INTERNATIONALSECURITYANDSTRATEGYSTUDIESCENTERFORINTERNATIONALSECURITYANDSTRATEGYTSINGHUAUNIVERSITYByCISSWorkingGrouponArti?cialIntelligenceGlossaryResearchWeapongenerallyreferstothecombatinstrumentsanddevicesusedtoattackordefendthemselves,whichareusedtodirectlykilltheenemy’seffectiveforces,destroytheenemy’sfacilities,ordefendagainstattacks,andcanbedividedintooffensiveweaponsanddefensiveweapons.Weaponsandequipment:Weaponsandequipmentisageneraltermforweapons,weaponsystemsandrelatedothermilitarytechnicalequipmentusedbycombatantstocarryoutandsupportcombatoperations.Itisanimportantmaterialbasisforconductingwar,animportantforceforgeneratingcombateffectivenessofthearmy,andanimportantsupportforcompletingvariousmilitaryoperations①.Armaments:Armamentsprimarilyinclude?rearms,artillery,ammunition,tacticalmissiles,opticalinstruments(reconnaissanceequipmentsuchasbinoculars,periscopes,gunsights,anti-aircraftcommandscopes,etc.),aimingdevices,angle-measuringinstruments,general-purposeradars,commandinstruments,mobilepowerstations,riotcontrolequipment,coldweapons②,andarangeofotherweaponsandequipmentusedformilitary①Thepeople,sLiberationArmymilitaryChina,《中國人民解放軍軍語》，中國②ChinaMilitaryEncyclopaediaEditorialCommittee.(2016).ChinaMilitaryEncyclopaedia.EncyclopaediaofChinaPublishingHouse.purposes.Ammunition:Ammunitionreferstoitemswhichcontaingunpowder,explosives,orother?llingmaterialsthatcanin?ictdamageontargetsoraccomplishothertacticaltasks.Thisincludesshells,artilleryshells,handgrenades,ri?egrenades,aerialbombs,rockets,missiles,torpedoes,navalmines,landmines,etc.,aswellassalutesandammunitionusedfornon-militarypurposessuchashuntingandshootingsports.③Equipment:Weapons,uniforms,equipment,technicalforces,etc.whicharmiesequipwith.④2.WeaponssystemTheWeaponssystemreferstotheorderlycombinationofseveralfunctionallyinterrelatedweapons,technicalequipment,etc.collaboratestocompleteanorganicwholeofcertaincombattasks.⑤Unacceptableautonomouslethalweaponsystems:AccordingtoChina’sWorkingPaperon“LethalAutonomousWeaponSystems”,Chinabelievesthat“unacceptable”autonomousweaponsystemsshouldmeet,butnotbelimitedto,thefollowing?vebasiccharacteristics:a.Lethality:Equippedwithpayloadsandmeanssuf?cienttocausedeath.b.Fullautonomy:Nohumaninterventionorcontrolduringtheentiremissionprocess.③ChinaMilitaryEncyclopaediaEditorialCommittee.(2016).ChinaMilitaryEncyclopaedia.EncyclopaediaofChinaPublishingHouse.④ChinaMilitaryEncyclopaediaEditorialCommittee.(2016).ChinaMilitaryEncyclopaedia.EncyclopaediaofChinaPublishingHouse.EconomyDictionary),(Beijing:MilitarySciencePress,200c.Irreversibility:Nomeanstoterminateonceactivated.d.Indiscriminatekilling:Automaticallyexecutinglethaltasksregardlessofconditions,scenarios,andtargets.e.Evolution:Capableofautonomouslearningandfunctionalexpansionthroughinteractionwiththeenvironment,exceedinghumanprediction.“Acceptable”autonomousweaponsystems:Thesehaveahighdegreeofautonomybutshouldalwaysremainunderhumancontrol,besafely,credibly,reliably,andcontrollablyused,andallowhumaninterventionatanytime.Inmilitaryoperations,theymustcomplywiththebasicprinciplesofinternationalhumanitarianlaw,includingdistinction,proportionality,andprecaution.⑥3.Non-lethalWeaponsNon-lethalweaponsarewidelyusedbythemilitaryandotherpowerfuldepartmentsintheduty,emergency,anti-terrorismandothertasks,whichcandisablethetargetquicklyandthedisablingconsequenceshavethemaximumprobabilityofreversibilityunderthepremiseofnon-lethal.Non-lethalweaponsprovidetheabilitytooperatewithaminimumofforce,controllingviolenceandpreventingunnecessarycollateralcasualtiesanddestruction.⑦⑥中國關(guān)于“致命性自主武器系統(tǒng)”問題的工作文件，2022年7月，https:///wp-content/uploads/2022/07/Working-Paper-of-the-Peoples-Republic-of-China-on-Lethal-Autonomous-Weapons-SystemsChinese.pdf⑦趙陜東，馬永忠.非致命武器與警用器材[M].北京：兵器工業(yè)出版社，4.RobotsRobots:Amachinethatautomaticallyperformswork.Itcanaccepthumancommands,runpre-programmedinstructions,oroperateaccordingtotheprinciplesmadebyartificialintelligencetechnology.Themaintasksofrobotsaretoassistorreplacehumanwork,suchasinmanufacturing,construction,ordangeroustasks.⑧5.IntelligentweaponsIntelligentweapons:Refertoweapons,weaponsystems,andothermilitarytechnicalequipmentequippedwithadvancedtechnologiessuchasartificialintelligence(AI),sensors,andautomationsystems,whichprovidethemwithacertainlevelofautonomousdecision-making,targetrecognition,tracking,andattackingcapabilitiestoimplementandsupportcombatactions.Theseweaponscanenhancecombatefficiencyandaccuracytodifferentdegreesincomplexbattle?eldenvironments,reducingrelianceonhumanoperation.Theso-calledintelligenceisanextensionanddevelopmentofhumancapabilities,asapartofthehuman-machineinteractiveenvironment.Humans,astheprimaryagentsofmilitaryandintelligentactivities,shouldretainthenecessarycapabilitytointervenewhenintelligentweaponsachieverelativelyhigherlevelsofautonomy.⑨ChineseTermsinComputerScienceandTechnology(ThirdEdition),Beijing:SciencePress,2018,p.460.⑨Compiledbasedonthebasiclogicofintelligentweapons,incorporatingtheviewsofChinesescholarsoverthepast20years.6.Autonomousweapons(systems)Intermsofautonomy,themainpurposeofautonomyistoreducethedependenceonhumanandexternalresourcesinmilitaryoperations,toimprovetheadaptabilitytocomplexdynamicenvironmentandsurvivabilityonthebattle?elds,andthustobetteraccomplishthebattle?eldmissionsassignedbyhumanbeings.Theuseofrelevantweaponssystemsshouldberegulatedspecificallyaccordingtodifferentscenariosanddifferentdegreesofautonomouscapabilities.Ifautonomouscapabilitiesarenotusedinthekillingchain(e.g.,dronesforintelligencecollectionandreconnaissance),evenifsomeweaponssystemshaveahighdegreeofautonomy,theautonomyinthoseweaponswillnotleadtohumanitarianconcerns.Thekillingchainofweaponssystemsincludesobservation,orientation,decision,actionandothercriticallinks.Weaponssystemswithautonomousfunctionsatcertainlinksmaynotnecessarilycauseindiscriminateeffects.Therefore,generalprohibitionsorrestrictionsmayunderminethelegitimatedefensecapabilitiesofcountries,andeventheirrightstouserelevanttechnologiespeacefully.Paritiesshouldconsiderclassifyingautonomousweaponssystemsintotwocategories:unacceptableandacceptable,andprohibittheunacceptablepartsandregulatetheacceptableparts,soastoensurerelevantweaponssystemsaresecure,reliable,manageableandinlinewithinternationalhumanitarianlawandotherapplicableinternationallaw.⑩⑩WorkingPaperofthePeople’sRepublicofChinaonLethalAutonomousWeaponsSystems,9August2022,/Convention_on_Certain_Conventional_Weapons_-_Group_of_Governmental_Experts_(2022)/CCW-GGE.1-2022-——WorkingPaperofthePeople,sRepublicofChinaonLethalAutonomous7.Intelligentcluster/swarmsystemsArti?cialintelligencecluster⑩:Acollectionofarti?cialintelligence(AI)computingfunctionalunitsthataresubjecttouni?edcontrol.Note1:TheAIcomputingfunctionalunitmaycontainAIaccelerationprocessors,AIservers,AIaccelerationmodules,etc.Note2:WhencomposedofAIservers,anAIclustermaybereferredtoasanAIservercluster,wheretheAIserversmaybereferredtoasnodes.SwarmIntelligencel:Swarmintelligence(SI)isalsoreferredtoascollectiveintelligence(CI)…Agroupcomposedofmanysimpleindividualsissaidtohave“swarmintelligence”ifitsmemberscanaccomplishacollectivetaskthroughsimplecooperation…Thecoreofswarmintelligenceisthatagroupofmanysimpleindividualscanachieveamorecomplexfunctionandcompleteamorecomplextaskthroughsimplecooperation.Swarmintelligenceprovidesthepossibilityofsolvingcomplexdistributedproblemswithoutcentralizedcontrolandwithlimitedglobalinformationandmodels.8.UnmannedsystemAnunmannedsystemisanelectro-mechanicalsystemcapableofexertingitspowertoperformdesignatedmissionswithnohumanoperatoronboard.⑩StateAdministrationforMarketRegulation(SAMR),StandardizationAdministrationofChina(SAC),StandardofthePeople,sRepublicofChina:InformationArtificialIntelligenceTerminology(GB/T41867-2022),ReleasedonOctober12,2022(EffectiveonMay1,2023),lZhangGuohui&WenXiaoyu,SwarmInPress,2022,pp.1-3.Thesystemoftenequippedwithnecessarydataprocessors,sensors,automaticcontrol,andcommunicationssystems.9.Unmanned(combat)platformUnmannedcombatplatformreferstoatypeofweaponsystemtoperformcombatmissionswithouthumanoperatoronboard.10.Unmannedaerialplatforms(drones,unmannedaerialvehicles(UAV),RemotelyPilotedAircraftSystems(RPAS))Unmannedaerialvehicles(UAVs)areaircraftthatdonothaveanonboardpilotandareequippedwiththeirownpowersystems.Remotelypilotedaircraft:Anunmannedaircraftoperatedbyaremotecontrolstation.Remotelypilotedaircraftareasubcategoryofunmannedaircraft.m11.UnmannedgroundplatformsUnmannedgroundplatformsisanimportantforceinlandcombat,capableofperformingvarioustasksinhigh-riskenvironmentswithouthumanoperatoronboard.12.UnmannedsurfacevesselAnunmannedsurfacevesselisakindofsurfacevesselthatiscapableofautonomousnavigationandperformingrelevanttasksbycarryingamissionload.13.UnmannedunderwatervehicleUnmannedunderwatervehicleisavehicleautomaticallyperformitstasksunderwaterwithouthumanonboard.mCivilUnmannedAircraftSystemsAirTraf?cManagementMeasures,/zwgk/zcwj/bj14.Arti?cialIntelligenceArtificialintelligencereferstotheories,methods,technologiesandapplicationsystemsthatutilizedigitalcomputersordigitalcomputer-controlledmachinestosimulate,extendandexpandhumanintelligence,perceivetheenvironment,acquireknowledge,andusethatknowledgetoobtainoptimalresults.——China,sWhitePaperonArti?cialIntelligenceStandardization(2018)15.MachineLearningItreferstoaprocessthatoptimizestheparametersofamodelthroughcomputationaltechniquestomakethemodel’sbehaviourreflectdataorexperience.nAsasubject,machinelearningreferstoaninter-disciplinethatinvolvesstatistics,systemidentification,approximationtheory,neuralnetworks,optimizationtheory,computerscience,brainscienceandmanyotherfields.Itstudieshowcomputerscansimulateorrealizehumanlearningbehaviourstoacquirenewknowledgeorskills,re-organizeexistingknowledgestructuresothatitcancontinuouslyimproveitsownperformance,anditisthecoreofarti?cialintelligencetechnology.o16.DeeplearningDeeplearning:amethodofcreatingrichhierarchicalrepresentationbytrainingneuralnetworkswithalargenumberofhiddenlayersn“Informationtechnology-Artificialintelligence-Terminology”(GB/T41867-2022)issuedonMay1,2023,ChineseElectronicsStandardizationInstitute,https://std./gb/search/gbDetailed?id=71F77o中國電子技術(shù)標(biāo)準(zhǔn)化研究院ChineseElectronicsStandardizationInstitute.(2018).Whitepaperonarti?cialintelligencestandardization,人工智能標(biāo)準(zhǔn)化白皮書Note:Deeplearningisasubsetofmachinelearning.“Informationtechnology-Artificialintelligence-Terminology”(GB/T41867-2022)issuedonMay1,2023Deeplearning:Basedonlearningapproaches,machinelearningcanbedividedintotraditionalmachinelearninganddeeplearning.Deeplearningisalearningapproachforbuildingdeepstructuralmodels.Typicalexamplesofdeeplearningalgorithmsincludedeepbeliefnetworks,convolutionalneuralnetworks,restrictedBoltzmannmachineandrecurrentneuralnetworks.Deeplearningisalsocalleddeepneuralnetworks(referringtoneuralnetworkswithmorethanthreelayers).Itisanemergingfieldinmachinelearningstudies,proposedbyHintonetal.in2006.Derivedfrommulti-layerneuralnetwork,deeplearninginessenceprovidesanapproachtocombiningfeaturerepresentationandlearningintoone.—WhitePaperonArti?cialIntelligenceStandardization(2018)17.NeuralnetworksNeuralnetworks:Anetworkofbasicprocessingelementsconnectedbyweightedlinkswithadjustableweights,whereeachcellproducesavaluebyapplyinganon-linearfunctiontoitsinputvalues,andtransmitsthisvaluetoothercellsorrepresentsitasanoutputvalue.Note1:Whilesomeneuralnetworksaimtosimulatethefunctionofneuronsinthenervoussystem,mostneuralnetworksareusedinarti?cialintelligencetoimplementconnectivitymodels.—NationalStandard(GB/T5271.34-2006)Neuralnetwork:anetworkconsistingofoneormorelayersofneurons,connectedbyweightedlinkswithadjustedweights,thatreceivesinputdataandproducesoutput.Note1:Neuralnetworkisaprominentexampleoftheconnectionistapproach.Note2:Althoughthedesignofneuralnetworkwasoriginallyinspiredbythefunctionofbiologicalneurons,mostneuralnetworkstudiesnolongerfollowthisinspiration.—NationalStandard(GB/T41867-2022)18.AutonomouscontrolAutonomouscontrolistheabilitytoorganicallycombinethesensingability,decision-makingability,synergisticabilityandactionabilityoftheautonomouscontrolsystemwithouthumanintervention,tomakeself-decisionaccordingtocertaincontrolstrategiesinanunstructuredenvironment,andtocontinuouslyexecuteaseriesofcontrolfunctionstoaccomplishpredeterminedtargets.Automaticcontrolp:referstoanautonomousoperationofmachines,equipment,orspecificoperatingstatesorparametersinproductionprocessesenabledbysupplementaryequipmentordevices,accordingtoapredeterminedpatternwithoutdirecthumanintervention.Autonomouscontrolsystemq:Asystemcapableofachievingself-stabilization,self-regulation,self-optimization,self-adaptation,andotherfunctionsthroughtheintegrationofcomputersystems,servoactuators,sensingandfeedbackmechanisms,stabilizationandcorrectioncomponents,digital-to-analogandanalog-to-digitalconversion,time-sharingsamplingchannels,andsupportingsoftware.pChinaNationalCommitteeforTerminologyinScienChineseTermsinComputerScienceandTechnology(ThirdEdition),Beijing:SciencePress,2018,p.451.qWuyiXiong,JiafaZhou,MingxinZhuo,XinguangLi,andJichangXu,MilitaryDictionary,FirstVolume,Beijing:19.AILifecycleTheAIlifecyclereferstotheentireprocessfromtheconceptphaseofanAIprojecttotheendofthedeploymentandmaintenancephase.IntheWhitePaperonStandardizationofArtificialIntelligence(2021edition),ChinacitedthedefinitionofAIlifecyclesysteminISO/IEC22989ArtificialIntelligenceConceptsandTerminology,mentionedbySub-CommitteeonArti?cialIntelligenceoftheFirstJointTechnicalCommitteeoftheInternationalOrganizationforStandardizationandtheInternationalElectrotechnicalOrganization(ISO/IECJTC1/SC42):AIsystemlifecyclemodelincludinginitialphase,designanddevelopment,verificationandvalidation,deployment,operationandmonitoring,re-evaluationandexitphases.ThislifecyclemodelisderivedfromthesystemandsoftwareengineeringsystemlifecycleandemphasizesthecharacteristicsoftheAIfield,includingdevelopmentandoperation,traceability,transparencyandexplainability,securityandprivacy,riskrmanagement,governance,etc.r20.ChainofcommandIntheChinesecontext,thetermmostsimilartothecommandchainisthecommandrelationship.Thechainofcommandorcommandrelationshipreferstoacontinuouslineofauthoritythatextendsfromthetopofthecommandstructuretothebasiccombatunit.usuallyfromthehigherlevelrChina,sArtificialIntelligenceStandardizationWhitePaper,NationalArtificialIntelligenceStandardizationGeneroftheNationalInformationTechnologyStandardizationCommitttothelowerleveltoformarelationshipofcommandandsubordination.Accordingtooperationalrequirementsorforceformationadjustments,commandrelationshipsmaybetransformed,i.e.,theoriginalcommandrelationshipischangedandanewcommandrelationshipisformed.21.CommandandControl(System)Commandandcontrol:Commandandcontrolisanimportantactionthroughoutmilitaryoperation.Commandersuseorders,planning,instructions,andcommandandcontrolsystemstodeploy,coordinate,superviseandrestricttroopsandoperationsinordertocompletethedecidedorgivenoperationalmissions.Jointoperationcommandandcontrolincludesmonitoringandcoordinationofvariouscombatactivities,aswellasregulationandconstraintsontheexerciseoftheauthorityandresponsibilityofvariousactors.CommandandControlSystem:Itisasystemthatguaranteesthecommandersandcommandorganstoexercisecommand,controlandrestrictionsonthetroops,operationsandweaponsystems.Accordingtotheapplicationlevel,commandandcontrolsystemscanbecategorizedintostrategic,operationalandtacticaltypes.22.DecisionpointAsimilarconcepttothedecisionpointintheChinesecontextisdecision-making,whichreferstothecommanderreceivedtheorderfromtheuppercommander,understandtheintentionsoftheuppercommander,combinedwiththejudgmentofthecombatand?eldsituationandmadethedecisionontheactionsofhisorhertroops.23.MeaningfulhumancontrolThisisnotanof?cialmilitaryterm;Itisaconceptdevelopedbyscholarsintheirpapers.OurunderstandingisthatautonomousweaponsystemsorlethalAI-enabledweaponsystemsarecon?rmedbyhumansatcriticalmissionpoints,suchastargetveri?cation,targeting,andbeforeexecutingastrike,ratherthanrelyingexclusivelyonthepredictionorjudgmentofsensors,autonomousweaponsystems,orlethalAIweaponsystems.24.Human-machineinteractionHuman-machineinteraction(orhuman-computerinteraction,HCI)referstotheinteractiveactivitiesbetweenhumanusersandthecomputersystem(acombinationofsoftwareandhardwareunits)involvingtheinputandoutputofinformation,withtheaimofcollaborativelyachievingtasksthatmeetshumanneeds.Inthecontextofarti?cialintelligencetechnologies,HCIaimstoresearchanddesigntheinteractivemodebetweenhumansandcomputersystems,improvetheinteracitveef?ciencyofhuman-machine-environmentthroughsmarttechnologies,andmakethecommunicationbetweenhumansandcomputersystemsmorenaturalandrealistic.ThosetechnologiesintheareaofHCIincludebutnotlimittonaturallanguageprocessing,speechrecognition,gesturerecognition,affectivecomputing,virtualrealityandaugmentedreality.25.TrustworthyAIApeople-centeredapproachshouldbeupheldindevelopingAI,withthegoalofincreasingthewellbeingofhumanityandonthepremiseofensuringsocialsecurityandrespectingtherightsandinterestsofhumanity,ensuring“AIforgood”andprotectingpersonalprivacyanddatasecurity,sothatAIalwaysdevelopsinawaythatisbene?cialtohumancivilization.Tothisend,weshouldcontinuouslyimprovethetransparency,stability,explainability,predictability,fairnessandreliabilityofAIsystems,increasedataauthenticityandaccuracy,ensurethatAItechnologiescanbereviewed,monitored,andtraced,andremainAIalwaysunderhumancontrol.Transparency:TheactivitiesofAIsystemssuchasthedecision-makingmechanisms,operationalprocesses,datause,andbehavioraloutcomescanbeclearlyunderstood,explainedandtrustedbyhumans.Thiscanmainlybere?ectedinthetransparencyofalgorithms,dataanddecisions,aimingtoachievetheexplainability,predictability,traceability,reviewabilityandaccountabilityinAItechnologies.TransparencyhelpsallstakeholdersclearlyunderstandhowAIsystemswork,butitshallnotbeharmfultonationalinterestsandtheinterestsofallstakeholders.Explainability:AIsystemsareabletorevealtheirreasoning,decisionmaking,andpredictionprocessestohumansinaclear,understandable,andmeaningfulmannerthatenableshumanstounderstandandtrusttheactionstakenorconclusionsdrawnbythosesystems.ExplainabilityfocusesonprovidingunderstandablereasonsforthedecisionsmadebyAIsystems,ratherthanattemptingtoarguethat“thecourseofactionthatwastakenwasnecessarilyoptimal”.Traceability:Thedecisiondatasets,processingandoutputsateachstageoftheentirelifecycleoftheAIsystemscanberecorded,trackedandtraced,sothatthedecisionoutcomesoftheAIsystemscanbeunderstoodandtracedbyhumans.Reliability:AIsystemsareabletoconsistentlyandeffectivelyoperateinalignmentwithitsintendedgoalsandachievedesiredoutcomesundervariousconditions,includingbothacertainlevelofresistancetomaliciousattacksandafallbackmechanismintheeventofsevereproblems.Predictability:HumansareabletopredictandunderstandtheoutputsandbehaviorsofAIsystemsunderdifferentconditions,enablingthemtomakereliableassumptionsabouttheexpectedoutputs,whichrequirestheoutcomesofAIsystemstobeaccurate,reliableandrepeatable,whileeffectivelyresistingthevulnerabilitiesandmaliciousattacks.Controllability:Alsoknownas“humancontrol”,meanshumanjudgmentshallbeintegratedintoprocessesofresearching,developing,deployingandemployingAIsystems,whichenableshumanstodirect,supervise,evaluate,intervene,suspendorterminatethebehaviorsofAIsystemseffectively,ensuringthattheycomplywithethical,legalandsafetystandards.Additionally,accountabilitymechanismsforAIshallbeestablishedtoensurehumanasthe?nalsubjectofresponsibility.26.ChainofDecisionTargetidenti?cation:Basedonthetargetinformationprovidedbyvarioussensors,theAIsystemdeterminesthetarget’senvironment,categoryandtype,aswellasthetarget’sfriendorfoestatus.Targetcon?rmation:TheAIsystemveri?esandvalidatestarget’sidentityandstatusfollowingthetargetidentificationtoensuretheaccuracyandreliabilityoftheidentificationresults,andtoconfirmthatthetargetisalegitimateobjectofattackunderthelawsofwarandrulesofengagement.OperationAuthorization:Thespeci?coperationalauthorityisgrantedtotheAIsystembasedonclearrules,purposes,andrestrictions,enablingittomakedecisions,takeactions,orperformtaskswithinade?nedscopeunderhumansupervisionandcontrol,ensuringthatthetasksareaccomp