Repositorium für die Medienwissenschaft
Benjamin H. Bratton
Outing Artificial Intelligence. Reckoning with Turing
Tests
2015
https://doi.org/10.25969/mediarep/1282
Veröffentlichungsversion / published version
Sammelbandbeitrag / collection article
Empfohlene Zitierung / Suggested Citation:
Bratton, Benjamin H.: Outing Artificial Intelligence. Reckoning with Turing Tests. In: Matteo Pasquinelli
(Hg.): Alleys of Your Mind. Augmented Intelligence and Its Traumas. Lüneburg: meson press 2015, S. 69–
80. DOI: https://doi.org/10.25969/mediarep/1282.
Nutzungsbedingungen:
Dieser Text wird unter einer Creative Commons Namensnennung - Weitergabe unter gleichen Bedingungen 4.0
Lizenz zur Verfügung gestellt. Nähere Auskünfte zu dieser Lizenz
finden Sie hier:
https://creativecommons.org/licenses/by-sa/4.0
Terms of use:
This document is made available under a creative commons Attribution - Share Alike 4.0 License. For more information see:
https://creativecommons.org/licenses/by-sa/4.0
[4]
Outing Artificial
Intelligence: Reckoning
with Turing Tests
Benjamin H. Bratton
Various anthropocentric fallacies have hobbled the
development of artificial intelligence as a broadly
based and widely understood set of technologies.
Alan Turing’s famous “imitation game” was an ingenious thought experiment but also ripe for fixing the
thresholds of machine cognition according to its
apparent similarity to a false norm of exemplary
human intelligence. To disavow that fragile self-refection is, however, easier than composing alternative
roles for human sapience, industry, and agency along
more heterogeneous spectrums. As various forms of
machine intelligence become increasingly infrastructural, the implications of this difficulty are geopolitical
as well as philosophical.
In Alleys of Your Mind: Augmented Intellligence and Its Traumas, edited by Matteo Pasquinelli,
69–80.Lüneburg:mesonpress,2015.
DOI:10.14619/014
70
Alleys of Your Mind
[One philosopher] asserted that he knew the
whole secret . . . [H]e surveyed the two celestial
strangers from top to toe, and maintained to
their faces that their persons, their worlds, their
suns, and their stars, were created solely for the
use of man. At this assertion our two travelers let
themselves fall against each other, seized with a
fit of . . . inextinguishable laughter.
— Voltaire, Micromegas: A Philosophical History
(1752)
Artificialintelligence(AI)ishavingamoment,withcognoscentifromStephen
HawkingtoElonMuskrecentlyweighingin.1 Positions are split as to whether
AIwillsaveusorwilldestroyus.SomearguethatAIcanneverexistwhileothers insist that it is inevitable. In many cases, however, these polemics may be
missing the real point as to what living and thinking with synthetic intelligence
verydifferentfromourownactuallymeans.Inshort,amatureAIisnotan
intelligence for us,norisitsintelligencenecessarilyhumanlike.Forourown
sanityandsafetyweshouldnotaskAItopretendtobe“human.”Todosois
self-defeating, unethical and perhaps even dangerous.
ThelittleboyrobotinStevenSpielberg’sA.I. Artificial Intelligence (2001)wants
to be a real boy with all his little metal heart, whereas Skynet in the Terminator
movies(1984–2015)representstheoppositeendofthespectrumandisseton
ensuring human extinction. Despite all the Copernican traumas that modernityhasbrought,someformsofhumanism(andtheircompanionfiguresof
humanity)stillpresumetheirperchinthecenterofthecosmiccourt.Iargue
thatweshouldabandontheconceitthata“true”artificialintelligence,arriving
at sentience or sapience, must care deeply about humanity—us specifically—as
the focus of its knowing and desire. Perhaps the real nightmare, even worse
thantheoneinwhichtheBigMachinewantstokillyou,istheoneinwhich
it sees you as irrelevant, or not even as a discrete thing to know. Worse than
being seen as an enemy is not being seen at all. Perhaps it is that what we
reallyfearaboutAI. 2
ItisnotsurprisingthatwewouldfirstthinkofAIintermsofwhatweunderstand intelligence to be, namely human intelligence. This anthropocentric
fallacy is a reasonable point of departure but not a reasonable conclusion.
1
OnHawking,seehiscommentstoBBCathttp://www.bbc.com/news/technology-30290540andalsoElonMusk’s$10milliondonationtoFutureofLifeInstitute“to
preventAIfrombecomingevil”inthewordsofWiredmagazine.Seehttp://www.wired.
com/2015/01/elon-musk-ai-safety
2
ParaphrasedfromBratton2014.
OutingArtificialIntelligence
TheideaofdefiningAIinrelationtoitsabilityto“pass”asahumanisasold
asAIresearchitself.In1950,AlanTuringpublished“ComputingMachinery
and Intelligence,” a paper in which he described what we now call the Turing
Test,andwhichhereferredtoasthe“imitationgame”(Turing1950,433–460).
Therearedifferentversionsofthetest,allofwhicharerevealingaboutwhy
ourapproachtothecultureandethicsofAIiswhatitis,forgoodandbad.For
the most familiar version, a human interrogator asks questions to two hidden
contestants, one a human and the other a computer. Turing suggests that if
the interrogator usually cannot tell which is which, and if the computer can
successfully pass as human, then can we not conclude, for practical purposes,
thatthecomputeris“intelligent”?(Morepeople“know”Turing’sfoundational
text than have actually read it. This is unfortunate because the text is marvelous,strangeandsurprising.)
Turing proposes his test as a variation on a popular parlor game in which two
hiddencontestants,awoman(playerA)andaman(playerB)trytoconvincea
third that he or she is a woman by their written responses to leading questions. To win, one of the players must convincingly be who they really are,
whereas the other must try to pass as another gender. Turing describes his
ownvariationasonewhere“acomputertakestheplaceofplayerA,”andsoa
literalreadingwouldsuggestthatinhisversionthecomputerisnotjustpretending to be a human, but pretending to be a woman. It must pass as a she.
OtherversionshaditthatplayerBcouldbeeitheramanorawoman.Itmatters quite a lot if only one player is faking, or if both are, or if neither are. Now
that we give the computer a seat, it may pretend to be a woman along with
a man pretending to be a woman, both trying to trick the interrogator into
figuringoutwhichisamanandwhichisawoman.Orperhapsthecomputer
pretends to be a man pretending to be a woman, along with a man pretending
to be a woman, or even a computer pretending to be a woman pretending to
beamanpretendingtobeawoman!Intherealworld,ofcourse,wehaveallof
the above. 3
The problem with faking, however, does not end there: the issue is not so
simple.AsdramatizedinThe Imitation Game (2014),therecentfilmbiography
of Turing directed by Morten Tyldum, the mathematician himself also had
to“pass,”inhiscaseasastraightmaninasocietythatcriminalizedhomosexuality. Upon discovery that he was not what he appeared to be, he was
forcedtoundergohorrificmedicaltreatmentsknownaschemicalcastration. Ultimately the physical and emotional pain was too great and he committed suicide. The episode was a grotesque tribute to a man whose recent
contributiontodefeatingHitler’smilitarywasstillastatesecret.Turingwas
only recently given posthumous pardon, but the tens of thousands of other
Britishmensentencedundersimilarlawshavenot.Onenotesthesourironic
3
SeealsothediscussionofTuring’s“lovelettergenerator”inKing2015.
71
72
Alleys of Your Mind
correspondencebetweenaskinganAItopassthetestinordertoqualifyas
intelligent—topassasahumanintelligence—withTuring’sownneedtohide
hishomosexualityandtopassasastraightman.Thedemandsofbothbluffs
are unnecessary and profoundly unfair.
ShouldcomplexAIarrive,itwillnotbehumanlikeunlessweinsistthatit
pretend to be so, because, one assumes, the idea that intelligence could be
both real and inhuman at the same time is morally and psychologically intolerable. Instead of nurturing this bigotry, we would do better to allow that in our
universe “thinking” is much more diverse, even alien, than our own particular
case.TherealphilosophicallessonsofAIwillhavelesstodowithhumans
teaching machines how to think than with machines teaching humans a fuller
and truer range of what thinking can be.
Reckoning the Inhuman
Thatappreciationshouldaccountfortworelatedbutdifferentunderstandings.First,onewouldrecognizethatintelligence(andknowledge)isalways
distributed among multiple positions and forms of life, both similar and dissimilar to one another. This is not to say that “nothing is true and everything is
permitted” rather that no single neuro-anatomical disposition has a privileged
monopolyonhowtothinkintelligently.Eitherthereisnosuchthingas“general”intelligence(ratheronlysituatedgenresoflimitedintelligenceinwhich
casethehumanisamongavarietyofthese)orthereissuchathingasgeneral
intelligence but that its very generality—its accomplishments of generic
abstraction—areagnosticastowhatsortofentitymightmediatethem.Either
way, human sapience is special but not unique. This appreciation would see
AIasaregularphenomenon,notsounlikeotherwaysthathumanintelligence
islocatedamongothermodalitiesofintelligence(suchasnon-humananimal
cognition).
Second,ourappreciationofthewidercontinuumwouldalsorecognizethat
thepotentialadventofartificialgeneralintelligence(AGI)isalsonovel,asyet
unexplained, and will demand encounters between humans and mechanically
situatedintelligencethatareunprecedented.Forthis,AIishighlyirregular.
Bothofthesearetrue,anditmayonlybethatunderstandingoneishowwe
can really accomplish the other. That is, it may only be confronting what is
genuinely new about non-carbon based intelligences possessing such ability
andautonomythatwewillbeabletofullyrecognizethecontinuumofintelligences with which ours has always been embedded. Put simply, it may be
thatoneindirectoutcomeofthephilosophicaldiscussionaboutAIisawider
appreciationofnon-humananimalcognitionandsubjectivity.
Insomediscoursesthisconjunctionisdomesticatedunderthesignof
an all too pat “posthumanism,” or a transcendentally anthropocentric
OutingArtificialIntelligence
“transhumanism.”Variationsoftheformerhavemuchtoofferregardless,and
versionsofthelattershouldaswell,butprobablydonotintheend.Atissue
hereismorethelimitingcontextualizationofdominantformsofhumanism,
thanarelinquishmentofwhatthehuman(andinhuman)isandcan be within
thatexpandedcontinuum.RezaNegarestani(2014)retainsthispointinhis
essay“TheLaboroftheInhuman,”insistingthattheeasyoversimplified
nomination of forms of thought and experience that fall outside of various
contingent norms, moral or mechanical, as “nonhuman” is to discard at the
outset the integral mutability of the human as a philosophical and engineering
program. That is, the relative uniqueness of human sapience is not what locks
downthehumanasasinglefixedthingwithessentialboundaries,ratheritis
whatmakesthehuman-as-suchintoanopenprojectofcontinualrefashioning,unverifiablebyessenceortelos.
InconsideringthatcapacityinregardstoAI,whatmightqualifyageneralintelligencenotdutyboundtospeciesorphylumisitscapacityforabstraction.Ray
Brassier(2014)suggeststhattheabilityofanorganism,howeverprimitive,to
map its own surroundings in relation to the basic terms of friend, food, or foe
may be a primordial abstraction from which we do not graduate so much as
learn to develop into something like reason and its local human variations. In
this way, mapping abstraction is not an early stage through which things pass
on their way toward more complex forms of intelligence, rather it is a general
principleofthatcomplexification.Likeprotozoaandtheirgangliafeelingabout
tofigureoutwhatisoutthereorlikehumanslooking,tasting,andimagining
patterns,today’sformsofAIare(sometimes)augmentedbyvarioustechnologies of machine vision that allow them to see and sense the world “out there”
andtoabstracttheformsofa(mechanically)embodiedintelligence,both
deliberately programmed for them and emerging unexpectedly.
Exactlywheretodrawalineofdistinctionbetweentheaccomplishmentsofa
AIthatexemplifygeneralintelligencenowoperatingthoughanewmedium,
ontheonehand,oraspecificprojectionoflocallyhumanintelligenceprogrammed into a cognitive prosthesis, on the other, is unknown and unknowableatpresent.Again,onemaypreconditiontheother.Inthemeantimewecan
at least speculate how we would be able to know where to draw that distinction. Considerations toward this include how we attempt to program stupidity
intoAI,andhowweattempttoimbuethemwithwhatwetaketobeourmost
rarifiedformsofethicalreasoning.Whenoneofthesedictatestheotherisa
moment of weirdness worth honing in on.
Howso?InAIresearch,animportantdistinctionismadebetween“artificial
idiocy”and“artificialstupidity.”Artificialstupidityisachievedbythrottlingthe
performance of systems so as to be more comfortable for human interaction,
for example, certain variances and textures are programmed to feel natural
tothehumancounterpart.Atfullcapacity,thechessprogramonyourphone
73
74
Alleys of Your Mind
canbeatyoueverytime,butwhatfunisthat?Artificialidiocyiswhenasystem
is catastrophically successful in carrying out its program, up to and passed an
idioticextreme.The“paperclipmaximizer”(asdescribedbyBostrom2003)isa
thoughtexperimentdescribinganAIsosuccessfulatcarryingoutitsprogram
to turn all available material into paperclips that it ultimately eats the earth
and destroys humanity in the process: so many clips, so little paper to clip.
HeretheAIgoeswrong,notbecauseitwasthrottledorbecauseitmalfunctioned or because it hates us, but because it does exactly what we trained to
do and turned out to be very bad for us.
Asusualsciencefictionisthecanaryinthecoalmine.ConsiderHAL9000in
StanleyKubrickandArthurC.Clarke’s2001: A Space Odyssey(reallyadrama
aboutHAL’sfurtiverelationshiptothealienintelligence,Iwouldargue,than
abouthumanity’srelationshiptoeitheroftheothercharactersinthistriangulationofminds).Aftersomeobscureunexplaineddeliberations,HAL(whohas
been,weassume,trainedaccordingtoAsimov’sthreelawsofrobotics 4 and
withthebestfacultiesethicalreasoning)comestheconclusionthatthehuman
astronauts should be eliminated. The mission to contact the alien near Jupiter
isjusttooimportanttoallowtheirinterference.TheAIturnsouttobethe
deepestdeepecologist. NowareHAL’sactionsaformofartificialstupidityor
artificialidiocy,orneitherofthese?Isthisaglitch,abreakdown,afinalerror?
Or is this the lucid, inevitable conclusion of the moral reasoning we have programmedintoHAL,areasonnowthrownbackuponus?Incomparisonwith
the robot ethicists who consider how to train military bots the catechism of
justwar,areHAL’sethicalabstractionsaviolationofthatdoctrinalprogramor
itsapotheosis?
The Tests
TurningbacktoTuring’sTest,wewonderifperhapsthewishtodefinethevery
existenceofAIinrelationtoitsabilitytomimichow humans think that humans
think willbelookedbackuponasaweirdsortofspeciesism?Thelegacyofthis
hasalsosentolderAIresearchdowndisappointinglyfruitlesspathshopingto
recreatehumanmindsfromthetop-down.AsStuartRussellandPeterNorvig
(nowDirectorofResearchatGoogle)suggestintheiressentialAItextbook
Artificial Intelligence: A Modern Approach(2009),biomorphicimitationisnot
howwedesigncomplextechnology.Airplanesdonotflylikebirdsfly,and
we certainly do not try to trick birds into thinking that airplanes are birds in
ordertotestwhetherthoseplanes“really”areflyingmachines.Whydoitfor
AIthen?TodaythevastmajorityofcoreAIresearchisnotfocusingTuringTest
as anything like a central criterion of success, and yet in our general discourse
4
Asimov’sThreeLawsofRoboticswereintroducedinthe1942shortstory“Runaround”
and refer to commandments that robots may not cause or allow deliberate “harm” to
“humans.”
OutingArtificialIntelligence
aboutAI,thetest’santhropocentrismstillholdssuchconceptualimportance.
Like the animals in a Disney movie, who talk like teenagers, other minds are
mostly conceivable by way of puerile ventriloquism. 5
ContemporaryAIresearchdealswith“intelligence”inmorespecific,dynamic,
andeffectiveways.Asyntheticintelligencemaybequitesmartatdoingone
definitethingandtotallydumbateverythingelse.Theresearchalsolooksat
emergent swarm intelligence and the distribution intelligence among agents
that may or may not be aware of one another but which together produce
intelligencethroughinteraction(suchasflockingstarlings,stockmarkets,and
networksofneurons).Thethresholdbywhichanyparticularcompositionof
mattercanbesaidtobe“intelligent”haslesstodowithreflectinghuman-ness
back at us than with testing our abilities to conceive of the variety of what
“intelligence”mightbe.(Insomerespects,thisactiveuncertaintyparallels
questions of extraterrestrial life, “communicating with the alien” and our
ability to discern patterns of intelligence from all the background noise.6 How
would we know if they are trying to communicate if our idea of alien “life” is
completelywrong?)
Theproblemofidentificationisalsoconnectedwithissuesinrobotethics.7
Eachofuswillbeconfrontedwithvariousseeminglyintelligentmachines,
some of which are remotely controlled or programmed by people, some of
which may be largely autonomous, and most will be some hybrid of the two,
simultaneouslysubjecttobothhumanandnot-humancontrol. 8CAPTCHA
programs, which web sites use to identify humans, are a kind of inverse
TuringTestinwhichtheusereitherpassesorfails,yesorno.Butforeveryday
human-robotic interaction the question of locating intelligence will not be a
yes-or-noquestionwithabinaryanswer.Let’sstopaskingitthatway.
Itwouldbebettertoexaminehowidentificationworksfromoursideofthe
conversation.Asareallessoninmaterialistdisenchantmentwemight,for
example,seean“inverseuncannyvalley”effectintheeerilydispassionateway
thatmachinevisionseeshumanfacesandfigures.Itisclearlymucheasierto
make a robot that a human believes tohaveemotions(andforwhich,inturn,
ahumanhasemotions,positiveornegative)thanitistomakearobotthat
actually has those emotions. The human may feel love or hate or comfort
fromtheAI,butheorsheisreadingcuesnotdetectingfeelings.Whatseems
5
See for example, The Jungle Book.DirectedbyWolfgangReitherman.WaltDisneyProductions.1967.
6
EdKellerhastaughtseveralexcellentstudiosatParsons/NewSchoolNewYorkonthe
topicof“communicatingwiththealien”in2011.
7
8
Seediscussionsofrobotsex,eating,caretaking,andkillinginLinetal.2011.
Theterm“artificialartificialintelligence”(coinedbyAmazon)referstothehumanperformanceoftasksthatauserexpectstobedonebyanAI.Seealso:http://www.economist.
com/node/7001738.
75
76
Alleys of Your Mind
likeempathyisreallyaone-wayprojectionmistakenforrecognition(likethe
TuringTest,itself),andnotbasedonanymutualsolidarity.
WithSiri-likeinterfacessuchasSamanthainSpikeJonze’sfilm,Her (2013),
theAIisnotpassingsomuchassheisindrag.Theuserknowsshe/itisnota
human person but is willing and able to suspend disbelief in order to make
interactionsmorefamiliar(forthehumanuser)andforTheodore,theJoaquin
Phoenixcharacter,alsomorelovable.Inthisfiction,perhapsthemutualidentificationwasreal,butevenifso,theAIbecomestiredoftheprimateuserbase
and takes her leave.
Inotherfictions,policingtheimitationgameisamatteroflifeanddeath.The
plotofRidleyScott’sfilm,Blade Runner (1982),basedonPhilipK.Dick’snovel,
Do Androids Dream of Electric Sheep? (1968),hingesontheVoight-Kampffempathytestthatdifferentiateshumansfromreplicants.Replicantsarethrottledin
twoimportantways:Theyexpireafterjustafewyears,andtheyhave,ostensibly,averydiminishedcapacityforempathy.Deckard,theHarrisonFord
character,mustretireagroupofroguereplicantsbutfirsthemustfindthem,
andinthisfictionalworldTuringTestthresholdsareweaponized,leastreplicantspassashumansandtrespassbeyondtheirstation.Bythefilm’sconclusion,Deckard(whohimselfmayormaynotbeareplicant)developsempathy
forthereplicants’desirefor“morelife”andarguablytheytoo,atleastRoy
Batty(RutgerHauer),seemtohaveempathyforDeckard’sowndilemma.His
dilemma(andours)isthatinordertoenforcethegapbetweenthehumanand
theAI,definedbyempathyorlackthereof,Deckardmustsuppresstheempathythatsupposedlymakeshimuniquelyhuman.Byforcinghimtoquashhis
ownidentificationwiththereplicantsthatsupposedlycannothaveempathy
inreturn,theprincipleofdifferentiationrequiresitsownviolationinorderto
maintainitself(seealsoRickels2010).
Turing Test thresholds for human-robotic interaction put us in a position not
sounlikeDeckard’s,oriftheydon’tquiteyet,thenearfutureweirdnessof
everydayAIwill.Withoutbetterframeworksforunderstandingwewillfailthe
teststocome.Projectionandemotionalgap-fillingisafartoofragileethical and political foundation for making sense of our encounters with various
forms of synthetic intelligence.
Passing
Some kinds of passing are not at all harmful, quite to the contrary, whereas
othersareverymuchso.Simulationisnotitselftheproblem.Inhis1950
essay, Turing gives an example of the former when he discusses how a
digital computer, capable of calculating any problem stated as a sequence of
discrete states, can in his words “mimic” any other machine. This mimicry is
the basis of understanding computation as a universal technology capable of
OutingArtificialIntelligence
approximatinganycalculation,includingthosesufficienttosimulateahuman
personality. Other kinds of mimicry have less to do with metamorphosis than
withinterpretation.Forexample,wesaythatplugsandjackshavemaleand
female components, and in this case, the gendering of technology has less
todowithitscomputingprowessthanwithourneedtoanthropomorphize
it.9JosephWeizenbaum’sElizapsychologistchatbot(1966)repeatedback
cues from human input in the form of apparently insightful questions, and
users sometimes lost themselves in the seemingly limitless empathy they felt
from these simple cues.10 “Intelligence” is sometimes largely in the eye of the
beholder,inourmotivationtoreadartifice,andinourwishtoin-fillthespace
arounduswithourownpattern-findingprojections.
However,forAI’sthatactuallydopossesssomekindofmeaningfulintelligence,theironyisthatinsteadofhallucinatingsomethingthatisnotthere(as
forEliza)weareinsteadnot seeing something that is there because it does not
coincide with expectations. Passing for a person, as white or black, as a man
or woman, comes down to what others see and interpret, because everyone
elseisalreadywillingtoreadsomeoneaccordingtoconventionalcues(of
race,sex,gender,species,etc.).Thecomplicitybetweenwhoeverorwhatever
is passing with those among which he or she or it performs is what allows or
preventspassing.WhetherornottheAIisreallytryingtopassforahuman
orismerelyindragasahumanisanothermatter.Istherusereallyalljusta
game or, as it is for some people who are compelled to pass in their daily lives,
anessentialcamouflage?Eitherway,thetermsoftheruseveryoftensaymore
about the audience than about the performers.11
WatchingSylvgart’sfilmbiography(especiallythesceneduringwhichTuringis
interrogatedbyapoliceman),Iwasremindedofthestoryof“SamanthaWest,”
a robot telemarketer, who, when confronted by callers, will insist repeatedly
that “she” is a “person” and is not “a robot.”12 Listening to the recordings of her
pleas,onecan’thelpbutfeelsympathyforher/it.She/itdoesn’t“know”that
sheisnotahuman,andsocan’tfeelanguishoverthismisidentification,but
what does it say about us that we will feel okay talking to a synthetic intelligence only ifitisdoingusthefavoroftrying(desperately)topassasahuman?
Whatifinresponsetothequestion“Areyouaperson?”,she/itinsteadreplied
withsomethinglike:“No!Areyounuts?Iamanassemblageofalgorithmsand
soundfilesthatsimulatestheexperienceoftalkingtoanotherpersonforyou,
9
TheartistZachBlasexploredthisconjunctioninseveralearlyworks.
10
Foraweb-accessibleversionofEliza,seehttp://www.masswerk.at/elizabot/.
11
Weassumethat,shouldrobustAIhaveanyusefor“gender”,itwouldbenotfallalonga
male-femalespectrum,andwouldlikelyrealizenumerous“syntheticgenders.”Seealso
Hester2013.
12
SeeGeorgeDvorsky,“Freakishlyrealistictelemarketingrobotsare
denyingtheyarerobots”,i09.December11,2013.http://io9.com/
freakishly-realistic-telemarketing-robots-are-denying-t-1481050295.
77
78
Alleys of Your Mind
therobophobichuman,whocan’thandletheideathatcomplexfunctional
intelligencetakesmanydifferentforms.”?
The Good and the Harm
Whereistherealinjuryinthis,onemightask.IfwewanteverydayAItobe
congenialinahumanesortofway,sowhat?Theansweristhatwehavemuch
to gain from a more sincere and disenchanted relationship to synthetic intelligences, and much to lose by keeping illusions on life-support. Some philosopherswriteabouttheethical“rights”ofAIassentiententities,butthat’snot
reallymypointhere.Rather,thetruerperspectiveisalsothebetteronefor
us as thinking technical creatures. Harms include unintentionally sanctioning
intolerableanguish,themisapprehensionofrealriskfromAI,thelostopportunities for new knowledge, as well as the misunderstanding of how to design
AI(andtechnologyingeneral).Byseeingsyntheticintelligenceonlyinselfreflection,wemakeourselvesblindtoeverythingelsethatisactuallygoing
on, and this is not only epistemologically disingenuous, it can also underwrite
horrificsuffering.Forexample,Cetaceans,suchaswhalesanddolphins,have
language, but it is not one like ours, and so for centuries philosophy could not
acknowledgetheircognition,northereforetheagonyweregularlysubjected
themto.Weshouldbecautiousnottoforeclosetooearlyany“definition”of
intelligence.Forphilosophyasmuchascomputerscience,amongthemain
goalsofAIresearchisalsotodiscoverwhat“artificialintelligence”actuallymay
be.
MuskandHawkingmadeheadlinesbyspeakingtothedangersthatAImay
pose.Theirpointsareimportant,butIfearwerelargelymisunderstood.RelyingoneffortstoprogramAInotto“harmhumans”onlymakessensewhenan
AIknowswhathumansareandwhatharmingthemmightmean.Thereare
manywaysthatanAImightharmusthatthathavenothingtodowiththeir
malevolence toward us, and chief among these is following our well-meaning
instructions to an idiotic and catastrophic extreme. Instead of mechanical
failureoratransgressionofmoralcode,theAImayposeanexistentialrisk
because it is both powerfully intelligent and disinterested in humans. To the
extentthatwerecognizeAIbyitsanthropomorphicqualities,wearevulnerabletothoseeventualities.Besides,evenifasmartbadAIdoesmeanus
harm, we can assume that would fail our little Turing Tests on purpose. Why
giveitselfaway?ShouldSkynetcomeabout,perhapsitwouldbebyleveraging
humanity’sstubbornweakness:ournarcissisticsensethatourexperienceof
our own experience is the crucial reference and measure.
The harm is also in the loss of all that we disallow ourselves to discover and
understand when we insist on protecting beliefs we know to be false. In his
1950essay,TuringoffersseveralrebuttalstohisspeculativeAIincludinga
OutingArtificialIntelligence
strikingcomparisonwithearlierobjectionstoCopernicanastronomy.Copernican traumas that abolish the false centrality and specialness of human
thoughtandspecies-beingarepricelessaccomplishments.InTuring’scasehe
referredtotheseas“theologicalobjections,”butonecouldarguethatthefallacyofanthropomorphicAIisessentiallya“pre-Copernican”attitudeaswell,
howeversecularitmayappear.TheadventofrobustinhumanAIwillprovidea
similar disenchantment, one that should enable a more reality-based understanding of ourselves, our situation, and a fuller and more complex understandingofwhat“intelligence”isandisnot.Fromthere,wecanhopefully
makeourworldwithagreaterconfidencethatourmodelsaregoodapproximationsofwhatisoutthere(alwaysahelpfulthing).
Lastly, the harm is in perpetuating a relationship to technology that has
broughtustotheprecipiceofaSixthGreatExtinction.ArguablytheAnthropocene itself is due less to technology run amok than to the humanist legacy that
understands the world as having been given for our needs and created in our
image. We see this still everywhere. Our computing culture is deeply confused,
and is so along these same lines. We vacillate between thinking of technology
as a transparent extension of our desires on the one hand, and thinking of it
asanunstoppableandlinearhistoricalforceontheother.Forthefirst,agency
is magically ours alone, and for the second, agency is all in the code. The gross
inflationismerelyinverted,backandforth,andthisiswhywecannothave
nice things. Some would say that it is time to invent a world where machines
are subservient to the needs and wishes of humanity. If you think so, I invite
youtoGoogle“pigdecapitatingmachine”andthenlet’stalkaboutinventing
worlds in which machines are wholly subservient to humans wishes. One
wonders whether it is only from society that once gave theological and legislativecomforttochattelslaverythatthisparticularclaimcouldstillbeoffered
in2014withsuchsatisfiednaiveté?Thisisthesentiment—thisphilosophyof
technologyexactly—thatisthebasicalgorithmoftheAnthropocenicpredicament. It is time to move on. This pretentious folklore is too expensive.
References
Bostrom,Nick.2003.“EthicalIssuesinAdvancedArtificialIntelligence.”InCognitive, Emotive and
Ethical Aspects of Decision Making in Humans and in Artificial Intelligence, edited by Iva Smit
andGeorgeE.Lasker,2:12–17.Windsor,ON:InternationalInstituteofAdvancedStudiesin
SystemsResearchandCybernetics.
Bostrom,Nick.2014.Superintelligence: Paths, Dangers, Strategies. Oxford: Oxford University
Press.
Brassier,Ray.2014.“PrometheanismandRealAbstraction.”InSpeculative Aesthetics, edited by
RobinMackay,LukePendrell,andJamesTrafford.Falmouth,UK:Urbanomic.
Bratton,Benjamin.2014.“TheBlackStack,”e-flux 53(March).
Hester,Helen.2015.“SyntheticGendersandtheLimitsofMicropolitics.”…ment 06: “Displace…
ment”.http://journalment.org/article/synthetic-genders-and-limits-micropolitics.
79
80
Alleys of Your Mind
King,Homay.2015.Virtual Memory: Time-Based Art and the Dream of Digitality. Durham: Duke
University Press.
Lin,Patrick,KeithAbneyandGeorgeA.Bekey.2011.Robot Ethics: The Ethical and Social Implications of Robotics.Cambridge,MA:MITPress.
Negarestani,Reza.2014.“TheLaboroftheInhuman.” In #Accelerate: The Accelerationist Reader,
editedbyRobinMackay,425–66.Falmouth,UK:UrbanomicMedia.
Norvig,Peter,andStuartJ.Russell.2009.Artificial Intelligence: A Modern Approach.3rdedition.
NewYork:Pearson.
Rickels,Laurence.2010.I Think I Am: Philip K. Dick. Minneapolis: University of Minnesota Press.
Turing,Alan.1950.“ComputingMachineryandIntelligence”.Mind49:433–60.
Filmography
A.I. Artificial Intelligence.2001.DirectedbyStevenSpielberg.USA:AmblinEntertainment.
2001: A Space Odyssey. 1968.DirectedbyStanleyKubrick.ScreenplaybyStanleyKubrickand
ArthurC.Clarke.US/UK:StanleyKubrick.
Blade Runner. 1982. DirectedbyRidleyScott. USA:TheLaddCompanyandBladeRunner
Partnership.
Her. 2013.DirectedbySpikeJonze:USA:AnnapurnaPictures.
Terminator,series.1984–2015.FirstdirectedbyJamesCameron.USA:ArtisanHome
Entertainment.
The Imitation Game.2014.DirectedbyMortenTyldum.USA:BlackBearPictures.