Dataset Viewer
Auto-converted to Parquet Duplicate
text
stringlengths
0
6.44k
--- Start of Document artificial_intelligence_0 ---
The Governance of Physical Artificial Intelligence
Yingbo Li*1,*, Anamaria-Beatrice Spulber2, and Yucong Duan*3,*
1HainanUniversity
2Visionogy
3HainanUniversity
*Correspondingauthor: xslwen@outlook.com,duanyucong@hotmail.com
ABSTRACT
Physicalartificialintelligencecanprovetobeoneofthemostimportantchallengesoftheartificialintelligence. Thegovernance
3202
ofphysicalartificialintelligencewoulddefineitsresponsibleintelligentapplicationinthesociety.
rpA Introduction
ArtificialIntelligence(AI)hasgrowntobethefundamentaltechnologyintoday’sworld. Overthelastfewyears,notonlyhas
AIbeenpopularlyappliedintypicalAIapplicationsofinformationandsignalprocessingsuchasNaturalLanguageProcessing
6
(NLP),butithasalsoempoweredalltheotherindustriessuchashealthcareandrobotics. MiriyevandKovac1 proposedto
]IA.sc[ definetheAIusedinRoboticsasPhysicalArtificialIntelligence(PAI)becausePAIinteractswiththephysicalworld,contrary
tothenotionoftraditionalDigitalArtificialIntelligence(DAI)appliedindigitalinformationprocessing. Fromthisperspective,
weproposetoextendthenotionofPAItoamuchwiderdomaintoalsoincludeInternetofThings(IoT),orautomaticdriving
cars. Tothebestofourknowledge,mostresearchonAIgovernanceislimitedtothedomainofDAI,sointhepresentpaperwe
proposetooutlinethegovernanceframeworkofPAI.
1v42920.4032:viXra
The application of PAI
IntheproposedconceptofPAIbyMiriyevandMirko1,PAIreferstothetypicalrobotsystem. While,weproposetoextendthe
conceptofPAItocoverallpotentialapplicationswiththebuilt-inAIperceivingandinteractingbetweenthecyberspaceandthe
physicalworld. BesidestherobotsystemwiththeAIworkinginanintegratedandlimitphysicalenvironment,thedistributed
intelligentsystemwithAIcapabilityisthetypicalDistributedPAI.AsshowninFig. 1,PAIcouldbeappliedinandinclude
multipledistributedindustries,suchasIoT,self-drivingcars,agriculture,healthcareandlogistics.
WeproposetoclassifyPAIintotwooverlappedkindsasshowninFig.1:IndependentPAIandDistributedPAI.Independent
PAIreferstotheintelligentdeviceandtherobot1. DistributedPAIbecomesmoreandmorepopularwhentheedgecomputing2
ismatureandeverydeviceisconnectedtothenetworkinthewiderspace. IoTandedgecomputingaretypicalDistributedPAI
subdomains. SinceitispopularforeveryintelligentsystemtobeonlineandindividualunitsinDistributedPAIhavestrong
computingcapabilitiesnow,IndependentPAIandDistributedPAIwilloverlapinmultipleapplications3.
TheIoTisatypicaldistributedsystemwithaspatialdistributionthatrangesfromasmallspacesuchasaroomtoawider
areasuchasacity. TheIoTisformedofvarioussensorsthatcapturethesignalsandchangesinthephysicalworld. ItsAI
powercouldhappeninbothserversideandtheedgeside. BasedontheAIanalysis,IoTcoulddirectlyorindirectlymake
predictionsinthecyberspacetoinfluencethephysicalworld. Forexample,aself-drivingcarneedstofirstperceiverealtime
roadsituationsandconnecttotheInternetfornavigation,thenadjustthedrivingbehavior. Theagricultureisoneofthemost
successfulPAIapplications. Thesensorsintheagricultureincludingcameras,temperaturemeter,hygrometer,etc,monitorthe
growthoftheplantandpredict,forexample,theoptimalpesticideinterventionandthebestharvestingtime. Inthehealthcare
industry,families,nursinghomesorhospitalscouldusethebiologicalsensorsandthechemicalsensorstomonitorapatient
andpredictpotentialriskssuchasfallingorunstablesituationsthroughamonitoringcenter. The"lastmile"istheexpensive
andhardprobleminthelogisticindustry. AtypicaldistributedAIapplicationcouldhelpwithdeliverytasksthroughdelivery
robotsanddronesconnectedtoandcommandedbythecenterserver. Anotherexampleistheautomaticsortingrobotthathas
beenusedinthesortingcenterofthelogistics. ThegeneralframeworkofDistributedPAIisdescribedinFig. 2.
DAImimicsthebraincapabilityoflogicalthinkingandinductioninhumanbrain,toprocessthedataandsignalsperceived
byhumaneyesandears. Thehumanbrainisonlyresponsibleforprocessingthesignalsandtransmittingcommandstoother
partsofthebodysuchasmovement,visionperception,soundperception,digestionandetc. Bycomparison,IndividualPAIis
likeanindividualhumanbody,whileDistributedPAIfurtherextendstheAIcapabilitiesjustlikethehumansocietyiscomposed
1
Figure1. TheapplicationsofDistributedPAI
ofmultiplehumans.
DuetotheprobabilisticdataprocessingofthecurrentDAI,decisionsfromDAIareunsubstantial,notbeingabletoreduce
theuncertaintyofapplications. OntheothersideofpromotingtheapplicationofDAI,theexplainablepropertyofcurrentDAI
posesincreasinggovernancechallengesagainstnegativeandmaliciouspracticeofDAI,includingdatabiasesandAIfrauds,
etc. ExplainableAItriestoexplainandunderstandtheinternaloperatingmechanismofAI.DistributedPAIinteractswiththe
physicalworldwithamuchlargerspatialareaandconsequentlyaccumulatesBigDataAIfootprintswhichincludesmuch
longerinteractiontrajectoriescrossingcyberspaceandphysicalworld. SoexplainableAIappliedinDistributedPAIhasthe
advantagetorevealtheinternalmechanismofDAIandtheintegratedhuman-cyber-physicalsocialphenomena.
PAIneedstocombinemultiplestreamsofinformationincludingmaterials,temperature,vision,sound,etc,crossingmultiple
modalsfrommultiplesensorsasshowninFig.1. Throughthemixofthemultimodalinformation,PAIbuildscompetitive
capabilitythatusesmultipletypesofinformationwhichallowsittomakebetterdecisionandbetterprecision,inthecontext
ofimprecisedatacollection,inconsistentinformationandincompleteknowledgescatteringovervariousabstractionlevels.
Thevarioussourcesofdataandinformationbringmultiplekindsofdata,whichoutperformasinglesourceofdata,tomake
real-timedecisionsandpredictions. ThisisasignificantfeatureofPAI.
The governance of PAI
ThegovernanceofDAIhasbeenchallengedbyresearchersfromafairnesstosocialimpactperspective. DAIhasbeenfacing
thechallengesofriskandgovernance4from,butnotlimitedto,thefollowingaspects: 1)Thestorageandtransfersecurity;2)
Thefakedata;3)Thesocialprivacy;4)Thebiasofthesex,gender,andracebecauseofthelimitedtrainingdatasets.
Asaconsequencetothemulti-sourceperceptionandmulti-dimensioninteractioninamuchlargerspace,PAI,especially
DistributedPAI,bringsmoreuncertaintyandriskfromthesocialimpacttothetechnologyinfluence:
• Theexistenceproblem. PAIespeciallyDistributedPAIsuchasIoTneedsmultiplekindsofsensorstointeractwiththe
physicalworld. IfPAIisdistributedinalimitedspacesuchasafactory,itwillnotencounterchallengingregulation
problemsbecauseitisinaninternalspace. However,ifthespaceisextendedtoalargerspacesuchasacitywhichisnot
underoneuniqueregulation,PAIwillfaceproblemsofsocialregulations.
• Thedataorganizationproblem. Themultiplesourcesofdatafromthephysicalworldofawiderspacewillincrease
thestructuralconstructionandintegrationcomplexityofthedataandinformation. TheKnowledgeGraphcouldbethe
potentialsolutionfortheinformationorganizationinthehierarchicalstructure.
2/4
Figure2. DistributedPAI
• CannikinLaw. ThedevelopmentofPAIdependsonatleast5disciplinesofmaterialsscience,mechanicalengineering,
chemistry,biologyandcomputerscience. Therefore,theslowerdevelopmentofonedisciplinewillcausetheproblemof
CannikinlawandprohibitthedevelopmentofPAI.
• Thesocialacceptance. SimilartothedilemmaofDAI,theubiquitousapplicationofPAIwillcausetheworryofthe
societyregardingtheincreaseofunemployment,broadeningofthegapinincome,theshrinkingofprivacyspace,etc.
TheacceptancefrommultipleaspectsofthelawandsocietywillinfluencetheapplicationofPAIintheresearch,the
industryandthesociety. So,theacceptanceofthesocietyandthecorrespondinglegislationisapotentialfactorforPAI.
WeillustratesabovegovernanceproblemsofPAIinFigure3. ThedevelopmentofPAIhastoresolvethesefourproblems.
Conclusion
WehavesuggestedextendingthenotionofPAItoalargerphysicalspacewiththedistributedapplicationswiththenotionof
DistributedPAI.ThespatialvarietyofDistributedPAIcouldvaryfromaroomspacetoacityspace,whileitsapplications
includeIoT,agriculture,andsoforth. SincePAI,especiallyDistributedPAI,perceivesandinteractswithdifferentphysical
entitiesindifferentspaces,thegovernanceissuesincludingtheexistenceproblemhasbeenchallenged. Wehaveputforwarda
3/4
Figure3. PAIgovernanceproblems
frameworkofgovernanceproblemsofPAIhoweverthisisopentofurtherdiscussionssinceitisaresearchtopicapplyingnot
onlytotheresearchbutalsothedevelopmentofthewholehumansociety.
References
1. Miriyev,Aslan,andMirkoKovacˇ."Skillsforphysicalartificialintelligence."NatureMachineIntelligence2.11(2020):
658-660.
End of preview. Expand in Data Studio

No dataset card yet

Downloads last month
9