Transcription
A survey of the European Union’sartificial intelligence ecosystem
Compiled by Charlotte StixResearch Associate, Leverhulme Centre for the Future of Intelligence, University of CambridgeAdvisor, Element AIThis report is by no means final or comprehensive. It will act as a living document regularly undergoing updatesto reflect the rapidly changing AI landscape within the European Union.Thanks for edits and comments to Seán Ó hÉigeartaigh, Stephen Cave, Matthijs M. Maas, Jess Whittlestone, EmmaBates, Rune Nyrup and Haydn Belfield.2 A survey of the European Union’s articial intelligence ecosystem March 2019
Contents:Executive SummaryIntroductionPart One: The overarching planThe European Union’s AI strategyDigital Day Declaration, ‘Declaration of Cooperation on Artificial Intelligence’European Commission’s ‘Communication on Artificial Intelligence’European Commission’s ‘Coordinated Plan on Artificial Intelligence’High-Level Expert Group on Artificial IntelligenceEuropean Artificial Intelligence AllianceThe European Union’s visionEuropean Group on Ethics in Science and New Technologies’ statement on ‘ArtificialIntelligence, Robotics and ‘Autonomous Systems’’High-Level Expert Group on Artificial Intelligence’s ‘Draft Ethics Guidelines forTrustworthy Artificial Intelligence’European Economic and Social Committee’s ‘Opinion on AI, Artificial intelligence –The consequences of artificial intelligence on the (digital) single market, production,consumption, employment and society’General Data Protection Regulation and ePrivacy RegulationDECODEAlgo AwarenessPart Two: The ingredientsFunding and financial supportVentureEUEuropean Fund for Strategic InvestmentEuropean Innovation CouncilHorizon 2020 (2018-2020)Digital Europe programme (2021-2027)A survey of the European Union’s articial intelligence ecosystem March 2019 3
Talent creationNew Skills AgendaDigital Skills and Jobs CoalitionPutting the pieces back together(i) Large-scale collaboration(i.a) Active collaboration: infrastructureAI4EU: AI-on-demand platformDigital Innovation Hubs(i.b) Active collaboration: R&D&I and hardwareSPARCElectronic Components and Systems Joint UndertakingEuropean High-Performance Computing Joint Undertaking(i.c) Proposals for collaboration: pan-European AI laboratoriesA ‘CERN for AI’European Lab for Learning & Intelligent SystemsConfederation of Laboratories for Artificial Intelligence Research in Europe(ii) Granular collaborationJoint European Disruptive InitiativeParis Artificial Intelligence Research InstituteEuropean Association for Artificial IntelligenceSummaryAppendix4 A survey of the European Union’s articial intelligence ecosystem March 2019
Acronyms:AAAI:Association for the Advancement of ArtificialIntelligenceEurAI:European Association for ArtificialIntelligenceAENEAS:Association for European NanoElectronicsActivitieSGDPR:General Data Protection RegulationAI:Artificial IntelligenceHBP:Human Brain projectAI HLEG :High-Level Expert Group on ArtificialIntelligenceARM:Acorn RISC MachineARTEMIS:Advanced Research & Technology forEMbedded Intelligent SystemsCERN:Conseil Européen pour la RechercheNucléaire (European Organization for NuclearResearch)EuroHPC JU: High-Performance Computing JointUndertakingHQ:HeadquartersH2020:Horizon 2020ICT:Information and Communications TechnologyIEEE:The Institute of Electrical and ElectronicsEngineersIJCAI:International Joint Conference on ArtificialIntelligenceJEDI:Joint European Disruptive InitiativeMAR:SPARC’s Multi-Annual RoadmapMILA:Montreal Institute for Learning AlgorithmsMOOC:Massive Open Online CourseMFF:Multiannual Financial FrameworkNATO:North Atlantic Treaty OrganizationPPP:Public-Private PartnershipPRAIRIE:Paris Artificial Intelligence Research InstituteCIIRC:Czech Institute of Informatics, Robotics andCyberneticsCLAIRE:Confederation of Laboratories for ArtificialIntelligence Research in EuropeDARPA:American Defense Advanced ResearchProjects AgencyDIHs:Digital Innovation HubsECAI:European Conference on Artificial IntelligenceECSEL:Electronic Components and Systems JointUndertakingEESC:European Economic and Social CommitteeR&D:Research and DevelopmentEFSI:European Fund for Strategic InvestmentR&D&I:Research, Development and InnovationEGE:European Group on Ethics in Science andNew TechnologiesSMEs:Small and Medium sized EnterprisesSPARC:Public-Private Partnership for robotics inEuropeEIB:European Investment BankEIF:European Investment FundSRA:SPARC’s Strategic Research AgendaELLIS:European Lab for Learning & IntelligentSystemsSTEM:Science, Technology, Engineering and MathS&T:Scientific and TechnologicalEPoSS:European Platform on Smart SystemsIntegrationVC:Venture CapitalEU:European UnionWG:Working GroupsA survey of the European Union’s articial intelligence ecosystem March 2019 5
Executive SummaryNarratives in international media, andincreasingly within governments, place greatimportance on nations achieving leadershipin artificial intelligence (AI). The EU1,2 israrely considered the leading player in thesediscussions.3 This report investigates thisassumption and outlines existing building blocksthat could form the basis for EU leadership inAI. The research is based primarily around EUlegislation, policy and strategy documents,publicly available databases of ongoing projects,and funding decisions. In aggregating thisinformation, the report aims to provide anintroductory overview of the EU’s AI ecosystem.The report is structured into the followingsections: (1) strategy and vision4,5 and (2.a)funding and financial support, (2.b) talentcreation 6,7 (2.c) collaboration.8 The primaryconclusions around each of these are as follows:(1) The EU, via the European Commission in theirCommunication on AI, sets out a “EuropeanInitiative on AI” as well as a Coordinated Plan onAI from the EU Member States.9 This initiative issupported with a declaration signed by all 28Member States, the Declaration of Cooperationon AI. The Coordinated Plan on AI outlineshow EU Member States could coordinate theirstrategies, financial commitments and otherresources to increase European competitivenessas a whole. Perhaps most notably, the EU’s visionfor AI is centered around ‘ethical AI’ in a way thatcould distinguish it from its American or Chinesecounterparts.(2.a) Although it is too early to judge theirimpact, the EU is undertaking active steps totackle funding bottlenecks.10 The EuropeanCommission will invest 1.5bn11 between 20182020 towards research and innovation in AI, a70% increase from current investments, andexpects to invest 2.5bn12 during the DigitalEurope Programme (2021-2027) alone. Thesenumbers exclude funding from other sourcessuch as the European Research Council and theEuropean Fund for Strategic Investment. TheEU is struggling to attract venture capitalists.They invested only 6.5bn in the EU in 2016,in comparison to 39.4bn in the US.13 Newlyestablished initiatives such as VentureEU aim toredress this imbalance.(2.b) The EU has a number of initiatives dedicatedto talent creation for the ICT and digital sector.However, there is mounting concern over braindrain and a loss of future-oriented entreprises,with researchers leaving the EU and a numberof promising companies being acquired byinternational companies, such as Acorn RISCMachine (acquired by SoftBank),14 KUKA(Midea),15 and Magic Pony Technology (Twitter)16.Talent creation through education, re- andupskilling may put the EU on a solid foundationfor future AI competitiveness, but brain drainremains a substantial concern.6 A survey of the European Union’s articial intelligence ecosystem March 2019
(2.c) The EU has several ongoing and upcominginitiatives between Member States and groupswithin those that are of collaborative nature.Large-scale infrastructural collaborations such asthe Digital Innovation Hubs and the AI4EU pilotsexist alongside prominent collaborations onresources, evidenced by the Joint Undertakingsfor electronic components as well as highperformance computing. Looking forward,the EU could build on a track record of majorcollaborative projects such as the Human BrainProject and CERN, which could provide a modelfor collaborative AI initiatives of similar scale.This report does not judge whether thesecomponents will suffice for the EU to achieveor maintain AI leadership. Future work will beneeded to assess the effectiveness of theseinitiatives and compare them to those of othernations. However, by providing an overview ofthe EU AI ecosystem this report hopes to opena broader discussion about what EU leadershipin AI could look like and highlight many of thecomponents that could provide its basis.A survey of the European Union’s articial intelligence ecosystem March 2019 7
IntroductionRecent progress in AI has led a variety ofcountries to set up governmental advisoryboards, propose AI strategies and establish newinstitutions that focus on the societal impact ofthis technology. Some countries and strategiesfeature more prominently in public fora thanothers. This has the potential to contributeto a misleading impression of the shape andstate of the global AI ecosystem. In contrast toother global powers, the EU appears at timesreluctant to trumpet achievements, initiatives orambitions, be that in AI or other fields. As a result,awareness about the work undertaken by the EUis often lacking, which, in turn, contributes to thenarrative that it is on the waiting bench when itcomes to AI.This report aims to rectify this impression byproviding a deeper insight into the EU’s AIecosystem through an investigative tour of past,current and proposed AI-relevant activities.Several areas are examined in a closer manner:(1) strategy and vision,17,18 (2.a) funding andfinancial support, (2.b) talent creation,19,20 and(2.c) collaboration.21Part 1 outlines the EU’s strategy and vision withregard to AI. It is divided into two sections.Section (a) looks directly at the EU’s AI strategy.Section (b) explores the accompanying vision.Section (a) outlines the EU’s AI strategy throughthe introduction of three key publications:the Digital Day Declaration ‘Declaration ofCooperation on AI’, the European Commission’s‘Communication on Artificial Intelligence’and the European Commission’s ‘CoordinatedPlan on AI’. It then presents the two groupsthat help steer the European Commission inthe implementation of several componentsof these strategy documents: the High-LevelExpert Group on Artificial Intelligence and theEuropean Artificial Intelligence Alliance. Thefirst group is composed of 52 subject expertswho act as an advisory group to the EuropeanCommission on ethics, investment and policymaking and work on outlining a long-term AIstrategy. The second group is a multi-stakeholderforum open to all EU citizens to share theiropinion, concerns and feedback on developingAI policy and strategy. Crystallising out of section(a) is the image of the EU as an actor that selfidentifies as a leader in ‘ethical AI’.22Section (b) explores this on the basis of recentreports, papers, regulations and EU-fundedprojects. It introduces three relevant documents,the European Group on Ethics in Science andNew Technologies’ statement on ‘ArtificialIntelligence, Robotics and ‘AutonomousSystems’’, the High-Level Expert Group’s ‘DraftEthics Guidelines on AI’, and the EuropeanEconomic and Social Committee’s ‘Opinion onArtificial Intelligence’.23 All three publicationsdemonstrate the importance that actors withinthe EU place on human-centric and ‘ethicalAI’. This is followed by a brief examination ofregulations that could contribute to ‘ethicalAI’, such as the General Data ProtectionRegulation and ePrivacy Regulation.Finally, two EU funded projects are discussed,concerning related issues such as data usage,ownership and algorithmic decision making froma non-regulatory perspective: DECODE and AlgoAwareness. Part 1 concludes with a summary ofthe EU’s AI strategy and vision.8 A survey of the European Union’s articial intelligence ecosystem March 2019
Part 2 then investigates the granular elementsthat will support the EU in the implementationof its strategy and vision. It is divided into threesections: (a) funding and financial support, (b)talent creation and (c) collaboration.Section (a) explores funding and financialsupport. It looks at the newly proposedVentureEU fund and the European Fund forStrategic Investment. It then broadly coversother funding sources for research, developmentand innovation such as the Horizon 2020framework programme for research andinnovation (2018-2020), Horizon Europe (20212027) succeeding Horizon 2020, the EuropeanInnovation Council and the upcoming DigitalEurope Programme.Section (b) explores talent creation. The EUtackles talent creation through education,upskilling and reskilling, as evidenced by e.g theNew Skills Agenda and the Digital Skills andJobs Coalition. Various national strategies, suchas Germany’s Eckpunkte der Bundesregierung füreine Strategie Künstliche Intelligenz,24 Finland’sAge of Artificial Intelligence25 and France’s Donnerun sens à l’intelligence artificielle26 propose moredirect next steps to halt brain drain.Section (c) investigates areas of collaborationand cooperation. It is divided into twosubsections: (i) exploring initiatives at macrolevel (i.e. collaborations between MemberStates) and (ii) exploring initiatives at a moregranular level. Subsection (i) highlights existingcollaborative infrastructure: the AI4EU project(formerly AI-on-demand platform) and theDigital Innovation Hubs initiative. Both aim toincrease access to AI, including to data, tools, andadvice, for small and medium sized enterprises(SMEs). This is followed by the introduction ofa number of other EU-wide collaborations thatare likely to play an important role for the EU’sAI leadership. In particular, it outlines SPARC, aPublic-Private Partnership between the EuropeanCommission and the robotics community,the Electronic Components and SystemsJoint Unde
(Midea),15 and Magic Pony Technology (Twitter)16. Talent creation through education, re- and upskilling may put the EU on a solid foundation for future AI competitiveness, but brain drain remains a substantial concern. A survey of the European Union’s articial intelligence ecosystem March 2019 7 (2.c) The EU has several ongoing and upcoming initiatives between Member States and groups .