressacquiescenceorlackofresistanceorfurtherscrutinytoanineluctableprocessofchangeinrelationto‘AI’(whereAIisatravellingfrontieroftechnologicaldevelopmentratherthananentityforwhichthoughtetchasbeendecisivelydemonstrated).Itreversesthemeaningofcautiontomean‘donotbehastyinimpedingchange’ratherthan‘considercarefullywhattheconsequencesofchangemaybebeforetheyactuallymanifest’.Moreover,itrestrictsthecapacitytoparticipate,andsobepowerful,tothosewhoarealreadypowerfulbyvirtueofpositionascontrolofinformationorresources.Itempowersandauthorises those who own rather than those who are subject to consequences.Concomitantly,itcreatesabarriertobroadeninganddemocratisingdeliberationandparticipation in the process bywhich change occurs and throughwhich change isshaped.AI iswhatAIdoesandAIwillbewhatAIresearchersdoacquiremoreofaproblematic set of connotations when considered in this way. AIw is not withoutstrengthifonestartstothinkaboutpowerrelations,andhereonemightalsogobacktoandrethinkthecontextofthePewresearchIpreviouslyreferredto.IntheUSat

42

least, the public feel uninformed and unable to effectively engagewith change inrelationtoAI(andTH).

Thenconsiderthenatureofcontemporarycapitalistsocietiesinwhichthesetofinjunctionsagainsthastyinjunctionaremade.Thevaluetohumanlivesisnotthemeasureofsuccessincapitalistprocesses,sostatingthisasthemeasureofsuccessofAI applications is potentially adverse arrogation rather than justified extension. Acamouflage of concerned language disguises a basic logic that requires that thespecific interests of some become the engine by which the interests of all willmanifest.Nowconsiderhowthetwoaresupposedlyintegrated,alignedormediated:itisnotinrelationtocitizensascitizensonly.Itiscitizensasconsumers,citizensasworkers, and then citizens as further recipients of services (state provided orotherwiseinthecontextofwelfare).Letusconsidertheseinorder.

Intermsofdominantideology,itisasconsumersthatthemanyexercisepowerthroughmarkets,andsoit isthroughthedynamicsofsuchbehaviourthatalightlyregulatedandmainlyself-regulatedsystemofcorporationsisshaped.Thewantsandneedsofthemanyarerespondedtobycorporations,andsocorporationsareshapedby what the many want and need. Corporations are disciplined in this primitivedemocraticexpressionofindividualpowerthatbecomescollectivepowerthroughitseffectontheprofitsofcorporations.However, thevalueofhuman lives isnot thefocusorgoalofthissystem, it isdeemedtobeanunintendedconsequenceoftheinteractionsof the system.Rather thanapublicdeliberationonwhat thevalue tohumanlivesisandhowthisisexpressedintermsoftheconceptsofwantandneed,wearelefttosimplyassumethatprocesseswilldeliverwhatwereallywantorneed,andthiswillbeofvaluetohumanlives.Thisimplicitlyentailsbenignorbenevolentcapitalism where technological change including AI is more-or-less conducive tohumanprogress. Italsodisguisestheasymmetryofpowerthat isheavilyweightedtowardscorporations.

Corporationsshapewhatwewantorneedandshapethemarkets inwhichindividuals supposedly exercise democratised marketplace power. Profit drivescorporationstocapturemarketsandlimitandpressurisechoices.TheeffectofrealAIherecanbemultiple.Forexample,intheabsenceof‘netneutrality’(aprohibitiononserviceprovidersmanipulatingaccesstotherangeofsourcesofservicestoencouragesomeoverothers--ahiddenmarketadvantagethroughconstitutingthemarketofchoices)thenAIcanbeusedtochannelaccessandactivitytoanythingthatrequiresinternet connectivity.Another example is that in the absenceof prohibition,AI assmart algorithms canproduceopaque artificial stupidity that is difficult to contestbecauseoftheapparentobjectivityofbigdataandquantifiedmetricbaseddecisions(affectingeverythingfromcreditaccessbecauseofcreditratings,towhogetsfiredbased on ‘performance’ measures). What both these examples illustrate is thatcorporationscancontroltheinfrastructureofcontemporarylifethroughAIinwaysthat preconfigure the social world of the consumer. Thereafter, specific AIs canbecomenecessary toparticipation in societyandsonecessary toemployability,oracceptancethroughsocialnormativity(apps,smartphonesandchatbotsallhavethispotential).Allofwhichsuggeststhatan informationandbestpracticeapproachtoregulation,devolvingtoself-regulation,astechnologiesdevelopundertheaegisofAI,cannotbeassumedtobeadecentredformthatleadstooutcomesthatvaluehumanlives.Thereisrarelyasimplesituationwhereconsumerscanchoosebetweeninfinite

43

options with no consequence to themselves but every consequence for thecorporation. An information and best practice approach favours those who arealreadypowerfulbyvirtueofpositionthroughcontrolof informationorresources,has the potential for human harms, and subordinates any concerns regarding thevalueofhumanlivestothevaluesofcorporations.

AsInotedatthebeginningofthispapertherearegravepotentialproblemshereintermsofAIwandtheproblemoffocusonfunctionasawaytodecentreormarginalise the important issue of what is important to human concerns orflourishing.Inadditiontothecontextofconsumption,AIhasmajorramificationsforwork. A discursive split is beginning to emerge between thosewho argue that animminentAI(androboticAI)revolutionwillbetransformativeandliberatingandthosewhoargueitwillbedevastating.Theformeristypicallyexpressedasanintentto‘taketherobotoutofhumanworkrather thantoput thehumanoutofwork’ (tedious,repetitive,andonerouslyphysicallabourwillnolongerbenecessaryforsomesub-setofhumans).Thelatteristypicallystatedasa‘thistimeisdifferentargument’(AIwillaffectalmostallpartsofeconomiesalmostsimultaneously,includingpreviouslynon-replicableskilledmiddleandupperincomejobssuchasaccounting,lawandmedicine,preventing awidespread responseof transition to someother kindofwork, sincetherewillbeinsufficientscopeforthatwork---capitalistcreativedestructionisthistime going to be destructively destructive). The problems hinge on the issue ofsubstitutionofAIforhumanswithinacontextwheretheeconomicsystemrequiressufficienthumanstobeemployedtoearntheincomethatthenbecomesthesourceofconsumptionthatmaintainsthecorporationsthatusethelabour(whetheritbeAIorhuman).Thebasicchallengeisthat individualcorporationsarerequiredtotreatwithcautionorresisttechnologythatcreatesa24houradayworkforcethatdoesnotgetsick,doesnotretire,doesnotstrike,anddoesnotearnwagesorseektermsandconditions, since if theydonot resist then the systemof corporations is adverselyaffectedcollectivelybytheself-interestofeveryindividualcorporation(thoughthereisnothingnewaboutthistension,sinceMarxwasabletopointitout150yearsago).The collective consequences of that self-interest is then socio-economic collateraldamage--thepotentialforwidespreadsocio-economicdisintegrationanddisruptionwithrealhumancosts,unlesssolutionsemergeoraredesigned(suchasauniversalAItaxonproductionofgoodsandservicesinconjunctionwithuniversalbasicincomeforhumans,orthemonetisationofstatespendingsystems--aradicalnewapproachtotheinstitutionsoffiscalpolicyviamoneycreation).Again,thereisaclearproblemofhowtoaddressthevalueofhumanlivesherebasedonprocessesthatarealreadyrecognized,butarenotsufficientlycentredasmattersofconcernforthepopulationsofsocietiesthatseemsettoexperiencetheconsequences(theyarecurrentlymattersof latent anxiety rather than front-and-centre urgent debate). So, in the case ofcitizensasconsumersthereseemstobeanadverseassumptionthatitisbyoptinginandoutofmarketsthatmostproblemswillbesolved,andinthecaseofcitizensasworkers, the consequences forworkerswait upon the capacity of corporations torecognizethecollectiveproblemoftheirindividualactivitywhereitistheirindividualactivitythatmarketsencourage(thebottomline).

AwarenessintheworldofacomingAIrevolutionintheformofsubstitutionandworkisgrowing.PeoplearealsoincreasinglyawarethatAIinvolvesanissueofdelegation.Thatis,thetakingoverofactivitybyAIonbehalfofhumans.Thisisthe

44

realm of citizen welfare (the nurturing of self and others). The way people arebecomingawareofthisisfragmented.ThereisarecognitionthattheuseofAImayfreeuptime:an‘internetofthings’cancoordinate,anticipateandundertaketasksonourbehalfthroughAI--everythingfromadjustingcentralheating,toshapingthroughsuggestionandpre-selectionwhatwemaywanttobeinformedabout,beinterestedin,andconsume,tomanagingacalendarandmaintainingcontactswithassociates,colleagues,friendsandfamily.

Clearly,thepossibilitiesherecanbepositionedaspotentialbenefitstohumanlives(andperhapsalsototheenvironment).However,muchasinthepreviouscases,onecannotassume thatpotential is realisedor that it iswithout thepossibilityofadverseconsequences.Delegationcreatesahostofpotentialproblems.Ifdelegationissubordinatedtoefficiencythentimeisfreedupfromthingswemayhavefailedtoappropriatelyvalueinordertocreatetimetopursueimposedfunctionsthatmaybeharmfultothenurturingofafullyrealisedhuman.Thingsforgonemayhavehadvalueinthemselves:thecraftofmakingorcompletingsomething,thedevelopmentofselfthrough the thinking through and doing of something on one’s own behalf, thepleasureandmeaningfulnessinengagementwithothers,andsoforth.Curtailment,convenienceandquicknessofendproductbasedondelegationarenotnecessarilythesameas‘better’.Thisisparticularlysoiftheprocessofdelegationinfantilisestheselfandleadstothefreeingoftimethatisthencaptured.OneshouldnotneglecttheTomorrow’sWorldfallacy:fordecadesmediahavebeenpredictingthatlabourandtimesavingtechnologieswouldresultingreatlyincreasedleisureandreducedworkbecausemorecanbedoneinlesstime.ThoughitmaybethatfutureAItransformsworkthroughsubstitution,inthemeantime,theobservedtendencyoftechnologicalchangeshasbeenthecapacitytocompelmorehoursofwork,andforsome,thishasbeen based on the capacity to work from anywhere at anytime (in connectiveemployment)ortobecalledintoworkatanytime(in‘gig’economies).Inthiscontextthe value of human life is subordinated to themeaning and practice imposed onefficiency.Benignorbenevolentcapitalismcannomorebeassumedherethanitcanintermsofthecitizenasconsumer,notleastbecausethetwoincreasinglyoverlapinconsumption-basedandfinancialisedsocietiesofdebt-dependence.

Intermsofcitizenwelfareanddelegationthereisalsorecognitionthatmanysocieties confront a demographic problem in the form of an aging populationcombinedwith reducedbirth ratesanddisaggregatedpatternsof living, creatingaproblemofcarefortheelderly;AIandsmartaccommodationbasedonaninternetofthings, combined with robotics, are now being considered likely solutions to thisproblem.Ostensibly, thismore than any other area seems one inwhich potentialbenefits will manifest. However, it still shares with all the other areas set out anultimate problem of being and doing. There is a basic challenge that needs to beaddressedintermsofhowbeingcanbenurturedregardingwhatisandwhatisnotdone.Thisisanissuethatafocusonfunctioncannotresolveunlesswealsoconsiderthenatureofthehumanthattheproblemoffunctionistoberesolvedfor.53

53 And so the problem ofwhat is a person and inwhat sense they flourish and suffer is centrallyimportant;thereisgreatscopefordevelopmenthereofanaturalisticethics(ifwhatapersonisaffectshowapersonflourishes).Searle,forexample,considersthisintermsofhumanrightsinMakingtheSocialWorld, Andrew Sayer provides a general account of embodied needy beings inWhy ThingsMattertoPeopleandChrisSmithprovidesasetofconstituentsofapersoninWhatisaPerson?--an

45

AsIhavearguedthroughoutandcumulatively,AIwdeferstheproblembutwithpowerful consequences in so far as it favours what already exists in terms oftendencies,interestsandpower.FollowingTuringandSearle,AIscreatesawholehostofissuesthatneverquitebringtogetheracentralconcernwithtechnologyasisandthehumanwho isaffected. Interaliaontology is rendered interstitial, andyet theissuesarequintessentiallyamatterofontologyandsocialontology.Intermsofthisfinalmatterof the realmofcitizenwelfareone insightful conceptual innovation insocialontologyisDonatiandArcher’srelationalgoods.Relationalgoodsprovideanimportant way to think about how the human is nurtured in and through socialrelations. For Donati and Archer relational goods are goods created and enjoyedthrough relations, they involvesomeactivitywhich is itsownrewardbut thatalsocreatescollectivesocialbenefits.Suchgoodsarediverseandareconstitutedasthequalityofarelationthatarisesbetweenpeople,suchastrust,aswellasthequalityofexperience of cooperation, coproduction or collaboration. Such relations can beintimatelyinter-subjectiveandinformalormoreassociativeandimpersonal,butinallcasesthegoodsarenotinterchangeablewithmaterialgoods,anddonotconsistintheproductoftheactivity(DonatiandArcher,2015:pp.199-200and207).Theyareconstitutedandenjoyedthroughtheactivity.Assuchtheyrequiredevelopmentandnurture and become the products of enduring relations.54 They cannot simply becreatedbylawordictate.Theycannotbecapturedorappropriatedbyanygivenpartyand cannot be commodified, bureaucratised or marketised without the relationsthemselves being subverted in ways that corrode the goods that are otherwiseconstituted. They are ‘pro-social’ in so far as they contribute to the integrationofsociety, but they also donot fit readily into traditional categories of the public orprivatesphere,sincetheformerisassociatedwithadministrativeprovisionofgoodsbythestateandthelatterwiththemarketisationofgoodsbycorporations,neitherofwhich captures the sense ofwhat relational goods are or provides unproblematicgroundsfortheconstitutionofrelationsfromwhichtheyarise.However,accordingtoDonati and Archer, relational goods ‘correspond to fundamental human needs’(2015,p215)and‘Ifthesegoodsareignored,dismissedorrepressed,theentiresocialorder is impoverished… with serious harm caused to people and the overallorganization[ofsociety]’(2015:p.203).

The concept of relational goods can appear amorphous, but this seems aconsequenceofwhattheconcept is intendedtoarticulate,ratherthanafailureofclarity. We intuitively recognize that there is something common to the positivequalityofexperienceofrelationsandthereisnothingmysteriousaboutthis.Whatismysteriousisthewaysocietycanbothrecogniseandyetfailtofostersuchrelations(whatelseisinstrumentality,alienation,ennuiorevenanomie,ifnotananti-human

approachthencritiquedbyArcherandalsoPorpora,sincethereisnoclearsenseintermsofwhichSmith’s set is the set of required constituents, rather thanmerely a list of possible characteristics.Lawsonalsoprovidesavarietyofnaturalisticethics.54Ingeneral,DonatiandArcherclaimthatrelationalgoodsrequire:1)apersonalandsocialidentityofparticipants (they cannot be anonymous for each other) 2) non-instrumental motivation of eachsubject;therelationmustinvolvemorethanachievementofsomeend3)participantsmustacquireorbeinspiredbyruleofreciprocityasasymbolicexchange4)sharing:goodscanonlybeproducedandenjoyed together by thosewhoparticipate 5) require elaboration over time; a single interaction isinsufficientfortherelations6)reflexivitythatoperatesrelationally-sharingisalsoofthesenseofwhatitisthatisshared.

46

relationalfailure).Forexample,followingmycommentsonAIsearlyinthispapertheEuropean Union Civil Law on Robotics both marginally recognizes (in two shortstatements)andyetdefersanymeaningfulcommentonthequalityof relationsasgoods,specificallyintheformofcare:‘the‘soft impacts’onhumandignitymaybedifficulttoestimate,butwillstillneedtobeconsideredifandwhenrobotsreplacehumancare,’(EP,2016:p.4)and‘humancontactisoneofthefundamentalaspectsof human care… replacing the human factorwith robots could dehumanise caringpractices,’(EP,2016:p.9).Whatisclearisthataconceptofrelationalgoodsprovidesa potentially insightfulway to examine the issue of delegation in terms of AI andcitizenwelfare.Insodoingitreturnsustoethics,aperennialissueforAI,butdoessoonabroadercanvas:

The proof that today’s public ethics do not involve a common good in arelationalsense is found inthecase inwhich, forexample,theproblemsofpeace,development,theenvironment,andalsoofnewformsofpovertyarenotconfrontedasproblemsofconcretehumanrelationsbetweenco-presentsubjectsbutaresimplytreatedas‘things’toeliminatebymarginalizingviolentpersons,punishingthosewhodonotsucceedincompeting,banningpolluters,and helping the poor withmeasures that promote passivity, Problems areconfrontedbyputtingpeoplewheretheycannotcausetrouble.Thesearefalsesolutionstoproblemsbecausetheyarenotinspiredbythecommongoodinthat they leave aside completely the necessity of involving poor andmarginalizedpeople…Inthearenaofsocialpolicies,itisnowveryclearthatthesemodalitiesforconfrontingdistress,poverty,andsocialmarginalizationare completely unsatisfactory. Peace, development, a clean and safeenvironment,adecentlifeforeveryone-theseareallgoodsthatcorrespondtotherelationalcharacteroftheseobjectives:thatistosaytheycanonlybeachievedtogether;theyarenotthesumofindividualpreferences…Relationalgoodsarethekeyformovingfromthewelfarestatetothewelfaresociety.’(DonatiandArcher,2015:p.217)

ConclusionAIisoneissueamongmanyandintheendmustbeconceivedasoneaspectofoneworld.Searle issurelycorrectthatsocialscience isan investigation intoasingle(ifmultiply produced and constructed and disputed)world (2016). This is a claim heshareswithrealistontologyandsocialontology.However,asIhavetriedtoestablishin thisessay theproblemofAIhasnot come together inany clear singular sense.Sophisticatedorigins(‘yesterday’s’)inphilosophyhavehadconsequences.Focihavedeveloped expressing bifurcations and marginalisations, and creating interstitialissues.Little ifanythinghasbeenresolved,andsofunctionhasdominated inwaysthataresignificantforwhatoccurswhilstdisputecontinues.Awholehostofcriticallyimportant issues based on actual technological potentials have arisen. Tomorrowcontinuestobeaffectedbytodaywithout(peoplein)todayhavinganyclearcollectiveideahowit(they)willproducetomorrow.Andyetthereiscontenttoprocessinsofarassomehaveaveryclearideaoftheiragenda,andtheyinthemainarenotthinkingascitizensorbeinginvitedto(orinvitingothersto)deliberateascitizens.

47

ReferencesAnderson,D.(1987)‘IstheChineseRoomtherealthing?’Philosophy62(241):389-393Archer,M.(1995)RealistSocialTheory:TheMorphogeneticApproachCambridge:

CambridgeUniversityPressArcher,M.(2012)TheReflexiveImperativeinLateModernityCambridge:Cambridge

UniversityPressBadmington, N. editor (2000) Posthumanism: Readers in Cultural Criticism

Basingstoke:PalgraveMacmillanBostrom, N. (2016) Superintelligence: Paths, dangers, strategies Oxford: Oxford

UniversityPressBostrom,N.andOrd,T.(2006)‘Thereversaltest:Eliminatingstatusquobiasinapplied

ethics,’Ethics116(4):656-679Braidotti,R.(2013)ThePosthumanCambridge:PolityPressBrockman,J.editor,(2015)WhattothinkaboutmachinesthatthinkNewYork:HarperCaliskan,A.Bryson,J.andNarayanan,A.(2017)‘Semanticsderivedautomaticallyfrom

languagecorporacontainhuman-likebiases,’Science356,April:183-186Calverley,D. (2007) ‘Imagining anon-biologicalmachine as a legal person,’AI and

Society22(4):523-537Cabrera,L.(2015)RethinkingHumanEnhancementBasingstoke:PalgraveMacmillanChalmers,D.(1996)TheConsciousMind:InsearchofafundamentaltheoryOxford:

OxfordUniversityPressChurchland,P.andChurchland,P.(1990)‘Couldamachinethink?’ScientificAmerican

262(1):32-37Clarke,S.,Savulescu,J.,Coady,C.,Giubilini,A.andSanyal,S.editors(2016)TheEthics

ofHumanEnhancement:UnderstandingtheDebateOxford:OxfordUniversityPress

Collier,A.(2003)IndefenceofobjectivityandotheressaysLondonRoutledgeCrockett, L. (1994), The Turing Test and the Frame Problem: AI’s Mistaken

Understanding of Intelligence, Norwood New Jersey: Ablex PublishingCorporation

Damasio,A.(1994)Descartes’error:EmotionreasonandthehumanbrainNewYork:Putnam

Dennett,D.(2013)IntuitionpumpsandothertoolsforthoughtNewYork:NortonDonati, P. and Archer, M. (2015) The Relational Subject. Cambridge: Cambridge

UniversityPressEconomist(2016)‘Fromnotworkingtoneuralnetworking,’TheEconomistJune25thElder-Vass, D. (2012) The Reality of Social Construction Cambridge: Cambridge

UniversityPressEllen,M. ‘Everybody lovesasadsong,said themasterof themall,’TheTimes12th

November2016European Parliament [EP] (2016) ‘Draft report with recommendations to the

Commission on Civil Law Rules on Robotics’ European Parliament2015/2013(INL)

Fodor,J.(1992)AtheoryofcontentandotheressaysCambridgeMa:MITPress

48

Fuller,S.(2011)Humanity2.0:Whatitmeanstobehuman,pastpresentandfutureBasingstoke:PalgraveMacmillan

Funk,C.Kennedy,B.Sciupac,E(2016)‘USpublicwaryofbiomedicaltechnologiesto‘enhance’humanabilities,’PewResearchCenter,July26thavailable:http://www.pewinternet.org/2016/07/26/u-s-public-wary-of-biomedical-technologies-to-enhance-human-abilities/

Gidley,J.(2017)TheFuture:AveryshortintroductionOxford:OxfordUniversityPressGodfrey-Smith,P.(2017)OtherMinds:Theoctopusandtheevolutionofintelligentlife

London:WilliamCollinsHarari,Y.(2016)HomoDeus:abriefhistoryoftomorrowLondon:HarvillSeckerHarnad, S. (1989) ‘Minds, machines and Searle’, Journal of Experimental and

TheoreticalArtificialIntelligence1(1):5-25Herbrechter,S.(2013)Posthumanism:AcriticalanalysisLondon:BloomsburyHauser, L. (1997) ‘Searle’s Chinese Box: Debunking the Chinese Room argument’,

MindsandMachines7(2):199-226Kurzweil,R.(2000)TheageofspiritualmachinesLondon:PenguinLawson,T.(2016)‘Comparingconceptionsofsocialontology:Emergentsocialentities

and/or institutional facts?’ Journal for theTheoryofSocialBehaviour46(4):359-399

Lassegue,J.(1988),‘WhatKindofTuringTestdidTuringHaveinMind?’,Tekhnema3:37–58.

Legg,S.andHutter,M. (2007) ‘ACollectionofDefinitionsof Intelligence’Technicalreport,IDSIA07-07Available:https://arxiv.org/pdf/0706.3639v1.pdf

Mason, D. (2016) ‘Human enhancement: The scientific and ethical dimensions ofstrivingforperfection,’PewResearchCenter,available:http://www.pewinternet.org/essay/human-enhancement-the-scientific-and-ethical-dimensions-of-striving-for-perfection/

Mays,W.(1952),‘CanMachinesThink?’Philosophy27(101):148–162.McGilchrist,I.(2009)TheMasterandhisemissaryLondon:YaleUniversityPressMillican,P.andClarke,A.editors.(1996)MachinesandThought:TheLegacyofAlan

Turing,Volume1Oxford:ClarendonPressMillar,P.(1973),‘OnthePointoftheImitationGame’,Mind82(328):595–597.Moor,J.(2000)‘AlanTuring1912-1954,’MindsandMachines10(4):461Morgan,J.(2013)‘Ahumanistnarrativelessthanthesumofitsparts,’Metascience

22(1):111-113Morgan,J.(2014)‘Whatisprogressinrealism?Anissueillustratedusingnormcircles,’

JournalofCriticalRealism13(2):115-138Morgan,J.(2016)‘Changeandachangingworld?TheorizingMorphogenicSociety,’

JournalofCriticalRealism15(3):277-295Nagel, T. (1979) ‘What is it like to be a bat?’ pp 165-180 inMortal Questions,

Cambridge:Canto/CambridgeUniversityPressNordmann,A.(2007)‘Ifandthen:Acritiqueofspeculativenano-ethics,’NanoEthics

1(1):31-46O’Connell,M.(2017)Tobeamachine:Adventuresamongcyborgs,utopians,hackers.

AndthefuturistssolvingthemodestproblemofdeathNewYork:DoubledayPinker,S.(1998)HowthemindworksLondon:Penguin

49

Pinker,S.(2015)‘Thinkingdoesnotimplysubjugating,’inBrockman,editor(2015),pp.5-8

Pinsky,L.(1951),‘DoMachinesThinkAboutMachinesThinking?’Mind60(239):397–398.

Rainie,L.Hefferon,M.Sciupac,E.andAnderson,M.(2016)‘Americanvoicesonwayshumanenhancementcouldshapeourfuture,’PewResearchCentre,July26thAvailable:http://www.pewinternet.org/2016/07/26/american-voices-on-ways-human-enhancement-could-shape-our-future/

Regis,E.(1991)GreatMamboChickenandtheTranshumanConditionLondon:BasicBooks

Rescher,N.(2011)RealityanditsAppearanceLondon:ContinuumSandel,M.(2007)Thecaseagainstperfection:ethicsintheageofgeneticengineering

CambridgeMass.:HarvardUniversityPressSaygin,A.Cicekli, L. andAkman,V. (2000) ‘TuringTest: 50 years later,’Mindsand

Machines10(4):463-518Searle,J.(1980)‘Minds,brainsandprograms,’BehaviouralandBrainSciences3(3):

417-457Searle,J.(1985)Minds,BrainsandScienceCambridge,Mass:HarvardUniversityPressSearle,J.(2002)‘Twenty-oneyearsintheChineseroom,’pp.51-69inPreston,J.and

Bishop,M. (2002)Views into the Chinese room: New essays on Searle andartificialintelligenceOxford:OxfordUniversityPress

Searle,J.(2010)MakingtheSocialWorld:TheStructureofHumanCivilizationOxford:OxfordUniversityPress

Searle,J.(2016)‘Thelimitsofemergence:ReplytoLawson,’JournalfortheTheoryofSocialBehaviour46(4):400-412

Shanahan,M(2010)Embodimentandtheinner life:CognitionandconsciousnessinthespaceofpossiblemindsOxford:OxfordUniversityPress

Sparrow,R.(2004)‘TheTuringTriageTest,’EthicsandInformationTechnology6(4):203-214

Stephan, A. (2006) ‘The dual role of emergence in the philosophy ofmind and incognitivescience,’Synthese151(3):485-498

Stone,P.,Chair,(2016)ArtificialIntelligenceandLifein2030:Reportofthe2015StudyPanel forOneHundredYearStudyonArtificial Intelligence (AI100) StanfordUniversity

Turing,A.(1950)‘Computing,MachineryandIntelligence’,Mind59(236):433-460Wilczek, F. (2015) ‘Three observations onArtificial Intelligence,’ in Brockameditor

(2015),pp.121-123Wolfe,C.(2009)WhatisPosthumanism?Minneapolis:UniversityofMinnesotaPressWorld Transhumanism Association (WTA) (2005) ‘Artificial Intelligence and

Transhumanism,’Availableat:http://itp.uni-frankfurt.de/~gros/Mind2010/transhumanDeclaration.pdf

Xie,L.Kang,H.Xu,Q.Chen,M.Liao,Y.Thiyagarajan,M.andO’Donnell,J.(2013)‘Sleepdrivesmetaboliteclearancefromtheadultbrain,’Science342(6156):373-377

50