1

(VERSIONPRELIMINAIREPOURDISCUSSION)

CHINA’DIGITALTRANSFORMATION

WHYISARTIFICIALINTELLIGENCEAPRIORITYFORCHINESER&D?

GuilhemFabre,UniversitéPaulValéryMontpellier3,IRIECEA740

BRICSSeminarFMSH

guilhem.fabre@univ-montp3.fr*

I.CHINA’SR&DPOLICY

Innovation requires human talent, financing, the capacity to discover,developanddistributenewproductsongrowingmarketswiththehelpofintellectual property. Eco-systems of innovation may vary according tonational economic conditions, i.e., the share of public versus privateenterprises, the degree of confidence to intellectual property (I.P), thedegree of interactions between the industrial-military complex, which isoften at the vanguardof innovation, and the civil sectors, the insertionofthe local economy in the global system. All these factors depend largelyfrom national policies, which are the framework of the innovationdynamics.

Besides, innovation is generally concomitant with creativity. Economic,social,institutionalinnovationstendtoenhanceartisticcreativity.Thebesthistoricalexampleisthe«quattrocento»(XVthcentury)period,whenItalybecame the centre of institutional, scientific, and artistic innovation. Thevalue given by education to the techniques of measurement and to thegeometric concepts, allowed the people of quattrocento to develop apeculiarattentiontothestructuresofforms,tothevolumeandsuperficiesofthebodies.ThisunionofscienceandartopenedthewaytotheEuropeanRenaissance.

In the case of China, the picture is full of contrasts. On one hand, theartisticscene,especiallysince2000,hasbecomeglobal, intheplasticarts,literature and film industry, with renowned artists who have sometimesemigratedabroad.Ontheotherhand,itisquitechallengingtomeasuretheactual degree of innovation. The 2006medium to long term plan for thedevelopment of Science and Technology (2006-2020) has embraced theconcept of «indigenous innovation» （自主创新) which guidespolicy tothisday.In2010,thegovernmentidentifiedsevenstrategicemerging

2

* I thank my colleagues Fajwel Fogel (SANCARE) & Stephane Grumbach (INRIA) fortheirveryusefulcommentsandrevisionofthispaper.

industries(SEI)supposedtoreceivespecialsupport.Finally,the13thfive-yearplan,adoptedinmarch2016,iscompletedbytheMadeinChina2025plan(2015)andInternetPlusinitiative.

All these plans are inspired by a techno-nationalist point of view, byimportsubstitutionandexportpromotionstrategies:themarketisopenedto attract foreign technology, to enhance technological progress andmanagement learning from foreign enterprises. Once Chinese companiesaremakingsignificantprogress inclosing the technologygap, theChinesegovernement seeks to increase theirmarket shareby erectingbarriers toforeignmarketsactivities1.(FollowingFigures)Inthesametime,Chinesepolicymakerssupportastrongwaveofoutwardinvestmentincutting-edgetechnologiesintheU.SandEurope.Despitethedifficultiesofthispath,theaimisnotonlytobetechnologicallyindependent,butalsotochallengetheU.Sscientificaswellastechnologicalsupremacy,afterhavingbypassedtheU.Sasthefirstindustrialcountryin2010.

1JostWubekke&al.,MadeinChina2025,Themakingofahigh-techsuperpowerandconsequencesforindustrializedcountries,MERICSPapersonChinan°2,Dec.2016,p.55:https://www.merics.org/fileadmin/user_upload/downloads/MPOC/MPOC_Made_in_China_2025/MPOC_No.2_MadeinChina_2025.pdf

3

However,themeasureoftheirsuccess,isquitechallengingfromamacro-economic point of view, as there is a lag between inputs and outputs,accordingtoarecentstudybyScottKennedy2.

Ifwefocusonhumancapital,educationishighlyvaluedinChinesecultureand society,with theConfucianheritage. Since theRealLeapForward ineducation, launched in the beginning of this century, China annuallygraduatestheworld’slargestpoolofscientistandengineers.Althoughtheyare not especially encouraged to be creative in school, or to developinitiatives in the national system of education, China has sentmillions ofstudentsabroad,attractforeigninvestorswhotendtosharetechnologyandcreateR&DcentresandtraintheirChineseemployees.Chineseentrerprisescreate representative offices in the top high-tech hubs like Silicon Valleyand acquire technology through mergers and acquisitions (M&A), hiringtalent from their western competitors, and also from Taiwan, which haslongbeenaworldhubininformationtechnology(I.T),andhasinvestedalotinthecontinentwitharoundonemillionsTaiwanese,mainlymanagersandengineers,nowlivinginChina.

Ifwefocusonfunding,with2.1percentofitsGDPinvestedinR&D,Chinahas alreadybypassed theE.U in absolute terms, from10.9US $ billion in2000toUS$232billionin2016,andisnowcatchingupwiththeU.S.A.Butthis does not mean that these massive inputs translate in comparableoutputs.Basicresearchrepresentjustover5percent,appliedresearch10.8percent and 84.2 percent of funding is directed towards development(2015). Most of the funding goes to companies (77 %), which decideaccording to commercial competitiveness criteria. By avoiding spendingonbasic researchand foundational technologies,whose shareareat leasttwoto threetimeshigher inotherdevelopedcountries(14percent in theU.S, 17 percent inRussia, 11 percent in Japan in 20153), income is less aresult of new technologies and more a result of new applications orbusinessmodels.

2ScottKennedy,TheFatTechDragon,BenchmarkingChina’sInnovationDrive,CenterForStrategicandInternationalStudies,August2017:https://www.csis.org/analysis/fat-tech-dragon3http://www.oecd.org/innovation/inno/researchanddevelopmentstatisticsrds.htmR&DbysectorofperformanceandtypeofR&D.

4

Thesamemaybesaidforotherindicatorsofinnovation.Thenumberofpatentsforexampleisveryhighonanationalbasis,withonly21%oninventions,butlowfromaninternationalperspective.Thedomesticaswellas international value of patents is still quite low. Triadic patents,successfulintheU.S,JapanandE.Uarein2014,thelatestavailableyear,sixtimeslessthanforapplicantsfromtheU.SandJapan(2582versus14,994for theUSand17,121for Japan). In2015,patent licensinghasgeneratedonly US $ 1.75 billion of local revenues and international sales of patentrights,US$1.38billion,whileIPlicensingrevenueintheU.SgeneratedforthesameyearUS$115.2billion.IfweturntothebalanceofI.Preceiptstopayments, China is still a massive importer of I.P: US $ 22 billion ofpaymentsversusUS$1billionofreceipts,accordingtotheIMF.

The ratio of receipts to payment in international licensing (FollowingFigure) indicate that China is still primarily an assembler andmanufacturerintheglobalsupplychain(FollowingFigure)4.

4ScottKennedy,op.cit.

5

StanShi’sfamoussmilecurveisstillvalidforalotofhightechexports:

Ifwenowfocusonoutput,thecommercialperformanceofinnovation,theofficial vision is that the indigenous innovation works very well withindicators such as the National Innovation Index measuring essentiallyquantitativeinputs(R&Dintensity,humancapital,etc.),andtheScienceand

6

TechnologyContributionRate, constructed insuchaway that itwouldbeimpossibletofallintheabsenceofastrongeconomiccontraction.Abetterapproach,accordingtoScottKennedy,istolookatthehightech’scontribution to manufacturing value added, rising from 8.8 percent in2010 to 14.2 percent in 2014, and the share of high tech exports fromdomestic companies, which has climbed from 15 percent in 2000 to 19percentin2015,withdivergingtendenciesbetweensectors:

7

Ifwe turn toamicro-economicpointofview,PriceWaterhouseCooper(PWC)ranksinitsannualreport«GlobalInnovation1000»5thecompaniesaccording to their investment in R&D and through a survey of businessexecutives asking them to identify the most innovative companies. Byeither method no Chinese company appears among the top 25 of thisgrading. The first Chinese Company in terms of R&D is Alibaba, whichoccupy the 56th rank. However, according to PWC,which underlines theglobaltendencytoeconomicnationalism,thereisnostatisticalrelationshipbetweenR&Dspendingandcommercialmetrics(intermsofsalesgrowth,profits,orshareholdersreturns).This is precisely the case of Chinese companies. The cross-findingbetween macro and micro data shows that prominent high tech Chinesecompanies(FollowingFigureinScottKennedy’sReport)havenotrisenontechnological innovation, but that their growth is the result of twomainfactors:

1. Incrementalinnovationsinapplicationsandcustomizationduetothescale of the market, from mobile phones (Vivo, Oppo, Xiaomi) andlaptops (Lenovo), to telecom equipment (Huawei, ZTE), homeappliances (Midea, Haier), E-commerce (Alibaba, Tencent, JD) or

5https://www.strategyand.pwc.com/innovation1000;https://www.strategyand.pwc.com/media/file/Strategy-The-2017-Global-Innovation-1000-Study.xlsx

8

search (Baidu). The scale also allows to pursue a low margin highvolumepricing strategyand togetanunprecedentedamountofbigdatafromcustomers.

2. The second factor is relative to entry barriers. Chinese companies,have forged their competitiveness under a wide display ofprotectionist measures provided by state policies, not only in thepublic procurement system (Goldwind, Trina Solar,DJI Innovations,ChinaRailwayCorp.),butalsointhefinancial,insurance,telecomandinternetmarkets,whicharebasedonprofitableoligopolies.

Scott Kennedy’s conclusions is that China’s drive to innovation may behighlyinefficientinthesensethatthereisalagbetweentopdownpoliciesandmassive inputs and themodest scale of inventions and technologicalbreakthrough.Butheaddsthatitmaynotbeaproblemaslargemistakesandwastemayfacilitate learningandinvestmentdowntheroad,andthatChina’s size,not comparable to JapanandotherNPI countries requires tojudgethecountry’sperformancenotonlyontheconsequencesforChinesecompanies,butforentiresectorsandtheglobaleconomy.Here is thepoint,as the lagbetweenmassive inputsand low immediatebreakthroughinnovationsortechnologicalGreatLeapdoesnotmakesensein a development aswell as historical perspective. From a developmentperspective, we can not deny, for instance that the high speed trainprogram,basedon retroengineering (FollowingFigure,Zhonghuwangluntan)implementedbeforeandduringtheglobalfinancialcrisis,hasbeenasuccessstory,sinceChinahasbuiltinlessthantenyearsmorehighspeedtrainslines(20.000km)thanthewholeworld6.

6«Chinasetsitssightondominatingsunriseindustries»,TheEconomist,Sept.23,2017.

9

Source:ZhonghuawangluntanAnother example of theChinese catchup capacity is themultiplication ofChinese participation in nuclear industry projects (in U.K Romania,Pakistan,Turkey,Argentina),inassociationwithU.S,CanadianandFrenchfirms:

10

Source:MercatorInstituteforChinaStudiesIf we rely on an historical perspective, technology and innovations maymovequicklyfromacountrytoanother.Italywasthecenterofinnovationandartisticcreationinthe15thcentury,butPortugalandSpainbecametheglobalpowersinthe16thbecausetheirstateshadthepoliticalwillandthefinancialcapacitytofinancefleetsandrecruitthemostenterprisingsailorssuch as the Genovese Christophe Colomb. At the end of the nineteenthcentury, Europe discovered the automobile, the cinema and the aviation,but theUnitedStates,withthescaleof theirmarket,andtheirmanagerialcapacity were able to develop Taylorism, Fordism, Hollywood and airtransportonanunprecedentedscale.

FROMINVENTIONSTOAPPLICATIONS:ARTIFICIALINTELLIGENCEASAPRIORITY

Thus, all the normative talk about the right path of innovationsmay bemisleading, inventions may not be a precondition for technologicalbreakthrough.Applicationsseemessential,inhightechindustriesaswellasinthenewfrontiersofscience,suchasQuantumTechnology.ArecentU.SCongressHearing(SubcomitteeonResearchandTechnology&Subcomitteeon Energy) has shown that although the United States retains globalleadership in the theoretical physics that underpins quantum computing

11

and related technologies, the countrymay be slipping behind others, andespecially China in developing the quantum applications – programmingknow-how,developmentofnationalsecurityandcommercialapplications.7Artificial intelligence may be the best example of the priority ofapplications versus inventions. Artificial Intelligence is the idea, firstlaunched by Alan Turing in 1950, that a computer system can performfunctionsassociatedwiththehumanmind, inwhathassincebeenknownastheTuringtest:canacomputercommunicatewellenoughtopersuadeahuman that it, too, is a human? 8 The term «intelligence» may bemisleading, as the real question seems to decouple successful problemresolution from any need to be intelligent, or to cite the US computerscientist John Mc Carthy and colleagues, who invented the term anddevelopedthefieldin1955,«theproblemistakentobethatofmakingamachinebehaveinwaysthatwouldbecalledintelligentifahumanweresobehaving». To paraphrase Clausewitz, with Luciano Floridi, from theDigital Ethics Lab, of Oxford University, «A.I is the continuation ofintelligencebyothermeans»9.Intelligencedoesnotmeanhereconsciousthoughtordeductivereasoningor «understanding», but the capacity of computer systems to performtasks normally demanding human intelligence, such as visual perception,voice recognition, translations between languages, the ability to makeassociations, or in an action domain to evaluate a situation and to actappropriately. In the past, computer systems had to be programmed toexecutedefinedtasks.Butnow,MachineLearning(M.L),atypeofartificialintelligence,providescomputerstheabilitytolearnwithoutbeingexplicitlyreprogrammed.MachineLearningfocusesoncomputerprogramsthatcanteachthemselvestoadapttonewdatainputs,andgrow.ManyM.Lsystemsare already developped for commercial use, such as Apple Siri, smartspeakers such asAmazon’s virtual assistantAlexa,which has partnershipwith more than 100 third party services, Google Home, which allowsshoppers to complete orders with 50 participating Google Express

7https://science.house.gov/legislation/hearings/american-leadership-quantum-technology

8A.M.Turing,«Computingmachineryandintelligence»,Mind,volume49,n°236,October1950.9LucianoFloridi,«Afallacythatwillhinderadvancesinartificialintelligence»,FinancialTimes,June1,2017.

12

retailers10. A real-time recognition of handwritten Chinese charactersdevelopedbyAppleisspanningalargeinventoryof30.000characters11.Other applications are still in the development stage, like the driverlesscar:thefirstcomputerchipdesignedtocontrolacompletelydriverlesscar,developed by US chipmakerNvidia,will be in operation for a trial in thestreetsofGermanyin201812.ThemostAdvancedformofMachineLearningisDeepLearning(D.L), abreakthroughinnovationinartificialintelligencefunctionsthatimitatestheworking of the human brain in processing data and creating patterns foruse in decisionmaking. AlphaGo, a self-taught computer,made sensationwhenitscreatorsfromDeepMind,aLondonbasedCo.acquiredbyGoogle,beatChina’sworld championGoplayer,Ke Jie inmay2017. Since, a newdevelopment, Alpha GoZero, dispenses form human expertise and startsjustbyknowing the rules andobjectivesof the game. It learns toplaybyplayingagainstitself.,andhasdefeatedthepreviouslypublishedversionofAlphaGoby100games to zero13. In the future,DeepMindwill focuson«developing advanced general algorithms that could one day helpscientists as they tackle some of our most complex problems, such asfindingnewcuresfordiseases»14.As the new generation of A.I applications is based on the formation ofdigitization,leadingsectorsindigitaltendtobeleadingsectorsinA.I.Hightech,telecomandfinancialserviceshavelonghistoryofdigitalinvestmentand are generallywell placed inA.I.Automotive and assembly, leaders intheuseofrobots,withnearly40%oftheworld’sstock,(FollowingFigure)are also developing A.Iwith the driverless car project. But A.I can highlyenhancetheproductivityandgrowthofothersectors15.

10«Google,AppleandAmazonsoundoutconsumersonsmartspeakers»,FinancialTimes,October7,2017;DominiqueBarton&al.ArtificialIntelligence,Thenextdigitalfrontier,MacKinseyGlobalInstituteDiscussionPaper,June2017,p.44.11AppleMachineLearningJournal,2017/09:https://machinelearning.apple.com/https://machinelearning.apple.com/2017/09/12/handwriting.html12«FirstchipsforfullcontrolofdriverlesscarstohittheStreets»FinancialTimes,Ocotber10,2017:https://www.ft.com/content/99121e62-adbf-11e7-aab9-abaa44b1e13013DavidSilver&alii,«MasteringtheGameofGowithouthumanknowledge»,Nature,550,354-359,19/10/2017:https://www.nature.com/articles/nature24270?error=cookies_not_supported&code=7a186490-bb5f-4f41-8082-a6eb7017455814LucianoFloridi,FinancialTimes,articlecited,June1,2017.15For furtherdetailsandacompletestudy, see:ArtificialIntelligence,Thenextdigitalfrontier, MacKinsey Global Institute Discussion Paper, June 2017.https://www.mckinsey.com/~/media/McKinsey/Industries/Advanced%20Electronics

13

In the retail sector, A.I may anticipate demand trends, automatewarehouses and store operations, (See Video,FinancialTimes, August 25,2017,https://www.youtube.com/watch?v=j9nv_0MOOcI;similarmachinesare used in China in the depots of Tmall, part of Alibaba Co.http://www.alizila.com/video/smart-robots-alibabas-intelligent-warehouse/), optimize pricing and personnalize promotions, offerassistancewithvirtualagent,andcompletelast-miledeliverybydrones.

Source:FinancialTimesInelectricutilities,A.Imayenhancesupplyanddemandpredictionwithsmartmeters,maximize theuseof renewablepower,optimizepreventivemaintenanceandpricing,andreduceenergywasteandtheft.DigitalWindFarms, launched by General Electric, could boost a wind farm’s energyproduction by 20 percent and China’s State Grid Corporation announced

/Our%20Insights/How%20artificial%20intelligence%20can%20deliver%20real%20value%20to%20companies/MGI-Artificial-Intelligence-Discussion-paper.ashx

14

plans to investUS$45billion insmartgrid technologies,andanother$45billionbetween2016and202016.MachineLearningcanalsoreduceenergylossesintransmissionanddistribution.In manufacturing, A.I can improve product design and efficiency,automatesupplierassessment, automateassembly lines, reducematerialdelivery time, predict sales of maintenance services or in the Aerospaceindustry,optimizeflightandplanningroutesandenhanceflightandpilotstraining.In the health sector, A.I may bring more intelligence to imaging andradiology,helptopredictdiseasesanddiscovernewtherapies,identifyhighrisks patients groups, automate and optimize hospital operations,diagnostic test andmake them faster andmore accurate, adapt therapiesanddrugformulationtopatients.In education, A.I may help to predict job market demand, automateteachers routine tasks, identify early disengagement signs, personalizelearningandbuildstudentsself-awareness.A lot of these applications are already in use such as image recognitionwiththeautomatictagsonFacebookorbyvideosurveillancecameras,therecommandation of films or musics (Netflix, Spotfire), the predictivemaintenance,etc.17.All these developments show thatwe are notwitnessing the boom andbustcyclethatmarkedthegrowingroleofI.Tatthebeginningofthe21stcentury. We are at the beginning of a new age. A.I is powered by newentrants,suchasUBER,whohavealreadydisruptedwholesectorssuchasthetaxiindustry.Dataisattheheartofthedisruptionsoccurringacrosseconomies,andisactuallyrecognizedasacriticalcorporateasset. It is the fuel thatpowersA.I. Theworld creates actually 2.2 exabytes or 2.2 billion of gigabytes ofdataeveryday.Moredatameansmoreexamples,moreinsightandabetteraccuracyforMachineLearningsystems,whichcanreducecomputererrorrates, insomeapplicationsuchasvisualrecognition,toaboutthesameasthe rate for humans. The scale of data determines the output of A.Iprograms.The second condition for A.I development is computer power. PowerfulGraphic Processing Units (GPU), Integrated Circuits (I.C) originallydevelopedforvideogames,canprocess images40to80timesfasterthanthe version available in 2013. Following the promotion of a nationalIntegratedCircuitIndustry, in2014,withinvestmentofmorethan$US20billion, in June 2016, China broke record with the world’s fastest

16Ibidem,p.44.17CommentsbyFajwelFogel,SANCARE.

15

supercomputer, Sunway TaihuLight, without US developed processors18.Chinaisnowplanningtoboostitscomputerpowertenfoldwithinacoupleofyears,bybuildinganexascalecomputer,whichwillequalthecalculationpower of all the world’s top 500 computers combined, andwill help thecountry ‘s maritime expansion by processing data collected from theworld’socean19.ThethirdconditionforAIdevelopmentisthevarietyandsophisticationofalgorithms. All these trends are growing rapidly, with a fast developingR&D investment, despite the scarcity of talent : on the 150 million U.Sworkers,only250.000aredatascientists.Hightechcompaniesareactivelybuying start-ups, not only to acquire technology and clients, but to hiretalents,asthepooloftrueexpertsisreduced.

CHINA’sA.ICRAZEThe OECD Science and Technology Industry Scoreboard 2017, providescomparative data on China’s digital transformation, but does not give anideaofthescaleofthemarket,andthedevelopmentofAI20.Withitsgianttechcompanies,China isoneof the leadingglobalhubofA.Idevelopmentand seeks dominance at the world level. Beijing’s blueprint for A.I,publishedinJuly2017,aimstocreatea$US150billionindustryby203021(FollowingFigure).

18DominiqueBarton&al.ArtificialIntelligence:ImplicationsforChina,McKinseyGlobalInstitute,April2017,p.8.19Stephen Chen, «The world’s next fastest supercomputer will help boost China’sgrowingseapower»,SouthChinaMorningPost,Hongkong,23August2017.20http://www.keepeek.com/Digital-Asset-Management/oecd/science-and-technology/oecd-science-technology-and-industry-scoreboard-2017_9789264268821-en#.WiGDurZ7RZ0;(Knowledgeeconomiesanddigitaltransformation)21 Financial Times, October 16, 2017; http://www.gov.cn/zhengce/content/2017-07/20/content_5211996.htm in Chinese : State Council , The next generation AIDevelopmentPlan.

16

ChinaandtheU.SarenowtheworldleadersinA.I.WhileChinaranksfirstforwidelycitedA.Irelatedpapers,USandUKresearchremainsmoreinfluential(FollowingFigure)

17

Source:ArtificialIntelligence,ImplicationsforChina,McKinseyGlobalInstitute,April2017.China ‘s A.I ecosystem is not as vibrant as the United States,which hascreatedalotmoreAIstartupcompaniesthanChina(FollowingFigures):

18

Source:Artificial Intelligence: Implications forChina,McKinseyGlobal Institute,April2017.In2016, theUS,moreattractive, absorbedaround66percentofexternalinvestment (Venture Capital, Private Equity andMergers & Acquisitions),Chinawasadistantsecondat17percent.TheSanFranciscoBayAreaandSilicon Valley were attracting alone 40 percent of global externalinvestment, followed by Boston, New York, and Beijing and Shenzhen AItecheco-systems22.MajorM&Adealswhicharealsoawaytorecruittalents,weremademainlybyU.Stechgiantsbetween2011andthebeginningof201723(Figure):

Source: McKinsey Global Institute, Artificial Intelligence, Implications for China, April2017

22McKinseyGlobalInstitute,ArtificialIntelligence,Thenextdigitalfrontier,op.cit.23DominiqueBarton&al.:ArtificialIntelligence:ImplicationsforChina,McKinseyGlobalInstitute,DiscussionPaper,April2017.

19

ButthesituationseemstochangeveryquicklyasChinesestart-upsleadthepack for software and self-driving cars, in terms of investment 24 :

InVentureCapitalInvestment,theyaresecondtotheU.Sinmostleadingtechnologies(FollowingFigure)

24MarkBergenandDavidRamli,«China’sPlanforWorldDominationinAIIsn’tSOCrazyAfterAll»,Bloomberg,August15,2017.

20

Source: JonathanWoetzel & al. China’sDigital Economy: A LeadingGlobal Force, McKinseyGlobalInstituteDiscussionPaper,August2017.Thesecretof thisrapidcatch-upmaybethe ideathatmarketsize is thekey driver of innovation25. As of December 2016, China had 731 millionInternetusers(morethanE.UandU.Scombined),95percentofwhichareSmartphonesusers.DigitalnativeInternetusers,definedasinternetusersaged 25 or under, number 280 million, practically the same as the totalnumber of U.S Internet users. 496 million internet users, or 68 percent,madedigitalpaymentwiththeirsmartphones.(FollowingFigures)

25Acemoglu,Daron&JoshuaLin,2004,«MarketSizeinInnovation:TheoryandEvidencefromthePharmaceuticlaIndustry»,QuarterlyJournalofEconomics,119(3):1049-1090.

21

Source:Ibidem,MacKinseyGlobalInstituteDiscussionPaper,August2017.

22

ItisinterestingtonotethatChina’sdigitalexpansionisstronglylinkedtolowcostsproductssuppliedbydomesticmanufacturers.Thepenetrationofcheap and high quality smart phones enabled the rapid spread of themobile internet. In just ten years, from 2007 to 2017, Chinese brandmobilesrosefrom5percentto90percentofthemarket,thebestexampleof the import substitution policy launched in 2006, with the Medium toLong-Term Plan for the Development of Science and Technology. Inaddition, the lax regulation of digital content in terms of intellectualproperty rights strongly enhanced the penetration and the use of thesedevices,withfreeandeasyaccesstomusic,booksandmovies.In2008,99percentof internationalmusicdownloads inChinawere freeof chargeor«illegal» according to the terms of the International Federation of thePhonographicIndustry(IFPI)whichsuedsearchenginelikeBaidu,beforeasecondphaseofstrongerenforcementofintellectualpropertyrights(IPR)ontheInternet26.

Source: Ibidem, China’s digital economy, A leading global force, MacKinsey GlobalInstitute,2017.

26China’sdigitaleconomy,Aleadingglobalforce,McKinsey2017,p.12-14;TheFrenchChamber of Commerce and Industry in China, «China, The digital revolution»,Connexions Magazine, february 2, 2017: http://www.ccifc.org/single-news-connexions/n/connexions-79-china-the-digital-revolution/

23

The strong penetration of the internet and smartphones enhanced theexpansion of e-commerce. China accounted for less than 1 percent of thevalueofworldwidedigital payments in2005, but raised tomore than40percent in2016, a value larger than thoseof France, Japan,Germany, theUK, and the US combined. The value of its mobile payments related toconsumptionbyindividualswas11timesthatoftheU.S.(FollowingFigure)ApartfromthesheervolumeofChina’sInternetusers,machine-generateddataalsoprovidesamassiveamountofrawmaterial for theexpansionofA.I,(FollowingFigurefortheworldtendency).

Computerpowerhasbeenstimulatedbygovernmentpoliciesaswellastheexpansion of e-commerce. During Singles Day on November 11 2016,whichhasbecomeahugeonlineshoppingday,withpromotionssuchastheUSBlackFridaywhichhitUS$3.34billionin2016,Alibabamade$US17.8billionsinsales,upfrom$14.3billionin2015(FollowingFigure)

24

25

26

, its payment platform processed 120,000 transactions per second, threetimes higher capacity than than one of the leading world Platform(FollowingFigure).DuringSinglesDay2017,Alibababrokeitsownrecordbyselling theequivalentofUS$25.3billion,agrowthof40percentoverthepreviousyear. Chinese cloudprovidersalsohold theworld recordofcomputingefficiency,since2014,withBaidu,Alibaba(2015),andTencent(2016) in the Sort Benchmark’s world competition, regarded as thecomputingOlympics27

Source: China’s Digital Economy, A Leading Global Force , McKinsey Global Institute,2017.TheBaiduAlibaba&Tencent(BAT)havediversifiedtheirapplications,byproviding solutions to sectors fragilised by inefficiencies. Alibaba forinstance, for itscorebusinessthroughitsonlineshoppingwebsiteTaobao(founded in2003) laterbet,with itsAlipaydigitalpaymentbusiness, forthe retailmarket in the small andmiddle sized towns,whichwere badlydeserved by traditional retailers. Now the Alibaba Distribution Platform(ADP), a retail management system established inMay 2016, which hasmorethan500,000clients,offersmarketing,trainingandlogisticsservicesto independentshopowners.Thegoal is torebrand10,000momandpopstoresunderthenameofitsonlinemarketplace,Tmall(foundedin2008).But the potential is enormous as they are more than 6 million of these

27China’sdigitaleconomy,op.cit.McKinsey2017,p.6.

27

stores,mostofwhicharefamily-run,accordingtoAlibaVice-Presidentandgeneral manager of ADP. Alibaba rival JD.Com京东, is also tapping themarket of traditional retailing. It announced plans to open 1 millionconveniencestoresinfiveyears,halfofwhichwillbelocatedinruralChina.JD.comwillfocusondecorationandinventorysupplywhileAlibabaaimstoconcentratemoreondata28. In the financial sector,Alipay launchedYueBao, which offers interest rates two to four percentage points higher fordepositors,thanthelowornegativeratestheyhadtoacceptfrompublicbanks. It also diversified in entertainment and other sectors (FollowingFigure) Alipay has now 90 functions. Alibaba decided to invest $ US 15billioninthethreeyearsfrom2018to2020inA.Iapplications,andtohirehundredofscientistsacrossthetechnexusof theUS,ChinaandIsrael. Itslabswill especially focusondata intelligence, internetof things, quantumcomputingandhuman-machineinteraction.29

Source:MacKinsey2017,China’sdigitaleconomy,op.cit.Tencent main business is social media, services such as WeChat, amessaging application which has more than 900 million active users in

28WangQionghui&CocoFeng,«AlibabawantstobringdatatoImillionMom-and-Popstores»,Caixin,August29,2017.29LouiseLucas,FinancialTimes,October11,2017.

28

2017. Tencent later diversified in payments (Tenpay), online banking(Webank), entertainment (videos) and on demand dining services(Meituan-Dianping).WeChathasnowmorethan40functions.

Source:MacKinsey2014,China’sdigitaleconomy,op.cit.FinallyBaidustartedwithitssearchengine,whichtodayaccountsfor80percentof themarketshare,and laterdiversifiedonmobileserviceswith660millionusersinmobilesearch,andonO2O(onlinetooffline)servicessuchasfooddelivery,groupbuying,andfinancialproducts.Withthestrongsupport of the National Development and Research Commission, Baidulaunched with partners Tsinghua, Beida, and Beihang, a NationalEngineering Laboratory, an open Platform on A.I which is focusing onmachine-learningbasedvisualrecognition,voicerecognition,newtypesofmachine Learning interaction, and Deep Learning. It is helping moreChinese researchers, companies and universities to access the mostAdvanced AI technologies in China. Baidu’s Artificial Intelligence Group(AIG) now focus on deep learning’s commercial applications in diversesectors,anAmazonAlexa-stylechatbot,DuerOS,andthedriverlessvehicleswith its Apollo project, described as a «complete and reliable softwarePlatform»designedtohelppartnersfromtheautomotiveandautonomous

29

drivingindustrybysupplyingthemwithdata».30BaiduBrainisdevelopingaplatformfothird-partyAIapplications.PartnershiphasalreadydevelopedwithChinesecarmakersBYD,CheryandBAIC,andwithBosch,theworld’sbiggestautomotivesupplier. If successful,BaiducouldbewithApollo, theautomotive equivalent of Google’s Android software for Smartphones.31Baidualsodevelopscollaborationwithaviationservices,bytestingitsfacialrecognitionsysteminBeijing’sCapital InternationalAirport. ThusChina’stechgiantsareincreasinglyblurringtheboundariesbetweensectors32.AlibabaandTencent,arefinancedintheUSandHongkongstockmarkets.With capitalizations as high as $US 400 billion each, they take shares orcontrolabouthalfoftheleadingstartups(FollowingFigure).

Source:MacKinsey2017,ChinaDigitalEconomy,op.cit.Arecentreport(inChinese)bytheBostonConsultingGrouptogetherwiththeresearchdivisionofAlibaba,BaiduandDidiunderlinesthespeediness30SouthChinaMorningPost,February24,2017;MacKinsey2017,China’sdigitaleconomy,op.cit.p.9-10.31Theriseofthemachines,HowChineseexecutivesthinkaboutdevelopmentsinArtificialIntelligence?,McKinsey,December2016,p.1;EmilieFengandPatrickMcGee,«BaiduforgesallianceswithGermanautosuppliers»FinancialTimes,June1,2017.YuanYangandYigzhiYang,«Baidubets its futureonAI revolution»,FinancialTimes,August30,2017.https://www.bloomberg.com/profiles/companies/BIDU:US-baidu-inc32Technode,October2,2017http://technode.com/2017/10/02/chinas-tech-giants-are-increasingly-blurring-the-boundaries-between-sectors-mckinsey-report/

30

of the interneteconomy’srise inChina,which isacomparativeadvantageon the U.S. China now equals the US for the number of top ten internetcompaniesbymarketcapitalization:

Chinese unicorns, (defined as start-up which reach US$ 1 billion marketvalue,bypublicorprivateinvestment)take4yearsonaveragetoreachthisstatus, while American unicorns take 7 years. The US has 112 unicornswhileChinahas63.But asChineseunicorns grow innumber andmarketcapitalization, their market capitalization could soon overtake their U.Scounterparts.

31

Finally, the Chinese internet ismore concentrated than theUS, in a fewcompanies that have reached over 50 percent market penetration, muchmorerapidlythantheirUScounterparts33.

33https://www.bcg.com/zh-cn/d/press/13sep2017-china-internet-170385http://technode.com/2017/09/18/fast-and-furious-chinese-unicorns-to-overtake-american-counterparts-says-bcg-report/

32

FROMFASTFOLLOWERTOFUTURECHALLENGER?

AsmuchofChina’sinternetsectorisoff-limittoforeigntechgiants,suchasGoogle, Facebook andAmazon, Chinese tech companies have exclusiveaccess to Chinese internet users’data, they can consolidate withoutcompetitionfromoutside.Apartfromtheircomparativeadvantageindataandincomputingpower,theyhaveaneasyaccesstoalgorithms.MuchofAIresearchisavailablewithopen-sourcesoftwares,asthebasictechniquesofmachine-learning–algorithmsthatbecomessmarterastheyaretrainedonamassiveamountofdata-arenowwell-known.ItmakesiteasierforChinato position itself as a «fast follower»34 . All the conditions, datasetcollection, computing power, and mass collaboration on open sourceplatforms which propel Deep Learning and other techniques are nowconvergingtoenhancetheriseofA.I(FollowingFigure)

34RichardWaters,«ChinareadytocashinonriseofmachineLearning»FinancialTimes,June1,2017.

33

Source:TheRiseofthemachines:HowChineseexecutivesthinkaboutthedevelopmentsinArtificialIntelligence,McKinsey&Co.December2016,P.4.AnothercomparativeadvantageofChinainthisregard,isitspositionasthe world leader in terms of industrial value-added, since 2010, when itbypassedtheU.S.(FollowingFigure)

34

Artificial intelligence is actually at the core of the fourth industrialrevolution, or «Industry 4.0», originating from a project in the high techstrategyof theGermangovernment,whichpromotes the computerizationofManufacturing.Thefirst,secondandthirdindustrialrevolutionsresultedfromthedevelopmentofmechanicalproductiondrivenbywaterandsteampower, the adoption of assembly lines for mass production driven byelectricity and a move toward automation through the utilisation ofelectronics and information technology (I.T) respectively. The emergingfourthrevolutionencompassincreasingdigitisationofmanufacturingwithcyber-physical systems of production, in which big data and cloudcomputing enable information to be shared and analysed along entireindustrial value chains, as connected networks of humans and robotsinteractandworktogether.35(FollowingFigure)

35ChinaManufacturing2025,EuropeanChamberofCommerceinChina,p.6.http://www.europeanchamber.com.cn/en/china-manufacturing-2025

35

Source:Wikipedia,Industry04.ThefirstandsecondprioritiesoftheMadeinChina2025Plan, launchedinMay 2015, are the Next Generation Information Technology andHigh-end Numerical Control Machinery and Robotics. They are basicallycompleted by The Internet Plus Action Plan, adopted in the same time,whichaims,accordingtoPremierLiKeqiangto«integratemobileInternet,cloud computing, big data and the Internet of Things with modernmanufacturing»36.TheTechnicalAreaRoadMapofMadeinChina2025fixesgoalstoreachadvancedinternationallevelinthesetwotopprioritiesbyselectingsectorssuch as integrated circuits (IC) and specialized equipement, informationand telecommunicationequipement,wirelessmobilecommunication,nextgeneration internet and high performance computers and servers, where(except forintegratedcircuits),thedomesticmarketshareissupposedtoreach50to80percentin2030(FollowingFigure37)andtheinternationalmarketshareissupposedtoraisedramatically.

36ChinaManufacturing2025,EuropeanChamberofCommerce,op.cit.,p.10.37JostWubekke&al.,MadeinChina2025,Themakingofahigh-techsuperpowerandconsequencesforindustrializedcountries,MERICSPapersonChinan°2,Dec.2016:https://www.merics.org/fileadmin/user_upload/downloads/MPOC/MPOC_Made_in_China_2025/MPOC_No.2_MadeinChina_2025.pdf

36

All thismeans thatChina carrieson its import substitutionandexportpromotionpolicydefinedinthe2006 MediumtoLong-TermPlanfortheDevelopment of Science and Technology, and refined in all the followingplans.

ThisaquestionofconcernforChina’seconomicpartnerssincethereisagrowing asymetry between China’s more and more restrictive marketaccess,andtheconditionsofmarketaccessfixedbythem.Forinstance,theCybersecurity Law, which is applied since June 2017, in the name of«building secure and controllable information technology systems»,requiresoperatorsofcriticalinfrastructure,withoutdefiningthemclearly,toundergoasecurityreview.ItseemsthatITproducerswillhavetosubmitinformation on their products’design and source code to governmentaffiliatedrevieworganisations,andthatcloudserviceoperatorswillhavetolocate their services facilitiesandnetworkdatawithinChineseterritory. Ifcloud computing is «the delivery of on-demand computing resources –everythingfromapplicationstodatacenters-overtheinternetonapay-for-usebasis»,foreigncompaniesareabletousethiscapacityonlyiftheyarefully confident that their cloud service provider is able to protect theirproprietarydataatalltimes,whichmaynotbethecasefollowingthenewofficialregulation38. Chinaforinstanceisnotoriouslysecretiveconcerninggovernmentdata,asitranks93rdfordataopeness(FollowingFigure).ThenextgenerationAIdevelopmentplan,publishedbytheStateCouncilinJuly2017 allows to revise the policies of data protection by launching pilotprojectstoopengovernmentdata, tomineitscommercialvaluewiththe

38MadeinChina2025,EuropeanChamberofCommerceinChina,op.cit.p.23-25

37

collaborationofthepublicandprivatesector,andpromotetheinnovationofAIapplications.

Source:McKinseyGlobalInstitute,ArtificialIntelligence,ImplicationsforChina,2017.

As all aspects of the AI value chain, from fundamental research andapplications development to hardware manufacturing involve globalcollaboration39, China’s AI strategy is progressive but resolute for threemainreasons.

Firstofall,beyondtheavailablepooloftalentwithinChina,anestimated43percentoftheworld’strainedAIscientists,Chinaistryingtoovercomeitsmainweakness in fundamental researchbyrecruitingmassively in theU.Sandelsewhere.Baidu,Tencent AlibabaandapleiadofChinesebasedstartups,suchastheonedirectedbyKai-FuLee,aformerseniorexecutiveforGoogleandMicrosoftinChinawhohasmorethan50millionfollowersontheChinesemicrobloggingPlatformSinaWeibo, havebeenlookingfor

39DominiqueBarton&al.ArtificialIntelligence:ImplicationsforChina,McKinseyGlobalInstitute,April2017,p.8.

38

the best AI experts in the Sillicon Valley and the Boston area, mostlyChinesescientistswhoworkintheUS40.

Secondly,ifAIisthekeytofuturegrowth,byenhancingtheproductivityof inefficientindustriesandservicesandhelpingtocreatenewbusinessesthatleveragethetechnology,asunderlinedbyWillKnight(previousarticlecited), itmay also be disruptive in termsof regulations and employment.ThenextgenerationAIdevelopmentplanaimstoadapt the frameworkoflaws,regulationsandethic,andthetrainingofreemploymentworkerswhomayloosejobsbecauseofAI.

Thirdly, the race towards preeminence in A.I, has considerablegeopolitical consequences, since it is a dual use technology with bothmilitaryand civilian implications41.Asmostof the researchandadvancesaremadebyprivatecompanies,theChineseleadershiphascreatedin2017a Central Commission for Integrated Military and Civilian Development,directedbythePresidentXiJinpinginordertoensurethatalltheadvancesin cloud computing, big data, internet of things and artificial intelligencemeetdefenserequirements42.Thedefenseobjective(国家安全),aswellastherecruitmentoftalentsasatoppriority,areclearlyprecisedinThenextgenerationAIdevelopmentplanpublishedbytheStateCouncil43.

ThePLAseekstocapitalizeonthetransformationoftoday’sinformatized(信息化) ways of warfare into future «inteligentized» (智能化) warfare.Lieutenant General Liu Guozhi, director of the Central MilitaryCommission’s Science and Technology Commission plan to integrate civiland military advances in A.I with an approach of «shared construction,sharedenjoyment,andshareduse»(共建， 共享， 共用)44.ThePLAbets

40Elsa Kania, «Beyond The Committee on Foreign Investment in the Unites States(CFIUS): The strategic challenge of China’s rise in artificial intelligence», Yale LawSchool, Paul Tsai China Center, June 20, 2017https://lawfareblog.com/beyond-cfius-strategic-challenge-chinas-rise-artificial-intelligence ; Will Knight, «China’s AIAwakening», MIT Technology Review, October 10, 2017:https://www.technologyreview.com/s/609038/chinas-ai-awakening/41HelèneLavoix,«Whenartificialintelligencewillpowergeopolitics,PresentingAI»,TheRedTeamSociety,Nov29,2017;https://www.redanalysis.org/2017/11/27/artificial-intelligence-will-power-geopolitics-presenting-ai-open-access/42Xinhua,Beijing,Sept22.43 http://www.gov.cn/zhengce/content/2017-07/20/content_5211996.htm?from=timeline&isappinstalled=044ElsaKania,articlecited.

39

on the asymetric tactic of drone swarms, first implemented by theUS in2013, a light attack force inspiredbyAttila and theHuns, to defeatmorepowerfulandsophisticatedopponentswiththeirsheernumbers45.SwarmintelligenceispresentinthenextgenerationAIdevelopmentplan.

The US have developed autonomous weapons systems, but the lethalautonomousweapons (LAWS) or «killer robots» raise a lot of questions.The founders of 116 robotics and artificial intelligence companies havecalled for their ban in an open letter, following the postponement of UNtalks on regulating killer robots. They estimate that LAWSwould permit«armedconflict tobefoughtonascalegreaterthanever,andatimescalefaster than humans can comprehend»… «If LAWS become part of amilitaryindustrialcomplextheywillendupbeingusedagainstcivilians»46.The robots created by Boston Dynamics, could be equipped with lethalweapons…. (Following video):https://www.youtube.com/watch?time_continue=81&v=M8YjvHYbZ9w

For theastrophysicistStephenHawking,«wespendagreatdealof timestudyinghistory,which-let’sfaceit- ismostlythehistoryofstupidity.Soit’sawelcomechangethatpeoplearestudyingthefutureof intelligence».ButinthesametimethecreationofAIwillbe«eitherthebestortheworstthing, ever to happen to humanity». A few forums of discussion andresearch have been created on the potential and risks of AI. TheLeverhulme Centre for the Future of Intelligence (LCFI), an international multi-disciplinary institute at Cambridge University 47 , and the UnitedNations Interregional Crime and Justice Research Institute are deepeningthe reflexion, in collaborationwithworldscientists48.For the first time inNovember 2017, the group of governmental experts on the lethalautonomousweapons (LAWS)met inGeneva, at the invitation ofUNODA(United Nations Office for Disarmament Affairs) and realised a

45EmilieFeng&al.«Droneswarmsvsconventionnalarms:China’smilitarydebate»FinancialTimes,August24,2017.https://www.ft.com/content/302fc14a-66ef-11e7-8526-7b38dcaef61446JamieSmyth,FinancialTimes,August21,2017.47TheGuardian,19October2016;http://www.cam.ac.uk/research/news/the-future-of-intelligence-cambridge-university-launches-new-centre-to-study-ai-and-the-future-of48UNICRI,TherisksandbenefitsofArtificialIntelligenceandRobotics,Aworkshopformediaandsecurityprofessionals,Cambridge,Februray6-7,2017:http://unicri.it/in_focus/files/Report_UNICRI_Cambridge_Workshop_Feb_2017.pdf

40

multidisciplinary report which may set the base for internationalregulations49.

ButtheUS-Chinaracegoeson,asAIwillnotonlybedecisive inmilitarycapacitiesbutalso for futureeconomiccompetitiveness. In June2017, theUS government commissioned six companies including IBM, Intel andHewlett Packard Enterprise to come up with countermeasures againstChina’s lead on high performance computer, which has reached an«inflection point», according to Horst Simon, the deputy director ofLawrence Berkeley National Laboratory in California. A US exascalecomputerisexpectedby202150.

Two factors may explain the US sense of urgency, apart from China’scutting edge advances in AI, high performance computing and quantumcomputing. As innovation is largely driven by consumers preferences51,China, as the most massively connected country in the world, withpractically 300 million digital natives, is a huge experimental lab for AIapplications.DevelopmentsareveryquickinChina,andmuchslower,aswehaveseen,intheU.S.Secondly,theclosemilitary-civilintegrationinChinaas regard to AI comes at a time when ties between the US military,university research and companies become strained and tech group arebecoming ambivalent about working with the military 52 . US-China’scompetition inthefieldofAIextendsfarbeyondthetraditionalcontroloftechnologytransfersmadebytheCommitteeofForeignInvestment intheUS(CFIUS)53astheyarenumerouscollaborationsbetweentheUSandtheChinese tech eco-systems, which imply flows of scientists and companyexecutives.

Secondly,asaccesstotheChinesemarket ismoreandmoreconditionedbytechnologytransfers,somecompaniesmaybemoreandmoreinvolvedin China’s high tech import substitution strategy. Qualcomm, theworld’s

49UNODA,Geneva,PerspectivesonLethalAutonomousWeaponsSystems,November2017;https://www.un.org/disarmament/publications/occasionalpapers/unoda-occasional-papers-no-30-november-2017/50SouthChinaMorningPost,Hongkong,23August2017.51ChinaManufacturing2025,EuropeanChamberofCommerce,op.cit.p.22.52HennySender,«AdvancedtechoutfitsgiveChinesemilitarythecuttingedge»,FinancialTimes,August8,2017.53ElsaKania,art.cit.June20,2017.

41

dominantmobilechipmaker,whichwasmakingmorethanhalfofitsglobalprofitsinChina,hasbeenfinedforanticompetitivebehavior($975millionin2015or8percentof itsannualrevenue inChina),andhadto lower itspricesandshiftmoreofitshigh-endmanufacturingfromTaiwanandSouthKorean contractors to partners in the country. It is led to competewithIntel bydeveloping leading edgemicrochips, tohelpChinese start-ups, tocollaborate with Chinese companies to develop drones, virtual realitygoggles, internet-connected devices, supercomputers, and mass-marketsmartphonechips54.

CONCLUSION:

FROMCHIMERICA’sCOOPETITIONTOCONCERNSABOUTAIUSE&ABUSE

Themixofcooperationandcompetition(coopetition)betweenthetwoAIgiants,will certainlyevolvewith thedevelopmentofquantumcomputing,another new frontier. If thenew computers that harness quantumeffectswill not only bemuch faster, but alsomore sophisticated, able to designmoleculestocreatenewdrugsorbreakthroughmaterials55,wemayexpectunintended applications in the civil as well as defense sector, whichunderlines the urgency of a global reflexion and regulation of these newdevelopments, certainly disruptive for the economy, and maybeconstructiveordestructiveforhumanity56.

Let us focus on the present. China for example, had 176 millionsurveillancecamerasinoperationin2016,andisexpectedtoseethatfiguremore than triple to reach 626 million by 2020. Various stories haveemerged onChina’s efforts to increase surveillance of its peoplewith theadded capabilities of AI facial and gait recognition. Beijing announced in

54DavidBarboza,«HowthisUStechgiantisbackingChina’stechambitions»,TheNewYorkTimes,August4,2017.55RichardWaters,«Quantumcomputingrivalsmustersoftwarepowerinnew‘armsrace’»,FinancialTimes,October2,2017;MorganStanleyResearch,Oct.4,2017:«Aquantumleaptowardacomputingrevolution»,http://www.morganstanley.com/ideas/quantum-computing56JonhThornhill,«AI’srapidadvancesparkscallforacodeforrobots»,FinancialTimes,August31,2017;AnjanaAhuja,«Robotbehaviouriscreepingbeyondourcontrol»,FinancialTimes,August2,2017;W.BrianArthur,«Whereistechnologytakingtheeconomy»,McKinseyQuartely,October2017.

42

October 2015 that it now had 100 percent coverage57. « Big Brother iswatchingyou!»ButwhoiswatchingBigBrother?

*

57Technode,November22,2017;http://technode.com/2017/11/22/china-to-have-626-million-surveillance-cameras-within-3-years/