Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
FINALVERSION
Strategy for the Development of
Quebec’s Artif icial Intel l igence
Ecosystem
Artificial Intelligence Cluster Steering Committee
A mandate from
May 2018
2
FINALVERSION
Université de Montréal’s role
InMarch2017,theQuebecgovernmentmandatedtheUniversitédeMontréaltosetupaSteeringCommitteetodevelopthestrategyforthedevelopmentofQuebec’sartificialintelligenceindustrialandscientificcluster.ThisCommitteewastosubmititsrecommendationsonthedevelopmentofthecluster'sresourcesregardingtheattractionofinternationalexperts,theavailabilityofcomputingresources,theconditionsmostfavourableforindustrialcollaborationsandthehiving-offofcompanies,aswellashumanresourcetraininganddevelopment.Inadditiontodevelopingastrategicdevelopmentplan,theCommitteewastoliaisewithrelevantfederalgovernmentbodies,theCanadianInstituteforAdvancedResearch(CIFAR)andtheCanadaFirstResearchExcellenceFundandencouragefutureprivateandpublicinvestments.
TheUniversitédeMontréalwasalsomandatedtosetupaQuebecArtificialIntelligenceInstituteasaseparatenon-profitentity.Moreover,thisInstituteshouldbeabletofosterthepartnershipsneededtofunditsactivities,whileaddressingtheeconomicissuesandsocialtransformationsresultingfromtheAI-drivendigitalrevolution.
3
FINALVERSION
Co-Chairs’ message
TheArtificialIntelligenceClusterSteeringCommittee(hereinafter,theSteeringCommittee)wasconstitutedinMay2017andhasasitsmandatetopositionQuebecasaleadingplayerinartificialintelligence(AI),throughresearchandinnovation,businesscreationorapplicationdeploymentwithinitsindustrialstructure.Oneyearlater,inthewakeofaseriesofimpactfulevents,theCommitteeandits12memberspresentthefruitoftheirreflections,researchandanalysisinthisStrategyfortheDevelopmentofQuebec’sArtificialIntelligenceEcosystem.
Inthepastfewyears,Quebechasexperiencedahistoricboominartificialintelligence,thankstotheexcellenceofacademicresearchinthefieldthathasledtothearrivalonthesceneoftechnologygiants,therapidproliferationofQuebecstart-upsandtheestablishmentofrenownedresearchinstitutes.Quebeccannowrelyonitsstrengthstodrivethedevelopmentandgrowthofarichanddynamicartificialintelligenceecosystemthatwillbetheenvyofitsinternationalpeers.Whilethisstrategydoesnotpresumetosolveallthepresentorfuturechallengesecosystemstakeholderswillface,weareconvincedthatitwillhelpthegovernmentshapeaprosperousfutureforQuebecinartificialintelligence.
Twelverecommendationsthatfallintofivestrategicapproacheslieatthecoreofthisstrategy.ThesearetoensurethegrowthandsustainabilityofQuebec’sartificialintelligence(AI)academicresearchcluster,developdigitalsciencetalenttomeetQuebec'sneeds,acceleratetheadoptionanddevelopmentofAIsolutionsbytheentireeconomicfabricofQuebec,developaninternationalresponsibleAIhubinQuebecand,finally,supportthedevelopmentofecosystemsupportstructures.
WewouldliketoextendoursincerethankstotheSteeringCommitteemembersandobserversfortheirdedication,andalsothankalltheadvisors,staffandexpertsweconsultedwhohavegenerouslyagreedtoparticipateinthisexercise.TheSteeringCommitteeanditsmemberswillnowturntheirattentiontotheimplementationoftherecommendations,astheyareconfidentthattherecommendationswillenableQuebectobecomeaworldleaderinartificialintelligence.
Wealsowishtoacknowledgethestrongvisionthegovernmenthaswiselyoutlinedforthisforward-lookingsectorandfeelcertainthat,thankstocontinuedandtargetedgovernmentsupport,thebestoutcomeforthebenefitofthegreatestnumberisassured.
PierreBoivin GuyBreton
May2018
4
FINALVERSION
Table of contents Introduction ..................................................................................................................................5
Recommendations........................................................................................................................21
! Recommendation1:IncreaseourabilitytoattractresearchersinareaswhereQuebechasthepotentialtobecomeaworldleader...............................................................................................24
! Recommendation2:Ensuretheinflow–inquantityandquality–oflocalandforeigngraduatestudentsinAIandAI-relatedfields.................................................................................................29
! Recommendation3:Developandmaintainthrougha“talentwatch”aclearandcomprehensiveperspectiveonneeds,gapsandopportunitiesregardingdigitalscienceskills..............................33
! Recommendation4:Supporteducationalinstitutionsinupdatingtheirtrainingoffering............37
! Recommendation5:Promotemathematicalliteracyandthedemocratizationofscience...........38
! Recommendation6:SupportQuebecbusinessesthatdesignAIsolutions,includingstart-ups,sotheycandevelop,growandremaininQuebec..............................................................................42
! Recommendation7:EnsurefutureprivatesectorAIusersreceiveadequatesupport..................49
! Recommendation8:PositiongovernmentandCrowncorporationsmodelusersthatbenefitfromAItechnology..................................................................................................................................55
! Recommendation9:EnsurethesustainabilityofongoingresponsibleAImobilizationeffortsandpreparetointegrateemergingconsensusesintoQuebecpublicpolicy........................................58
! Recommendation10:EnsureaccesstocomputingpowerandcomputationalexpertiseforAIanddatasciencestakeholders...............................................................................................................64
! Recommendation11:CreateapermanentAI.QuebecplatformwithitsownadministrativestructuretoplayamonitoringroleandfacilitatecoordinationbetweenQuebecAIecosystemstakeholders...................................................................................................................................66
! Recommendation12:FacilitateaccesstoinformationonavailableresourcesforAIecosystemstakeholdersandpromoteconnectionsbetweenthemthroughacommonshowcase................68
Conclusion ................................................................................................................................70
Appendices ................................................................................................................................76
FINALVERSION
Introduction
6
FINALVERSION
Foreword 1. Steering Committee mandate
Forsomeyears,Canada,andmoreparticularlyQuebec,hasexperiencedanever-before-seeneffervescenceintheartificialintelligence(AI)sector.AsshowninFigure1,technologygiantslikeGoogleandFacebookhaveopenedAIlabsinMontreal,start-upslikeElementAIhaveraisedimpressiveamountsofprivatefundingandresearchinstituteshaveralliedaroundQuebec’shistoricalstrengths.
Thecurrentgeneralizedeffervescenceisrootedintheuniversityenvironment,aspherewhereQuebecstandsoutontheworldstage.ItisalsodueinparttothesupporttheCanadianInstituteforAdvancedResearch(CIFAR)providedin2004forYoshuaBengio'screationoftheMachineLearningProgram.1AsanotherproofoftheexcellenceofQuebec'sendeavoursinAI,thefederalgovernmentawardeda$93.6MCanadaFirstfederalgranttotheInstituteforDataValorization(IVADO)–agroupof1,000datascience,operationalresearchandartificialintelligencescientists–inthefallof2016.
TheQuebecuniversitymilieu,fromFrance’sperspectiveCédricVillani’srecentreporttotheFrenchgovernmentstatesthatQuebeccurrently“has,withtheCityofMontreal,thelargestconcentrationofAIresearchers—250doctoralresearchersatMcGillandtheUniversitédeMontréal—whoalsobenefitfromalargecohortofroughly9,000AIstudentsincomputerscience(6194),computerscienceandcomputerengineering(1149),mathematics(1360),appliedmathematics(113)andprobabilitiesandstatistics(141)programs.”2
InMarch2017,theministèredel’Économie,delaScienceetdel’Innovation(MESI)choosetosupporttheaccelerationandoptimizationofdevelopmentofQuebec’sAIecosystemwithatotalof$100Moverafive-yearperiod.Atthesametime,itgavetheUniversitédeMontréalthemandatetoformaSteeringCommitteetoidentifythemeasuresneededtoleveragethismajorinvestmenttostrengthenthecompetitiveadvantageinAIQuebechasgainedthankstotheuniversitysector.
TheMESItaskedtheArtificialIntelligenceClusterSteeringCommitteewithfourmissions: 1. EnsurethelaunchoftheQuebecArtificialIntelligenceInstitute.2. DevelopastrategicplanforthecreationoftheAIclusterby2018-2022.3. Studytheethical,economicandsocialimpactsofAI.4. Takealeadershiproleindefiningapan-CanadianAIstrategy.
InSeptember2017,theConseilconsultatifsurl’économieetl’innovation(CCEI)haspraisedtheQuebecgovernment'sfocusedambitiontoseeAIbecomearealdriverofQuebec'seconomy.Initsreport,the
1CIFARalsospearheadedCanadianandglobalAIresearchbysupportingtheworkofGeoffreyE.Hinton,ProfessorofComputerScienceattheUniversityofToronto,inthe1980s.2Seehttps://www.tresor.economie.gouv.fr/Articles/28619068-9771-411c-9a43-20fdaeb0adc8/files/5a885cf0-0af1-4243-be03-8857b5319fae,webpagevisitedonApril30,2018(aswereallthepagesreferencedinthisdocument).
7
FINALVERSION
CCEIurgedthegovernmentandecosystemstakeholderstopushevenfurtherto“createthelargestartificialintelligenceecosysteminNorthAmericainQuebecanddevelopaworkforceinthisareasothatthisnichetakesrootandpositionsQuebecasoneoftheleadingdesigners,usersandexportersofartificialintelligencesolutions.”(translatedfromtheFrenchoriginal)3
3Seehttp://www.ccei.quebec/fr/rapport-integrateur.
8
FINALVERSION
Figure1–ChronologyofQuebecAIecosystemstructuringevents
9
FINALVERSION
Moreofanecosystemthanacluster…Theterm“cluster”hasoftenbeenusedtodescribetheAIecosystemdevelopinginQuebec.Toavoidanyconfusionwithindustryclusters–whicharesectorspecific(e.g.,AéroMontrealfortheaerospacesector)–werecommendusingtheexpression“AIecosystem.”Indeed,AIwillbeusedinallindustrialsectorsandbecomeaubiquitoustechnology.Asaresult,closecollaborationbetweenexistingQuebecindustrialclustersandtheAIecosystemisadvocated.
2. Last year’s achievements
ThisreportistheSteeringCommittee’sresponsetothesecondofthefourrequeststheMESImadeinspring2017.
BeforepresentingtheQuebecAIecosystem’sdevelopmentstrategy,itisimportanttomentionhowtheSteeringCommitteeaddressedtheMESI’sthreeotherrequests.
Initially,theSteeringCommitteewastaskedwithensuringthelaunchoftheQuebecArtificialIntelligenceInstitute.Thecreationofthisentityprogressedapace,and,inJuly2017,theMILA–QuebecArtificialIntelligenceInstitutewassetupasanot-for-profitorganization(NPO).ThisinstituteistheresultofacollaborationbetweentheUniversitédeMontréalandMcGillUniversityandalsohasclosetieswithPolytechniqueMontréalandHECMontréal.WorktosecurepropergovernancefortheInstituteisunderway,anditsboardofdirectorsshouldbefullyconstitutedshortly.AlthoughitisnowanNPO,theMILA–QuebecArtificialIntelligenceInstituteintendstohonouritscommitmentstoIVADOandtheCanadaFirstResearchExcellenceFund.
Theorganization’smissionextendsbeyonditsuniversityvocationandincludesthefollowingelements:
a. Training–Incollaborationwithuniversitiesandassociatededucationalinstitutions,train,attractandretainmachinelearningtalent.
b. Academicresearch–Incollaborationwithuniversitiesandassociatededucationalinstitutions,contributetobasicresearchbybuildingonitsstrengthsindeeplearningandreinforcementlearning.
c. Technologytransfer–ContributetotheeconomicdevelopmentofQuebecthroughtechnologytransferandbusinessinnovation.
d. Socialdialogue–Promotetheresponsibleuseofartificialintelligenceaswellassocialdialogueregardingissuesrelatedtoit.
TheMILA–QuebecArtificialIntelligenceInstitutewillalsobeoneofthemainattractionsofthenewinnovationhubintheprocessofbeingsetupinoneofMontreal’semergingneighbourhoods.Thespaceinquestionislocatedinacomplexcalled“OMile-Ex”at6650-6666St-UrbainStreetandwillbehometoresearchersfromvarioussectors,IVADO,companylaboratories,start-upsandventurecapitalinvestors.Itwillbringtogethervariousstakeholdersinasingleplace,whichisacrucialstepincreatingatrueinnovationecosystemontheground.Theproximityofthedifferentstakeholderswillcreateanenvironmentconducivetocollaboration,innovationandtechnologytransfer.ManyAIecosystemstakeholdershavealreadyconfirmedtheirpresenceeitherintheneighborhoodorinthecomplexitself,
10
FINALVERSION
orexpressedtheirinterestinsettingupshopthere,includingElementAI,QuantumBlack,ThalesandtheRBCInstituteforResearch,BorealisAI.
TheSteeringCommitteewasmandatedtostudytheethical,economicandsocialimpactofAI.ItshouldbenotedthatQuebechasalreadypositioneditselfasanactiveplayerinthedesignanduseofresponsibleAI.TheMontrealDeclarationforaResponsibleDevelopmentofArtificialIntelligenceco-developmentprocesswaslaunchedonNovember3,2017atthe“ForumAIresponsable”heldatthePalaisdesCongrèsinMontreal.Thisprocessaimstogeneratepublicdebateandproposeaprogressiveandinclusiveorientationforthedevelopmentofartificialintelligence.ThefirststepinthepublicconsultationprocessthatwillleadtothepublicationoftheDeclarationinDecember2018essentiallyendedinApril.
TheSteeringCommitteewasalsotaskedwithtakingaleadershiproleindefiningapan-Canadianstrategyforartificialintelligence.TheSteeringCommitteefulfilledthistaskbysupportingthecreationoftheCanadianAIConsortium(CAIC),whichbringstogethertheMILA–QuebecArtificialIntelligenceInstitute,Ontario’sVectorInstituteandAlberta’sAMIIInstitute.ThemissionofthisconsortiumistoexchangebestpracticesinAI,ensuresynergybetweenthethreeorganizationsandformulatecommonrequeststothefederalgovernment,suchasthoserelatedtoCanadiancomputinginfrastructures.Inthefuture,membersoftheconsortiumwillbeabletojoinforcestoestablishCanadaasaninternationalleaderinAIversuscompetitorswithgreaterfinancialresources.TheCanadianInstituteforAdvancedResearch(CIFAR)hasalsopromotedtheCanada-widestrategiceffortinartificialintelligencebyawardinga$125Mgranttothesethreeinstitutes,including$44MtoMILA–theQuebecArtificialIntelligenceInstitute.
Finally,thelastyear’sachievementsincludethecreationofSCALE.AIsuperclusterthankstofederalandQuebecfunding.Thisconsortium—whoseoperationwillbebasedoncollaborationsbetweentwoacademicleaders(IVADOinQuebecandtheUniversityofWaterlooinOntario)andorganizationsthatwillworkcloselywithQuebecstart-upsinAI(suchastheCreativeDestructionLabandNextAI)—willshape“anewglobalsupply-chainplatform,strengthenCanada'sartificialintelligence(AI)leadership,andaccelerateindustryadoptionofAI-enablingtechnologiesthroughcollaborativeandinnovativeprojects.”(translatedfromtheFrenchoriginal)4
SCALE.AIsuperclusterannouncementInMay2017,thefederalgovernmentheldanationalbusiness-ledinnovationsuperclustercompetition.TheSCALE.AI(SupplyChainsandLogisticsExcellence.AI)supergroup,ledbyAIleadersIVADOandtheUniversityofWaterloo,wasselected.
Thefederalgovernmentwillallocate$234Mtothissupercluster,whiletheQuebecgovernmenthasannouncedaninvestmentof$60Minits2018-2019budget.Takingintoaccountthecloseto$700Mincommitmentsoftheparticipatingcompanies,thisstructuringprojectrepresentsatotalinvestmentcloseto$1B.
4Seehttps://aisupplychain.ca/fr/.
11
FINALVERSION
Report Summary
12
FINALVERSION
Overview Artificialintelligence(AI)ispoisedtotransformQuebec’seconomyinaradicalwayoverthenextfewyears.Artificialintelligenceapplications,suchasvoicerecognitionandimageprocessing,arealreadywidespreadintheserviceindustryandcontinuetogainground.Foritspart,themanufacturingindustryisonthecuspofaprofoundtransformationwithAIasoneofthemaindrivers.Frompredictivemaintenancetoreal-timechemicalprocessefficiencyoptimization,AIapplicationswillgraduallybutinevitablyhelpourindustriestodomorewithless.
Inits2017report,theConseilconsultatifsurl’économieetl’innovation(CCEI)hasidentifiedthedevelopmentofAIasoneofQuebec’s12priorities,citingtheprovince’s“scientificlegup”andextraordinarypotentialforimplementingAIsolutions.Thiscomparativeadvantage,mentionedandexplainedinmanyreports,isbasedonthreemainpillars:basicAIresearch,massivedatavaluationandappliedresearch.
Indeed,oneofthelargestcohortsofAIresearchersintheworldiscurrentlyinQuebec.Theseresearchersareveryprolific(havingpublishedmorethan2,000articlesinthepastfiveyears),andtheirwritingshaveaninternationalinfluence(inthefieldofdeeplearning,somehavebeencitedseveralthousandtimes).Alsoofnote,moreandmorestudentslikelytoengageinAIresearchinthefutureareenrollinginQuebeceducationalinstitutions.Asanexample,attheÉcoledetechnologiesupérieureduQuébecandPolytechniqueMontréal,therearetwiceasmanystudentsininformationtechnologyastherewerefiveyearsago.
However,Quebec'sstrategicpositioningisnotlimitedtoresearch.Theprovince,andmorespecificallyMontreal,alsohasacriticalmassofstart-ups,investors,potentialusers,grantingagenciesandliaisonandtransferorganizations.Thesestakeholdersformadynamicandcomplexecosystemthatextendswellbeyonduniversitywallsandhaveattracted$150MinAIventurecapitaloverthepasttwoyears.TheyhavealsospurredlargecompanieslikeThalesorMicrosofttoopensevenAIresearchcentresinrecentmonths.
However,theQuebecAIecosystemalsohassomeweaknesses,asshowninthetablebelow,whichtheSteeringCommitteedrewupafterpreliminarybenchmarkingwork.Often,coordinationbetweenuniversitiesandindustrystakeholdersremainsdifficult.Forexample,companiesreportthattheydonotwhichdoortoknockontoundertakeacollaborativeappliedresearchproject.Similarly,inQuebec,aselsewhere,theindustry’shighwagesencourageprofessorswhoareinapositiontotrainfuturegenerationsofAIresearcherstoleaveacademiatopursueacareerinbusiness.Fortheirpart,AIproducerssometimeshaveadifficulttimefindingtheirfirstQuebeccustomers(amongotherreasons,becauselargefirmsarerelativelyrareintheprovince),whowouldserveasaspringboardforaccessingforeignmarkets.AsforpotentialAIusers,theyoftenmisunderstandthetechnology’spotentialandimpactsincreatinganewbusinessmodelorimprovingtheirprocesses.
Therefore,vigorousactionisrequiredtoensureQuebec'scompetitiveedgeismaintainedoverthelongterminthefaceofinternationalcompetition.And,actingquicklyisessential!
13
FINALVERSION
Strengths and weaknesses of Quebec AI ecosystem stakeholders5
Trainingandresearch Industrialfabric Supportecosystem
5SeeArtificialIntelligenceClusterSteeringCommittee,Balisagedel’écosystèmequébécoisenIA,Montreal,2018,startingonp.66.
14
FINALVERSION
OtherjurisdictionshaveannouncedsignificantcontributionstothedevelopmentofAI.Forexample,considerthefollowingfacts
• ChinahasdeclareditsintentiontobecometheworldleaderinAIby2030.Thisplanisalreadybeingcomparedtothecountry’s2014nationalstrategyonsemiconductors,towhichBeijingallocatedUS$15Bayearovera10-yearperiod.6
• Singaporehascreatedtwoorganizations–AISingaporeandtheSingaporeDataScienceConsortium–toboostitsAIexpertiseandbettercoordinateitsacademicandindustrialstrengths(thefirstentityhasalreadyreceivedUS$115Minfundingoverfiveyears).
• Finlandannouncedthelaunchofits“AIBusiness”program,withabudgetofUS$250Moverfouryears,tofinanciallysupportstart-ups,SMEsandlargecompaniesinthedevelopmentanduseofAI.
• InMarch2018,Franceannounceditsdecisiontoinvest∍1.5Bby2022tocreateartificialintelligenceresearchchairs,significantlyincreasethenumberofAIstudentsandachievemajorprojectsinthefield.
• Britainhopestoadd£630Btothetotalsizeofitseconomyby2035byinjecting£300MintoAIresearch,matchingbusinessAIinvestmentsinsectorssuchasagricultureandlifesciencesandadoptingothermeasures.
Statesarealsotakingnon-financialmeasurestobecomemorecompetitiveinAI.Manyofthemhaveequippedthemselveswiththemeanstoquicklyleveragetheirdata.Finlandhassetupapublicdatauseinitiative,emphasizingdatauseratherthanprotection,whichisacompleteparadigmshift.
TheQuebecgovernmenthasfarfromstoodstillinthefaceoftheseglobaldevelopments,asevidencedbythelistofdirectandindirectAIinvestmentsithasmadeinthelastyear.Inits2018EconomicPlanandotherinstruments,thegovernmenthas,inadditiontothe$100MoverfiveyearsreportedbytheSteeringCommitteeinthisreport,alreadyinvestedorannouncedthatitwillinvest$102MinAIprojectsorprograms,asfollows:
• $60MoverfiveyearstoSCALE.AIfortheimplementationofthissupercluster’sprojects• $10MoverfiveyearstohelplaunchCreativeDestructionLabofMontrealandNextAI,two
centreswhosemissionwillbetohelpstart-upsspecializedinAIdevelopandsupportstudents,professionalsandentrepreneurswhowishtodevelopideasthatrelyontheuseofAI
• $5MtohelpestablishanewinternationalAIorganizationinMontreal• $1MsoappliedmathematicalsciencecanbeleveragedtosupportAI-relatedfieldsinQuebec• $12.5Mtoimprovethestate-of-the-artcomputinginfrastructureavailabletotheQuebecAIfield
andothersectors• $5MforIVADOtocarryoutaCanadaFirstresearchprojectonbigdata,aswellastoincreasethe
state-of-the-artcomputingpoweravailabletoresearchers• $8.5MtosupportcollaborativeAIresearchprojectsthroughPrompt.
ThegovernmenthasalsoannouncedplanstomakefurthermajorinvestmentsthatcouldpotentiallyhaveanimpactonthedevelopmentoftheAIecosystem.Thus,the2018EconomicPlanprovidesforinvestmentsof:
• $15Mtoencouragedigitalownershipbytourismindustrystakeholders• $23MtotheCRIQ(CentrederechercheindustrielleduQuebec)toensurethetransitionof
manufacturingcompaniestoIndustry4.0
6Seehttps://www.cfr.org/blog/chinas-artificial-intelligence-strategy-poses-credible-threat-us-tech-leadership.
15
FINALVERSION
• $60MovertwoyearsincallsforprojectsforindustrymemberstousetechnologiessuchasAIto
meetthechallengesfacingtheindustry,testnewbusinessmodels,marketnewservices,etc.
• $47Mtoincreasethegraduationrateinkeyareas,suchasscienceandappliedmathematics,for
thedevelopmentofAIinQuebec.
Inconclusion,toleadtheinternationalracecurrentlyinfullswingandachievethegoalofcreating“the
largestartificialintelligenceecosysteminNorthAmerica”(Recommendation12oftheCCEIreport),
Quebecwillneedtofocusitsresourcesonthestrategicleversatitsdisposal.Threeimperativeslieatthe
heartofthestrategyoutlinedherein.
1. WemustguaranteethepositioningandempowermentoftheMILA–QuebecArtificial
IntelligenceInstitute,anepicentreinmachinelearningwhosemissioncoversuniversityresearch,
training,technologytransferandsocialdialogue,whichwillprovidecross-disciplinarysupport
withinQuebec’sAIdevelopmentstrategy.
2. WemusthelpQuebecasawholecontributefullytothisshiftinthesectorsthathavebuildits
historicstrengthandenabletheQuebeceducationsystemasawholetoproducemoreAIand
digitalsciencespecialists.
3. WemustcontinuetosupportIVADOandSCALE.AIandbuildontheircurrentmomentum.
Thesethreeimperativesarecenteredaroundfivemajorstrategicdirectionsonwhichthisstrategyrests:
! MaintainQuebec'sacademicleadership
! Developtalent
! Supporttechnologytransferandmarketing
! DevelopacentreofexpertiseinresponsibleAI
! Developrobustsupportstructures
Thisreportpresentsthesefivestrategicdirectionaswellasthe12detailedrecommendationsrelatedto
them.Itshouldbenotedthatsomeoftheserecommendationscould–andshould–beimplemented
withoutdelay,whileotherswillrequireamorein-depthinteractionbetweenecosystemstakeholders.
16
FINALVERSION
Steering Committee views 1. AI is a tremendous economic driver
Inarecentstudy,McKinsey&CompanyestimatesthattheuseofAIisexpectedtogenerateglobalproductivitygainsof0.8-1.4%peryearby2065(asapointofcomparison,theinventionofsteamproduceda0.3%increaseinproductivitybetween1850and1910).7Theseincreaseswillultimatelyincreasetheoveralllevelofwealthinoursocieties.
ThankstoAI,theplanet’scountrieswillenjoymuchhighereconomicgrowthinthefuturethantheywouldhaverecordedwithoutthistechnology(seechartbelow).Asaresult,by2030,AI'scontributiontotheglobaleconomyisexpectedtoreachUS$15.7T,asif10countriesthesizeofCanadaappearedontheplanetandinjecteditwithwealth.8Forexample,theUnitedKingdomestimatesthat,onitsown,AIhasthepotentialtogeneratemorethan£600Binitseconomyby2035.9
AI’simpactonrealgrossaddedvaluegrowth(%)10
Thesemajorgainswillmostlytaketheformofsavings.Thatistosay,thankstoAI,itwillcostlessto:
• respondtocitizensorcustomers(conversationalagents,orchatbots,willbeabletoautomaticallyprocessthemostfrequentwrittenrequestssubmittedtopublicauthoritiesorcompanies).
7Voirhttps://www.mckinsey.com/global-themes/digital-disruption/harnessing-automation-for-a-future-that-works.8Voirhttp://www.businessinsider.com/infographic-ai-effect-on-economy-2017-8.9Voirhttps://newsroom.accenture.com/news/artificial-intelligence-poised-to-double-annual-economic-growth-rate-in-12-developed.10Voirhttps://goo.gl/v1GWVM.
17
FINALVERSION
• preventtheemergenceofseriousdiseases(aMontrealcompany,Imagia,hascreatedsoftwarethat
canpredicttwoyearsbeforeanyradiologistifalungnoduleismalignant).11
• transportgoodsfromMontrealtoQuebec(driverlesstruckswillbeabletotransportgoodsfromone
citytoanother).
• createnewcharactersfromscratchfromexistingimagesorphotosinvideogamecompanies.12
• discoverneworedeposits(aminingcompanyusedAItopredictthelocationof86%ofAbitibi’sgold
depositsbyreviewingdataforonly4%oftheregion).13
• manageacompany'sinventory(Amazoniscurrentlytestingrobotsthatareabletoprepareapackage
andreplenishitswarehouseshelves).14
Thesegainswillalsocomeintheformofinnovations.AIwillallow:
• learnerstoputtheirknowledgeintoactionortakepersonalizedcourses(e.g.,Duolingogives
InternetuserstheopportunitytopracticetheirSpanishwithaconversationalrobotandprovides
lessonstailoredtotheirneeds,strengthsandweaknesses).15
• travelerstoplantheirtrip(themobileappofQuebeccompanyHopperletsInternetusersknow
whentheyshouldbuytheirplanetickettosavemoney).
• researcherstomakestillunimagineddiscoveries(byautomaticallyanalyzingmillionsofscientific
publications,IBM’stoolWatsondiscoveredinafewweeksalmostallthekinases–keyproteins
inthehumanbody–thatthebiologistsacrosstheworldhadtakennearly15yearstoidentify,as
wellastotallyunknownkinases).16
• olderpeopletolivelonger,independentlyandcomfortablyathomebecausearobotcan
completecomplexdomestictasks,suchascookingorcleaning,etc.
Insomecases,companiesusingAIwillevenbeabletocreatenewbusinessmodelsthatwillrevolutionize
theirindustry.
Infact,accordingtoresearchersattheUniversityofToronto,AIfundamentallyreducesthecostofmaking
predictions,and“whenmakingpredictionsbecomescheap,morepredictionsaremade.”Initially,the
authorsstress,thealgorithmswillbeabletoperformcertainpredictivetasksnowentrustedtohumans,
therebyenablingorganizationstosavemoney.However,“apredictionmachinemaybecomesoaccurate
andreliablethatitchangeshowanorganizationdoesthings.SomeAIswillaffecttheeconomicsofa
businesssodramaticallythattheywillnolongerbeusedtosimplyenhanceproductivityinexecuting
againstthestrategy;theywillchangethestrategyitself.”17
Insummary,AIisamajortechnologythatQuebeccompanieswillneedtostartintegratingasquicklyas
possibletoremaincompetitiveattheprovincial,Canadianandinternationallevels.Tothisend,theywill
needtheclosesupportofmanystakeholders,includingthoseinresearchandtechnologytransfer.
11Voirhttp://ici.radio-canada.ca/nouvelles/special/2017/02/intelligence-artificielle/apprendre-creer-predire-cancer-go-
science.html.12Voirhttps://futurism.com/these-people-never-existed-they-were-made-by-an-ai/.
13Voirhttps://www.techemergence.com/ai-in-mining-mineral-exploration-autonomous-drills/.
14Voirhttps://www.nytimes.com/2017/09/10/technology/amazon-robots-workers.html.
15Voirhttp://bots.duolingo.com/.
16Seehttp://ici.radio-canada.ca/nouvelles/special/2017/02/intelligence-artificielle/apprendre-creer-predire-cancer-go-science.html.
17Agrawal,A.,J.GansetA.Goldfarb(2018).PredictionMachines–TheSimpleEconomicsofArtificialIntelligence,HarvardBusiness
ReviewPress.
18
FINALVERSION
2. The importance of having a common language
Theconceptofartificialintelligenceemergedin1950,withAlanTuring'seponymoustest:Canacomputer
communicateconvincinglyenoughtopersuadeahumanbeingthatitishuman?Coinedin1955to
describethefirstlectureonthesubjectatDartmouthCollege,ingeneraltermsandfromanapplication
pointofview,theterm“artificialintelligence”isusedtodescribemachinesorprogramscapableof
exhibitingbehaviourscommonlyattributedtohumans,suchaslearning,problemsolvingornatural
languageunderstanding.
Becauseitstimulatestheimagination,arousesavarietyofinterestsandhasevolvedthroughscientific
progress,thetermAIcancreateconfusion.Toavoidmisunderstandings,measureoutefforts,andabove
all,demystifyartificialintelligence,webelieveitiscrucialtoestablishacommonAIlexicalfieldbyfirst
makingaconceptualdistinctionbetweentwolevelsartificialintelligence:“weak”artificialintelligence,
whichdesignatesamachineorprogramcapableofperformingaspecific,potentiallyverycomplexbut
veryspecializedtask,suchasplayingGoefficiently,and“general”artificialintelligence,whichreferstoa
machineorprogramthatcanlearnamultitudeoftasksandtransferknowledgefromoneproblemto
another,asahumanwould.Todate,only“weak”artificialintelligenceformsexist,eventhoughthey
involvemoreandmorecomplextasks.
Whiledifferenttechniques–suchaslogicprogramming–havehelpedadvanceartificialintelligencesince
itsbeginningsinthemiddleofthetwentiethcentury,muchattentionhasnowturnedtomachine
learning,whichoffersthebest-knownwaytoadvanceartificialintelligencethroughalgorithmsthatlearn
fromdataorexperiences.Threecategoriesoflearningalgorithmscanbecited:supervisedlearning,which
involvesextrapolatingrulesfromasetofwell-definedexamples;unsupervisedlearning,whichinvolves
findingunderlyingstructuresinadataset;andreinforcementlearning,whichisaboutimprovingdecision-
makingbyrepeatingexperiencesmultipletimes.
Someofthesemethodsinvolvemodelsandalgorithmsthathaveexistedforalongtime,butwhosescope
hasincreasedtenfoldinrecentyearsduetotheexponentialgrowthofcomputingpoweranddata
availability(BigData).Moreover,significantbreakthroughsindeeplearning–aformofmachine
learning–haveledtotheAIshiftweareexperiencingduetotheabilitytosolvenewtypesofcomplex
problemsthatwereoutofreachwithtraditionalapproaches,forexample,thedevelopmentof
autonomousvehicles.
Breakthroughsindeeplearningandotherformsofmachinelearningarewhatisdrivingthecurrentboom
inAIandarewhatmanypeoplerefertoasAIinanarrowersense.Machinelearning,afieldinwhichMILA
isawell-knownworldleader,isgarneringextraordinaryinvestments.Itisalsotheareathatseemstohave
thebestchanceofeventuallyleadingtoaformofgeneralartificialintelligence.Wemustmaintain
Quebec'sleadershipinthisfieldandmarketthistechnologytocreatewealth.
MassiveamountsofdataarerequiredtodevelopAIsolutionsandupdatealgorithms.Thesedatamustbe
prepared,structured,transformed,analyzedandinterpreted,expertisewhichfallsunderthepurviewof
datascience,amultidisciplinaryfieldthatmoregenerallyaimstoextractknowledgefromdatasets.IVADO
–theInstituteforDataValidation–hasworld-classexpertiseindatascienceandisrecognizedforits
operationalresearchhub.Operationalresearchaimstodevelopconceptualmathematicalmodelsto
optimizedecision-makingthatinvolvescomplexissues.Therearesynergiesbetweenthisdisciplineand
machinelearning,asevidencedbytheCanadaExcellenceResearchChairinDataScienceforReal-Time
Decision-Makingresearchprogram,whichispositionedatthecrossroadsofthesedisciplines,aswellas
bythefundamentalscientificobjectiveofIVADO’sCanadaFirstprogram,whichistobuildonthe
synergiesbetweenoperationalresearchandmachinelearning.
19
FINALVERSION
Also,itshouldbenotedthatAIispartofawiderdigitaltransformationonwhichitdependsandwhose
componentsarecloselyconnected,inparticulartheInternetofThings,digitizationandautomationof
processes.Allofthesetechnologiespositiondatanotasaby-productbutasan“asset,”anew“raw
material”thatmustbegenerated,exploited,analyzedandvalued.
Finally,othertechnologiesthatarestillattheresearchstagearepoisedtoacceleratethegrowthof
artificialintelligence.Quantumcomputers,forexample,coulddrasticallyincreasethecomputingpower
availableforAIoptimizationproblemsandapplications.Quebechasrecognizedexpertiseinthisfield,as
shownbythe$33.5MCanadaFirstResearchGrantawardedtotheUniversitédeSherbrookein2015.
3. The need to make choices and focus our efforts
WiththeriseoftheAI,Singaporehaschosentofocusitsattentiononcertainaspectsofitseconomy.
WiththecreationofAISingapore,thecountryhasannounceditsambitiontorevolutionizethreesectors
bydevelopingadvancedAItechnologies.Thesethreesectorsarehealth,financeand“smartcity”
technologies,whichincludeurbanmobility.
Moreover,withthecreationoftheDataScienceConsortium(DSC),Singaporeisattemptingtocapitalize
onitsexistingstrengthsinadvancedanalyticsanddatasciencesthroughacademicandindustry
collaborations.Specifically,thegoaloftheSDCistofacilitatetheadoptionandmarketingofmoremature
technologiesinthreeadditionalsectors:sales,logisticsandthemanufacturingsector.
TheSteeringCommitteeisconvincedQuebecmustalsoacquiresectorialexcellenceifitwantstowinthe
AItechnologyrace.Inotherwords,toensureitsleadership,Quebecmustmakechoicesandsticktothem.
Thisconcentrationofeffortsisallthemorecriticalinacontextwheretalentsarelimitedandpublicand
privatesectorinvestmentsinQuebeccannotmatchthoseofotherstates,suchasChina,ortechnological
giantssuchasGoogleandAmazon.
Inits2018-2019budget,theQuebecgovernmenthastaggedcertainprioritysectorssuchaseducation
andhealth,inwhichmorethan$1.6Band$5.4Binadditionalinvestmentsareplanned,respectively.The
governmentisalsoplanningmajorinvestmentsofmorethan$650Mtoprotecttheenvironmentand
promoteenergytransition,particularlyinthemobilitysector.Thesemainprioritiesarecomplementedby
otherkeysectorsoftheeconomythatthegovernmentplanstosupport,includingcleantechnologies,
biofood,lifesciencesandthefinancialsector.Thedevelopmentofsupplychainprojectsisalsooneofthe
government’sgoals,asisthestrengtheningofQuebec'sforestrysector.MILA,IVADOandSCALE.AIhave
alsoidentifiedkeyareasofintervention:health,mobility,18logistics,
19resourcesandgreenenergy,as
wellastradeandfinance.
BasedontheprioritiesofMILA,IVADOandSCALE.AI,theSteeringCommitteeconcludesthatthesectors
thattheQuebec’sAIecosystemwillneedtofocusonoverthenextfewyearswillmostlikelyinclude
health,energy,mobility,logisticsandfinance.
Notethatthefinalchoiceofsectorstobefavouredwillbemadeintheimplementationplanofthe
strategydefinedinthisdocument,afterfurtherconsultationwithecosystemstakeholders.Itshouldalso
beemphasizedthattheresultsachievedintheimplementationofthebusinesscoachingprograms
referredtoinRecommendation7mayalsoaffectfuturepriorities.Forexample,mappingdoneinvarious
18Nousdéfinissonslamobilitécommelacapacitédedéplacementoulavueholistiquedutransportetdudéplacementdes
personnesetdelamarchandise,avecousansvéhicule.19Par«logistique»,nousentendonsl’ensembledesprocessusetdesactivitésliésàlagestiondeschaînesd’approvisionnement.
20
FINALVERSION
industriesmayshowthatAItechniquesinwhichQuebecstakeholdersspecializewillhaveagreaterimpactinsomesectorsornichesthaninothers.20
Itisimportanttoemphasizethatthispositioningdoesnotrepresentexclusionandisnotaquestionof“choosingwinnersandlosers.”Rather,itisaboutfocusingtheeffortsofgovernmentandecosystemstakeholdersinareaswhereQuebeccantrulyplayaleadershiprole.
4. The need to elevate and make Quebec’s ambition sustainable as well as to invest with conviction
ThepoliticalwilltomakeAIadrivingforceintheCanadianandQuebececonomiesisobvious.Inadditiontothesignificantinvestmentsalreadyannouncedbybothlevelsofgovernment,anumberofstructuringinitiatives–includingthecreationofSCALE.AI–havebeenlaunched.
However,theannouncedinvestmentswillnotbeenoughtopositionQuebecasaleader.Atmost,theseinvestmentswillonlyallowittostayintherace.Moreover,thecontinuationoftheinitiativeslaunchedbythegovernmentsinplacedependslargelyontheelectoralcycle.Therefore,whilenotable,theseeffortsarenotenough.
Therefore,webelievethatthegovernmentshouldsustainitseffortsbymakingamuchlargerinvestmentover10yearsinsteadoffive.TogeneratetheimpactdesiredinQuebec,thisinvestmentmustbe$500Matminimum.Weareawarethatthisisabigask,butwebelievethatAIhasthepotentialtobecomearealdriveroftheeconomyifwemakethecollectivechoicetoinvestinit.
Theadditional$300Maddedtothe$202Mcommittedtodatewouldservethreepurposes:
1) Sustaintheanchoringinstitutes,MILAandIVADO,overa10-yearhorizon.2) AcceleratethedevelopmentofAIanddigitalsciencetalentwithintheecosystem.
20Mentionnons,parexemple,queselonMcKinsey,ilyaplusd’avantagesàutiliserlestechniques«avancées»del’IA(commel’apprentissageprofond)quedestechniquesdetraitementdedonnéesplustraditionnellesdansdesindustriescommecellesduvoyage,dutransportoudelaventeaudétail.Voiràcesujethttps://goo.gl/8guv5W,p.13.
21
FINALVERSION
3) Acceleratethetechnologytransferproce
Recommendations
Figure2—Theinterdependenceoftheecosystem’scomponents
22
FINALVERSION
Whilethestrategicdirectionspresentedinthisreportcanbeunderstoodandreadindependently,theyconstituteawholewithinwhicheachcomponentisessentialtoensurethegrowthandsustainabilityofQuebec'sleadershipinartificialintelligence.
Figure2showsthelinksbetweentheAIecosystemelements.Themainissueisonewhichdrivesalltheothers:talent.Atpresent,thesupplyofskilledlabourinthisareaisfailingtomeettheever-increasingdemand.ToprovideleadershipinAIresearch,designanduse,Quebecmustadoptmeasurestodramaticallyincreasetalentinthisfield.TheseindividualsareresearchersandstudentswhoarebuildingtheAIacademicresearchanddatasciencehub.TheyareyoungentrepreneurswhoareactivelyinvolvedinthedesignofAIapplicationsandwillcontributetotheadoptionanduseofAIandAI-relatedfieldsinQuebec.Whentalentdrivesthesethreeareas,marketingandtechnologytransferbecomepossible.TrainingnewtalentboostsAIresearch,designanduseand,throughmarketing,generatessignificanteconomicspin-offsforQuebec.
SomeactionsmustbetakentostrengthenQuebec’sAIecosystem.Expandingthetalentpoolis,ofcourse,closelylinkedtouniversityresearchleadership,whichlargelydeterminesourabilitytotrainandattractnewtalenttotheprovince'seducationalinstitutions.Thisleadershipwillfosterthedevelopmentofnewnationalstart-upsandtechnologytransferintheprovince.Therefore,thisreport’sfirststrategicdirectionproposesvariousrecommendations,whichrunfromhowtoattractrenownedresearcherstorecruitingqualitylocalandforeignstudentsatthegraduatelevel,toensurethegrowthofanAIresearchhub.
ThesecondstrategicdirectionfocusesonbarriersthatspecificallylimitAIanddigitalsciencedevelopmentinQuebec,suchasthelackofvisibilityoftheecosystem’sneedforskills,theconstraintsofreviewingcollegeanduniversityprograms,butalso,thebroaderissuesofmathematicalliteracyandthepromotionofscience.
Foritspart,thethirdstrategicdirectionproposesmeasurestosupportthemarketingprocesstoacceleratetheadoptionanddevelopmentofAIsolutionsacrosstheQuebececonomy.ThesemeasureswillimpactAIdesignorganizations(i.e.start-ups)thathavethepotentialtotransformtheprovince’seconomiclandscapeaswellastechnologytransferandappliedresearchcentres.Themeasureswillalsoimpactuserorganizations(i.e.SMEs,largecompaniesandgovernmentbodies).
TheuseofAIinoursocietynotonlygeneratesasetofsocial,ethicalandeconomicreflectionsincludingthoseregardinggoodgovernanceessentialtothetechnology’sacceptance,butitcanalsobecomearealassetforQuebecattheprovincial,federalandinternationallevels.ThefourthstrategicdirectionfocusesonthedevelopmentofaresponsibleAIhubinQuebec.
Finally,thefifthorientationaimstostrengthentheecosystem’senginetransverselyand“addoiltoitsgears”byupgradingITinfrastructuresandestablishingapermanentorganizationtopromotecollaborationbetweenstakeholders.
.
FINALVERSION
Strategic direction 1
Ensure the development and sustainabil ity of Quebec’s AI
academic research hub
24
FINALVERSION
> Recommendation 1: Increase our abi l ity to attract researchers in areas where Quebec has the potential to become a world leader
FINDINGS
Atpresent,Quebecisrecognizedworldwideforitsexpertiseinartificialintelligenceresearch.Indeed,ithasthelargestconcentrationofmachinelearningacademicsintheworld–about250experts–andhostsoneofthemostwell-regardedmachinelearningresearchcentres,theMILA–QuebecArtificialIntelligenceInstitute.Moreover,severalworld-renownedscientists,includingoneofthethreedeeplearningpioneers,ProfessorYoshuaBengio,arebasedinQuebec.
Inadditiontothismachinelearninghub,QuebecisalsorecognizedforitsknowledgeinAIapplicationsresearch.Forexample,theMcGillCentreforIntelligentMachines(CIM),createdin1985,bringstogether21researchersand150graduatestudentswhoseresearchinAIfocusesontheuseofalgorithmsinrobotics,computervisionandperson-systeminteraction.
Alsoworthmentioning,istheInstituteofDataValorization(IVADO),whichbringstogethermorethan1,000scientistsaffiliatedwiththeUniversitédeMontréal,HECMontréalandPolytechniqueMontréal,whoaredevelopingcutting-edgeexpertiseinthefieldsofdatascience,operationalresearchandAI.AmongIVADO’smembersaretheCentreinteruniversitairederecherchesurlesréseauxd’entreprise,lalogistiqueetletransport(CIRRELT)andtheGrouped’étudesetderechercheenanalysedesdécisions(GERAD),twointeruniversityresearchpillarsinQuebecfoundedmorethan30yearsago,whicharehometotheCanadaExcellenceResearchChairinDataScienceforReal-TimeDecision-Making.
Finally,wemustmentionthepresenceoftheCentrederechercheendonnéesmassivesdel’UniversitéLaval(CRDM),whichhasmorethan35permanentmembersfromfivedifferentfacultiesworkingonbioinformatics,unstructureddataandtheconfidentiality,ethicsandsocialacceptabilityofdatause.ThemapofQuebecinFigure3showstheextentofQuebec'suniversitynetworkinAIandAI-relatedfields.
Quebec’sacademicresearchhubandhistoricaladvantagearefragile,however.Ontheonehand,thereisamovementtodemocratizeAItechnology,especiallywiththeriseofopensourcecodes.Privatecompanies,includingFacebook,willinglysharealgorithmsthey’vedevelopedinternally.Ontheotherhand,manyjurisdictionsandlargeforeigncompaniesinvestheavilyinAIandAI-relatedfields.Theseinvestmentsareusedtoattractresearchersinparticular.In2016,investmentsbytechnologygiantsamountedtobetween$20Band$30B.21In2017,BeijingannounceditsplantoincreasetheeconomicweightoftheAIsectorby$25Bby2020.22QuebecnowhastheopportunitytostrengthenitsacademicresearchhubinAIandensureitsgrowth.Let'sseizethisopportunityandactquickly.
SUGGESTEDAPPROACH
Phaseno1:Ensurethedevelopmentof,andreinforce,amachinelearninganddeeplearninghub
Asafirststep,werecommendthattheMESIgrantsubstantialfundingtotheMILA–QuebecArtificialIntelligenceInstitutetoensurethedevelopmentofthemachinelearninganddeeplearninghubcurrentlyfocusedatthisInstitute.
ThefundingprovidedtotheMILAwillmakeiteasiertoattractandretainhigh-caliberresearchersinmachinelearning,deeplearningandreinforcement
21McKinseyGlobalInstitute(2017).Artificialintelligence:Thenextdigitalfrontier.22Voirhttps://www.challenges.fr/high-tech/le-plan-de-la-chine-pour-devenir-leader-mondial-en-intelligence-artificielle_488717.
$100Mfundfinancingpriority
25
FINALVERSION
learning,throughwhichAIwillbecomeapowerfuldriverofeconomicandsocialdevelopmentforQuebec.Morespecifically,theamountspaidtoMILAwillbeusedto:
• developyoungtalents(scholarshipsforgiftedstudents,updatedAItrainingprogramsforQuebecers,mathematicalliteracypromotionandthedemocratizationofscience).
• createresearchpartnershipsthatwillallowtheemergenceofnewAIcompaniesandtheadoptedofAIbyQuebecSMEsoperatinginkeysectorsoftheprovince’seconomy.
• improveaccesstothecomputingpowerandcomputationalexpertiseforresearchersandsomeAIfirmstomaketheirworkmoreefficient
• incubatebusinesses.
Alltheseelementswillhelpcreatearichandstimulatingecosystemuniversityresearcherswillappreciate.Highlyqualifiedpersonnelwillalsoberesponsibleforthesupervisionofappliedprojectsandinternshipsforstudentsenrolledinprofessionalmaster'sprogramsandadvancedgraduatediplomaprograms,whichwillrapidlytrainmorethan100graduatestudentseachyear.
ItshouldbenotedthattheMILAwillensurethecloseinvolvementofstakeholdersfromdifferentQuebeceducationalandresearchinstitutions,bothinitsactivitiesandthegovernancestructureitwillsetup.Forexample,itisalreadyplannedthatarepresentativefromUniversitéLavalwillsitontheMILABoardofDirectors.Similarly,discussionsareunderwaytoclarifyhowthisInstitutewillcollaboratewithresearchersoutsideUniversitédeMontrealandMcGillUniversity.
InvestinginMILAandinresearchtogrowtheeconomyisasensiblechoiceThedecisiontomakeMILAoneofthedrivingforcesbehindthegrowthofQuebec'sAIsectorandeconomicdevelopmentisinlinewiththatofothercountries.Forexample,theU.K.'snewAIstrategyaimstoprovidetheAlanTuringInstitutewithresourcesthatwillallowittoexpandrapidly,becomeanationalAIresearchcentreandcollaboratewithforeigncentres.GermanyisamajorplayerinthesectorwiththeDeutscheForschungszentrumfürKünstlicheIntelligenz(DFKI),anorganizationinwhichsome900researchersandgraduatestudents,withstateandindustrysupport,participatein300researchprojects.TheimplementationofSouthKorea'sAIplanislargelybasedonthecreationofanationalresearchcentreaspartofapublic-privatepartnership.Recently,Japanhasusedasimilarapproachbyinvesting$340MtofoundtwonewbasicAIresearchcentres.
Phaseno2:Facilitatepan-QuebecmobilizationtofosterthedevelopmentofAIapplicationsresearchanditspotentialimpact
WeproposethreestepstoensurethegrowthanddurabilityofAIapplicationsresearchexcellence:
1) QuebechasstrengthsotherthanMILAinacademicresearch.Thesestrengths,whichtheprovincecouldbuildontobolsteritssectoralpositioning,includeresearchersworkingonAIapplicationsthatwillunderpinthedevelopmentofQuebec'smaineconomicsectors.ThesestrengthsalsoincluderesearchersfromdifferentdisciplineswhoareassessingtheimpactofAIonindividuals,organizationsandsocieties,thinkingaboutthesocialinnovationsneededforAIpenetrationtoachieveexpectedresultsandconcernedaboutresponsibleAIdevelopment.
26
FINALVERSION
2) WethereforerecommendthattheQuebecgovernmentparticipateinthecreationof20to30
sectoralexcellencechairsinAIapplicationsinQuebec'stop-prioritysectors,suchashealth,mobilityandlogistics,aswellasthesocial,societalandhumanaspectsofthedevelopmentanduseofAI.Whileitisessentialtoensurethedevelopmentofbasicresearchintheareasofadvancedmachinelearning,itisalsocrucialtosupportresearchintotheapplicationsofAIwithinindustrialandacademicsectorsthatareQuebec’sstrength.ThecreationofthesesectoralchairswillencouragethecontributionofallQuebecresearcherstothedevelopmentofAI,basedonthehubsalreadypresentinQuebecuniversities.Itisimportanttodistinguishtheabove-mentionedsectoralchairsfromtheMILA–QuebecArtificialIntelligenceInstitute’sCIFARchairsandtohighlighttheircomplementarynature.TheCIFARchairsensurethegrowthofthefundamentalresearchhubinmachinelearning,deeplearningandreinforcementlearning.ThesectoralchairswillhelpdevelopAIapplicationresearchinsectorsinwhichQuebechasalwaysbeenverystrong.Theywillalsointensifyworkonsocial,societalandhumanaspectsofAIpenetration.Moreover,itisexpectedthataportionoftheCIFRAfundstheMILA–QuebecArtificialIntelligenceInstitutereceiveswillberedirectedtothird-partyuniversities(e.g.,UniversitéLaval)tosupportthedevelopmentofexpertiseinmachinelearning,deeplearningandreinforcementlearningthroughouttheprovince.OneofthethreeFondsderechercheduQuebec(FRQ),theFRQ-NT(NatureetTechnologies),theFRQ-S(Santé)ortheFRQ-SC(SocietéetCulture),couldberesponsibleforthecreationandmanagementofsectoralchairsofexcellence.TheFRQ’sthreeresearchfundscouldcollaboratetomakeitpossibletoawardchairsinvariousfieldsrangingfromhealthtotransporttothesocietalimpactofAI.Thedesignatedfundwillneedtoappointacommitteeofexpertstoevaluateapplicationsforthegrantingofthechairs.TheGovernmentofCanada,throughfederalinstitutionssuchastheCanadaResearchChairsProgram,willalsobeaskedtoparticipateinthecreationofthe20to30AIchairs.Wedonotrecommendaspecificprocessforgrantingthechairs.Thedesignatedfundwilldecidewhetheraquotaofchairswillbeallocatedperuniversityorwhetherallchairswillbeaccessibletoalluniversities.Similarly,thefundwillhavetodecideonthenatureanddurationofthechairs.Amongthecriteriaforevaluatingchairapplicationswillbetheabilitytoshowhowtheresearchgroupwillfitintotheecosystemandcollaboratewithcurrentresearchlaboratories.Asanindication,seven-yearchairs,withanannualvalueofabout$250,000each,wouldrepresentatotalcostofapproximately$35to$52.5M,dependingonthenumberofchairsawarded.Therewillalsobeadditionalcostsinvolvedinoperatingandimplementingtheappropriateinfrastructure.Thegovernmentcouldalsoconsiderpartneringwithprivatecompaniesthatcouldfund,atleastinpart,someofthesesectoralchairs.Chairsareanexcellenttoolandcouldevenbedescribedasnecessaryforattractingandretainingtopresearchers.Nevertheless,wemustadmittheyarenotalwaysenough.Thereareadditionalbarrierstoresearcherrecruitment,someofwhicharestructuraloneswithintheacademiccommunity.Beyondtheestablishmentofchairs,wemustthereforemakesomerelatedproposalstoovercometheseobstacles:
27
FINALVERSION
• Firstofall,letusconsidertheissueofjobcreationforrecruitedresearchers.Theprocessofcreatinganewpositionisnotonlycomplex–theapprovalofboththefacultyandthedepartmentisrequired–butalsoextremelydemandingfortheuniversity,giventhelong-termnatureofthepositionscreated.Wethereforerecommendthattheministèredel’Éducationetdel’Enseignementsupérieursupportthefacultiesanddepartmentsinvolvedincreatingnewpositionsthatwouldresultnotonlyfromthechairsawarded,butalsofromhiresinAI-relatedstreams,whichwillprovidebasiccomputerandmathematicseducation,atypeofeducationthatwillcertainlyincreaseinpopularitybyattractingmorelocalandforeignAIstudents(seeRecommendation2).
• Theretentionofresearchersbecomesanissuewhenachairexpires.Universitiesfindthemselvesinadelicatesituationwheretheycannotaffordtopaytheequivalentsalarytoresearchers.Wethereforerecommendthatthegovernmentputinplace,overthelongterm,achairprogramthatwillpromotetheemergencenotonlyofAI,butofallthesectorsandtechnologiesoftheneweconomy.Overthedecades,thethemesofthechairscanchangedependingontheareasQuebecwishestopromote.ThisprogramwillcontinuetosupportandretainthemostpromisingresearchersinQuebec.
• TheissueofresearcherretentionisalsocrucialforholdersoftheMILACIFARchairs,whichlastfiveyearsandwhosepurposeistofinancethedevelopmentofaresearchhubinmachinelearning,deeplearningandreinforcementlearning.ToretainresearchersastheCIFARchairsareinstituted,werecommendthattheQuebecgovernmentintercedewiththefederalgovernmenttoensuretherenewalofthefundingforthesechairsorthegrowthandsustainabilityoftheCIFARbasicAIresearchexcellencechairs.
• Also,weinvitethegovernmenttoengageinadeepreflectionregardingthesystemicissueofcompensationofferedtouniversityresearchers,whichoftencannotcompetewithhigh-techsectors.
• Lastly,beyondthechairsandgovernmentsupportforthecreationofnewresearcherpositions,Quebecneedstobecreativeifitwantstoattracthighlytalentedresearchers.Forexample,China(seethecasestudybelow)createstailoredrecruitmentpackagesinitsThousandTalentsPlanandrecentmeasuressuchasthecreationoftheAICity(salariesof$1Mormore,grantsforchildren’seducation,guaranteedemploymentforspouses).Wethereforerecommendthatagovernmentearmarkabudgetaryenvelopetosupportuniversitiesinrecruitingworld-classAIandAI-applicationsresearchers.Giventhatuniversitiesdonotallhavethesameresearcherrecruitingissues,wesuggesttheysubmittheirneedstothegovernment,whichwillmanagetheawardingofgrants.
2) Inadditiontosupportinguniversities,werecommendthatthegovernmentsupportQuebec
CEGEPsintheirtrainingandresearchcapacity.Atpresent,manyprojectsareunderwayinthecollegenetwork.Forexample,thecreationofanAttestationofCollegialStudiesinAIandaDiplomaofCollegeStudiesinAIatCollègeBois-de-Boulogne.Toensurethesuccessoftheseprograms,itwillbenecessarytorecruitprofessorsandresearcherswhohaveexcellentknowledgeofAI-relatedfields.However,itisdifficulttoattractfacultytotheseinstitutions,giventheoverwhelmingdemandforAIspecialistsacrosstheecosystem.IfQuebecwantstoattractexcellentprofessor-researcherswithinthecollegenetworktonotonlyeducate,butalsodoresearch,itbecomesimperativetosupporttheCEGEPsinthisprocess,forexample,bycreatingabudgetaryenvelopethatwillbeusedtosupportthecollegenetworkinitsrecruitmentofprofessorsinAI-relatedfields.
28
FINALVERSION
3) Finally,werecommendapplyingtoEmploymentandSocialDevelopmentCanadaandtheministèredel’Immigration,delaDiversitéetdel’InclusionduQuebectoenrollfacultyandacademicresearchersinAI-relatedfieldsonthelistofprofessionsforthe“Volettalentsmondiaux”pilotproject,launchedinJune2017.Thisadditionwouldmakeitpossibletoprocessapplicationsforaworkpermitinjust10workingdays.Already,thelistoftheseprofessionscoversmanyareasoftheinformationandcommunicationtechnologysector.Sinceitisupdatedannually,itisimperativetoactquicklytoimplementthisrecommendation.BetweenJuneandDecember2017,10-dayprocessingwasrespectedfor95%ofthe850applications.Inadditiontoprofessorsandresearchers,thegovernmentshouldalsotakeadvantageoftheupdated“Volettalentsmondiaux”listtoincludeotherAI-relatedprofessions,suchasresearchersvisitingforafewmonthsandAItraineesandprogrammers.
AttractinghighlyqualifiedresearchersisofvitalimportancetothedevelopmentofQuebec’sAIacademicresearchhub.Indeed,thegrowthofacriticalmassofrenownedAIandAI-applicationsresearchersleading-edgesectorsinQuebecwillhelpincreasetheprovince'sprestigeinthisfield,whichwillinturnattractstudentsandotherresearchers.AttractinghighlytalentedresearchersisthereforethekeyissuethatwillpropelthevirtuouscircleofgrowthandsustainabilityofQuebecAIacademicresearchleadership.
CASESTUDY
In2008,ChinalauncheditsThousandTalentsPlan,aninitiativetorepatriate,overa10-yearperiod,2,000Chinese“globalexperts”workinginternationally. These“globalexperts”aredefinedasPhDswhohaveattainedthelevelofAssistantProfessoratarecognizeduniversityorscientif ic institution.In2011,theprogramexpandedtoincludeattractingforeignscientiststoChina.Todate,thisprogramhasalreadyexceededitsobjectives,sincemorethan7,000scientistshavebeenrepatriated.23
TheThousandTalentsPlanprovidesChineseuniversitieswithareservefundtoattracthighlyqualif iedtalentinternationally. Thesefundsareinadditiontothefundsalreadyplannedbyuniversitiestocreateresearcherpositionsandmakeit possibleforuniversitiestooffercompetitivepackagestoworld-classresearchers.
Moreover,Chineseuniversitiesreceivegovernmentgrantsfortheir internationalrecruitmentactivities(travel, promotions,etc.). Thesegrantsarenotasub-componentoftheuniversities’budget,butanindependentbudgetaryenvelope,whichencouragesuniversitiestoexpandtheirrecruitmentbeyondChina.
Finally,universitiesareinvitedtooffercustomizedrecruitmentpackagestothevariousscientiststheywishtoattract.Anysalarysurplusisnegotiated,and,veryoften,otherbenefitsareoffered,suchasarelocationallowance,subsidizededucationforchildren,paidreturnairfaretoreturntovacationinthecountryoforigin,guaranteedemploymentforspouses,mealal lowances,housingsubsidies,etc.Someuniversitiesevenallowrecruitedresearcherstospend50%oftheirt imeintheircountryoforigin.24
23Voirhttps://www.nature.com/articles/d41586-018-00538-z.24ConferenceBoardofCanada(2016).TheThousandTalentsProgram–LessonsFromChinaAboutFacultyRecruitmentandRetention.
29
FINALVERSION
Figure3–Quebecuniversities,AI-relatedresearchcentresandCEGEPnetworkofferinganITprogram
> Recommendation 2: Ensure the inflow – in quantity and quality – of local and foreign graduate students in AI and AI-related f ields
FINDINGS
Currently,universitiesareexperiencingaslightincreaseingraduatestudentenrollmentacrossallfields,especiallyatthemaster’slevel.Ontheotherhand,itismoredifficult,especiallyinareassuchasdatascienceandAIwhereprofilesarehighlysoughtafterintheindustry,toconvincemanystudentstoengageindoctoralstudies.Thejobmarketoffersareattractiveandrarelyrequireadoctorate.Infact,thelowvalueofaPhDintheCanadianlabourmarketisthemainreasongiveninanOECDreport25forthelackofstudentinterestinpursuingstudiesatthedoctoratelevel.Yet,ifQuebecwantstobeaworldleaderinAIresearch,weneedtoattractmorestudentsandsupportthemthroughouttheirPhDstudies.
25Seehttps://www.oecd-ilibrary.org/docserver/5k92pghq4247-en.pdf?expires=1527308148&id=id&accname=guest&checksum=366E9D971EE22CE6576967EB89DDD5E3.
30
FINALVERSION
Tofillpositionsintheirresearchgroups,researcherstodayturntorecruitingforeignstudents.InQuebec,in2017,foreignstudentsrepresented20%ofthestudentpopulationatthegraduateleveland40%ofthestudentsenrolledindoctoralprograms.26Also,QuebecrankedlastinCanadaforthegrowthinthenumberofinternationalstudentsbetween2000and2014.BritishColumbia,theCanadianleaderonthisfront,experiencedagrowthratefivetimeshigherthanthatofQuebec.27Currently,aforeignstudent’spathfromrecruitmenttointegrationintheworkplace,isstrewnwithvariousobstaclesthathinderimmigrationandpermanentsettlementinQuebec.
Tostart,tuitionfeesforinternationalstudentsarehigh,andinternationalstudentsdonothaveaccesstoallthescholarshipsavailabletoCanadianstudents,whichcanmakefindingfundingmoredifficult.
Moreover,thefiguresshowthatin2015,Quebechostedabout43,000foreignstudentsatthepost-secondarylevel,butonly3,500selectioncertificateswereissuedtostudents.28AccordingtoMontréalInternational,thethreemainobstaclestoretainingforeignstudentsareaccesstoemployment(disparityinservicesofferedwithinuniversities,reluctanceofemployerstohirestudentswithatemporarystatus),theimmigrationprocess(unknownprocessthat’sconsideredcomplex,increasedtimeforgrantingpermanentresidence)andtherequiredlevelofproficiencyinFrench(difficulttoattainthelanguageproficiencylevelrequiredtoobtaintheQuebecselectioncertificate).29Yettoday,Quebecisadestinationofchoiceformanyinternationalstudents,accordingtotheQSBestStudentCities2017list,whichranksMontrealasthebeststudentcityintheworld.30Moreover,accordingtotherecentNestpickstudy,31Montrealisthesecondlargestcityintheworldformillennials.ConsideringthathalfoftheinternationalstudentswouldliketostayinQuebecaftertheirstudies32(accordingtoastudyconductedbyMontréalInternational),wemustconcludethatalargenumberofthemreturntotheirhomecountryaftergraduation,whilewecouldhavekeptthemhere.ItwillbeimportanttoensurethatthisdoesnothappeninAIandAI-relatedfields.
ThegrowthanddurabilityoftheQuebecAIhubshouldencouragemanyforeignstudentstoimmigratetoQuebectopursuetheirstudies.Thisisonemorereasontoensurethattheirjourneyissmooth,fromrecruitmenttointegrationintotheworkplace.
SUGGESTEDAPPROACH
WeproposefourmeasurestoensuretheinfluxoflocalandforeignstudentstograduatestudiesinAI-relatedfields:
1) First,werecommendmakingconcertedeffortsandrampingupcurrentinitiativestopromotescienceandtechnologyandmathematicalliteracy,andtoincreasedigitalscienceandAItrainingfromthehigh-schoolleveltotheuniversitylevel.SupplyingandgrowingthepipelineofgraduatestudentsinAI-relatedfieldsbeginswiththedemocratizationofscienceandthepromotionofcareersinthemathematicalsciencesfromthegrade-schoollevelandalsobygrowingtheAItrainingoffering.Recommendations4and5coverthistopicinmoredetail,inparticularbyrespondingtotheneedtosupporteducationalinstitutionsinupdatingtheirtrainingofferingandpromotingscienceandtechnologytoyoungpeopleandwomeninparticular.
26Ministèredel’Éducationetdel’Enseignementsupérieur,Directiondesindicateursetdesstatistiques.27StatisticsCanada.28Seehttp://www.institutduquebec.ca/docs/default-source/recherche/8606_attirer-et-retenir-d-%C3%A9tudiants-intern_br_fr.pdf?sfvrsn=0.29Seehttp://www.montrealinternational.com/wp-content/uploads/2016/08/MI-Consultation-Immigration-2017-2019.pdf.30TopUniversities.QSBestStudentCities2017.31Seehttps://www.nestpick.com/millennial-city-ranking-2018.32Seehttp://www.montrealinternational.com/wp-content/uploads/2016/08/MI-Consultation-Immigration-2017-2019.pdf.
31
FINALVERSION
SeeRecommen
dation12
2) Inordertoattractmorelocalandforeignstudents,wesuggestincreasingthevisibilityandtransparencyoftheresearchofferavailabletothem.Forexample,sharedandcentralized
informationwouldhelpstudentsfindprofessorstosupervisethemduringtheirgraduatestudies.
Atpresent,manystudentswhowishtodotheirgraduatestudieswithQuebecprofessors
recognizedworldwideinartificialintelligencearedeniedtheirapplicationsbecauseofthelimited
availabilityoftheseresearchers.Asharedshowcasewouldhavetheadvantageofdirectingthese
studentstootherexcellentresearchersinAI-relatedfieldsinQuebecand,ultimately,toenrich
theQuebecpoolofhighlyqualifiedtalentinAIandAI-applicationsresearch.
Recommendation12ofthisreportoutlinestheestablishmentofsuchashowcaseinmoredetail.
3) IfQuebecwantstoattractmorehighlyskilledlocalandforeignstudentswhoaredoingresearch
inAI-relatedfieldsandbecomeaworldleaderinsaidfields,werecommendincreasingthefundingavailabletostudentspursuingmaster’sdegree(master'sdegreeinresearch,professionalmaster'sdegree,diplomaofhigherspecializedstudies)anddoctoralstudiesintheseareas.Thisinitiativecouldincludethecreationofabursaryprogramaccessibletothebest
localandforeignstudentsstudyinginAI-relatedfields.Master’sdegreeanddiplomaofhigher
specializedstudies(DESS)studentscurrentlyreceivelittlefunding,buttheseprogramshavethe
advantageofrapidlyincreasingthenumberofstudentsenrolled,asstudentsarenotdirectly
supervisedbyaresearcher.Theobstacleofinsufficientresearchersthenbecomesmuchless
important.Theseprogramsalsoofferinternshipsthathelpsforeignstudentsdeveloprootsin
Quebec.Finally,bursariesofferedtoPhDstudentsinAI-relatedfieldsshouldservetoencourage
themnotonlytobegintheirstudies,butalsotocompletethem.
4) Finally,toretainforeignstudents,werecommendupgradingcurrentprogramssuchasthosedevelopedbyMontréalInternationalandQuébecInternationalaspartofthestrategy
establishedbytheministèredel’Immigration,delaDiversitéetdel’InclusionduQuebec,which
aimstoincreasethenumberofforeignstudentswhosettlepermanentlyinQuebecattheendof
theirstudies.Fortheseprograms,MontréalInternationalandQuébecInternationalreceived
grantsof$1.6Mand$0.2M,respectively,overthreeyears.Theenhancementoftheseprograms
couldincludetheofferoffreeFrenchcoursesforforeignstudentstofacilitateandpromotetheir
integrationandrootinginQuebec.
Asweconcludethisfirststrategicorientation,wemustkeepinmindthatattractinghighlyqualified
researchersinAIanditsapplicationswillfurtherenhancethereputationofMontrealandQuebecasa
wholeandattractmorelocalandforeignstudents.Theaimistocreateavirtuouscirclewheretheeaseof
accesstotalentedstudentswillattractrenownedresearchers.
FINALVERSION
Strategic direction 2
Develop digital science talents to meet Quebec’s needs
33
FINALVERSION
See
Recommen
-datio
nno
12
Foreword
ThepreviousstrategicorientationaddressedtheneedtoensurethegrowthandsustainabilityoftheQuebecAIacademicresearchhub.Theriseofsucharesearchhubisparamountsinceitwillattractresearchersandhighlytalentedstudents.ItwillalsodrivethedevelopmentofatalentpoolindigitalsciencetomeettheneedsofQuebecemployers.SincemostresearchersandPhDandpostdoctoralstudentsteach,growthinthepoolofstudentsandfacultydoingresearchwillincreasethenumberandrelevanceofAI-relatedcoursesofferedatalllevels(college,university,in-housecorporatetrainingandtechnologytransferstakeholders),therebyhelpingtotraintheskilledworkforcethatQuebecrequires.
Inaddition,othermeasuresmustbeputinplacetomatchtrainingtoemploymentandsupporteducationalinstitutionsinupdatingtheirtrainingoffer.
> Recommendation 3: Develop and maintain through a “talent watch” a clear and comprehensive perspective on needs, gaps and opportunities regarding digital science ski l ls
FINDINGS
TheareasofAIsolutionandapplicationdevelopmentaregrowingandrequireanincreasinglyabundant,qualifiedanddiversifiedworkforce.However,wecurrentlyhavelittleknowledgeoftheexactsizeofthisworkforce,thespecificneedsoftheindustryintermsofjobprofilesandthenumberoftalentedpeopleinthedigitalscienceQuebecinstitutionsproduce.Afirstapproximationhasmadeitpossibletoestimatethatthetalentgapisatleast1,000graduatesperyear,anumberthatispotentiallysettoincreaseoverthemediumterm.Theneedfortalentedpeopleisobvious,but,currently,itisdifficulttodeterminetheexactnumberandpreciseskillsrequiredwithinbusinesses.
SUGGESTEDAPPROACH
Wedonotrecommendtrainingaspecificnumberofstudentsorcreatingaparticularprogram;weareoftheopinionthattheexactmagnitudeandnatureoftheneedsaretoounknowntogetspecificaboutnumbers.
Instead,wesuggestgettingabetterunderstandingoftheneedsoftalentedpeoplethroughouttheecosystemsothattheycanmobilizeandadaptquickly.Thistransparencyofneedsiscriticaltoenableeducationalinstitutionstomodifytheirtrainingofferingsaccordingly,employersadapttheirrecruitmentstrategyandstudentsinterestedininformaticsormathematicslearnmoreaboutopportunitiesintheseareas.
Toimprovethisperception,werecommendsettingupa“talentwatch”thatwouldhavethefollowingmandate:
1) Conductaprofessionalstudytoassessindustryneeds,currentsupplyandtrainingprogramsandidentifyfutureprofilesandestimatethecurrentandprojectedtalentgap.ToensuretransparencyaboutfutureprofilesinareasrelatedtoAI,wesuggestprofilesbemadeaccessibletostudents,thuspromotingtheirinterestinsectorswheretheneedisgreatest.
$100Mfundfinancingpriority
34
FINALVERSION
SeeRecommendation12
Common
platform
2) Provideaquickoverviewofthecurrentofferofshort“conversion”programsandcontinuing
educationprograms;alertecosystemstakeholdersandthegovernmentifashortfallisidentified.
Thissecondpointisessentialtorapidlyincreasethemarket’slabourforcewithouthavingtowait
thethreeorfouryearsrequiredtotrainacohortofundergraduatestudents.Inotherwords,to
ensurethatnewgraduatesinscience,technology,engineeringandmathematics(STEM)
programs–whoalreadyhaveexcellentmathematicalliteracy–canquicklyconvertandoptfora
careerrelatedtoAIanddatascience.ThishasbeenSingapore’sapproachwithitsAI
Apprenticeship,anine-monthprogramthatincludesasix-monthindustryinternshipaimedat
providingpracticalknowledgeandAIworkexperiencetorecentgraduatesincomputerscience
andmathematics.Studentsarepaidbetween$2,000and$3,000amonthforthefulldurationof
theprogram.InitsAIstrategy,theU.K.hasalsooutlinedplanstostudythepotentialforaone-
year“conversion”master’sdegreeforSTEMgraduates,sothattheycanapplyAItotheir
respectivespecialties(e.g.,biomedicalsciencesandAI).Asthesegraduatesalreadyhavethe
fundamentalskillstoworkonAIapplications,theycouldbequickly“converted”toworkintheAI
field.
Inaddition,areviewofalltheAIandAI-relatedtrainingprogramsavailableshouldbeeasily
accessibletoecosystemstakeholderstohelpthemtoquicklyfindtheprogrambestadaptedto
theirneeds.Thislistingwouldincludetrainingprogramsatthecollegeanduniversitylevelbut
alsotheentirecontinuingeducationandonlinecourseofferings.Onthislastpoint,Quebec,as
partofitsPland’actionnumériqueenéducationetenenseignementsupérieur,hasalreadyexpressedinterestinsettingupan“eCampus,”anonlineplatformthatwouldlistalltheonline
coursesofferedbyQuebec’spost-secondaryinstitutionsandcouldpotentiallybecomeatrue
onlineeducationcollaborativecampus.Moreover,the“talentwatch”willneedtotieinwith
currentinitiatives,particularlytheeCampusproject,andoptimizeitsmarketingtomakeQuebec
educationmorevisibleinternationally.
3) Performcompetitiveintelligencebyobservingwhatisbeingdoneinternationallyintermsof
trainingoffersandmatchingtrainingtoemployment.
Werecommendthatthis“talentwatch”beplacedundertheresponsibilityofAI.Quebec(see
Recommendation11),supportedbyanexistingorganization(e.g.,TECHNOCompétences)whoseprimary
rolewillbetostudyneeds.AI.Quebecwillbesupportedbyacommitteeofstakeholdersfromeducational
institutionsandindustry,whowillprovidesupportandadvice,whilerigorouslyvalidatingthe
methodology,resultsandanalysis.Thepresenceofgovernmentleadershipwithinthewatchisessential
todemonstratetheseriousnessoftheprojectandensureactivememberparticipation.
35
FINALVERSION
TalentprofilediversityinAIandAI-relatedfieldsInthisreport,wemakenorecommendationsastothenatureofthetrainingtoimplement,butwehave,throughinterviewswithecosystemstakeholders,identifiedvarioustypesofneeds.Figure4givesanon-exhaustiveviewofthedifferentprofilesweidentifiedduringthisfirstinitiative;theobjectivehereistoillustratethediversityofprofilesrequiredtomeetthegrowingneedsoftheartificialintelligenceecosystem,andmorebroadly,inthedigitalsciences.
Profilesfallintotwobroadcategories:artificialintelligenceanddigitalsciencespecialists,thatis,thosewhorequiretechnicalknowledgeintheseareas(e.g.,researchers,datascientists,machine-learningengineers)and“interface”stakeholders,whomusthaveanunderstandingoftheapplicationsandtransformationpotentialofartificialintelligenceandadvancedanalytics(e.g.,interfacescientists,translators).AlloftheseprofilescanbespreadoveraspectrumthatrunsfromR&DtotheuseandapplicationofAIsolutions.
Inlightoftheseefforts,itappearsthatsomeprofileswillrequirethecreationormodificationofcertainprograms,suchastheadditionofanintroductoryAIcourseforMBAstudents,forexample.Onthislastpoint,theFrenchgovernment,initssynthesisreportonFranceandAI,hasalreadytakenstepstoshiftcoursesbyannouncingtheintroductionofAImodulesinmanagementschoolstomakefuturebig
decisionsmakersawareofAIissues.33
33Seehttps://www.economie.gouv.fr/files/files/PDF/2017/Rapport_synthese_France_IA_.pdf.
36
FINALVERSION
Figure4–IllustrationoftheprofilediversityrequiredintheAIecosystem
37
FINALVERSION
> Recommendation 4: Support educational institutions in updating their training offerings
Establishinga“talentwatch”willbebeneficialforallecosystemstakeholders,includingeducationalinstitutions,whichcanuseittoreviewtheirprogramsaccordingly.Itisimportanttoensurealinkandatransparentdialoguebetweenthe“talentwatch”andeducationalinstitutions.Atthesametime,certainbarrierspreventuniversitiesandCEGEPsfrombeingflexibleandagileinupdatingtheirtrainingoffer.
FINDINGS
Currently,Quebec'scollegeanduniversityinstitutionsarestrugglingtoobtainthefinancialresourcestheyneedtoquicklyupdatetheirtrainingofferingstomeetchangingneeds.34However,theyhavebeguntorevisesomeoftheirtrainingprogramstoaddressthetalentgapinAI-relatedareas.UniversitédeMontréal,HECMontréal,PolytechniqueMontréalandMcGillUniversity,togetherwiththeMILA,arecommittedtocreatingnewprogramsbyofferingaprofessionalmaster'sdegreeandadiplomaofhigherspecializedstudiesinartificialintelligenceinSeptember2018.Thelatterofthetwoprogramsrunsforoneyear,includingafour-monthin-companyinternship.UniversitéLavalwillalsoofferaprofessionalmaster'sdegreeinAIbynextfall.CEGEPsarealsoparticipatinginthisshiftwiththedevelopmentofanAttestationofCollegeStudies(ACS)diplomaandaDiplomaofCollegeStudies(DCS)inAIatCollègeBois-de-Boulogne.
Quebecwillbenefitgreatlyfromtheseprograms,whichwillquicklytraintalentindigitalsciences;however,theyarenotlikelytomeettheprovince'sfulldemandforskilledlabour,whichissorelyneeded.EducationalinstitutionsmustactquicklytoaddintroductorycoursesinthedigitalandAIfieldstoalloftheirscience,technology,engineeringandmathematics(STEM)programsandcreateonlinecoursestosupportregionalstudenttraining,todevelopAIorientationsorconcentrationsincurrentundergraduateprogramsandbuildnewAIundergraduateprograms.Moreover,CEGEPsmustconsiderexpandingtheirtrainingofferingstomeettheoverallneedsofQuebec.Finally,asmentionedinthepreviousrecommendation,itiscrucialtoevaluatecurrentcontinuingeducationprogramsandshortconversionprogramsfornewSTEMgraduateswhocouldquicklyacquireknowledgeindatasciencesandAIandenterthejobmarket,allinlessthanayear.
ToensurethecreationofapoolofAItalent,QuebeceducationalinstitutionsmustreceivefinancialsupportfromtheQuebecgovernment.
Also,thetimerequiredtocreateorupdateAI-relatedandallSTEMprogramscurrentlyrangesfromtwotothreeyears,whichmeansthatitcanfiveyearsfromthebeginningofthereviewtothefirstcohortenteringthelabourmarket.
SUGGESTEDAPPROACH
1) Asafirststep,werecommendthatfinancialsupportbeprovidedtocollegesanduniversitiestosupporttheminupdatingtheirtrainingoffering.GiventheexponentialgrowthintheAIanddigitalindustries'needfortalent,itisimportanttoensurethattheinstitutionsthattrainthistalentwillhavethemeanstodoso.ThissupportcouldtaketheformofagovernmentbudgetaryenvelopeearmarkedtosupportQuebec'seducationalinstitutionsinupdatingtheirtrainingoffering.Theseinstitutionscouldsubmittheirneedstothegovernment,whichwouldmanagetheawardingofgrants.Thisgovernmentbudgetaryenvelopemustincludeadedicatedbudgetto
34Entrevuesavecdesacteursdel’écosystème.
38
FINALVERSION
supportonlinecoursecreation,thebenefitsofwhichforQuebecareincludedinthisdocument’sAppendices.
2) WealsorecommendsimplifyingandspeedingupthegovernmentapprovalprocessfortrainingprogramsofferedbyQuebeceducationalinstitutionstoensuregreaterflexibilityandagilityinupdatingtheirtrainingofferingtomeetchangingneeds.ThegovernmenthasalreadymadecommitmentsonthisissuebyredesigningtheapprovalprocessforvocationalandtechnicaltrainingprogramsfortheQuebeccollegenetwork.Wealsosuggestthattheuniversityapprovalprocessbereviewed,asuniversitieswantandneedtorapidlyamendtheirtrainingofferingstomeetthegrowingneedforaskilledAIlabourforce.ToensurethatthetrainingprogramsmakesthebestpossibleuseoftheQuebecuniversityecosystem,wesuggestthatthisacceleratedprogramapprovalprocessbeappliedfirstandforemosttoprogramssubmittedjointlybytwouniversitiesormore.Interuniversitycollaborationshouldbepromotedtoreduceredundancies,ensurebaselinecoordinationbetweenuniversityprogramsandoffersubstantivetrainingprogramsthatmakeitpossibleforstudentstoaccessthedepartmentsandexpertiseatseveraluniversities.
> Recommendation 5: Promote mathematical l iteracy and the democratization of science
FINDINGS
Currently,thereisasignificantglobalshortageofdigitalandmathematicalsciencetalent.However,adirectlinkexistsbetweentheseskillsandartificialintelligence:excellentknowledgeofmathematicallanguageisnecessarytounderstanddatascienceandAIbasics.Therefore,feedingthepipelineinthesetwoareasmeansfeedingthepipelineinthemathematicalsciences.Tofillthistalentpool,wemustfirstandforemostmakeaQuebec-wideefforttodemocratizescienceandpromotetrainingandcareersinscience,technology,engineeringandmathematics(STEM)toyoungpeople.ThismovementwillreapbenefitsnotonlyinthedigitalandAIfields,butalsoinallSTEMtrainingprogramsandfuturecareers.
Theunder-representationofwomenisanotableissueinSTEMstreamsthatmustbeaddressed.AccordingtoareportbyTDBank’sChiefEconomist,althoughwomenarebetterrepresentedinthephysicalandlifesciences,theyaccountforonly20%ofengineeringgraduatesand25%ofcomputerscienceandmathematicsgraduates,despitethefactthatthesefieldsaccountformorethanthree-quartersofSTEMjobs.35Indeeplearning,theproportionofwomenisevenlower,ataround10%.36
Anotherfinding:Today,inQuebec,about21%ofthepopulationdisplaysthelowestlevelofmathematicalliteracy,alevelcorrespondingtotheabilitytoperformsimplemathematicaloperationsusuallyinvolvingasinglestep,andonlyonein10Quebecershavehighmathematicalliteracyskills.37GiventherapidgrowthofAIanddigitaltechnologies,theworkersoftomorrowwillbecalledupontousemoreandmoreAI-baseddigitaltools.Therefore,itisessentialtoensurethattheentireQuebecworkforcehasasufficientlyhighlevelofmathematicalliteracytobeabletoacceptandusethesetools.
35Caranci,B.(2017).LesfemmesenSTGM–Comblerl'écart.36Interviewwithecosystemstakeholders.37Desrosiers,H.etautres(2015).Lescompétencesenlittératie,ennumératieetenrésolutiondeproblèmesdansdesenvironnementstechnologiques–DesclefspourreleverlesdéfisduXXIesiècle.RapportquébécoisduProgrammepourl’évaluationinternationaledescompétencesdesadultes(PEICA),Quebec,InstitutdelastatistiqueduQuebec.
39
FINALVERSION
WiththesupportofstakeholdersliketheGovernmentsofQuebecandCanada,manyorganizationsandinitiatives(e.g.,theBoîteàscienceoftheCapitale-NationaleandChaudière-Appalachesregions,theAssociationpourl’enseignementdelascienceetdelatechnologieauQuebec,theAmbassadeursscientifiquesduCentrededéveloppementdelarelèvescientifiqueettechnologiqueproject)arehelpingQuebecers,particularlyyoungpeople,betterappreciateandunderstandSTEM,aswellaspromotingsuccessionplanninginthiscriticalareaforthefutureofAI.Forexample,inthecomingyears,thelatesteditionoftheMESI'sNovaScienceprogramwillmakepossible36projectsaimedatenrichingscienceandtechnologyactivitiesforvocational,collegeandundergraduatestudents,raisingawarenessamongelementary,secondaryandadulteducationstudentsanddevelopingprofessionalpracticesamongscienceandtechnologyteachersinelementary,secondaryandpre-universitycolleges.In2016-2017,NovaSciencelaunchedmorethantwodozenawarenessandtrainingprojects,threeofwhich,conductedaspartnerships,werespecificallydesignedtoincreaseinterestinSTEMingirlsandwomen.
Thatbeingsaid,whereSTEMpromotionisconcerned,thereisaclearneedforcollaboration,scalingupandsharingofbestpracticesacrossQuebec
SUGGESTEDAPPROACH
TheGovernmentsofQuebecandCanadaalreadysupport,throughtheiractionsandvariousprograms,thedevelopmentofafutureworkforcethatcanbetterunderstandandmeetthechallengesofSTEM-relatedissues.Aspartofthe2017-2022QuébecResearchandInnovationStrategy–DaringtoInnovate(Stratégiequébécoisedelarechercheetdel’innovation20172022–Oserinnover),theQuebecgovernmenthasdecidedtostrengthentheNovaScienceprogrambyinjectinganadditional$7.55Mby2021-2022.
Ourrecommendationisnottoreinventthewheel,butrathertobetterhighlightandcollaborateonthe
maininitiativesundertakeninmathematicalliteracyandSTEM-careerpromotionandincreasetheir
impactbyensuringoptimalconsultationbetweenexistingorganizations.
Wesuggestthatanecosystemstakeholderbegiventhetaskofcoordinatingcurrentscience,technologyandmathematicalliteracypromotioninitiativestoincreasefirepoweracrossQuebec.Quebecneedssuchanentitytoleveragetheappropriatemeansandmeettheneedsofthoseresponsibleforexistinginitiativesintermsofintegration,consultation,evaluation,expansionandbestpracticesharing.Australiaestablishedsuchabodyin2002withtheAustralianMathematicalSciencesInstitute(seecasestudy).
WestressthatthesupportofsuchanentityinQuebecwouldhaveapositiveimpactontheAIanddatasciencefieldaswellasonallSTEM-relatedfieldsinwhichQuebechasopportunitytoexcel,forexample,videogamesandvirtualreality,statisticalbiology,preventativemedicine,finance,economicsandengineering.
40
FINALVERSION
CASESTUDY
TheAustralianMathematicalSciencesInstitute(AMSI)wascreatedtostrengthen
collaborationandconsultationregardingthepromotionanddemocratizationof
mathematicsandsciencewithinAustraliansociety.TheAMSI'smissionistodramatically
improvethemathematicalskil ls andcompetenciesofAustralians, includingsupportfor
high-levelmathematicseducationforyoungpeople,bygrowingthenumberofstudents
inthefieldwhoareproficientinmathematical l iteracyatthepost-secondarylevel,
supportingresearchinthemathematicalsciencesandimprovingtheexperienceof
mathematicsstudentsattheuniversitylevel.
AMSIoperatesonthreefronts:researchandhighereducation,schooleducationand
industrycollaboration.AmongthemanyprogramsdevelopedbytheInstitutesinceits
creation,ChooseMathsisanationalprogramthathasreceived$22Minfundingoverf ive
yearsfromtheBHPBil l itonFoundation.ChooseMathsaimstochangethepublic's
perceptionofwomeninmathandsciencecareers.Toachievethis, theprogramplansto
developmathematicseducationin120Australianschoolsandaddresourcesinall of
theseinstitutions,establishanationalcareerpromotioncampaigninmathematical
sciences,createanetworkof“inspiringwomeninmathematics”andhandouttheBHP
Bil l itonFoundationforExcellenceinTeachingandLearningMathematicsannualawards.
FINALVERSION
Strategic direction 3
Accelerate the development and adoption of AI solutions
within Quebec’s economic fabric
42
FINALVERSION
See
Recommen-
dation12 Common
platform
Foreword
QuebeccurrentlyhastheopportunitytohelpcreatecompaniesdestinedtomakeamarkintheAIsector.
However,manybarriershinderthelaunchofsuchstart-upsabletoexpandrapidlyorsuccessfully
integrateAIintoSME’smore“conventional”ITproductsandservices.Forexample,thecurrentmarketing
ecosystemdoesnotallowQuebecstart-upstoeffectivelymarkettheirAIproductstoQuebeccompanies.
Similarly,manyobstaclescurrentlypreventmanyQuebeccompaniesactiveintheprimary,secondaryand
tertiarysectorsfromtakingadvantageofemergingAIinnovations.Forexample,manycompaniesdonot
fullyunderstandthepotentialofAIontheiractivitiesandprocessesandhavedifficultydefiningtheirown
needs.ThechallengeisevengreaterforSMEsthatdonothaveeasyaccesstoacademicresourcesand
consultingfirmsthatcanhelpthemunderstandthepowerofAIandmakelogicaltechnologychoices.
Below,weoutlinethreerecommendationstofacilitateboththemarketingandadoptionofAI
applicationsacrossQuebec'sindustrialfabric.
Lastly,astheecosystemiscomplexandconstantlyevolving,itisdifficultforcompaniestoworkwithinit
andfindsuitablepartners.Facilitatingstakeholdernetworkingbetweenresearchers,designers,usersand
businessadvisersiscriticalandwillbecoveredinthefifthstrategicorientationofthisreport.
> Recommendation 6: Support Quebec businesses that design AI solutions, including start-ups, so they can develop, grow and remain in Quebec
FINDINGS
Inthecurrenttechnologicalenvironment,thekeytoeconomicsuccessistoensurethatnewbusinesses
taketheplaceofbusinessesthatwillbephasedoutandthatthe“winners”enjoyconditionsthatpromote
theirgrowthinQuebec.
Quebecmustcreateandretaintheeconomicvaluegeneratedbynewtechnologiestooffsetjoblosses
associatedwiththeproductivitygainsthesetechnologiesgenerate,whilealsoattractingforeign
investment.
ThecurrentecosystemofAIapplicationcreatorsincludesanumberofpromisingstart-upsthathave
securedfundinglocallyandinternationally.QuebeccompanieslikeElementAIandHopperraised$102M
and$82Mrespectivelyintheirlastroundoffinancing.Quebec’s(andmorespecificallyMontreal’s)
venturecapitalecosystemisdoingwell.In2017,Montrealbusinessesinallsectorsraised~$800Mwithin
thescopeof63transactions,aheadofToronto($779M).Start-upfundsandbusinessincubatorsthat
favourAIstart-upsarebeingsetup.Forexample,theQuebecgovernmenthassupportedCreation
DestructionLab(CDL)andNextAI,twogovernment-backedorganizations,tothetuneof$10Minits2018EconomicPlan.Montreal’sCreationDestructionLabwantstosupportninecohortsofabout25AIstart-
upsoverthenextthreeyears.NextAIisa225-hourtrainingprogramthatenablesstudents,professionals
andentrepreneurstodevelopandimplementAI-basedideas.
ThefundinggapduringgrowthphasesinCanadanotedbytheAdvisoryCouncilonEconomicGrowth
seemstobeclosing.InitiativessuchastheVentureCapitalCatalystInitiative(CCRI)andtheCanadian
43
FINALVERSION
BusinessGrowthFund(BGF),withsomeoftheirfundingwhichwillbeinvestedinQuebec,attemptto
meetthisneed.Bothofthesefundsareexpectedtocontributemorethan$1Bnationally(overafive-to
10-yearperiod).Also,investmentfundsdedicatedtotechcompanygrowthcapitalaresettobe
announcedshortly,includingtheNordaCapitalfund.
However,therearestillmajorchallengestoovercome:
1) Accesstodata:AIapplicationdevelopmentrequiresmassiveamountsofgooddata.Today,the
largestmassesofdataareheldbylargecorporationsandthegovernment.Andyetdataisthe
mostvitalinput–therawmaterial–oftheemergingeconomy.AIcompanies’lackofaccessto
thesedatabasesisafactorthatlimitsthedevelopmentoftheirbusinessactivities.
2) Accesstocomputinginfrastructures:AIapplicationdevelopmentrequiresconsiderable
computingpower.WhileQuebechasmajorcomputinginfrastructures,suchasCalculQuébec,
theyaremainlyusedforuniversityresearch,andaccesstothemislimitedforstart-ups(thisissue
willbediscussedlaterunderRecommendation10).
3) LackofAItalent:Asisthecaseforotherstakeholdersintheecosystem(universities,research
centres,etc.),accesstotalentiscurrentlyamajorchallengeforstart-ups.Theaddedchallenge
forentrepreneursistheneedtofindsalesandmarketingtalent;theserolesrequirebothabasic
knowledgeofAIandathoroughknowledgeofthepotentialclientbase.
4) Lackofalocalmarket:Localstart-upsmustoftenseekinternationalcustomersbecauseofthe
lackofmajor“firstcustomers”inQuebec,thatis,anchoringorganizationswiththeinterest,
meansandmaturitytoworkwithlocalstart-ups.
SUGGESTEDAPPROACH
Phaseno1:EnsureAIdevelopmentcompanieshaveaccesstoworkabledatabasesfromthepublicandprivatesectors
Quebec'sresearchcentres,start-upsandSMEswillrequirelargeamountsofdatatoproducetheAItools
thatwillhelpdevelopQuebec’seconomy.
TheissueofAIdataisdecisivesinceapplicationdevelopmentinthisfieldrequiresmassiveamountsof
gooddata.AsaChinesefinancierpointedout,inAI“themoredatayouhave,thebetteryourproduct;the
betteryourproduct,themoredatayoucancollect;themoredatayoucollect,themoreyoucanattract
talent;themoretalentyoucanattract,thebetteryourproductwillbe.It'savirtuouscycle[...].”38
Unfortunately,thisindispensablerawmaterialisoftenverydifficulttoaccess.Fewcountrieshavelarge
companiesthat,likeGoogle,Amazon,FacebookorBaidu(theChineseGoogle),haveaccesstoalmost
infiniteamountsofdatathankstotheirhundredsofmillionsofcustomer-users.Start-upsandSMEsthat
areactiveinAIwouldbenefitgreatlyfromtheopennessofdatacollectivelyheldbycompaniesinsectors
suchastransportationorinsurance.However,forfearofweakeningtheircompetitiveadvantage,these
sectorsaregenerallyreluctanttosharetheirdata.
AIfirmsaroundtheworldalsooftenhavedifficultyaccessingdatafromnationalorsub-national
governments,whichareamongthe“richest”datasources.Sometimesgovernmentagencieswouldliketo
makepublicdataavailabletocompanies,suchasthoseintheAIsector,todevelopproductsandservices
ofbenefittosocietyasawhole,buttheycannotdosowithoutadequatelytakingintoaccounttherights
38Seehttps://www.nytimes.com/2017/06/24/opinion/sunday/artificial-intelligence-economic-inequality.html.
44
FINALVERSION
andwishesofitscitizens.TheBritishpublichealthauthorityandGoogleDeepMindlearnedthislessonthehardwayaftersharingtherecordsof1.6millionpatientswithouttheirknowledge.39
Intheend,therelativescarcityofdataavailabletoQuebec'sstart-upsandSMEscouldaffecttheirinternationalcompetitivenessandtheirabilitytohelpdevelopQuebec’seconomy.
Therearewaysaroundthismajorhurdle,however.Forexample,countrieslikeGreatBritain(seecasestudybelow)haveconsideredcreatingdatatruststoensurebetterdisseminationanduseofthedatacompanieshold.Inshort,thesecountriesplantosupport:
• Thepoolingofdataheldbyaparticularsetoforganizations(e.g.,companiesandorganizationsactiveinaspecificsectorofactivitysuchashealth)ontheiractivities,“customers”(whetherthesearecitizens,patientsorconsumers)orotherpoints.
• Themanagementandcirculationofthesedataundercertainconditions(e.g.,theirprioranonymisationandexclusiveusefornewmodeltraining),whichwouldnotonlybeestablishedbytheparticipatingorganizations,butalsobythepeoplewhosedataisconcerned(e.g.,transitortaxiserviceuser,hospitalorclinicpatient).
Trustsshouldhavetheresourcesnecessaryfortheeffectivemanagementanduseofthedataundertheircontrol,aswellastheprotectionoftheinterestsofallpartiesinvolvedinthesharingofsuchdata(e.g.,healthfacilities,patients,researchersandAIcompanies).
CASESTUDY
IntheBritishreportGrowingtheArtif icial Intell igenceIndustryintheUK , theissueofdataaccessibil ity isdeemedimportant.Thereport’sthreerecommendationsonthispointare:
–Developdatatrusts,whicharesecuredatabanks,tofacil itatetheexchangeofsensitivedatawhileensuringtrust inthedatasharingprocess.
–Producemoredatafromresearchactivitiesthatmachinesareabletoprocess.
–Promotedatamininginall industrial sectorstoencouragecompaniestoleveragedata.
Thedatatrusts,whichconsistofsecuredatasharingbetweenmultipleparties,hasthegreatestpotential. Atpresent,thesesharingagreementsareadhoc, andorganizationsdonotalwaysknowhowtogoaboutsharingsensitivedata.TheUKreportthereforerecommendedthatthegovernmentandcompaniesformoneentity,theDataTrustSupportOrganization,whoserolewouldbetoestablishbestpracticesforsharingdataandprovideall publicandprivateorganizationswithausermanualguaranteeingasecureprocess(templates,securedatatransferprocess,etc.)
IntheirAIstrategy,allcountrieswanttomakebetteruseoftheStateasasignificantassettohelpbusinessesaccessthedatatheyneedtobuildsuccessfulapplications.Asmentionedabove,FinlandhasmadeopendatafromFinnishministriesandagencies,anditsuse,apriority.
39Seehttps://www.courrierinternational.com/article/royaume-uni-google-acces-aux-dossiers-medicaux-de-16-million-de-patients.
45
FINALVERSION
Currently,datacollectedbytheQuebecgovernmentandQuebecmunicipalitiesisoftenverydifficulttoaccess.A2016CEFRIOstudyconductedbyÉcolenationaled’administrationpublique(ENAP)researchersshowsthat“despitetheenthusiasmithasgeneratedoverthepastfewyearsattheinternationallevel,opendatamostlyremainsanemergingphenomenoninQuebec.”40Atthetimeoftheliaisonandtransfercentre’sstudy,only17%ofmunicipalitieshadreleaseddata,comparedwith60%oftheQuebecgovernment’sministriesand28%ofitsagencies.
However,thesepercentagesconcealimportantfacts.First,inthe12monthspriortotheCEFRIOsurvey,noneoftheQuebecministrieshadreleasedmorethanfivedatasetsandonly10%oftheagencieshaddoneso.41Second,amongthegovernmentorganizationsthathadnotyetreleaseddata,veryfewwereplanningtodosowithin12monthsofthestudy.Moreover,manygovernmentcomputersystemsdonotallowformachineaccessorautomaticaccesstoministryoragencydata,andfewofthesedata(lessthan15%oftheoveralltotal)areautomaticallyupdated,twocharacteristicsthatmakethemlessattractiveandlessusefulforcompaniesandresearchersworkingintheAIsector.
Accordingtothesurvey,Quebecgovernmentministriesandagenciescitethelackofgoodorrelevantopendataasthereasontheyhavenotyetreleaseddata.Insomecases(especiallyintheministries),theproblemisthedataistoosensitive.“Weobserve,”sayENAPresearchers,“thattheassessmentofthevalueandrelevanceofopendataisnotalwaysasstrongonthesideoftheoriginalownerastheprospectiveuser.”42 ».
AccordingtotheauthorsoftheQuebecreport,thereleaseofgovernmentdataisnotyetapriorityforallministries,agenciesandmunicipalities,afactthatisreflectedinthelackofinvestmentsandeffortsmadeinopendata.Thus,inthe12monthsprecedingthesurvey,58%ofdepartmentsand83%oforganizationshadinvestedlessthan$25,000(excludinghumanresourcescosts)intheopendataproject.And,onaverage,theannualinvestmentsinopendataforministriesandagencieswerelessthan50person-days.
Therefore,dataisthekeyinput–the“rawmaterial”–oftheneweconomythatisemerging.LackofaccesstoqualitydatabasesforQuebecAIdesigncompaniescouldbeafactorthatlimitsthedevelopmentoftheircommercialactivities.ThegovernmentwillhavearoletoplayincreatingfertilegroundforAicompanies.
Thedevelopmentofinitiativestoremovesomeobstaclesfromtheprocess,suchassettingupaworkingcommitteewhich,underthecoordinationoftheministryofFinance,haslookedatthegeneraldatabankaccessprocessforresearchpurposes.However,wewillhavetodomoreinthecomingyearstosupportthedevelopmentoftheAIsector.
Therefore,wesuggesttwomajorcoursesofactiontosignificantlyincreaseaccesstodataforQuebec'sAIstakeholders:
1) First,werecommendthatthegovernmentparticipateinthecreationofapoolofanonymizeddatainspecificsectors,suchashealthortransportation,incollaborationwithlargecompaniesandotherdataholders,suchasmunicipalities.AccesstothesedatabasescouldbegiventoAIstart-upsandecosystemresearchers,whichwouldinturnbenefitlargecompaniesandthegovernment,astheywillbeabletoimprovetheirproductsandservicesthroughtheapplicationsdeveloped(seetheUKdatatrustcasestudyabove).
40Seehttps://cefrio.qc.ca/media/uploader/RAPPORTNETGOUVDONNESOUVERTES-Final.pdf.41Ibid.42Seehttps://cefrio.qc.ca/media/uploader/RAPPORTNETGOUVDONNESOUVERTES-Final.pdf.
46
FINALVERSION
Thegovernanceofthesedatatrustswillbeapriorityissuetobeaddressedduringtheplanning
phase.Itisimperativetoquestionthesovereigntyofourdataandtheresponsibilityofthe
Quebecgovernmenttomakethemaccessibleanddeploythemtoincreasethecollectivewell-
being.Beforeimplementinglarge-scaledatatrusts,itwillbevitaltostudytheinsandoutsofthe
conceptandvalidatetheirapplicationinoneortwoindustrialsectors.Thesetaskscouldbe
entrustedjointlytoAI.QuebecandtheobservatoryonresponsibleAIandtheimpactsofAI(see
Recommendations9and11).
2) Then,werecommendthattheQuebecgovernmentmakethereleaseofopendatacollectedbyallpublicandparapublicsectororganizationsarealpriorityforeconomicdevelopment.Inparticular,thegovernmentshould:
a)Intheshortterm,forsharingandusepurposes,reinforceQuebec'sopendatapolicies
sothatitappliestoallitsministries.ThemodeldevelopedbytheSOQUIJ(seethecase
studyunderRecommendation8)couldserveasabasisforthisreinforcementwork,
sinceitreflectstherealitiesofanorganizationthatreportstoaQuebecministryand
wishestoinnovatethroughtheuseofAI.
b)Overthelongerterm,amendtheActRespectingAccesstoDocumentsHeldbyPublicBodiesandtheProtectionofPersonalInformation,notonlytopromotetheadoptionof
AIbythegovernment,butalsotogivetheecosystemtheopportunitytouse
anonymizedopendatainitsownapplicationdevelopment.Quebechasrichdatabases
inthehealthsector.Addressingquestionsregardingtheiropeningmustbemadea
priority.Clearly,arigorousbutagileprocessneedstobeimplementedtoavoid
problemsandpotentialleaksofsensitiveinformation(seeStrategicdirection4).
Thegovernment’sdualrole…Thegovernment,asanimportanteconomicandsocialplayerinQuebec,mustadoptbestpracticesand
actboldlytoopenupitsdatatopromotetheiruseforthebenefitofallQuebecers,whichwillinturn
drivethedevelopmentofnewapplicationsbyandforQuebec.Thisaccesstodatadoesnotinanyway
removetheQuebecgovernment’sresponsibilityregardingprivacyprotection.However,inadditionto
privacyconsiderations,thegovernmentisnowresponsibleforrollingoutandpromotingdataforthe
commongood.
Phaseno2:Developa“firstcustomer”marketforecosystemstart-upsbysubsidisingQuebecbusinessAIprojects
ToencourageQuebeccompaniestodobusinesswithstart-upsthatare
developingAIsolutions,werecommendtheQuebecgovernmentearmarka
budgetenvelope,inthespiritofSingapore'sSMEsGoDigitalprogramandthe
FinnishGovernment’sInnovationVouchers,43whilerespectingQuebec'strade
agreements.Thisgovernmentassistancewouldhelpsupportcompaniesthat
developAIsolutions,includingstart-ups,whilepromotingthedigitalshifttocompaniesthatarepartof
themoretraditionalsectorsofQuebec’sindustrialfabric.ForlargerQuebecstart-upprojectswithaprice
43Seehttps://www.businessfinland.fi/en/for-finnish-customers/services/funding/sme/innovation-voucher/.
$100Mfund
financing
priority
47
FINALVERSION
tagover$500,000,projectexpensesmaybeeligibleforR&Dtaxcredits,whichwouldcreatesignificant
IncentiveforcompaniesthathaveaheadofficeinQuebecandareofaminimumsize(tobedetermined
whentheprogramisdevelopedindetail).SuchinitiativescouldalsoenticeentrepreneurstoQuebec.
Sometaxcreditsarealreadyavailable,likethosethatsubsidizeprivatepartnershippre-competitive
researchortheintegrationofITinSMEs.
AsubsidyprogramtopromotetheshifttoAIwouldevolveasfollows:
• First,thegovernmentwouldallocateabudgetenvelopeforthisprogramthatcouldbe
managedbyanNPOthatcanfacilitatenetworking(e.g.,Prompt).Thisenvelopecould
beusedtoimproveonexistingtaxcredits.
• Next,theNPOwouldactasafacilitatorwithintheindustrialnetworktohelpcompanies
identifythebestpartnerstomeettheirneeds.Forthesakeoftransparency,thelistof
alltheAIcompaniesthatdesignsolutionscouldbeconsultedonthegovernment
programwebsiteortheAI.Quebecplatform.TheNPOwouldalsoofferfinancial
structuringservicesthatmeetthegovernmenttaxcreditcriteria.
• AIcustomerslookingtomodernizeorimprovetheirproductsandservicescouldcontact
thesecompaniesdirectlytoassesstheirneeds,knowingthatsomeoftheprojectcosts
wouldbecoveredbyataxcredit.
Atthefederallevel,werecommendthattheIndustrialResearchAssistanceProgram(IRAP),managedby
theNationalResearchCouncil(NRC),beenhancedtoincreasethenumberofAIprojects.Throughthis
program,NRCpromotesthedevelopment,adoptionandadaptationoftechnologieswithinCanadian
SMEs.IRAPhasteamsofIndustrialTechnologyAdvisors(ITAs)thatassesstheneedsofSMEs.ThisSME
consultinganddiagnosticservicebranchshouldhavecloselinkswithinstitutionssuchastheMILA–
QuebecArtificialIntelligenceInstituteandIVADO.
Phaseno3:Makeiteasierforinnovativedigitalscienceand,especially,AIcompaniestoaccesspubliccontracts
Positioningthegovernmentorapublicorparapublicsectorcompanyasachampionoramodeluser
wouldhaveacatalyticeffectonQuebec'sAIecosystem.Tothisend,webelieveitisimportantforthegovernmenttocontinuemakingstrategicprocurementimprovements(asperRecommendation6oftheCCEI),tomakeiteasierforinnovativeserviceproviderstoaccesspubliccontracts.
Despitewhatissometimesheard,itseemsobviousthatthecurrentnormativeframeworkdoesnot
includea“lowestbidder”policyandthatQuebecpublicbodieschoosetheadjudicationmethodbest
suitedtotheirneeds.Moreover,inrecentyears,thisframeworkhasbeenenhancedtoallowpublic
organizationstousenewapproachesortoolsinthedigitaldomain(e.g.,cloudcomputing,totalnet
acquisitioncost,competitivedialogueprocedures)thatofferincreasedprocurementflexibility.The
Quebecgovernmenthasrecentlyimplementedmeasuresthatpromotethedevelopmentoftechnology
companiesandinnovationintegrationinministryandagencyprocurementprocesses(e.g.,Passeport
Innovationprogram,theStratégiegouvernementaleentechnologiedel’informationGovernmentandthe
reversetechnologyshowcase).TheAIecosystemanditsfuturecoordinatingbodywillneedtoworkin
closecollaborationwithQuebec’snewpublicprocurementsupportoffice,whichQuebecgovernmentset
upinaccordancewiththe2017reportoftheComitédetravailsurl’entrepreneuriatdesjeunes
entreprises.
However,theQuebecgovernmentshouldcontinuetoexplorenewavenuestohelpitsministriesand
agenciesgetthesupporttheyneedtomeettheoften-complexchallengestheyface,andforwhichno
single,ready-madesolutionexists,fromstart-upsandSMEsactiveinemergingdigitalsectorslikeAI.In
48
FINALVERSION
particular,theQuebecgovernmentshouldstudythemeritsofadaptingtheU.S.SmallBusinessInnovationResearch(SBIR)programorthefederalInnovativeSolutionsCanadainitiative.
InnovativeSolutionsCanadaisa$100millionprogramthataimstostimulatetheexpansionandgrowthofCanadianinnovatorsandentrepreneursbymakingthefederalgovernmentthenumberoneclientforCanadianinnovations.Intotal,20federaldepartmentsandagencieshavesetasideaportionoftheirbudgettofundinnovativesolutionscreatedbyCanadiansmallbusinesses.ThisfederalprogramshouldbeactivelypromotedtoinnovativecompaniesinQuebecasagatewaytothepublicsector.
Phaseno4:SupportthedevelopmentofAIstart-upsbysubsidizingcertainmarketingexpenses,ratherthanR&Dalone
Existinggrantprogramsmustincludecertainmarketing-relatedexpenses,whichareexcludedfromthescientificresearchandexperimentaldevelopmenttaxcredits(R&D)taxcredits.Forexample,aportionofthesalariesofkeymarketingpersonnelmaybeeligible.Inanexportcontext,someoftheseexpensesarecoveredbygovernmentprograms.Forexample,theExportProgram,wherethehiringofanexportspecialistorthecostofattendinginternationaltradefairsiseligibleforfinancialassistance.Thesemeasurescouldbeadoptedwithinabroadermarketdevelopmentframework.Therefore,werecommendthatthegovernmentreviseitsgrantprogramstoincludecertainexpensesrelatedtomarketingactivities.
Ingeneral,thegovernmenthasaplethoraofhigh-qualityprogramsforinnovativecompaniesandbusinessesinallsectorslookingtomakeadigitalshift.TheseprogramsincludetheAuditIndustrie4.0Auditformanufacturingcompanies,thePerforMEstrategyandthePasseportInnovationProgram,whichbringtogetherseveralmeasurestohelpQuebecbusinesses.However,mostcompaniesdonotseemawareoftheseprograms,sotheyarenotbeingleveragedtotheirfullpotential.Toincreasethevisibilityofalltheseprograms,wesuggestthattheybelistedonacommonplatformaccessibletoallbusinesses,suchasEntreprisesQuebec.44
CASESTUDY
InSingapore,thegovernmenthasincreaseditsSMEdigitalprojectf inancingprogramstosupportincreasinglysophisticatedneeds, includingAI. Thistypeofprogramwouldbeidealforcreatinga“firstcustomer”marketforourAIdesignfirms,whileenablingtraditionalbusinessestogodigital.
TheiSPRINTprogram,launchedin2010,aimedtoencourageSMEmodernizationbyofferingagrantofupto70%ofthecostofbuyingandimplementingdigitalsolutionspre-approvedbythegovernment,uptoamaximumof$1MperSME.Start-upsorcompanieswithinthelocalecosystemprovidethesesolutions,therebygettingaccesstotheirf irstcustomers.ThisnetworkingbetweenprivatecompaniesalsoreducesSME’sdependenceontechnologytransferNPOs.
Thenewversionofthisprogramcalled“SMEsGoDigital”waslaunchedin2017andhelpsthesecompaniesreachamoreadvancedlevelofdigitization.Theprogramnowcaterstocompaniesbeyondthefirstdigitalprojectstage.Pre-approvedsolutionsnowincludeadvancedformsofadvancedanalytics,suchasAI. Thisprogramhasalreadyenabledmore44Seehttps://www2.gouv.qc.ca/entreprises/portail/quebec.
49
FINALVERSION
than100SMEsadoptdigitalsolutions,whichhavegeneratedmorethanUS$2.4Mfortheprogram’spre-approvedlocalsuppliers.Moreover,tohelpcompaniesnavigatemorecomplexdigitalnotions,thegovernmenthassetupaSMEDigitalTechHubtoofferconsultinganddiagnosticservicesaswell asSMECentres,pointsofcontactforSingaporeanSMEs.Finally,thisnewprogramprioritizeslarge-scaleprojectsinkeysectorsoftheir industryaswell ascorporatecollaboration.
IAandintellectualpropertyTodeveloptheQuebecAIsector,theissueofintellectualpropertyneedstobecarefullyconsidered.Inthecomingyears,manystart-upsandinnovationswillemergefromresearchconductedinuniversitylaboratoriesorcolleges.
However,thetransferofintellectualpropertytotheindustrialsectorpresentssignificantchallenges.TheauthorsoftheU.K.AIstrategynotethatspin-offisapracticesomeresearchinstitutionsarebetteratthanothers(amongotherthings,notallinstitutionshavethecapacitytoworkatthepacecompaniesexpect).Theyaddthatstakeholdersexperienceothercommonproblems.Asaresult,determiningthereturnforeachofthepartiesinvolvedinatransferagreementisusuallyacomplexandcontentiousprocess.
Therefore,itwillbeimportanttoensureitbecomeseasierforexistingentrepreneursorbusinessestomakeefficientandrapiduseoftheintellectualpropertydevelopedbyresearchersanduniversityandcollegestudentsinQuebec.
Newstrategiesandapproachescouldbedevelopedtomakethetransferrulesforintellectualpropertyheldbyuniversitiesandcollegesmoretransparentandconsistent.ItmightalsobeagoodideatocreateacentralizedgroupofexpertswhoserolewouldbetohelpresearchcentresandcompaniesinQuebec,particularlysmallones,understandintellectualpropertyissuesandcarryouttransactionsthatwouldbeofbenefittothevariousparties.
ToimproveQuebec'sIPenvironment,ecosystemstakeholderswillneedtoleveragetherecentinjectionof$85.3MoverfiveyearsannouncedbytheGovernmentofCanadainitsIntellectualPropertyStrategy,“tohelpCanadianbusinesses,creators,entrepreneursandinnovatorsunderstand,protectandaccessintellectualproperty(IP)throughacomprehensiveIPStrategy.”45
> Recommendation 7: Ensure future private sector AI users receive adequate support
FINDINGS
Contrarytowhatsomepeoplemaythink,gettingthemostoutofAIisdifficult,inthesamewayit’shardtotakefulladvantageofIT.
AsresearchersattheUniversityofTorontohavenoted,“AIisageneral-purposetechnology.Ithasthepotentialtoaffecteverydecision,becausepredictionisakeyinputtodecisionmaking.Hence,no
45Seehttps://www.ic.gc.ca/eic/site/108.nsf/eng/home.
50
FINALVERSION
managerisgoingtoachievelargegainsinproductivitybyjust“throwingsomeAI”ataproblemorintoan
existingprocess.Instead,AIisthetypeoftechnologythatrequiresrethinkingprocesses.”46
However,reviewinganorganization'sprocessesindepthrequiresexpertiseandtheinvolvementofall
staffmembers.“Datascientistscandevelopusefulalgorithms,butdomainexpertsareneededtohelp
trainthemachinetorecognizeimportantpatternsandunderstandnewdata.Domainexpertsincludetop
analysts,contractmanagers,salespeople,recruiters,andotherspecialistswhoarenotonlyexpertsat
theirjobsbutwhoareacutelyawareofhowtheydeliverexcellence.”ForAItotrulybealever,says
technologyspecialistJeanneRoss,“itmustimproveontheskillsofthesespecialists.Responsibilitiesmust
alsoberedefinedsothattheyareempoweredandmotivatedtorolloutmachineswhentheybelieve
doingsowillimproveresults.Ultimately,youneedpeoplewhocanuseprobabilisticoutputtoguide
actionsthatmakeyourcompanymoreeffective.”47
ManyQuebeccompanieswillrequiresupporttomakethisshift.Indeed,fewofthem(especiallySMEs)
currentlyhavetheknowledgeneededtoimplementthetechnicalandsocio-organizational
transformationsonwhichtobuildaneffectiveuseofAI.
DirectandindirectsupporttocompaniesthatdesignAIsolutionswillbethefirststeptoensurethe
transferoftechnologiesdevelopedinourresearchcentresandleverageQuebec’sdata.Inadditionto
beingneweconomicdriversintheirownright,thesestart-upswillhelplocalSMEsandlargecompanies
integrateAIintotheirprocessesandbusinessmodels.
OthercompaniesarealsomobilizingtosupportpotentialAIusers,withconsultingfirmsincreasingly
involvedintheAIsectortohelptheircustomerstakeownershipofthetechnology.Moreover,leadersof
companiessuchasThalesCanadawanttoestablishpracticecommunitiestopromotepeerlearningand,
onanotherlevel,developcommontechnologyroadmapsthroughwhichtocommunicateneedstothe
localAIecosystem.
Nevertheless,othertechnologytransferchannelswillhavetobeconsideredtoensureAI’sadoption
throughoutthelocalindustrialsector.Forexample,IVADO,MILAandSCALE.AIwillallparticipateinthis
technologytransferprocesswithexistingQuebeccompanies.BecauseAIisanewtechnology,fewSMEs
andlargecompaniesareadequatelyinformedandequippedtoundertaketheAIshift.
Avastecosystemofgovernment-subsidizedtechnologytransferagenciesexistsinQuebectoprovide
SMEsandlargebusinesseswithaccesstonewtechnologiesandsupportservices.Theseorganizations
include,amongothers,theCentrescollégiauxdetransferttechnologique(CCTT),theindustryresearch
clustersandtheothermembersoftheQuebecInnoveplatform.Today,theseorganizationsdonot
necessarilyhavetheexpertiserequiredtoprovideAIservicestobusinessesanddonothavetheoverall
strengthtorespondtodemandacrossQuebec.Indeed,alargepartoftherequestsandquestionsrelated
toAIcomingfromQuebecbusinessesaresenttoresearchinstitutes(e.g.,IVADOorMILA),whicharenot
abletoanswerthemfully.
Quebec'smodelofappliedresearchcentresandinnovationvaluingislargelydecentralized.Acrossthe
province,thereareintermediatetechnologytransferorganizations,butnoindustrysectorhasacritical
massofresearchersandengineersdedicatedtoappliedresearch.Forexample,theTransTechnetwork,
whichincludesallcollegetechnologytransfercentres,employs1,300expertsinnearly50centres.While
manyorganizationsnowplayaroleinthetransferofcross-sectoraltechnologiessuchasdigital
technologies,theformasubsetofthesestakeholdersmighttakeinthefuturemustbeconsidered.
46SeeAgrawal,A.etal.,op.cit.
47Seehttps://sloanreview.mit.edu/article/the-fundamental-flaw-in-ai-implementation/.
51
FINALVERSION
SUGGESTEDAPPROACH
WeseetwopotentialfieldsofactiontoimproveQuébec'stechnologytransferinfrastructure:aseriesofactionstoberolledoutquickly,andanothertobeimplementedoverthenexttwoyears.
Phaseno1:UpgradetheAIknowledgeofQuebececosystemliaisonandtransferstakeholdersintheshortterm
1) First,werecommendupgradingtheAIknowledgeandskillsofasmallgroupofsupportplayerswhoalreadyprovideconsultingservicestobusinessesandareideallypositionedasa“gateway”forthem.Tothisend,about10“champion”organizationsshouldbechosen,fromtheroughly70existingorganizations,todedicatepartoftheirresourcestoAIanddatasciencetechnologytransfer.Academicinstitutions(i.e.MILA,IVADOanduniversities)wouldprovidetrainingusingatrain-the-trainerapproach.Followingtheecosystembenchmarkanalysis,adozenorganizationsappearasnaturalcandidates:thefollowingcentrescollégiauxdetransfertdetechnologie(CCTT):NovikaSolutions,theCentrederobotiqueetdevisionindustrielles(CRVI),theCentrededéveloppementetderechercheenimagerienumérique(CDRIN),theCentreenimagerienumériqueetmédiasinteractifs(CIMMI),theCentredegéomatiqueduQuebec(CGQ)andtheCentredeproductionautomatisée(CPA),aswellastheCentrefacilitantlarechercheetl’innovationdanslesorganisations(CEFRIO),theCentrederechercheinformatiquedeMontréal(CRIM),theAssociationpourledéveloppementdelarechercheetdel’innovationduQuebec(ADRIQ),theCentrederechercheindustrielleduQuebec(CRIQ),ProductiqueQuebecandtheNationalOpticsInstitute(INO).TheseorganizationsarementionedinFigure5.
2) Also,itisimportantthegovernmentsupporttheseintermediaryorganizationsindevelopingconsultingteamsabletoassessacompany’smaturityorreadinesstoadoptAIandAI-relatedapplicationsbyidentifyingthemostpromisingusecasesandhelpingthesecompaniesfindthebestpartnersfortheirneedswithintheecosystem.Inthiscontext,acommunicationchannelbetweentheseorganizationsandeducationalinstitutionsisneededtoensureproximitytothelatesttechnologicaldevelopments.TheQuebecInnoveplatformandtheAuditIndustrie4.0initiativecouldserveasalinkbetweenthevariousorganizations.
$100Mfundfinancingpriority
52
FINALVERSION
Figure5–GeographicaldistributionoftechnologytransferorganizationsnaturallypositionedtoreceiveAItraining
Phaseno2:CreateamajorsupportprogramtoenablehundredsofcompaniestoexecuteanAIproject
Severaltechnologytransferstakeholdersoperateinspecificsectorsandhaveaspecialrelationshipwiththeindustrialclustersandthenetworkofcompaniesthatmakethemup.OncetheAIknowledgeoftheirstaffisupgraded,thesestakeholderscanparticipateintheimplementationofprogramstosupporttheseclustersandorganizationsintheirAIshift.
TohelpSMEsmaketheAIshift,werecommendcreatinganimportantcoachingprogramthatwillallowseveralhundredcompaniesexecuteanAIproject,withthesupportofliaisonandtransferagenciesandrepresentativesofQuebec’smainindustrialclusters.48
TheAIcoachingprogramcouldbemodelledonpastgovernmentprograms,suchasPME2.0,whichvariouspartnerspilotedinthe2010stohelpbusinessesinQuebec'saerospace,fashionandretailindustriesleveragethepotentialofinformationandcommunicationtechnologies.49
48TheAIshiftiscontributingtodataenhancementinbusinessesandisalogicalextensionoftheprogramsalreadystartedunderthePland’actionenéconomienumériqueetdelaStratégiequébécoisedelarechercheetdel’innovation.CompanieswithhighdigitalmaturitycanuseAItoleveragetheirdataandimprovetheirproductivityandcompetitiveness.Forexample,companiesthathaveparticipatedinexistingdigitalshiftgovernmentprogramsmayultimatelybenefitfromtheprogramsrecommendedinthisreporttomaketheAIshift.49From2012to2015,theMESImandatedCEFRIOtoimplementthePME2.0programandhelp30aerospaceandfashionandclothingSMEsaffectadigitalshift.Thesecompaniesreceivedsupportfromindustrypartners,academicresearchersandtechnologyexperts,andhaveinvestedcloseto$10Moftheirowntodiagnosetheirlevelofdigitaladvancement,plantheirdigitalstrategy,implementtechnologyadoptionprojects,suchasERP(EnterpriseResourcePlanning)packages,andensuretheacceptanceandeffectiveownershipbytheirstaffofthetoolsimplemented.CEFRIOanditspartners(e.g.,AéroMontréal)thenwidelydisseminatedtheknowledgegainedfromthese30experimentsthroughthewww.pmenumerique.caportalandothertransferactivities.ThepopularityofthisprogrampromptedtheMESItosupportasecondeditionofPME2.0,whichwillendin2018.Thisnewphaseaimstosupportthedigitalshiftofretailandmanufacturingcompaniesaswellassocialeconomyorganizations..Formoreinformation,seehttps://cefrio.qc.ca/projets-recherches-enquetes/numerique-entreprise/pme-20/.
$100Mfundfinancingpriority
53
FINALVERSION
Initially,callsforapplicationsopentoallorganizationsinvolvedintechnologytransferandknowledge
transferontheuseoftechnologieswouldbelaunchedtoselect“champions”tosupportcompanies
throughthisprogram.Underthisinitiative,governmentfundswouldthenbeusedto:
• assessQuebec’smainindustrialclusterstodeterminehowcompaniesactiveintheprovince
coulduseAItocreatenewbusinessmodels,launchinnovativeproductsorservicesorincrease
theirproductivity,therebybecomingmorecompetitive.
• determinewhichindustrialclustersrequirepriorityintervention,particularlyinviewofAI’s
potentialforeachsector.
• providecustomizedsupporttosomeofthecompaniesintheseclusters(i.e.10to15SMEsin
eachclusterableandwillingtoundertakeanAIprojectandinvestsomeoftheirresources)so
thattheycanintegratetheAIeffectively.
• codifytheknowledgeandknow-howdevelopedbytheaboveSMEsandtransferthem,toallthe
companiesintheselectedclusters.
Thiscoachingprogramcouldalsoincludeaphasetohelpexistingcompanieswithintheinformation
technologycluster(suchas,SMEsthataremembersoftheAssociationquébécoisedestechnologies
(AQT))acquireadvancedAItechnicalknowledgewithaviewtoimprovingtheircurrentproductorservice
offering,byinjectingmoreintelligenceintocustomerrelationshipmanagement(CRM)software,a
documentmanagementtooloraninventorymanagementsystem,forexample.
AcoachingprogramthatcomplementsothergovernmentactionsThecoachingprogramoutlinedabovefitsinperfectlywithoneoftheStratégienumériqueduQuebec’s
objectives,whichis“tomakedigitalizationanengineforbusinessgrowthandapowerfulleverfor
industrialdevelopment,especiallythroughmajorinvestmentsinIndustry4.0andthereinforcementof
ecosystemsthatfacilitateprofitableinteractionsbetweenthedigitalsector’ssupplyanddemandfrom
othereconomicsectors.”TheQuebecgovernment's2018EconomicPlanembodiesthisobjectiveina
numberofways.Overthenexttwoyears,thegovernmentwillinvest$60Mtolaunchacallforproposals
sothatcompaniesinvariousindustrialsectorscanbidonmajorprojectstomarketnewproducts,
processesandservices,addresscommonchallengesandtestnewbusinessmodels.Evenifitdoesnot
focusontheuseofspecificsolutions,thisnewprogramwillbebeneficialtoQuebecSMEslookingtouse
AItosolvetheproblemstheyarefacing.
Moreover,theAIcoachingprogramwouldcomplementexistingorfutureQuebecSMEprograms,suchas
theActionsconcertéesdesFondsderechercheduQuebec,theMITACSprograms,Investissement
Québec’snewprogram,whichaimstohelpsecondarysectorcompaniesfollowthroughwithapplied
trainingprojectsbasedonthe“industrystudies”modelandtheCRIQstimulusprogramsdesignedtohelp
manufacturingcompaniestransforminto4.0companies.
Phaseno3:Consolidateseveraltechnologytransferorganizationsintoafundamentalandappliedresearchinstitutewithstrongindustryroots
Theoverallgoalistore-establishtherelationshipbetweencompaniesandappliedresearchcentresby
drawingoninternationallyprovenmodels,suchasthoseoftheFraunhoferandIMECinstitutesin
54
FINALVERSION
GermanyandBelgium(seecasestudybelow).Inconcreteterms,itwouldbenecessarytoconsolidateseveralexistingorganizationsandaligntheeffortsofthedigitalscienceecosystemwiththoseofQuebecclustersandexistingsectoraltables.Thisconsolidationshouldfocusonappliedresearchandtechnologytransferactivities.Tothisend,werecommendthattheQuebecgovernmentbeginthinkingaboutprocessandadditionalfundingneededtocompletetherestructuringoftherelationshipbetweenbusinessesandappliedresearchcentres.ThisrecommendationisdirectlyrelatedtotheCCEI’sRecommendation11.
CASESTUDY
LocatedinLeuven,Belgium,theInstitutdemicroélectroniqueetcomposants(IMEC)isaresearchinstitutespecializinginmicroelectronics. Itshighlyqualif iedstaff includesmorethan3,500people(includingresearchersaswell asprojectengineers). Duetoitspopularitywithcompaniesthatarelookingtocarryoutbasicandappliedresearchprojects, IMECis80%self-f inancing.Itsmainactivitiesaredividedintotwobranches:
R&Dsolutions:
• Collaborativepre-competitiveresearchprojectswithmultiplestakeholders
• BilateralresearchprojectsbetweenacompanyandtheIMEC
• Government-fundedresearch
Innovationservices:
• Fromideatoconception,theIMECoffersnumerousinnovationvehicles(“LivingLabs,”start-upincubationandaccelerationprograms,etc.).
• Prototypingandtestingservicesformoreadvancedprojects
• Productioncapacityandadviceaboutgrowth
TheIMEC’sproximitytoindustry,crit icalmassofresearchersandinnovationservicesmakesit amodelresearchcentrethatstronglycontributestothemarketofthetechnologiesdevelopedwithinit, asevidencedbythe60orsostart-upsthathavesprungfromtheIMECsinceits inception.
Inshort,overtime,theIMEChasbecomemorethanaresearchcentre.Nowit is atrueinnovationhub.TheIMEC’s24,400-square-meterspaceincludesaccelerators, incubatorsandtrainingroomsandholdsnumerousannualconferencesandtechnologyroundtables.
55
FINALVERSION
> Recommendation 8: Posit ion government and Crown corporations as model users that benefit from AI technology
FINDINGS
Giventhelargeamountofstructureddataitsvariousministriesgenerate(especiallyinthehealthsector),thegovernmentistheperfectpotentialcustomerforAIapplications.Undoubtedly,theQuebecgovernmentistheleaderincentralinformationmanagementsystemsinthepublicsector.
However,thegovernmentmustovercomesomeobstaclestobenefitfromAIsolutions.Inadditiontoissuessimilartothoseoflargenon-technologycompanies,suchasaccesstotalentandtheneedforgreaterawareness,structuralissuesneedtoberesolvedifthegovernmentwantstopositionitselfasalegitimateuserofAIsolutions.CEFRIO,initsNetGouvreportonopendata,seesonlyslightprogressintheevolutionofopengovernmentinQuebec.Indeed,fewerthan5%ofministriesreleasedmorethan10datasetsin2016.Sincethen,thesituationhasimprovedsomewhat,buttheissuepersists.KeepinmindthatpublicdataaccessibilityandsharingisanessentialconditionforAIuseinministries.
CASESTUDY
TheSOQUIJhasestablishedadatamanagementpolicywithinthecontextofthedevelopmentofAI-basedlegal informationsolutions.Thispolicyrespectsthecurrentlegalframeworkandproposessecuritymeasuresinl inewithindustrybestpractices( includingencryptionandsecuredatahosting). Thepolicyalsoevokesthenotionofbenefitandriskanalysisregardingdatacommunicationanduseanddefinesmitigationmeasuresforidentifiedrisks.WhenthebenefitsofusingdatainanAIcontextaresignificantlygreaterthantherisks,useispermitted.Internalandexternalethicalcommitteesevaluatetherecommendationswhenrisklevelsarehigh.Suchamodelcouldbeusedacrossministries.
SUGGESTEDAPPROACH
WeproposetopositiontheQuebecgovernmentasanAIuserthroughvariousimpactfulactions,including:
1) ClearpositioningintheStratégiedegouvernementnumériqueregardingtheadoptionofAIsolutionswithinthegovernmentapparatus.Thisstrategyshouldleadtothedevelopmentofinternalexpertise,startingwithteamsintheChiefInformationOfficer’sdepartmentwithintheTreasuryBoard(Conseildutrésor).Thisdepartmentwouldthencoordinateanassessmentinthefiveto10ministriesorgovernmentagenciesthatpossessthelargestamountofdatatoidentifycommonusecasesofbenefittothegovernmentandcitizens(e.g.,taxevasionpreventionatRevenuQuebec).Afterthisassessment,thegovernmentwillneedtosetcleargoalsonanumberofpriorityissues.
2) Themodernizationoftheregulationsgoverningtheuseofpublicdata(seealsoPhase1underRecommendation6).
56
FINALVERSION
ImplementingagovernmentAIstrategyFromamore tactical perspective, the operationalization of a governmentAI strategy could involvesettingupanAIanddatasciencesquadabletosupportmultipleministries.UpgradingtheskillsofkeystaffalreadyinplacecouldbedonethroughtrainingofferedbythevariousAIeducationalinstitutions,notably Université Laval’s Centre de recherche en données massives because of its geographicproximitytotheTreasuryBoardteam,whichincludestheChiefInformationOfficer.
Strong and experienced leadership is also crucial for the technological shift. In 2014, the ObamaadministrationhadateamofformerSiliconValleyexecutives, includingsomefromGoogle, leadthegovernment'smostinnovativeprojects.Itwouldbeanexcellentideafortheprovincialgovernmenttofollow this example and attract leaderswho have experience inmajor technology projects.On thedata side,we recommendanAI-solutiondevelopmentdatamanagementpolicybeadopted. In thisrespect,theworkdonebySOQUIJcouldserveasajumpingoffpoint(seetheSOQUIJcasestudy).
FINALVERSION
Strategic direction 4
Develop an international responsible AI and AI impact
analysis hub in Quebec
58
FINALVERSION
> Recommendation 9: Ensure the sustainabil ity of ongoing responsible AI mobil ization efforts and prepare to integrate emerging consensuses into Quebec public policy
FINDINGS
Aswementionedatthebeginningofthisreport,thedevelopmentofAItechnologiesandtheir
implementationshouldhavepositiveeffectsonsocietyoverall.Thatsaid,thereisnodoubtthatAI’s
increasingprevalencecouldalsohaveadverseeffects,intentionalornot,ifnotwellmonitored.
Nodoubt,somepeople,businessesandgovernmentsmaybeusingAItools – and,asFacebookandCambridgeAnalyticahaveshowns,thedatatheyneedtocreatethem
50– forillegalormorally
reprehensiblepurposes.Sometechnologies,suchasfacialrecognitiontools,couldbeusedtomonitor
consumersorcitizensandsurreptitiouslygatherinformationabouttheirsexualorpoliticalpreferences.51
Othertechnologiesmaybeusedtodeceivethepublic,asautomaticimagegeneratorscancreateever
moreconvincingfalsenews.52
Inothercases,organizationswilluseAItoachievetheirgoalslegally,butwillhaveanegativeimpacton
societyasawhole.Forexample,arecentWallStreetJournalsurveyshowedthatYouTubetendstoshow
extremeandcatchyvideostointernetuserslookingfor“mainstream”videostomaximizeitsadvertising
revenue.Therefore,conspiracy,anti-vaccinepropagandaorultra-marathonvideosmayappearas
suggestionstopeoplelookingforinformationabouttheU.S.President'sStateoftheUnionaddress,the
fluvaccineorjogging.53
UsingAItechnologiescouldalsohaveunwantedsideeffects,becausetheymaycodifyharmfulactions,
suchasdiscriminatorypractices.Forexample,someofthealgorithmsusedintheAmericanlegalsystem
showthatpeoplewhohavehadearlycontactwiththepolicearemorelikelythanotherstocommita
crimeafterbeingreleasedonparole.Thistypeoffinding,however,resultsintheunjustifiedextensionof
theincarcerationofmanyAfricanAmericans,sinceAfrican-Americanteensaremoreoftenbotheredby
thepolicefornogoodreasonthanarewhiteteenagers.54
AIapplicationswillalsohaveasignificantimpactontheQuebecandglobaleconomies.Indeed,thereisno
doubtthattheriseofAIwillresultintheeliminationortransformationofasignificantportionofthejobs
offeredinmanyindustrysectors.Automatedtextanalysisandgenerationtools,autonomousvehiclesor
high-performancechatbotsshouldhaveapositiveoveralleffectontheproductivityofbusinessesand
50Seehttps://www.nytimes.com/2018/03/19/technology/facebook-cambridge-analytica-explained.html?action=click&module=
Intentional&pgtype=Article.51Seehttps://www.theguardian.com/technology/2017/sep/07/new-artificial-intelligence-can-tell-whether-youre-gay-or-straight-
from-a-photograph.52Seehttp://www.lefigaro.fr/international/2016/10/04/01003-20161004ARTFIG00094-etats-unis-le-pentagone-a-debourse-des-
millions-de-dollars-pour-de-fausses-videos-djihadistes.php.53Seehttps://www.wsj.com/articles/how-youtube-drives-viewers-to-the-internets-darkest-corners-1518020478.Thisphenomenon
doesnothold,stressesacolumnistoftheNewYorkTimes,thatFacebook'sengineershavethe“desiretoleadtheworldtoruin”.It
ismoreplausible,accordingtoher,thatthisresulthastodowiththebusinessmodelofGoogle,theownerofYouTube.“YouTube's
algorithmseemstohaveconcludedthatpeopleareattractedtomoreextremecontentthantheywerelookingforatthebeginning
oftheirresearch-orbygenerallyinflammatorycontent.“
Cephénomènenetientpas,souligneunechroniqueuseduNewYorkTimes,aufaitquelesingénieursdeFacebookontla«volontédemenerlemondeàsaperte».Ilestplusplausible,selonelle,quecerésultataitàvoiraveclemodèled’affairesdeGoogle,le
propriétairedeYouTube.«L’algorithmedeYouTubesembleavoirconcluquelesgenssontattiréspardescontenusplusextrêmes
queceuxqu’ilsontcherchésaudébutdeleursrecherches–oupardescontenusgénéralementincendiaires.»54SeeO’Neil,C.(2017).Weaponsofmathdestruction–Howbigdataincreasesinequalityandthreatensdemocracy,Broadway
Books.
59
FINALVERSION
workersbutcouldquicklyputmanyprofessionals,workersandtechniciansoutofworkorforcethemtoacquirenewskillstoremainemployableormaintaintheirincomelevel.
AI’simpactontheenvironmentIn2016,theelectricityconsumptionoftheworld’sdatacentresexceededthatofalltheinhabitantsandcompaniesofGreatBritain.55TheriseofAIcouldcauseamajorincreaseinelectricityconsumption,becauseAIrequiresenormousamonthsofdataandcomputationalpower.BecomingaleaderinresponsibleAImeansthatQuebecshouldalsoseektominimizetheenvironmentalfootprintofitsactivitiesinthisarea.
Inresponsetothesechallenges,manyofthecountriesthathaveadoptedAIdevelopmentstrategieshavealsobeguntoreflectontheethical,socialandhumanissuesassociatedwiththedevelopmentanduseofAI.
Forexample,in2017,theStateCouncilofFrancecommissionedastudyentitledPuissancepubliqueetplateformesnumériques:accompagnerl’ubérisation,whichfocusesspecificallyontheeffectsofAIonFrenchsocietyandtheeconomy.IntheUnitedStates,theFutureofArtificialIntelligenceActof2017addressestheimportanceofdevelopingAIwhileprotectingtheprivacyrightsofcitizensandcreatingunbiasedtools.Movingfromdeclarationofintenttooperationalization,actorslikeAINowintheUnitedStates,ortheFutureofHumanityInstituteinBritain,begandesigningframeworksandmethodologiesthatbusinessesandgovernmentscouldadopttoframethedevelopmentofAItoolsandtheiruse.
ManyelementsdemonstrateQuebec'suniquepositioningwithrespecttothesocial,ethicalandlegislativeissuesofAIanddatasciencemoregenerally.
Inthefieldofmachinelearning,thereisabroadconsensusinQuebecwithinresearchteamsregardingtheneedtodevelopastrongdiscoursetosupporttheethicaldevelopmentofAI.Severalresearchersinthefieldhavebeenpersonallyinvolvedinthisinitiative,andorganizationssuchasIVADOandMILAarecloselyinvolvedinadvocacyandknowledgemobilizationactivitiesaboutresponsibledatauseandAI.Moreover,Quebechasstrongresearchexpertiseinthelegalsector(bothommonlawandcivillaw),publicpolicyandethics,andresearchersactiveintheprovincehavemobilizedaroundethicalandlegalAIissues.Finally,thedemocratictraditionandrespectforthefundamentalrightsinQuebecmakesitfertilegroundfordevelopingausefuldialoguewithcitizensregardingAIissuesandAI’ssocialimpact.
Inthisspirit,theUniversitédeMontréalandtheFondsdelarechercheduQuébecorganizedthe“ForumAIresponsable”in2017,inpartnershipwithMILA,IVADO,theCanadianInstituteforAdvancedResearchandthePalaisdescongrèsdeMontréal.
TheForumandthecitizenparticipationprocesswillconcludeinDecember2018withthepublicationoftheMontrealDeclarationforaResponsibleDevelopmentofArtificialIntelligence.Therecommendationsmadeinthisdocumentcomebothfromacommitteeofscientificexpertsandthepublicconsultationprocess.TheimplementationoftheDeclaration’srecommendationswillhelpbenchmarkAI’sfuturedevelopmentsothatitisdoneinamannerconsistentwiththevaluesofQuebecsocietyandallowsnotonlyforeconomicprogress,butalsoforsocialprogress.
55Seehttp://www.independent.co.uk/environment/global-warming-data-centres-to-consume-three-times-as-much-energy-in-next-decade-experts-warn-a6830086.html.
60
FINALVERSION
Giventherapidlyevolvingnatureoftechnologiesandthesocialchangestheybringabout,itisessentialtoanticipatethattheserecommendationswillneedtoberevisedregularly.Indeed,itwillbenecessarytoensurethattheycontinuetobeimprovedbasedonthetwopillarsoftheDeclaration,namelythecontributionofexpertsandongoingpublicconsultation.
Finally,thankstotheMontrealDeclarationforaResponsibleDevelopmentofArtificialIntelligenceandtheeffortsthatwillleadtoit,thestrengthofitsexperts,thequalityofitscitizens’participationinthemajordebates,itsculturalassetsandthestrengthofitsAIecosystem,QuebecwillbeperfectlypositionedtotakealeadingroleinthedevelopmentofresponsibleAIacrossCanadaandinternationally.Theprovincemustbegiventhemeanstofullytakeupthisleadershiprole.
SUGGESTEDAPPROACH
TopositionQuebecastheleaderinresponsibleAI:
1) WerecommendthattheQuebecgovernmentsupportthecreationofaninternationalprospectiveobservatoryontheresponsibledevelopmentofAIandthesocietalimpactsofAIanddigitaltechnology.Thisobservatorywouldbeanon-profitorganization.56Toavoidanyperceptionofaconflictofinterest,thegovernanceofthisentityshouldbeindependentfromtheAIanddatascienceresearchorganizations(e.g.,IVADOandMILA)aswellascorporationsandAI.Quebec(seeRecommendation12),whilestillbeinglinkedoperationallywiththeecosystem’svariouscomponents(inparticular,theobservatoryshouldbringtogetherseveralmajoruniversityorcollegestakeholders).Moreover,thisobservatoryshouldbeincontactwithallthestakeholdersconcernedbythesocietal,social,economic,culturalandethicalissuesofAIuseanddevelopment,i.e.interestedstakeholdersfromacademia,government,industryandcivilsociety.Governanceensuringadiversifiedrepresentationwhereinnoneofthestakeholdershaveamajoritycontrolwillensurethenecessaryindependence.Theobservatory'smandatewillbedividedintothreecomponents:
a. GeneratenewknowledgeonAIethicalandsocialissuesandimpactsbymonitoring,listingandanalyzinginitiativesinthisareaatthenationalandinternationallevels,andbyworkinginpartnershipwiththeFondsderechercheduQuebec(FRQ),throughtargetedcallsforinter-institutionalandintersectoralresearchprojectsinresponsibleAI.
b. Continuethelong-termeffortsundertakenwiththedraftingoftheMontrealDeclarationforaResponsibleDevelopmentofArtificialIntelligencebysettingupapermanentforumwhereallstakeholdersinQuebecsocietycandebatetheethicalandsocialissuesrelatedtothedevelopmentanduseofAItools,aswellastheirimpact.Thiscitizenforumwill,withinputfromAIethics,lawandtechnologyinnovationexperts,makeitpossibletoupdatetheDeclarationonaregularbasis,anessentialelementinacontextofacceleratedanddisruptivetechnologicaldevelopment.
c. MakerecommendationsregardinglegislativeframeworksandpublicpoliciestobeadoptedtoensuretheresponsibleuseofAIandpromotethesuccessfuladaptationofQuebecandCanadiansocietytotheincreasingpenetrationofAIinallspheresoflifeandactivity,includingsocialjusticeandthesharingofprosperity.
56TheCommitteeispleasedtonotethat,onMay4,theChiefScientistandthethreeFondsderechercheduQuebec(FRQ),inpartnershipwiththeMESI,invitedthescientificcommunitytorespondtoacallforproposalsaimedatsettinguptheObservatoireinternationalsurlesimpactssociétauxdel'intelligenceartificielleetdunumérique,whosemissionandmodeofgovernancecorrespondbroadlytothedescriptionpresentedabove.Governmentassistancewillconsistofanoperatingsubsidyofatleast$1M,andupto$2Mannuallyforfiveyears,withthepossibilityofrenewal.
$100Mfundfinancingpriority
61
FINALVERSION
ThismandatewillbeexecutedinclosecollaborationwithCanadianandforeignjurisdictionschargedwith
overseeingthedevelopmentofresponsibleAI.Thisobservatory,bymaintaininganactivedialoguewith
citizensandbybringingtogetherallAIstakeholders,willpromoteeducation,responsibilityand
empowermentwithinQuebecandCanadiansocietyandensurethatthevoiceofcitizensistakeninto
accountinthedevelopmentofAI.Therefore:
1) WerecommendthattheQuebecgovernmentadoptvigorousmeasurestopositionQuebecasaninternationalleaderinthefieldofAIandresponsibleAI.Itwoulddothisbyfacilitating,asannouncedinits2018-2019budget,theimplementationbyMontréalInternationalofanew
internationalAIorganization,aswellasbysupportingthestrongparticipationofQuebecAI
ecosystemmembersinthemainexistingCanadianandinternationalforumsregardingthesocial
issuesofAIuseandthestandardizationbodiesconcernedbytheseissues.Itshouldbenoted
thattheestablishmentofanewinternationalAIorganizationisdistinctfromtheproposed
observatoryinitiative,butthepresenceofanobservatory,inadditiontothescientificand
industrialAIassetsinQuebec,representsapullfactorforaninternationalorganization.
InthecontextofitseffortstobecomeanexemplaryAIuser:
2) WerecommendthattheQuebecgovernmentadoptaframeworkfortheresponsibleuseofAIinwhichitundertakestothefollowing:
a. Increasetheinternalexpertiseofministriesandagenciesregardingtheresponsibleuse
ofAI.
b. Ensurethattheanonymiseddataitgeneratesaremadeavailableinanefficientmanner
thatrespectstheprivacyofcitizens.
c. MakepubliceveryuseitmakesofAIalgorithmstocometodecisionsthatmayaffect
thelivesofcitizens.
d. Giveresearcherstheopportunitytoexaminethesealgorithmstodetectproblemsthat
mayaffecttheirfunctioning.
TheQuebecgovernmentmustapplythislastrecommendationquickly,toreassurecitizens
regardingtheuseitintendstomakeofthedatathatwillbemadeaccessible(see
Recommendation6)andtheAItoolsthatwillbecreatedinQuebec,insomecases,withthese
data
.
FINALVERSION
ActinternationallyonissuesrelatedtoartificialintelligenceDuringavisithemadetoFranceinMarch2018,thePremierofQuebecproposed
creatingaglobalorganizationthatwouldworkonsocietalandethicalissues
relatedtothedevelopmentanduseofAI.AccordingtotheQuebecgovernment,
thiswouldbeasortof“controltower”thatcould“reassurethepublic”aboutthe
evolutionofthisnewtechnology,amongotherthings.
Inits2018-2019budget,theQuebecgovernmentfolloweduponthis
announcementbyearmarking$5MforMontréalInternational“tofacilitatethe
establishmentofanewinternationalartificialintelligenceorganizationin
collaborationwithkeystakeholdersinthefield,”57
includingtheSteering
Committee.
TheSteeringCommitteeisoftheopinionthatthesettingupofsuchan
organizationanditspresenceinQuebecwouldhaveverypositiveimpacts.It
intendstoprovideallpossiblesupportMontrealInternationalneedstomakethis
initiativeasuccess.
TheCommitteealsobelievesthatitwillbeimportantforthegovernmenttogive
Quebec'sAIstakeholders,fromcompaniestoresearchcentrestoministriesand
agencies,themeanstointerveneforcefullyininfluentialCanadianorinternational
forums,suchasthePartnershiponAI(agroupfoundedbymajorU.S.digital
companiesthattodayincludessuchplayersastheAssociationforthe
AdvancementofArtificialIntelligence,theFraunhoferInstituteforIndustrial
Engineering,ElementAI
,McKinsey&CompanyorAccenture),aswellas
standardizationbodiesinvolvedintheeffectiveandresponsibledevelopmentofAI
(see“AIstandards”onpageXXoftheAppendices).
57
Seehttp://www.montrealinternational.com/a-propos/salle-de-presse/de-nouvelles-mesures-pour-renforcer-le-bassin-de-talents-
internationaux-et-mettre-en-place-une-organisation-internationale-en-intelligence-artificielle/.
FINALVERSION
Strategic direction no 5
Advance the development of ecosystem support structures
64
FINALVERSION
> Recommendation 10: Ensure access to computing power and computational expertise for AI and data science stakeholders
FINDINGS
IntheAIsector,manyresearchandtechnologydevelopmentsrelyheavilyonitsresearchers’andapplicationdevelopers’accesstocomputinginfrastructuresthatareasrobustastheyaresafe,inshort,onconnectivity,computingpowerandhigh-qualitydatastorageservices.
ComputeCanadaistheorganizationresponsibleforcoordinatingthedeliveryofadvancedcomputingservicesacrossCanada.ThisNPOhasfourregionalpartners:ACENETintheAtlanticProvinces,CalculQuébec,ComputeOntarioandWestGrid,eachsupportedbygovernmentpartnersandotherstakeholders.
CalculQuébecisagroupthatbringstogetherQuebecuniversitiesthatshareadvancedcomputinginfrastructure.CalculQuébechasacoregroupofhighlyqualifiedexpertswhocansupportAIresearchersintheirwork.
DespitetheeffortsmadebyCanadaandQuebectoensureabalancebetweenthesupplyofcomputingpowerandtheneedsofresearchersandcompanies,significantobstaclesoftenpreventthemfromhavingaccesstoalltheresourcestheyneedtocarryouttheirAIprojects.
First,itseemsonly30%ofgraphicsprocessingunits(GPUs)and50%ofCPUs(centralprocessingunits)neededbyQuebecresearchcentresandgroupsarecurrentlyavailable.58Theinvestmentsannouncedbythefederalgovernmentinits2018-2019budget(seethetextboxbelow)shouldpreventtheextensionofthegapbetweenQuebecandCanadaandtheirinternationalrivalsintermsofcomputingpower.ThisalonewillnotmeetalltheAIecosystemstakeholders’futuredemandsincetheirneedswillgrowrapidly.
Thefederalgovernmentandcomputingpowerinits2018-19budget,thefederalgovernmentannounced$572Moverfiveyearstoimplementanewstrategyfordigitalresearchinfrastructureandbigdatauseincollaborationwiththeprovinces,universitiesandplayerslikeComputeCanadaandtheCanadaFoundationforInnovation.ThedetailsofthisstrategyandtheamountsthatwillbeinjectedintoQuebecwillbeknownatalaterdate.
Second,currently,theactivitiesofcomputingpowerproviders,suchasCalculQuébec,cannotalwaysbeorganizedinawaythatmeetsalltheGPUorevenCPUrequirementsofAIplayers.Forexample,accordingtoAIplayers,thereportedwaittimetoaccesstheGPUrequiredforaprojectisoftentoolong.
Moreover,Quebec'sAIstakeholders,especiallySMEs,donotalwayshavethenecessaryknowledgetostructuretheirITassetsandactivities(e.g.,theircloudcomputingordataanonymizationandprotection58ResultsofthelastComputeCanadaawardprocess,estimatedbyCalculQuébec.
$100Mfundfinancingpriority
65
FINALVERSION
activities)andproperlyleveragethecomputationalpowerCalculQuébecandotherproviders(e.g.,
AmazonWebServices)currentlyoffer.
Atthemoment,CalculQuébecwouldnotbeabletorespondinaflexiblewaytotheemergingand
growingneedsofresearchers(inAIaswellasinsectorsliketransportation,healthandlaw,whereAItools
areincreasinglybeingused),aswellasthoseofstart-upsandestablishedSMEswhowouldliketoget
coachingtounderstandthefullrangeofcomputingresourcesavailableonthemarketandmakewise
choices.
Thatbeingsaid,challengesarethesameelsewhereinCanada.Tomeetthesechallenges,Ontariodecided
in2014toconferNPOstatusonComputeOntarioandinvest$75Moverfiveyearstoimplementits
strategy.
SUGGESTEDAPPROACH
Tocorrecttheshortcomingsmentionedabove:
1) WerecommendthatQuebec'scomputinginfrastructureneedsbeclearlydefinedandthatappropriaterepresentationsbemadetothefederalgovernmenttoensurethatQuebecreceivesanappropriateshareofthefundingprovidedbythefederalbudget.Inits2018-19budget,theGovernmentofCanadahasproposed$572.5Moverfiveyearsand$52Mperyear
thereaftertoimplementadigitalresearchinfrastructuredevelopmentstrategy.Thisstrategy
aimstoprovideCanadianresearcherswithmoreopenandequitableaccesstoadvanced
computingresourcesandbigdata.TheCanadianMinisterofSciencehasannouncedhisintention
toworkwithstakeholders,includingtheprovincesanduniversities,toproducethisstrategyand
definetherolethatplayersliketheCanadaFoundationforInnovation,ComputeCanadaandthe
CanadianNetworkfortheAdvancementofResearch,IndustryandEducationwillplayinits
application.
2) WealsorecommendthatQuebec'sAIresearcherswhouseAItechniquesintheirfieldof
research,aswellasthecompaniesthatworkwiththeseresearchersoncollaborativeresearchprojects,haveaccessthroughCalculQuébectotheincreasedgraphicsprocessingunits(GPU)andgreaterstoragecapacitiestheyrequire.
3) Finally,werecommendsupportingCalculQuébecinstrengtheningitscapacitytoeffectivelysupportresearchersandbusinesseswithintheAIecosysteminparticular.Amongotherthings,
itwouldbeappropriateto:
a. ActivelyparticipateinongoingdiscussionsregardinggovernancemodelsforCanada's
digitalresearchinfrastructuretoinfluencethedevelopmentsthatwillhappenacrossthe
countryandultimatelyincreaseCalculQuébec’sflexibilityandagility.
b. EnsureCalculQuébechasthemeanstosetupandrolloutalargerteamofadvisers
whoserolewouldbetosupportresearchcentresandgroups,aswellasQuebec
companies,inthedesignandimplementationofoptimalstrategiesforhigh
performancecomputing.Forinstance,thisteamcouldprovideinformationtoAI
stakeholdersaboutsolutionsofferedbyotherproviders(e.g.,helpthemworkwith
privatestorage,computinganddataprotectionserviceproviders).
c. SupportCalculQuébecinrevisingitsmissionorstrategytohelpitdeterminehowbest
toensureitslong-termfunding.
NotethatthepossiblereorientationofCalculQuébecandthereinforcementofitscapacitytohelpAI
companiescouldhelpboostQuebec'shigh-performancecomputingecosystem,particularlythroughthe
66
FINALVERSION
growthofcomputing,dataprotectionandotherserviceproviderswithwhomCalculQuébecwillworkcloselytogether.
> Recommendation 11: Create a permanent AI.Quebec organization with its own administrative structure to play a monitoring role and faci l itate coordination between Quebec AI ecosystem stakeholders
FINDINGS
InQuebec,apromisingsupportecosystemforthedevelopmentofAIalreadyexists,whichincludesplayersinprivatefinance,technologytransferandtraining,aswellasbusinessassociationsandsectorclusters.Asnotedinthisreport,manyongoingandupcomingnon-governmentalinitiativesareexpectedtoallowthissupportecosystemtogrowtoadequatelysupportAIgrowthintheprovince.However,failuretoimplementtheseprojectscouldseriouslyjeopardizeQuebec'sfutureleadershipinAI.
Moreover,whilethemanystakeholdersintheQuebecAIecosystemisamajorasset,italsoraisessomeproblems.Despiteitsmanystrengths,QuebechaslimitedresourcescomparedtoplayerslikeChinaortheUnitedStates.Therefore,itisallthemorecriticaltooptimizetheseassetsandcoordinatetheeffortsoftheecosystem.
Also,inasectorthatisevolvingasrapidlyasAI,andindigitalsciencesingeneral,theecosystem’sdirectionsmustbeflexible.Wedon’tknowwherethenextscientificbreakthroughorthenext“Facebook”willcomefrom,butit’spossibletosetupaflexibleandagilesystemthatcanadjustquicklytoprovideadequatesupporttonewplayerswithstrongpotentialorimprovetheecosystem’sstrategy.
Finally,notonlyisitessentialtosupportthegrowthoftheQuebecAIecosystem,butsignificantmarketingeffortsmustbemadetoensureitsvisibilityonaglobalscale.Someactorsarealreadymobilizedandareengagedinpromotioninternationally,buttheywillneedadequatesupporttohavesufficientstrikingforce.
SUGGESTEDAPPROACH
Werecommendthecreationof“AI.Quebec”,anadministrativeentitywithastrategicreflection,rallyingandpromotionrole,whichwillworkinclosecollaborationwithallthegovernmentalandnon-governmentalstakeholdersconcernedbythedevelopmentofAIinQuebec,fromtheQuebecInnovenetworktoTechnoMontréal,MontrealInternational,QuébecInternational,PromptortheministriesofEconomy,ScienceandInnovation,EducationandHigherEducation,InternationalRelationsandLaFrancophonieorLabour,EmploymentandSocialSolidarity.Morespecifically,AI.Quebecwouldhavethefollowingresponsibilities:
1. Continue,throughcollaboration,thestrategiceffortstocreate,inQuebec,thelargestartificialintelligenceecosysteminNorthAmerica.
a) Tothisend,wesuggestAI.Quebecplayamonitoringrolewiththeaimofcontinuouslyevaluatingtherobustnessofthesupportecosystem.Theorganizationcanrecommendspecificactionstothegovernmenttoaddressanyshortcomingsobserved.Areasinneedofmonitoringincludeaccesstofinancingcapital,accesstocomputinginfrastructure,
$100Mfundfinancingpriority
67
FINALVERSION
upgradingoftechnologytransferactors,technologytransfermechanisms,awareness
withinlargecompaniesandtheavailabilityofconsultingservicesforSMEs.
b) AI.QuebecwouldalsobemandatedtomonitorinternationalAIdevelopmentsandthen
informtheecosystemandthegovernmentoftheissuesandopportunitiesforresearch
andtheQuebececonomy.
c) Finally,AI.Quebecwouldberesponsibleforcontinuingtheongoingstrategiceffort,
regularlyreviewingtheecosystem’sfive-yearplanandsustainingthevision’s
implementationoverall.
2. Leadthe“talentwatch”mentionedinRecommendation3toensuretheecosystem’srobustness
intermsoftalentandtraining.
3. Ensuringsolidarity,collaborationandcooperationofQuebecAIstakeholders.Tothisend,AI.Quebecwouldbeinchargeoffacilitatingcurrentandfutureecosystemplayernetworking,act
asamediatorintheeventoflitigationandserveasacentralbodyaroundwhichecosystem
playerscanrallytoresolvetheissuestogether.Assuch,AI.Quebecwouldalsoberesponsiblefor
settingupandupdatingthecommontechnologyshowcasepresentedinRecommendation12.
4. BuildbridgesbetweenQuebec'sAIecosystemandotherAIecosystemsaroundtheworld,
startingwithotherCanadianecosystems.AI.Quebecmuststayinconstantcommunicationwith
theCanadianAIConsortiumandthefederalgovernmenttoensureapan-Canadiannetworkof
initiativesaffectingtheprovinces.
5. PlayasupportingroleinpromotingQuebec'sAIecosysteminCanadaandabroad.Wesuggest
thatAI.QuebecsupportexistingpartnersintheireffortstopromoteAI,e.g.,Montreal
International.AI.Quebec’sadministrativebodycouldalsosupportthegovernmentinits
representationrolesatthefederalandinternationallevels.
6. SupportthecoordinatedparticipationofQuebecinsomeofthekeyCanadianandinternational
forumsonthefutureofAIandthisnewindustry’sdevelopment.Inparticular,Quebecshould
ensure,throughAI.QuebecandthenewinternationalobservatoryonresponsibleAIandits
impacts,tohaveastrongsayatCanadianandinternationalstandardizationnegotiationtables
implementedovertherecentyearsandmonthstoprovideaframeworkforthedesignofAItools
andtheirethicaluse(seethetextboxbelow).
AI.Quebec’sprecisegovernancemodelwillbeadoptedduringtheimplementationofthisstrategy,in
closeconsultationwithallecosystemstakeholders.
QuebecandAIstandardsQuebecwillbenefitfromparticipatinginthestandardizationeffortsthatwillbeundertakenintheAI
sector.TheprocesswillenableQuebecresearchersandcompanies:
• TobeamongthefirsttoknowwhichstandardswilldirecttheevolutionAIsothattheycan
quicklyadapttothefuturerequirementsoftheinternationalmarket.
• Toinfluencestandardizationprocessessothatcertainissuesimportantforthedevelopmentof
theQuebecAImarketandindustriesthatuseAI,aswellasforthedevelopmentofQuebec
societyandcultureareaddressedadequately(e.g.,speechrecognitionorsynthesissoftwareable
toprocessQuebecFrencheffectively).
Therefore,itwillbeimportantforQuebectoensureastrongCanadianrepresentationonISO'snew
68
FINALVERSION
ArtificialIntelligenceCommittee(SC42JTC1),aswellasotherstandardizationjurisdictions.ThisrepresentationshouldalsotakeQuebec’sinterestsintoaccountand,aboveall,ensuretheeffectivetransferofacquiredknowledgetoallQuebecstakeholders.IncoordinationwiththeAlbertaandOntarioAIclusters,QuebeccouldworktodefineaninterventionplanintheISOSC42Committeeandotherstandardizationbodies.
QuebecandCanadacouldtakestepstoimplementamodestconsultationanddisseminationstructuretogatherinputfromAIecosystemstakeholdersandquicklyinformthemoftheworkbeingdonebystandardizationbodiesandtheimpactofthiswork.Finally,itmaybeagoodideatofostercollaborationbetweenQuebecanothercountries,suchasFrance,toensurethatissuesofculturalandlinguisticdiversityareproperlyaddressedinAI.
> Recommendation 12: Faci l itate access to information on avai lable resources for AI ecosystem stakeholders and promote connections between them through a common showcase
FINDINGS
ArichandcomplexecosystemofintermediariesexistsinQuebec,butthisecosystemiscurrentlyfragmented,withnoclear“gateway”fornon-technologycompaniesofallsizes.Whileaportalforbusinesses,QuebecInnove,exists,thisplatformisnotcurrentlyabletoreceive,filterandrouteAIrequests,whichaddsalevelofgranularity.Lackofclarityregarding“whodoeswhat”ensuresthatmanyrequestsareroutedtoMILAandIVADOwithouttheirhavingthecapacitytorespond.Therefore,itiscrucialtosetupaneffectivenetworkingmechanismtailoredtoend-userneeds.
Asmentionedinthisreport’srecommendations,increasedvisibilityforalltheAIecosystem’sresourceswouldmakethejourneyeasierforthefollowingstakeholders:localandinternationalstudents,businessleaders,professionalslookingtoconverttheirskills,etc.
SUGGESTEDAPPROACH
WerecommendthecreationofanonlineshowcaseplatformthatwouldbeaccessibleontheAI.QuebecandEntreprisesQuebecsites.Onthesites,thefollowingwouldbeaccessible:
• Eachplayer’sofferinginmarketingandtechnologytransfer
• AlistofAIstart-upsintheecosystemandtheirrespectivesectors
• Alistoftheexistingprovincialandfederalgrantprogramsandtheirtargetaudiences
• AlistoftheacademicresearchtopicsinAIavailableinQuebec,includingalistofresearchersworkingineachfield
• Alistofinternshipsavailableforlocalandforeignstudents
• Thescholarshipandfundingofferingsavailabletolocalandforeignstudents
• FutureprofilesinAI
• AlistofallexistingAIanddigitalsciencetrainingcourses,eithercredited(college,university,continuingeducation)andnon-credited(e.g.,onlinecoursesonAI’stransformationpotential).
$100Mfundfinancing
priority
69
FINALVERSION
Thisplatformmustmeetseveralcriteriatoachieveitsmainobjective,whichistobringtogetherinvestors,creatorsandpotentialusers,andincreasethevisibilityofavailableacademicresearchandtrainingprograms.Theplatformshouldreflecttheimageoftheecosystem,i.e.modernandeasytouse.ItmustbeinclusiveandapplyacrossQuebecasawholeandshouldhighlightexistingtechnologytransfermodels.Theplatformwillalsoneedtobeupdatedfrequently,ifnotautomatically.Finally,theplatformmustpromoteawarenessandeducation,i.e.explainwhatAIis,describetheapplicationsofresponsibleAIandprovidetheappropriateeducationalmaterialtoinformtheQuebecpopulationaboutthevarioustypicalAIusecasesaccordingtosector.
FINALVERSION
Conclusion
71
FINALVERSION
Concluding remarks
ThedevelopmentofQuebec’sAIstrategyistheresultofcollaborationbetweentheSteeringCommittee
membersandQuebececosystemstakeholderswho,inadditiontotheirrespectiverolesasinvestors,
entrepreneurs,employers,researchersorprofessors,cametogetherwiththecommongoalofensuring
futureprogressandprosperityforoursociety.Indeed,manymembersoftheacademic,industrialand
governmentalsectorswereinvolvedinthedevelopmentandvalidationofthisreport’s12
recommendationsinthehopeofhelpingQuebecbecomethemostimportantartificialintelligence
ecosysteminNorthAmerica.Ourcommunitywillbenefitgreatlyfromthecommitmentofalltheleaders
whohavesharedtheirpointofviewontheissues,challengesandopportunitiesforQuebecandallofits
regionsinthedevelopmentofanAIhubinQuebec.
Talentisthesinewsofwar.Itisafinding,anissue,evenanopportunitythatisunanimouslyacceptedin
theQuebecAIecosystem.
Quebec’ssuccesswilldependdirectlyonitsabilitytotrainandattractthetalentneededforbasicAI
research,thedesignofapplicationsthatuseAIalgorithmsandtheadoptionofAIandAI-related
technologiesbyQuebecbusinesses.Also,thecreationofapooloftalentisnecessary,butinsufficientto
enableQuebectotakeadvantagefromthesignificantpotentialeconomicbenefitsgeneratedofthe
commercializationofAIsolutions.Themainstrategicdirectionsofthisreportreflectthisvisionbyfirst
emphasizingthegrowthanddurabilityofanAIresearchhubinQuebec,thedevelopmentoftalentin
digitalsciencesthroughoutQuebecandtheacceleratedadoptionanddevelopmentofthesenew
technologiesbytheentireQuebececonomy.Itwillbeessentialtoproactivelyaddresstheissueof
responsibleAIwiththecommunity,giventhesocietalchangethatwillresultfromtheuseofAI.Finally,
thesefourstrategicorientationsrequirefacilitatingelements,suchasaccesstosuitablecomputing
power,whichwillalsoneedsupporttodevelop.
TheSteeringCommitteewillcontinueitsworkinthecomingweeks.Amongotherthings,itwillsollicit
governmentinputandworkcloselywithgovernmentandecosystemstakeholderstosupportthe
implementationoftherecommendations.
Someoftheapproachessuggestedinthisreportcan,andmust,becarriedoutquicklytoallowQuebecnotonlytostayintheglobalraceforAIexcellence,butalsotobecomealeaderinthisfield.
Hereisanoverviewoftheactionswerecommendthegovernmentimplementwithinsixmonthsofthe
tablingofthisreport.
72
FINALVERSION
Proposed plan for the next s ix months
Asmentionedintheintroductiontothisreport,someofourrecommendationsinvolvelong-terminitiatives,whileothersareachievableintheimmediatefuture.Toensurethefastandeffectiveimplementationoftheelementsthatcanemergeinthenearfuture,hereisalistofannouncementstheMESIshouldmakeandconcreteactionsitshouldtakewithinsixmonthsofthetablingofthisreport.
First,theMESIshould:
1. Publiclyannouncethereport’smaindirections.2. Decideonanincrementaltimelineandfundingscheme(toachievefundingsustainabilityover10
yearsandincreasethebudgetenvelopeforthedevelopmentoftheAIecosystem.3. AnnouncethefundsgrantedtotheMILA–QuebecArtificialIntelligenceInstitutetosupportits
development.4. AnnounceitssupportofSCALE.IA.5. AnnouncethecreationoftheAI.Quebecgroupbytheendof2018.
Subsequently,theMESIwillneedtoactuponthereport’sstrategicdirectionsand:
1. AllocatetheappropriategrantstotheMILA–QuebecArtificialIntelligenceInstitute.2. AllocatetheappropriatelevelofgrantsfortheoperationalizationofSCALE.AI.3. SupportthecreationofapermanentAI.Quebecadministrativebodytoserveasamobilizing
force.4. GiveAI.Quebecthemandatetosetupa“talentwatch”withthesupportof
TECHNOCompétences.5. LaunchtheprogramforthecreationofAIsectoralchairs.6. WorkwiththeMinistryofImmigration,DiversityandInclusiontoaddAIfacultyandresearchers
andotherAIspecialiststothe“Volettalentsmondiaux”listofoccupationstospeeduptheimmigrationprocess.
7. SetupagovernmentbudgetenvelopetosupportQuebec'seducationalinstitutionsinupdatingtheirtrainingofferings.
8. SimplifyandacceleratethegovernmentapprovalprocessfortrainingprogramsofferedbyQuebeceducationalinstitutions.
9. ApproachtheChiefInformationOfficeroftheConseildutrésorduQuebectoundertakeastrategicplanningeffortforAIadoptionbytheQuebecgovernment.
10. Setupmarketingandtechnologytransfersupportprograms(private-privateprojectfinancingandsectoralsupportprograms).
11. AnnouncethelaunchoftheobservatoryonsociallyresponsibledevelopmentofAIandontheimpactsofAI,andmandateathinktankontheinternationalpositioningofQuebecinthisarea.
12. FollowuponthePremier’sannouncementregardingtheestablishmentofaglobalresponsibleAIorganizationinMontreal.
73
FINALVERSION
Proposal for the al location of the $100M from the MESI and detai ls regarding the need for an additional $300M
Inlightofthestrategicdirectionspresentedinthisreport,belowareourallocationrecommendationsforthe$100MbudgetenvelopeearmarkedforthedevelopmentoftheQuebecAIecosystem.
First,theSteeringCommitteeisconvincedthatthedevelopmentoftheQuebecAIecosystemwillbedrivenbytheriseoftheMILA–QuebecArtificialIntelligenceInstituteandthereforerecommendstheallocationof:
• $80MtosupportthedevelopmentofMILA–QuebecArtificialIntelligenceInstitute,withaninitialamountof$65Mandaconditionalbalanceof$15Muponmeetingcertainmilestones(tobeestablishedwiththeMESI).ThedevelopmentofthisInstituteisthespearheadofthestrategyforthedevelopmentoftheAIinQuebecandservesmultiplestrategicdirectionsoutlinedinthisreport:theestablishmentofaresearchhub,talentdevelopment,thehivingoffofAIstart-ups,theadoptionofAIskillsbytheQuebectechnologytransfernetwork,makinganimportantcontributiontothedebatesregardingresponsibleAIandprovidingbetteraccesstocomputingpowerandcomputationalexpertisetoAIresearchers.
Also,weconsidermanyoftheinitiativesmentionedintherecommendationsaspriorities.Theseinitiativeswillrequiretimelyfinancingtotaling$20M,forwhichwerecommendthefollowingbreakdown:
• $10MtosupportthemarketingandadoptionofAIsolutions,including:o FinancingofsectoralsupportprogramsforcompanieslookingtomakeanAIshift.o Financialsupport(viataxcredits)forQuebecbusinessesthatmeettheirAIneeds
throughlocalstart-upsorSMEs.• $4MtosetupapermanentAI.Quebecadministrativebody
o Thisestimatetakesintoaccountthesalariesofateammadeupofafewpeopleandabudgettocarryoutthevariousactivitiesoftheorganization.
• $3.5MfortheestablishmentandoperationofaninternationalobservatoryonresponsibleAIandAIimpacts.
• $2.5Mforaseriesofimpactfulecosysteminitiatives:o Implementationofthe“talentwatch”thatwouldquicklyconductastudytodetermine
theindustry'sneedsintermsofAIandAI-relatedprofilesandskills,withaviewtomatchingtrainingtoemployment.
o UpgradingAItechnologytransferorganizationsusingatrain-the-trainerapproach.o Establishmentofthe“AI.Quebec”commonplatform,whichwilllistthevarious
ecosystemstakeholdersandtheirserviceofferings.o Financialsupporttocollegesanduniversitiestosupporttheminquicklyupdatingtheir
trainingoffering.
74
FINALVERSION
Therefore,asoutlinedinthetablebelow,theMILAshouldreceiveupto$80Mofthe$202MtheQuebecgovernmenthassofarearmarkedforAIdevelopmentinQuebec,or39.6%oftheavailablebudgetenvelope
QuebecgovernmentAIinvestments
Actions Amountallocated(in$M)
Percentageoftotalinvestment
$100MfundfinancingprioritySupportthedevelopmentoftheMILA
80.0 39.6%
$100M fund financing priority Creation of aninternational observatory on responsible AI and AIimpacts
3.5 1.7%
$100MfundfinancingpriorityImplementationofAI.Quebec
4.0 2.0%
$100MfundfinancingprioritySupport for AI marketing and adoption, including thefinancing of sectoral support programs for companieswhowanttoshifttoAI.
10.0 5.0%
$100MfundfinancingprioritySupportforotherinitiatives,namelytheestablishmentofa“talentwatch,”modernizeAItransferorganizations,setup a common platform and support rapid updating ofeducationalinstitutions’trainingofferings
2.5 1.2%
Other AI investments – Launch of SCALE.AI project, AI-optimizedprocurementchains
60.0 29.7%
Other AI investments – Creation of the International AIOrganizationinMontreal
5.0 2.5%
Other AI investments – Applied math to support AI-relatedfieldsinQuebec
1.0 0.5%
Other AI investments – Digital economy action plan,measure1; support for the IVADOproject (Canada First)andCalculQuébec
5.0 2.5%
Other AI investments – Digital economy action plan,measure3;supportforcollaborativeresearchprojectsviaPrompt
8.5 4.2%
Other AI investments – Development of the computinginfrastructuresrequiredforAI’sdevelopment
12.5 6.2%
OtherAIinvestments–LaunchofHECMontréal’sCreativeDestructionLabMontréalandofNextAIMontreal
10.0 5.0%
TotalQuebecgovernmentfundingforthedevelopmentofAIinQuebec
202 100.0%
Inadditiontothe$202Mdetailedabove,westronglybelievethatanadditionalinvestmentofapproximately$300MisneededtohelpQuebecbecomealeaderinAI.Thisadditionalamountwillservethreemajorpurposes:
1) About50%ofthesefundsshouldgotowardsthesustainabilityoftheecosystem’sanchoringinstitutions(i.e.tosupporttheMILAandIVADOovera10-yearperiod),strengthenCalculQuébec’sabilitytomeettheneedsofitspartners(particularlySMEs)andensurethecreationandoperationofAI.QuebecandtheinternationalobservatoryinresponsibleIA.
75
FINALVERSION
2) 20%oftheadditionalinvestmentwouldbeusedtoincreasethedevelopmentofAIanddigitalsciencetalentintheecosystem,includingthecreationofsectoralchairstodevelopanAIapplicationresearchhub,fundingsupportforadditionaluniversityandcollegefacultypositions,aswellasbringingtogetherandrampingupdemocratizationofscienceandmathematicalliteracyinitiatives.
3) And,30%ofthefundswouldbeusedtoacceleratethedevelopmentandmarketingprocessofQuebecAIsolutionsand,especially,theiradoptionbyorganizationsthatmakeupQuebec’seconomicnetwork.Inconcreteterms,thiswouldmeanearmarkinganadditionalbudgetenvelopetodirectlysupportcompaniesinmarketingtheirAIinnovations,rolloutsectoralprogramstosupporttheuseofAIandtodevelopthecapacitytoadoptAIofQuebec’sministriesandagencies.
FINALVERSION
Appendices
77
FINALVERSION
Summary of recommendations and suggested approaches
STRATEGIC DIRECTION 2: Develop digital science talents to meet Quebec’s needs
STRATEGIC DIRECTION 3:
Accelerate the development and adoption of AI solutions within Quebec’s economic fabric
STRATEGIC DIRECTION 4: Develop an international responsible AI and AI impact analysis hub in Quebec
STRATEGIC DIRECTION 1: Ensure the development and sustainability of Quebec’s AI academic research hub
STRATEGIC DIRECTION 5: Advance the development of ecosystem support structures
Recommendation 1: Attract researchers ▪ Support the development of the MILA – Quebec Artificial Intelligence Institute▪ Create excellence chairs by sector▪ Support the creation of academic positions (universities)▪ Set up a permanent chair program▪ Support the recruitment of university researchers ▪ Support the recruitment of professors/researchers within the college network▪ Add AI professors and researchers to “Volet talents mondiaux” professions list
Recommendation 2: Attract master’s and doctoral students ▪ Increase the visibility of research opportunities in AI and related fields▪ Increase financing for master’s and doctoral ▪ Increase foreign student retention programs
Recommendation 4: Support the updating of training offerings ▪ Support for the creation and modification of programs (does not include future online
courses)▪ Simplify and speed up government approval of training programs educational
institutions propose
Recommendation 3: Develop and maintain a “talent watch” ▪ Implement and roll out a “talent watch”
Recommendation 7: Support future private sector AI users ▪ Modernize the selected intermediary organizations▪ Support intermediaries in implementing consulting teams▪ Set up sector-specific AI shift coaching programs▪ Consider consolidating many technology transfer sector organizations into an applied
research institute
Recommendation 11: Creation of a permanent AI.Quebec administrative body ▪ Provide financing of AI.Quebec
Recommendation 6: Support AI design businesses ▪ Create data trusts ▪ Provide start-ups with priority access to Calcul Québec and Compute Canada▪ Subsidize partnerships between start-ups and private entreprise ▪ Review the government’s strategic procurement mode à▪ Implement marketing tax credits or funding programs
Recommendation 8: Government and Crown corporations as model AI users ▪ Implement the action plan on AI solution adoption within the government apparatus▪ Modernize regulations regarding the use of public data
Recommendation 9: Sustainability of mobilization efforts regarding responsible AI ▪ Creation of an observatory on responsible AI development and AI impact ▪ Position Quebec as an international leader in responsible AI
Recommendation 10: Access to computing power and computational expertise ▪ Support the development of Calcul Québec
Recommendation 12: Access to information on a common platform ▪ Create the AI.Quebec platform
Recommendation 5: Promote mathematical literacy and the democratization of science ▪ Set up an organization to coordinate and promote the striking force across Quebec
78
FINALVERSION
Steering Committee members
Businesssector
PierreBoivin
(CommitteeCo-Chair)
PresidentandCEO
Claridgeinc.
ChrisArsenault
ManagingPartner
iNovia
ThomasBirch
Vice-President,FundsandTechnologies
CaissededépôtetplacementduQuebec
MagalyCharbonneauPrincipaliNovia,entrepreneur,(PasswordBoxacquiredbyIntel)
PatrickPichette
CorporateDirectorandStrategicAdvisor
ValeryZamuner
Vice-President,Mergers,AcquisitionsandStrategicInitiatives
Stingray
CharlesBoulanger
CEO
LeddarTech
79
FINALVERSION
Universitysector
GuyBreton
(CommitteeCo-Chair)
Rector
UniversitédeMontréal
FrançoisBertrand
Director,Research,InnovationandInternationalAffairs
PolytechniqueMontréal
MarthaCrago
Vice-President,ResearchandInnovation
McGillUniversity
RobertGagné
Director,ResearchandTransfers—ChairoftheGoverningBoard
HECMontréal–IVADO
Marie-JoséeHébert
Vice-Rector,Research,DiscoveryandInnovation
UniversitédeMontréal
Observers
GillesSavard
GeneralManager
IVADO
YoshuaBengio
ScientificDirector
MILA,
UniversitédeMonréal
DoinaPrecup
AssociateProfessor,ITSchool
McGillUniversity
Marie-JoséeBlais
AssistantDeputyMinister
MinistryofEconomy,ScienceandInnovation
MarioBouchard
AssistantDeputyMinister
MinistryofEconomy,ScienceandInnovation
80
FINALVERSION
List of people met during our work
Amyot,Robert–CAE
Anglade,Dominique–Ministèredel’Éducation,delaScienceetdel’Innovation
Baillet,Francis–Ubisoft
Barette,Gaétan–MinistèredelaSantéetdesServicesSociaux
Bastien,Frédéric–MILAetUniversitédeMontréal
Bastien,Frédéric–Mnubo
Beaudoin,Josée–CEFRIO
Béliveau,Louise–UniversitédeMontréal
Bellemare,Alain–Bombardier
Benoit,Suzanne–AllianceNumérique
Béraud,Frank–MontrealInVivo
Bernstein,Albert–CIFAR
Bertrand,François–PolytechniqueMontréal
Bérubé,Vincent–McKinsey&Compagnie
Blais,François–MinistèreduTravail,del'EmploietdelaSolidaritésociale
Boel,Nicolas-Louis–Altissia
Boivin,Benoit–SecrétariatduConseilduTrésor
Bolduc,Hubert–MontréalInternational
Borrelli,François–Prompt
Boufaden,Nariés–Keatext
Boulet,Benoît–UniversitéMcGill
Brodeur,Pierre–RegroupementdescollègesduMontréalmétropolitain
Brouillet,Eugénie–UniversitéLaval
Brunel,Benoit–BonjourSanté
Brunelle,Eric–DruideInformatique
Brunelle,Martin–Desjardins
Bruno,Richard–CIRANO
Cai,Changtian–HuaweiThechnologies
Carignan,François–Purkinje
Carle,Sylvain–AeroMontréal
Chagnon,André–SOJECCI
Chapados,Nicolas–Imagia
Chenette,Madeleine–Accenture
Clark,James–UniversitéMcGill
Clerk,Maxime–Prompt
Cléroux,Louis–BlockchainSuperclus
Coiteux,Martin–MinistèredesAffairesmunicipalesetdel'Occupationduterritoire
Corbo,Jacomo–QuantumBlack
Côté,Diane–MEDTEQ
Côté,Myriam–MILA
Cournoyer,Jean-Sébastien–ElementAI
Courville,Aaron–MILAetUniversitédeMontréal
D’Amours,Sophie–UniversitéLaval
Debbabi,Mourad–UniversitéConcordia
Deguire,Charles–Kinova
Delage,Jean-Nicolas–Notman–OSMO
Desbiens,Denis–IBM
Desmarais,Hélène–Centred’entreprisesetd’innovationdeMontréal
Divry,Lidia–TechnoMTL
Dumouchel,Pierre–ETS
ElAsri,Layla–Microsoft
Émond,Marie-Claude–AllianceNumérique
Faubert,Denis–CRIAQ–Most21
Fortier,Suzanne–UniversitéMcGill
Gagné,Christian–UniversitéLaval
Gagné,Jean-François,ElementAI
Garinis,Tom–Deloitte
Gauthier,Johanne–MTESS
Girard,Anne-Marie–UniversitédeMontréal
Gravel,Vivianne–B-citi
GrégoireRousseau,Jean-Sébastien–McKinsey&Compagnie
Halinaty,Mark–ThalesCanada
Hamel,Manon–CaissededépôtetplacementduQuébec
Houle,Sébastien–Productique
Huneault,Michel–ETS
Jacques,Pierre-Étienne–UniversitédeSherbrooke
Jaskolka,Norman–Aldo
81
FINALVERSION
Klugman-Iain–Communitech
Labonté,Françoys–CRIM
Lacoste-Julien,Simon–MILAetUniversitédeMontréal
Lajoie,Gilles–SOQUIJ
Lalonde,Fred–GERAD
Lamothe,Marie-Josée–GoogleCanada
Lanoue,Josée–TechnoCompétences
Laportevation,Réal–Hydro-Québec
Larochelle,Hugo–MILAetGoogleCanada
Laviolette,François–UniversitéLaval
Lavoie,Alain–Irosoft
Lavoie,Valérie–Fédérationdescégeps
LeBouthillier,Alexandre–Imagia
Leblanc,Martin–Caprion
LeCun,Yann–Facebook
Létourneau,Stéphane–UniversitédeMontréal
Lévesque,Louis–FinanceMontréal
Lodi,Andrea–ChaireCERC,PolytechniqueMontréal
Martel,Nicole–AQT
Masson,Alexandra–QuébecInternational
Mauro,Andy–Automat.a
Molaret,Philippe–ThalesCanada
Montini,Emmanuel–CégepdeBois-de-Boulogne
Moravei,Amir–Botler/Nomino
Musbah,Mohamed–Microsoft
Noël,Jean-François–MEES
Pal,Christopher–PolytechniqueMontréal
Parent,Marc–CAE
Park,Thomas–BDCCapital
Patry,Michel–HECMontréal
Perreault,Jean-Pierre–UniversitédeSherbrooke
Pilon,Pascal–LANDR
Pineau,Joëlle–MILAetUniversitéMcGill
Poulin,Daniel–LEXUM
Poulin,Geneviève–Ubisoft
Prodromou,Evan–Fuzzy.ai
Raimondi,Patrick–Accenture
Rigazio,Luca–Panasonic
Roy,Louise–UniversitédeMontréal
Sabia,Micheal–CaissededépôtetplacementduQuébec
Schlesinger,Adam–BanqueNationaleduCanada
Schuler,Cameron–AMII
Sherry,Dave–Lightspeed
Sirois,Luc–Prompt
Smirnov,Alexis–Dialogue
Talon,Suzanne–CalculQuébec
Tarnowski,Marcos–McKinsey&Compagnie
Têtu,Louis–Coveo
Thérien,Denis–UniversitéMcGill
Tremblay,Jean-François–Institutdel’électrification
Trépanier,Martin–CIRRELT
Trischler,Adam–Microsoft
Usal,Siefried–ThalesCanada(CortAIx)
Vachon,Louis–BanqueNationaleduCanada
Vallée,Nathalie–CollègeAhuntsic
Veerman,Alan–VectorInstitute
Viel,Carl–QuébecInternational
Vigeant,Daniel–CollègeLionel-Groulx
Vigneault,Mathieu–Transtech
Vinet,Luc–CRM
Wallis,Jodie–Accenture
TheCommitteewouldalsoliketothankallthepeoplewhoweremetduringtheworkontheStrategyfortheDevelopmentofQuebec’sArtificialIntelligenceEcosystemthatarenotmentionedinthislist.
82
FINALVERSION
Overview of public and corporate f inancing
Figure6–AIorganizationssupportedandsourceoftheinvestments
83
FINALVERSION
Online courses
QuebecwillbenefitfromitssupportofonlinecoursedevelopmentasitwillmeetthetrainingneedsofalltheAI
ecosystem’shumanresources:pre-universitygraduateswhowanttobecomefamiliarwithdatapreparation
techniques,computerscientistseagertoacquiremachinelearningskillstocomplementtheirbasicskills,
agriculture,healthortransportationexpertslookingtouseAIintheirnicheofexpertiseormanagerswhowantto
meetthestrategicchallengesofAIandAI-applicationuseintheirbusiness.
Firstofall,thecreationofFrenchonlinecourses–asubjectinwhichUniversitéLaval,HECMontréalorTéluq
possessenviableexpertise–willmakeitpossibletoreachallQuebecersconcernedbyAInomatterwherethey
live.WhileitwillberelativelyeasyforworkersandfutureworkersinMontrealandothermajorurbancentresin
QuebectotakecoursesatMILAorothercentresorin-persontraininginstitutions,someAIcoursesshouldalsobe
offeredvirtuallysothattheyareaccessibletothegreatestnumberofpeople,fromGaspésietoOutaouaistoLes
Bois-Francs.TheriseofAIwillaffectallofQuebec’sregions,andallofthemshouldhaveeasyaccesstohigh-
qualityonlinecourses.
Somealsobelievethat,giventherightconditions,onlinecoursescanbeaneffectivetoolforattractingwomento
thefieldofAI.Theflexibilityofferedbythistrainingapproachcouldhelpsomewomen,includingmothers,
continueorresumetheirstudies.59
ThecreationofFrenchonlinecourseswillalsobeagreatwaytoraiseawarenessandtrainmembersofLa
FrancophoniewhowanttopursueacareerinAIorbetterunderstandthetechnologiesproducedbythissector.In
somecases,theonlinecoursesdevelopedinQuebecwillevenencourageforeignstudentstocompletetheir
traininglocally,eitherinQuebecuniversitiesorcollegesacrosstheprovince.Infact,QuebeconlineAIcourses
couldserveasagreatcallingcardforourresearchers,educationalandresearchinstitutionsandbusinesses.
Intheshortterm,thecreationofFrenchonlinecoursesmayevenencourageorganizationstoadoptAI,since
someofthemarecurrentlyreluctanttoprovideinstructor-ledtrainingfortheiremployeeslesttheyberecruited
bythecompetitiononcetheircourseisover.
ForQuebec,thechallengeoverthenextfewyearswillnotbetodirectoroverseetheeffortsalreadyundertakenbymanyoftheprovince'sacademic,collegeandtechnicalinstitutionstocreatecreditedanduncreditedonlineAI
coursesthattheywillofferaspartoftheirnormalorcontinuingtrainingprograms.
However,theprovinceshouldmakeresourcesavailablesothattrainingstakeholderscandevelophigh-quality
onlinecoursesonstrategicAItopicswhereneedsdonotmatchthetrainingoffering.Thismayoccurbecause:
• Producingaveryhigh-qualitycoursecanbeexpensiveforaninstitution(anestimated$80,000to
$150,000).
• Theinitialdemandforacoursemaybeinsufficient(toofewstudentsenrollinthefirstyears).
• Thedemandforacoursecanbeconsiderablewithoutanyreturnoninvestment(e.g.,inthecaseoffree
orlow-costbulkcourses,theexpenseswillobviouslybegreaterthantherevenues).
Therefore,ifthe“talentwatch”revealsalackofonlinecourses,theSteeringCommitteewillrecommendthe
QuebecgovernmentcreateabudgetenvelopetosupportthedevelopmentbyQuebec'seducationalinstitutions
ofonlineAIcoursestomeetthevariousneedsofQuebecorganizations.
ThemanagementofthisbudgetenvelopecouldbeentrustedtoAI.Quebec,which,incloseconsultationwiththe
industryandAIeducationalinstitutions,wouldissuecallsforproposalstosetuponlinecoursesinFrench,and
possiblyinotherlanguages,suchasEnglishandSpanish,totrainthehumanresourcescompaniesthatdevelop
anduseAIwillrequireatalllevels
59Seehttps://www.al-fanarmedia.org/2017/03/can-moocs-womens-entryway-stem/.
84
FINALVERSION
TheamountsavailableshouldbeusedtocreatecoursesthatwillattractQuebecandforeignworkersandstudents
totheQuebecAIecosystem,particularlyuniversityandcollege-basedinstitutionsthatofferprogramsthatwill
enablethemtoperfecttheirAItraining.
Inthisrespect,itwouldbeagoodideatoapplytotheAIsectiontheJeunesleadersdelaFrancophonie
frameworkthatHECMontréalanditspartnersadvocate.SupportedfinanciallybyGlobalAffairsCanada,this
initiativeaimstoencouragestart-upsinBenin,BurkinaFaso,HaitiandSenegal.TenthousandFrancophoneswill
beginbyfollowingaseriesofonlinecoursesforpotentialentrepreneurs.Then,themostpromisingstudentswill
receivesupportfromtutorsandmentorstoproduceabusinessplanandstarttheirownbusiness.Theproject’s
promotershopethat,inthelongrun,itwillgenerate240newstart-upcompaniesinthesefourselected
countries.60
DevelopAI’sFrenchlexiconTheDataFrancaproject(www.datafranca.org),issupportedfinanciallybytheOfficequébécoisdelalangue
française,ledbyateamthatincludesanartificialintelligenceandlinguisticsspecialistandendorsedbycompanies
likeGoogleBrainMontréal.TheprojectaimstodevelopFrenchAIterminologythroughaparticipatoryprocess.
TheterminologyproducedbytheexpertswillbeusedtocreateFrenchWikipediaarticlesonAIandimprove
existingWikipediaarticles.DataFrancamanagerswillalsodistributetheresultsofthisprojectonnetworkssuchas
LinkedInandreferenceresourcessuchastheGranddictionnaireterminologique.Bysupportingthisinitiative,theOfficequébécoisdelalanguefrançaiseishelpingQuebecsocietyembraceawholenewindustry(inthewordsof
thepromoters,tomakeit“asQuébécoisasmaplesyrup”),promotethetransferofAIknowledgetonon-experts
(French-speakingAIusers,journalists,teachers),providetheframeworkforfutureFrench-languageAIcourses
(e.g.,forCEGEPstudentsandAIusers)andpromotethecominginQuebecof“French-speakingbrains”interested
inAI.
60Seehttp://fr.allafrica.com/stories/201710110712.html.
85
FINALVERSION
AIstandards
TheInternationalOrganizationforStandardization(ISO)definesitsstandardasa“document,establishedbyconsensusandapprovedbyarecognizedbody,thatprovides,forrepeatedandcommonuse,rules,guidelinesorcharacteristicsforactivitiesortheirresults,aimedattheachievementofanoptimumdegreeoforderinagivencontext.”TheISOaddsthat“[s]tandardsshouldbebasedontheconsolidatedresultsofscience,technologyandexperience,andaimedatthepromotionofoptimumcommunitybenefits.”61
StandardsareusuallypublishedbynationalstandardizationbodiessuchastheStandardsCouncilofCanadaorinternationalorganizationssuchasISO.
Standardscanalsobegeneratedbyacompanythatholdsastrongcompetitiveposition(suchasMicrosoft),aprofessionalassociation(suchastheInstituteofElectricalandElectronicsEngineersorIEEE)oraconsortiumofcompanies(liketheCanadianWirelessTelecommunicationsAssociation).
Standardsareimportantbecausetheyleadtoreducedrisksfordevelopers,promotersandusersofaproductorservice.Forexample,settingstandardsonrailspacingsupportedthedevelopmentoftheU.S.railindustry.Indeed,beforethespacingwassetto4’8½”attheendofthe19thcentury,trainscouldnotnecessarilymovefromonerailnetworktoanother.Similarly,settingrulesabouthowmuchmethylmercuryisacceptableinfishthatissoldprotectsthehealthofconsumers.
Inthetechnologysector,standardsensurethesmoothrunningoftheInternet,sothatthesamecomputercanseamlesslyconnecttothewebfromanycountryandanyWi-Finetwork,helpuserslearnnewsoftware(it’snocoincidencethattheGoogleDocsmenubarlookslikeMicrosoftWord's,theleaderinwordprocessingsoftware)andletsthemuseFrenchorChinesecharactersinadomainname.
ThecreationofspecificAIstandardshasintensifiedinrecentyears.In2017,theISOsetupSub-Committee42(SC42)to“[s]erveasthefocusandproponentforJTC1's[thereferencebodyforthestandardizationofinformationtechnologies]standardizationprogramonArtificialIntelligence”and“[p]rovideguidancetoJTC1,IEC,andISOcommitteesdevelopingArtificialIntelligenceapplications.”62Foritspart,theIEEEhasjustpublishedthesecondeditionoftheEthicallyAlignedDesignReport:AVisionforPrioritizingHumanWellbeingwithAutonomousandIntelligentSystems(A/IS),whichcontainsrecommendationsonstrategiesfordesigningintelligentandautonomousITtoolsthattakeintoaccounthumanvaluesandwhichcangraduallybetranslatedintostandards.63
InAI,standardscanbeusedtoensure:
• ThatAIproducersandusersusecommondefinitionsandvocabularytofacilitatetheirexchangeonthesubject.
• ThatitiseasierandcheapertoexchangeandintegratethedatarequiredforAIdevelopment.• ThatAIpractitionersuseexemplarydesignandproductionpracticesand,asaresult,avoidsuboptimal
waysofdoingthings(suchasover-learningoroverfitting).
• ThattherisksassociatedwiththeemergenceofnewAItechnology(e.g.,self-drivingvehicles)andmethodstomitigatetheserisksbeidentifiedusingspecificandrigorousapproaches;64and
• Thattheintegrationoffundamentalethicalprinciples(e.g.,equity)bedoneattheinitialstagesoftooldesign.65
61Seehttps://www.cairn.info/revue-francaise-de-gestion-2003-6-page-49.htm.62Seehttps://www.iso.org/fr/committee/6794475.html.63Seehttps://ethicsinaction.ieee.org/.64Seehttps://arxiv.org/pdf/1709.02435.pdf.65Seehttps://www.techrepublic.com/article/ieee-announces-3-ai-standards-to-protect-human-well-being-in-the-robot-revolution/.