生成式AI展望-斯坦福人类-AI中心 -Generative AI - Perspectives from Stanford HAI - How do you think generative AI will affect your field and society going forward_第1页
生成式AI展望-斯坦福人类-AI中心 -Generative AI - Perspectives from Stanford HAI - How do you think generative AI will affect your field and society going forward_第2页
生成式AI展望-斯坦福人类-AI中心 -Generative AI - Perspectives from Stanford HAI - How do you think generative AI will affect your field and society going forward_第3页
生成式AI展望-斯坦福人类-AI中心 -Generative AI - Perspectives from Stanford HAI - How do you think generative AI will affect your field and society going forward_第4页
生成式AI展望-斯坦福人类-AI中心 -Generative AI - Perspectives from Stanford HAI - How do you think generative AI will affect your field and society going forward_第5页
已阅读5页,还剩36页未读 继续免费阅读

下载本文档

版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领

文档简介

March2023

GenerativeAI:

Perspectivesfrom

StanfordHAI

HowdoyouthinkgenerativeAIwillaffectyourfieldandsocietygoingforward?

1

TableofContents

3

4

Introduction

6

UpendingHealthcare,fromPatientCaretoBilling,CurtLanglotz

AI’sGreatInflectionPoint,Fei-FeiLi

ThePotentialsofSyntheticPatients,RussAltman

7

8

10

PoetryWillNotOptimize:CreativityintheAgeofAI,MicheleElam

13

TheNewCambrianEra:‘ScientificExcitement,Anxiety’,PercyLiang15

ACalltoAugment–NotAutomate–Workers,ErikBrynjolfsson

AnAIWindowintoNature,SuryaGanguli

TheNewToolsofDailyLife,JamesLanday

11

GenerativeAIandtheRuleofLaw,DanielE.Ho

16

18

TheReinventionofWork,ChristopherD.Manning

InEducation,a‘DisasterintheMaking’,RobReich20

SolvingInequalitiesintheEducationSystem,PeterNorvig21

Introduction:

ThecurrentwaveofgenerativeAIisasubsetofartificialintelligencethat,

basedonatextualprompt,generatesnovelcontent.ChatGPTmightwrite

anessay,Midjourneycouldcreatebeautifulillustrations,orMusicLMcould

composeajingle.MostmoderngenerativeAIispoweredby

foundation

models

,orAImodelstrainedonbroaddatausingself-supervisionatscale,

thenadaptedtoawiderangeofdownstreamtasks.

Theopportunitiesthesemodelspresentforourlives,ourcommunities,and

oursocietyarevast,asaretheriskstheypose.Whileontheonehand,they

mayseamlesslycomplementhumanlabor,makingusmoreproductiveand

creative,ontheother,theycouldamplifythebiaswealreadyexperienceor

undermineourtrustofinformation.

Webelievethatinterdisciplinarycollaborationisessentialinensuringthese

technologiesbenefitusall.ThefollowingareperspectivesfromStanford

leadersinmedicine,science,engineering,humanities,andthesocialsciences

onhowgenerativeAImightaffecttheirfieldsandourworld.Somestudy

theimpactoftechnologyonsociety,othersstudyhowtobestapplythese

technologiestoadvancetheirfield,andothershavedevelopedthetechnical

principlesofthealgorithmsthatunderliefoundationmodels.

3

4

AI’sGreatInflectionPoint

Fei-FeiLi,SequoiaCapitalProfessorintheComputerScienceDepartment;DenningCo-DirectorofStanfordHAI

540millionyearsago,thenumberofanimalspeciesexplodedinaveryshorttimeperiod.Therearemanytheoriesastowhathappened,butonehascapturedmyattention:thesuddenonsetandensuingevolutionofvision.Today,visualperceptionisamajorsensorysystemandthehumanmindcanrecognizepatternsintheworldandgeneratemodelsorconceptsbasedonthesepatterns.Endowingmachineswiththesecapabilities,generativecapabilities,hasbeenadreamformanygenerationsofAIscientists.Thereisalonghistoryofalgorithmicattemptsatgenerativemodelswithvaryingdegreesofprogress.In1966,researchersatMITdevelopedthe“SummerVisionProject”toeffectivelyconstruct“asignificantpartofthevisualsystem”withtechnology.Thiswasthebeginningofthefieldofcomputervisionandimagegeneration.

Recently,duetotheprofoundandinterconnectedconceptsofdeeplearningandlargedata,weseemtohavereachedaninflectionpointintheabilityofmachinestogeneratelanguage,image,audio,andmore.WhilebuildingAItoseewhathumanscanseewastheinspirationforcomputervision,weshouldnowbelookingbeyondthistobuildingAItoseewhathumanscan’tsee.HowcanweusegenerativeAItoaugmentourvision?Thoughtheexactfigureisdisputed,

deathsduetomedicalerror

intheU.S.isasignificantproblem.GenerativeAImodelscouldassisthealthcareprovidersinseeingpotentialissuesthattheymayhaveotherwisemissed.Furthermore,ifthemistakesareduetominimalexposuretorare

GenerativeAI:PerspectivesfromStanfordHAI

situations,generativeAIcancreatesimulatedversionsofthisraredatatofurthertraintheAImodelsorthehealthcareprovidersthemselves.

Additionally,beforeweevenstartdeveloping

newgenerativetools,weneedtofocusonwhat

peoplewantfromthesetools.Inarecentprojectto

benchmarkroboticstasksbyourlab,beforeeven

startingtheresearch,theprojectteamdidalarge-

scaleuserstudytoaskpeoplehowmuchtheywould

benefitifarobotdidthesecertaintasksforthem.

Thewinningtaskswerethefocusoftheresearch.

Endowingmachineswith

thesecapabilities,generative

capabilities,hasbeena

dreamformanygenerations

ofAIscientists.

Tofullyrealizethesignificantopportunitythat

generativeAIcreates,weneedtoalsoevaluatethe

associatedrisks.JoyBuolamwiniledastudytitled

“GenderShades,”

whichfoundAIsystemsfrequently

failtorecognizewomenandpeopleofcolor.Study

resultswerepublishedin2018.Wecontinuetosee

similarbiasingenerativeAImodels,specificallyfor

underrepresentedpopulations.

5

AI’sGreatInflectionPoint(cont,d)

GenerativeAI:Perspectives

fromStanfordHAI

Theabilitytodeterminewhetheranimagewas

generatedusingAIisalsoessential.Oursocietyisbuilt

ontrustofcitizenshipandinformation.Ifwecannot

easilydeterminewhetheranimageisAIgenerated,

ourtrustofanyinformationwillerode.Inthiscase,we

needtopayspecialattentiontovulnerablepopulations

thatmaybeparticularlysusceptibletoadversarialuses

ofthistechnology.

Theprogressinamachine’scapabilitytogenerate

contentisveryexciting,asisthepotentialtoexplore

AI’sabilitytoseewhathumansarenotable.But

weneedtobeattentivetothewaysinwhichthese

capabilitieswilldisruptoureverydaylives,our

communities,andourroleasworldcitizens.

6

ThePotentialsofSyntheticPatients

GenerativeAI:Perspectives

fromStanfordHAI

RussAltman,KennethFongProfessorintheSchoolofEngineering;ProfessorofBioengineering,ofGenetics,ofMedicine,andofBiomedicalDataScience;AssociateDirectorof

StanfordHAI

Itisoftendifficulttogetlargenumbersofpatientsinclinicaltrialsanditiscrucialtohavearealisticgroupofpatientswhodonotreceiveatherapyinordertocompareoutcomeswiththosewhodo.ThisisoneareawithinbiomedicalresearchwheregenerativeAIoffersgreatopportunities.GenerativeAIcouldmakeclinicaltrialsmoreefficientbycreating“synthetic”controlpatients(i.e.,fakepatients)usingdatafromrealpatientsandtheirunderlyingattributes(tobecomparedwiththepatientswhoreceivethenewtherapy).Itcouldevengeneratesyntheticoutcomestodescribewhathappenstothesepatientsiftheyareuntreated.Biomedicalresearcherscouldthenusetheoutcomesofrealpatientsexposedtoanewdrugwiththesyntheticstatisticaloutcomesforthesyntheticpatients.Thiscouldmaketrialspotentiallysmaller,faster,andlessexpensive,andthusleadtofasterprogressindeliveringnewdrugsanddiagnosticstocliniciansandtheirpatients.

Inthepast,wehaveused“historicalcontrols”whicharepatientswhodidnothavethebenefitofthenewdrugordiagnostic–andcomparedtheiroutcomestopatientswhoreceivedthenewdrugordiagnostic.Syntheticpatientscouldmatchtherealpatientsmorerealistically;theyarecreatedusingknowledgeofcurrentmedications,diagnostictools,andstandardsofpracticethatwerelikelydifferentinthehistoricalsituation.

Inthesettingofmedicaleducation,generativeAIcouldallowustocreatepatientsthatareveryrealisticandcouldallowmedicalstudentstolearnhowtodetect

diseases.Theabilityforgenerativemodelstocreatemanyvariationsonathemecouldallowstudentstoseemultiplecasesofthesamediseaseandlearnthewaysinwhichthesepatientscanvary.Thiscouldgivethemmoreexperienceinseeingadiseaseandprovideanearlyunlimitedsetofcasesforthemtopracticeiftheyfindthatcertaindiseasesaremorechallengingforthemtorecognizeanddiagnose.Thesesamegenerativemodelscouldalsointeractwiththestudentsandgivethempracticeelicitingsignsandsymptomsthroughconversationalinteraction.

Thiscouldmaketrials

potentiallysmaller,faster,and

lessexpensive,andthuslead

tofasterprogressindelivering

newdrugsanddiagnostics.

Withopportunitycomesworry.Ifsyntheticpatientsaregeneratedfromdatathatdoesnotreflectthepopulationofpatientsreceivingthedrug,thepatientsmaybebiased.Moreworrisome,however,isthateventherealpatientsreceivingthedrugwillnotreflectthefullpopulation,andsosyntheticcontrolscouldjustimprovetheuseofthedrugsforasubsetofpatientsandnotall–leadingtoinequity.

Whilegenerativetechnologiescanbeveryusefulin

acceleratingscientificdiscoveryandprogress,care

mustbetakeninselectingthedatausedtogenerate

patientsandthemodelsmustbeexaminedvery

carefullyforbiasesthatmayleadtodisparateimpact.

7

GenerativeAI:PerspectivesfromStanfordHAI

PatientCaretoBilling

UpendingHealthcare,from

CurtLanglotz,ProfessorofRadiology,ofBiomedicalInformaticsResearch,and

ofBiomedicalDataScience;DirectoroftheCenterforArtificialIntelligencein

MedicineandImaging(AIMI);AssociateDirectorofStanfordHAI

Oneofthebenefitsofourhealthcaresystemisthatpatientscanseeavarietyofspecialistphysicianswhoareexpertsinspecificmedicaldisciplines.Thedownsideofoursystemisthatthesespecialistsoftenaren’tacquaintedwiththepatientstheyareseeing.Imagineaworldinwhichaspecialistyouareseeingforthefirsttimehasalreadyreadasuccinctsummaryofyourhealthcareneeds,createdbygenerativeAI.Duringthepatientvisit,achatbotbasedonafoundationmodelcouldserveasthephysician’sassistanttosupportmoreaccuratediagnosisandtailoredtherapyselection.Agenerativemodelcoulddraftaclinicnoteinrealtimebasedonthephysician-patientinteraction,leavingmoretimeforface-to-facediscussion.Inthebackoffice,generativemodelscouldoptimizeclinicschedulingorsimplifygenerationofmedicalcodesforbilling,diseasesurveillance,andautomatedfollow-upreminders.Thesenewcapabilitiescouldimprovetheaccuracyandefficiencyofpatientcarewhileincreasingpatientengagementandadherencetotherapy.

Recentfederallegislationgivespatientstherighttoaccesstheirentiremedicalrecordindigitalform.Asaresult,patientsareincreasinglyencounteringcomplexclinicaldocumentsthatcontainobscuremedicalterms.Whenapatientreturnshomefromaclinicvisit,afoundationmodelcouldgeneratetailoredpatienteducationmaterialsandexplaintheircareplanattheappropriatereadinglevel.

Machinelearningmodelsinmedicinearecritically

dependentonlargemedicaldatasetsthatcontain

examplesofdisease.

Wehaveshown

howdiffusion

models,atypeoffoundationmodel,canbemodified

tocreaterealisticclinicalimagesfromtextprompts.

Ourresultsdemonstratethatsynthetictrainingdata

producedbythesemodelscanaugmentrealtraining

datatoincreasediagnosticaccuracy.Thisformof

syntheticdatacouldhelpsolvemachinelearning

problemsforwhichtrainingdataisscarce,suchasthe

detectionandtreatmentofuncommondiseases.

Duringthepatientvisit,a

chatbot…couldserveasthe

physician’sassistanttosupport

moreaccuratediagnosisand

tailoredtherapyselection.

Finally,generativeAI’swell-reportedchallenges

withfactualcorrectnessareparticularlyproblematicinmedicine,whereinaccuraciescancauseseriousharm.Recentproblemsinmedicineincludeincorrectdifferentialdiagnosisandinvalidscientificcitations.Weareworkingto

improvethefactualcorrectness

ofmedicalexplanationsfromthesemodelssotheycanachieveanaccuracythatissuitableforsafeclinicaluse.

8

AnAIWindowintoNature

SuryaGanguli,AssociateProfessorofAppliedPhysics;AssociateDirectorof

StanfordHAI

Scientificideasfromthestudyofnatureitself,intheformofnonequilibriumthermodynamicsandthereversaloftheflowoftime,leadtothecreationatStanfordofthefirst

diffusionmodel

,akeykerneloftechnologythatformsthebasisofmanysuccessfulAIgenerativemodelstoday.Now,inavirtuouscycle,AIgenerativemodelsarewellpoisedtodeliverconsiderableinsightsintonatureitself,acrossbiological,physical,andmentalrealms,withbroadimplicationsforsolvingkeysocietalproblems.

Forexample,

generativemodelsofproteins

canallowustoefficientlyexplorethespaceofcomplexthree-dimensionalproteinstructures,therebyaidinginthesearchforproteinswithnovelandusefulfunctions,includingnewefficaciousmedicines.Generative

AIisstartingtobeexploredinthe

quantumrealm

,enablingustoefficientlymodelstronglycorrelatedstatesofelectrons,withthepotentialofadvancingourunderstandingof

materialsscience

and

quantum

chemistry

.Theseadvancescouldinturnleadtothecreationofnewmaterialsandcatalyststhatcouldplayaroleinefficientenergycaptureandstorage.Simplegenerativemodeling,intertwinedwithclassicalnumericalsolvers,hasalsomadekeyadvancesinaccurateandfastlargescale

fluid

mechanicalsimulations

,whichwhenscaledup,couldaidinclimatemodelingandweatherforecasting,therebycontributingtoadeeperunderstandingofourchangingclimateanditsramifications.

GenerativeAI:PerspectivesfromStanfordHAI

Inabeautifulrecursion,thegenerativeAImodelsthat

wehavecreatedcanalsoactasscientificwindows,

notonlyintothephysicalworldbutalsointoour

own

minds

.Forthefirsttime,wehaveAIsystems

thatcanmodelhigh-levelcognitivephenomena

likenaturallanguageandimageunderstanding.ManyneuroscientistsandcognitivescientistshavecomparedtheneuralrepresentationsofbothdeepnetworksandAIgenerativemodelstoneurobiologicalrepresentationsinhumansandanimals,oftenfindingstrikingsimilaritiesacrossmanybrainareas.Examplesincludethe

retina

,the

ventralvisualstream

,

motor

cortex

,

entorhinalcortex

fornavigation,

cortical

language

areas,andneuralgeometriesunderlying

few

shotconceptlearning

.Theoftensimilarstructureofartificialandbiologicalsolutionstogenerativetaskssuggeststheremaybesomecommonprinciplesgoverninghowintelligentsystems,whetherbiologicalorartificial,modelandgeneratecomplexdata.

AIgenerativemodelsarewell

poisedtodeliverconsiderable

insightsintonatureitself,

acrossbiological,physical,

andmentalrealms,with

broadimplicationsforsolving

keysocietalproblems.

9

AnAIWindowintoNature(cont,d)

GenerativeAI:Perspectives

fromStanfordHAI

AnexceedinglyinterestingandprofoundquestionarisesintheforthcomingageofscientificcollaborationbetweenhumansandAIsystemsastheyworktogetherinalooptoanalyzeourcomplexbiological,physical,andmentalworlds:WhatdoesitmeanforahumantoderiveaninterpretableunderstandingofacomplexsystemwhenanAIprovidesasubstantialpartofthatunderstandingthroughpredictivemodels?Issuesregarding

explainableAI

willlikelyrisetotheforewhenafundamentallyhumanscientificendeavor,namelyunderstandingourworld,ispartiallyachievedthroughtheuseofAI.HumanscientistswillnotbecontentwithuninterpretableAI-generatedpredictionsalone.Theywilldesirehumaninterpretableunderstanding,inaddition.

Finally,todreamevenbigger,whiletoday’sgenerativeAIhasaccesstoimmenseglobalscaletrainingdataspanningimages,text,andvideofromtheinternet,itdoesnothavedirectaccesstoourownthoughts,intheformofneuralactivitypatterns.However,thisneednotalwaysbethecase,givenremarkablenewneuroscientificcapacitiestorecordmanyneuronsfromthebrainsofanimalswhiletheyviewimages,aswellastoperformMEG,EEG,andfMRIfromhumansastheyexperiencetheworldthroughrichmultimodalsensoryexperiences.Suchcombinedneuralandreal-worlddatacouldthenpotentiallybeusedtotrainnextgenerationmultimodalfoundationmodelsthatnotonlyunderstandthephysicalworldbutalsounderstandthedirectimpactthephysicalworldhasonourmentalworld,intermsofelicitedneuralactivitypatterns.Whatmightsuchhybridbiological-artificialintelligencesteachusaboutourselves?

Overall,thefutureofgenerativeAIasawindowinto

nature,andtheuseofthiswindowtosolvesocietal

problems,isfullofpromise.Wecertainlydolivein

interestingtimes.

10

TheNewToolsofDailyLife

JamesLanday,AnandRajaramanandVenkyHarinarayanProfessorintheSchoolofEngineeringandProfessorofComputerScience;ViceDirectorofStanfordHAI

GenerativeAI:Perspectives

fromStanfordHAI

Asweallknow,AIistakingtheworldbystorm.Wewillbegintoseemanynewtoolsthataugmentourabilitiesinprofessionalandpersonalactivitiesandworkflows.Imagineasmarttutorthatisalwayspatientandunderstandsthelevelofknowledgethestudenthasatanypointintimeonanysubject.Thesetutorswillnotreplaceteachers,butinsteadwillaugmentthestudentlearningexperience–givingstudentsamorepersonalizedinteraction,focusinginareaswheretheymightbeweaker.

Indesign,pictureatoolthatassistsaprofessionaldesignerbyriffingofftheirinitialdesignideasandhelpingthemexploremoreideasorfillindetailsontheirinitialideas.GenerativeAIwillalsounleashlanguage-basedinterfaces,whetherwrittenorspoken,asamorecommonwayofinteractingwithoureverydaycomputingsystems,especiallywhenonthegoorwhenoureyesandhandsarebusy.ImagineanAlexa,Siri,orGoogleAssistantthatcanactuallyunderstandwhatyouaretryingtodoratherthanjustansweringsimplequeriesabouttheweatherormusic.

WhilegenerativeAIcreatesmanyexcitingopportunities,weknowfrompastAIdeploymentstherearerisks.In2016,anAI-basedsoftwaretoolusedacrossthecountrytopredictifacriminaldefendantwaslikelytoreoffendinthefuturewasshowntobebiasedagainstBlackAmericans.Weneedtoensurewearedesigningthesetoolstogetthemostpositiveoutcomes.Todothis,weneedtodeeplydesignand

analyzethesesystemsattheuser,thecommunity,andsocietallevels.Attheuserlevel,weneedtocreatenewdesignsthataugmentpeoplebyaccountingfortheirexistingworkflowsandcognitiveabilities.Butwecan’tjustdesignfortheuser.Weneedtoconsiderthecommunitythatthesystemimpacts:thefamilies,theinfrastructure,andthelocaleconomy.But,eventhatisnotenough,weneedtoanalyzetheimpactstosocietyatlarge.Weneedtobeabletoforecastwhathappensifthesystembecomesubiquitousandfromthestartdesignmitigationsforpossiblenegativeimpacts.

Changesthatareunderpinned

bygenerativeAIareonlynow

startingtobeimaginedby

designersandtechnologists.

Ouruserinterfacetocomputinghasbeenfairlystatic

overthelast30years.Inthenext5–10years,we

willseearevolutioninhuman-computerinteraction.

ChangesthatareunderpinnedbygenerativeAIare

onlynowstartingtobeimaginedbydesignersand

technologists.Nowisthetimetoensurethatweare

criticallythinkingabouttheuser,thecommunity,and

thesocietalimpacts.

11

PoetryWillNotOptimize:CreativityintheAgeofAI

MicheleElam,WilliamRobertsonCoeProfessorintheSchoolofHumanitiesandSciencesandProfessorofEnglish;AssociateDirectorofStanfordHAI

In2018,theprofessionalartworldwasupendedwhentherenownedChristie’sauctionhouse

soldanAI-

augmentedwork

,“PortraitofEdmondBelamy,”forthewildlyunexpectedsumof$435,000.Thatsale,whichcamewiththetacitimprimaturoftheestablishedartcommunity,generatedmuchgnashingofteethandhand-wringingintheartssectoroverwhatartificialintelligencemeansforthecreativeindustry.

Sincethen,thegeniehaslongfleditslamp:GenerativeAIhasenabledvisualartofeveryknowngenreaswellas

AI-augmentedpoetry

,fiction,

filmscripts

,

music

and

musicals

,

symphonies

,AI-curatedarthistories,andmuchmore.

ThefurorovertheChristie’ssalemaynowseemquaint–itoccurredbeforeDALL-E,LensaAI,ChatGPT,Bing,tonamejustafew–butitheraldedmanyoftoday’sincreasinglyferociousdebatesoverthenatureofcreativityandthefutureofworkforthecreativeindustry.Itanticipatedthecurrenthornet’snestofethical,political,andaestheticconcernsthatgenerativeAIposesforthearts.

Someoftheseconcernshavebeenproductive:GenerativeAIhasencouragedmanyofthosewhoselivelihoods,andinmanycasestheiridentities,dependontheirartisticproductionstoconsideranew–andinnewways–perennialquestionsaboutfoundationalaestheticnormsandvalue:Whatdo

GenerativeAI:PerspectivesfromStanfordHAI

weidentifyas“art”?Whatcountsas“good”art?Isartistrydefinedbyhumanagencyorautomation?Justwhoorwhatcanmake“art”?Andwhodecides?GenerativeAIraisesimportant,thornyquestionsaboutauthenticity,economicvaluation,provenance,creatorcompensation,andcopyright.(TheGettyImageslawsuitagainstStableDiffusionisjustthetipofaniceberg.)Italso,arguably,normalizesextractiveandexploitativeapproachestocreatorsandtheirwork;amplifiesbiasesofeverykind;exacerbatesalreadyurgenteducationalandnationalsecurityconcernsarounddeepfakesandplagiarism,especiallyintheabsenceofcongressionalregulation.

Shouldtheprinciplesof

efficiency,speed,and

so-calledblessingsofscale

applysounequivocallyto

creativeprocesses?Afterall,

poetrydoesnotoptimize.

Perhapsthemostpressingconcern,intermsofnationalsecurity,isthatgenerativeAImighttakeadvantageofthefactthattheartshavealwaysshaped–forgoodorill–thecivicimagination,thatstories,films,plays,imagesshapeourperceptionofourselves,ofourphysicalandsocialrealities.OneofthemostfamousdisagreementsbetweenPlatoandhisstudentAristotlewasoverthepotentiallydangerouspowerof

12

PoetryWillNotOptimize:CreativityintheAgeofAI(contd)

GenerativeAI:Perspectives

fromStanfordHAI

poesytoinfluencebeliefsandworldviews.Thispoweriswhyfascistregimesfirstdoawaywiththeartistsandintellectuals:becausetheyholdswayoverourmindsandthusouractions.

SomeclaimthatgenerativeAIisdemocratizingaccesstocreativeexpressiontothosetraditionallybarredfromitbylackofstatusorwealth.Butdoclaimsto“democratization”and“access”function,ineffect,asindustrycoverfor

rushingacommercialapplication

“intothewild”

(i.e.,tothepublic)withoutthetime-intensiveworkofensuringethicalguardrails?

IsAIsimplyaneutralifpowerfulassistivetoolforthearts–akintopen,paintbrush,orphotography?Isit“

blitzscaling

”creativity,orinEmadMostaque’s

choice

description

,relievingour“creativelyconstipated”worldwithAItechnologiesthatcanhaveusall

温馨提示

  • 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
  • 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
  • 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
  • 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
  • 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
  • 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
  • 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。

评论

0/150

提交评论