Analysis:Artificialintelligenceisadeviloranangel?(ResearchpapersDownloadNews)May11,1997,chesschampionGarryKasparovandsupercomputer‘DeepBlue’inthechess.Kasparovintheyearofthe‘man-machinewar’toonewinandtwodefeatsandthreerecorddefeat.March10thisyear,‘AlfaGo,’theExecutivesub-humanHuangShijie(frontleft)Laziinthegame.Onthatday,AI‘AlphaGo’beatSouthKoreanplayersLeeSedol9ptotalscoreto2-0lead.March12,worldchampionLiShishiofKoreaGowithGoogle’sartificialintelligenceprogramAlphaGothirdchesswillbeginafterShishihastwosuccessivedefeat.SomepeopleLiShiShicalled‘TheLastSamurai’thathewouldbecomethetragichumanandrobotchess‘lastemperor.’AlphaGobecauseAIprogramsinthisareaknownas‘thefortressofhumanwisdom,’Gosuccessivevictory,itseemsthatfurtherconfirmstheAIhasbeendevelopedtoeasilysurpassthe1humanlevelstage.Theywillbeanangeloradevil?Whatisthestatusofthedevelopmentofartificialintelligence?LargesamplecalculationwinsupportAlphaGoAccordingAlphaGoresearchwerepresented,ithasbeenabletohave‘powerof3361’kindofsituationinthisultraGochessgamedifficulttowin,becauseitbreaksthroughthetraditionalprogram,setuptwosetstomimichumanthinkingdepthwaytheneuralnetwork,thefirstcalled‘networkstrategy’,whichallowsacomputerprogramtolearnhumanplayersdismount,pickoutthewinningpercentagehavemorechess,chessabandonsignificantdifference,sothatthetotalamountofcalculationcanbemaintainedatacontrolledtherange;theotheriscalledthevalueofnetworks,primarilyusedtoreducethedepthofthesearch,itwillnotsearchforagameonceallthesteps,butwhileatthesidethenexttenstepswerecalculated,sothereisasignificantreductionintheamountofcalculation.Infact,theneuralnetworkisnotAlphaGounique.AccordingtoSu,directorofresearchincomputingandBigDatacognitiveIBMChinaResearchInstitute2introducedthismodelfirstappearedin1943,butbecauseofthespeedofthecomputeritisdifficulttomeetitshugecomputationalsuffered‘criticized’andthereforesilencealongtime.Inrecentyears,therapiddevelopmentofcomputertechnology,asubstantialincreaseinspeed,themodelwasre-radiatevitality.Evenso,withShishichess,AlphaGoCPUwillstillbeincreasedtomorethan1200inordertomeetitshugeamountofcalculation.AccordingtooneAlphaGoresearchersdeveloped-DavidXierWeiintroduced,thisprogramwillplaychesswithhimself,ordinarypeoplemaybeabletounderonethousandayear,butcanAlphaGounderthreemillionperdaychess,throughalotofexercise,itmayfailtoabandontheprogram,selectedfinefine,thisistheso-called‘deeplearning’capability,thatis,byalargesamplesizechessgamechess,chessconstantlypickandchoosethebestprogramandpreserved.Thehumanbrainisstillabiggapbetweenlearning‘Bigdataonthedevelopmentofartificialintelligenceisa‘tricky’,providesanumberofdata-depthstudy.’In2015ChinaConferenceonArtificial3Intelligence,theChineseAcademyofSciencesTanTieniudepthstudyhasbeendoneontheinterpretation.Hesaiditwouldbejustlikehumansinformedaftersomeexperienceasmachinelearningalsorequiresavarietyofcontent.Justthishugeamountofcontentdata,requiredtostudythehumanbrainisnotanorderofmagnitude.So,learnthehumanbrainartificialintelligencehasaverybigdifference.‘GiveachildlookatAppleonthecard,hecanrecognizetheapple,butthemachinetorecognizeanapple,ApplemayneedtoputallthephotosontheInternetagainrecognize,identifyallthefeaturesofApple,willitbepossibletoidentifyasuccessfulApple.‘Susaid.Atthesametime,verydifferentfromthehumanbrainandartificialintelligenceisal...