Europe has finally a united and official position about Artificial Intelligence. On April 25 the European Commission released the communication Artificial Intelligence for Europe, outlining the plans for the coming years and finally breaking the embarrassing silence that followed the publication of China AI plan almost one year ago. The strategy is an indication, since the document is not legally binding.
European strengths and strategy in a nutshell
“Just as the steam engine and electricity did in the past, AI is transforming our world,” said Andrus Ansip, Vice-President of the European Commission in support of the strategy release. In the presentation occurred at the European Parliament on April 26, MEP Marietje Schaake spoke of a novel challenge for European policymaking: more than any previous technology paradigm, AI questions European institutions on defending the public interest proactively, by taking into account the wider societal impact while supporting the development of the technology and economy dimensions. The MEP also highlighted that AI is a rare case where people ask for more technology regulation, while Europe is often criticised for putting barriers to innovation.
The strategy aims at leveraging European excellences in research and industry, while ensuring that society isn’t left behind. More precisely, the European strategic assets, as presented in the communication and on April 26 by Juha Heikkila (Head of Unit of DG Connects Robotics and Artificial Intelligence Unit), are:
Research excellence (the EU accounts for one-fifth of the world’s R&D investment, and 23% of the global public R&D. With more than 8 million researchers, the EU is the economy with the largest number of researchers). Moreover, most big tech companies have already or are setting research labs in the EU
The Digital Single Market, whose common rules facilitate business and technology adoption across borders;
A large number of startups (Startup Hubs maps almost 1 million startups, with a €426 billion revenue);
An advanced robotics sector (32% of current world market: 1/3 of Industrial robotics market, 63% of professional service non-military robot market);
The availability of public data (the report Economic Benefit of Open Data calculates a €325 billion direct market size for the period 2016-2020 and 25.000 jobs in Open Data per year).
Policymakers are aware that the EU can capitalise on these assets only with an integrated approach including all Member States. The strategy is based on three axes:
- Strengthening the tech and industrial side of AI
- Promoting the uptake of AI by private European companies
- Preparing the change via the social and cultural framework (non-tech aspects like impact on jobs, ethics, diversity…)
The means deployed so far at the EU level are the building of an AI Alliance ensuring a broad consultation mechanism, which will be led by a Hi-Level AI Expert Group (the EC is currently analysing the applications). The first concrete task of the group will be to come up with ethical guidelines by the end of the year. Secondly, the Digital Innovation Hubs launched in 2017 will act as multipliers on the European territory, and facilitate the adoption of AI.
During the presentation EU officers described an effort to “democratise AI”, meaning that they want to make sure that AI is broadly available, so that companies can embed it in their processes. Central in this distribution of AI is the idea to build an “AI on-demand platform” starting next year, to concentrate in one place all the AI-related resources (resources are described as “knowledge, data repositories, computing power, tools and algorithms”).
The AI strategy sees the post-2020 as an era of upgrade and consolidation of research, facilities and uptake mechanisms: proposals range from upgrading the pan-European network of AI excellence centres and Digital Innovation Hubs to exploring joint innovation procurement for the use and development of AI; it’s worth reporting also “a support centre for data sharing, which will be closely linked with the AI-on-demand platform to facilitate development of business and public sector applications”.
What about money?
In 2016 European private investments in AI amount to around €2.4-3.2 billion, paling in comparison to Asia almost €10 billion euros and America €18 billion. It is evident that without massive investments Europe will simply be a data provider for somebody’s else AI.
The European Commission considers that between now and the end of 2020 both private and public sector in Europe should invest €20 billion in AI: to start with, the topic will be reinforced by an additional investment of €1.5 billion for 2018-2020 under the Horizon 2020 research and innovation funding framework. Such investment is expected to trigger another €2.5 billion from public-private partnerships.
On top of H2020, the European Fund for Strategic Investments will work towards attracting private investments. The Commission will work with the European Investment Bank Group aiming for at least €500 million in total investments in AI in the period 2018-2020. Finally, the EC mentions the launch with the European Investment Fund of VentureEU, a €2.1 billion Pan-European Venture Capital Funds-of-Funds programme, which aim is to boost investment in innovative startup and scale-up companies across Europe (but does not precise how much of it will be dedicated to AI).
The new financial framework: 2021-2027
After 2020 – the end of the current 7-year financial framework – the strategy numbers are vague. On May 2 the EC presented the proposal for the next EU multiannual financial framework (2021-2027), which will have to deal with the budget hole left by the UK (€13 billion loss per year). Cuts will concern mainly agriculture and regional funds, and proposals for alternative sources of budget have been made, including a redirection to the EU of revenues from selling carbon pollution permits under the EU emissions trading system, and a new tax on non-recyclable plastics. It’s worth noting that a total budget of €13 billion would be dedicated to the European Defence Fund, whose aim is to build interoperable defence technology and equipment. Such investment would place the European Union among the top 4 of defence research and technology investors in Europe, and continue the tradition of letting the military sector lead innovation. After all, the subtitle of the financial proposal is “a Union that protects, empowers and defends”.
On the positive side, Commission proposes to increase investment in research and innovation by allocating €114.8 billion (the current framework does not reach 80 billion). Moreover, Digital Transformation is set as a priority, equipping the Digital Europe programme with an overall budget of €9.2 billion to shape and support the digital transformation of Europe’s society and economy. Here, AI is finally mentioned: its adoption is presented as the only way to capture the data economy value, and big attention is placed on the Digital Innovation Hubs, which are supposed to ensure the availability and proximity of facilities and expertise. Supercomputing and digital skills are also part of the programme.
Member States initiatives
A few days before the release of the strategy, 25 Member States (Austria, Belgium, Bulgaria, Czech Republic, Denmark, Estonia, Finland, France, Germany, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Poland, Portugal, Slovakia, Slovenia, Spain, Sweden, UK, Norway) signed a Declaration of Cooperation on AI, where they commit to boost Europe’s technology and industrial capacity in AI and its uptake, including better access to public sector data, while addressing socio-economic challenges and ensuring an adequate legal and ethical framework.
Always in April, an international group of scientists from key institutions (including Max Planck Institute and Cambridge University) pledged with an open letter the creation of a European AI hub to be named Ellis (European Lab for Learning and Intelligent Systems) so that existing excellence can be streamlined and kept in Europe.
Meanwhile, individual Countries are interpreting in their own way the AI challenge. Macron made a stand in March by presenting Cédric Villani’s study and promising to invest €1.5 billion over five years, while opening up public data to support research as well as startups. Facebook, Google, Samsung, IBM, DeepMind, Fujitsu choose Paris to create AI labs and research centres, which makes the President particularly optimist on making France a lab for AI. Angela Merkel in April has highlighted the importance of winning the AI race against China, but did not make any formal investment commitment. A sense of urgency comes from the French and German industry representatives who are behind the JEDI (Joint European Disruptive Initiative), a non-governmental disruptive innovation funding scheme inspired by the US DARPA that promises to be agile and bet on ambitious ideas. The 2018 provisional budget is estimated at €235 million, of which €7 million will go into operational costs, aiming for a total €1 billion budget per year in the future. Technology verticals priorities are identified as AI, cybersecurity, components, computing power, biotechnologies, energy storage and nanotechnologies. The idea to launch a European DARPA comes from Macron: at the time the EU Research commissioner Carlos Moedas said it was too top-down and a duplication of the European Innovation Council (budget: €2.7 billions for three years). Nonetheless, the EIC focuses on close-to-market innovation, while JEDI aims at financing fundamental research and innovation with an eye on quick prototyping.
Elsewhere, national AI working groups and reports are becoming the norm. In March the Italian Digital Agency (AGID) released an AI White Paper, where the human part and the role of creativity are quite central (the subtitle is “at the service of citizens”). The aim is to improve the public administration via AI: the Agency has promised to invest €5 millions from its public procurement and open innovation fund in pilot projects following the White Paper guidelines. In Germany, the Plattform Lernende Systeme, which was launched by the Federal Ministry of Education and Research (BMBF) in 2017, has as a goal “designing self-learning systems for the benefit of society is the goal pursued by”. Similarly, the Finnish Minister of Economic Affairs Mika Lintilä appointed a steering group to prepare a proposal for an artificial intelligence (AI) programme for Finland.
Acknowledging that UK can’t compete with larger nations when it comes funding AI, a House of Lords report claims brexiters can lead the way with ethics. Quite ironic for a Country dismantling public services. Private investment is indeed key: only £300 million of the £1 billion-plus deal “to make the UK a global leader in AI” is public funding. The rest comes from more than 50 private companies and other organisations, including Microsoft and IBM. Similarly, JEDI will tap into both French governmental funding and private donors.
On the short run, the plan is penalised by a lack of legal coercion, and it appears vague compared to the Chinese Three-Year Action Plan to Promote the Development of New-Generation Artificial Intelligence Industry (2018-2020) released the past December, where research avenues are clearly identified. Moreover, the €20 billion investment targeted for 2020 is an indication based on optimistic uncertainty: “If Member States and the private sector (beyond established partnerships) make similar (referring to the additional 1.5 million poured into H2020) investment efforts, the total investments in the EU will grow to around EUR 7 billion per year, totalling more than EUR 20 billion by the end of 2020.” But what if they don’t? The Cooperation Agreement does not identify any way to measure progress besides regular meetings and overall reads as a declaration of intent. The post-2020 is even more blurred.
But let’s not try to define innovation results beforehand: as much as China relies on ‘moonshot’ (without an evident business case) projects and investments, Europe is legitimate to get some creativity and open-ended processes in the game. The rationale of the Digital Innovation Hubs is precisely to support bottom-up innovation via a strong territorial presence across the Union. Unlikely China, the EU is not an authoritarian state which can rule out top-down directives: Member States individual initiatives and private sector investments will be crucial for the competitiveness of the EU. Hence it’s important to not lose a collective focus and sense of purpose. The closest to Chinese moonshots made-in-Europe framework we have is the concept of mission-oriented innovation coined by Prof. Mazzucato in February. Europe strategy should be moved by ambitious and inspiring goals, to be materialised in an increasingly concrete set of projects and applications. Imagine complementing and speeding this framework by having EU countries taking ownership of different pilots and sharing the results, instead of focusing all at once on implementing the same approach.
“Knowledge, data repositories, computing power, tools and algorithms”: this is an enormous variety of sources and formats to be stored in a single place. The establishment of a European AI platform is a boundless challenge, both in terms of technology infrastructure and navigation (anybody having experience with the EC Participant Portal can see this) and governability. Who will implement and maintain it? What is the degree of openness and reusability of the documentation stored there? What about security? Will there be incentives for contributing to it? The platform is presented quickly and with little detail, while alone it could guarantee the cohesion needed. If the European Commission wants to go for it, it will have to put in place before 2020 dedicated massive resources and the best brains to make it work.
Reactions to the idea of establishing Europe as a sort of moral compass for AI are quite ironic. But as a matter of fact, similarly to the blockchain phenomenon, everybody is trying to patch AI everywhere. Giving technology such a salvation role is precisely what Silicon Valley behemoths have been doing so far, with evident societal disaster. Blockchain can be simplistically described as an infrastructure supporting exchanges, so the ethic fundamental question there is what do we want to be subject of an extremely rational and irreversible exchange. AI, instead, is the ultimate version of programming, intended as computational problem solving. The challenge is then which problems can be described and solved by an equation or code: where is the limit? What can be sensed by sensors and what shall be sensed by human soft skills? And how do the two blend together without hampering the social infrastructure? Ultimately, it is a matter of understanding how to be competitive on rational processes (in industry, commerce, research) while making sure that citizens – in their quality of private persons and members of society – are empowered by the widespread of AI. Europe is left with two main challenges: the pragmatic one is ensuring to make the best out of the resources invested in AI, by coordinating efforts at Union, national and local level; in terms of vision, it is urgent to substantiate the ethics claims with business cases and policies supporting the commons and digital sovereignty.