European Network of AI Excellence Centres: Pillars of the European AI lighthouse (RIA)

Topic description

ExpectedOutcome:

Proposal results are expected to contribute to the following expected outcomes:

• Establishing a new pillar of the European AI lighthouse^[1]
• Reinforcing a leading unified European AI community
• Scientific progress in AI, addressing major challenges hampering its deployment.

To ensure European open strategic autonomy in critical technologies such as AI, with huge socio-economic impact, it is essential to reinforce and build on Europe’s assets in such technologies, including its world-class research community, in order to stay at the forefront of technological developments.

Europe has undeniable strengths with its many leading research centres, but the efforts are scattered.

As stated in the communication from the European Commission on Artificial Intelligence for Europe^[2] and the coordinated action plan between the European Commission and the Member States and Associated Countries^[3], joining forces will be crucial for international competitiveness. Europe has to scale up existing research capacities and reach a critical mass through tighter networks of European AI excellence centres. The proposals should develop mechanisms to reinforce and network excellence centres in AI, bringing the best scientists from academia and industry to join forces in addressing the major AI challenges hampering its deployment; therefore, reinforcing excellence through collaboration in AI throughout Europe.

Such networks are expected to mobilise researchers to collaborate on key AI research challenges and to progress faster in joined efforts rather than working in silos, leading to fragmented and duplicated efforts. Such networks, are essential to reach critical mass and overcome the present fragmentation of AI research in Europe.

To complement and extend the initiative started in H2020 to develop a vibrant European network of AI excellence centres, and a vibrant AI scientific community, the proposals should create a network of excellence for the following topic:

Safe and secure AI addressing safety and security by design: foster the collaboration between AI with safety engineering communities, security, cybersecurity and related communities, to develop safe-by-design and secure-by-design AI, data and robotics systems, contributing to safety and security of citizens and any users of such AI-powered systems. The focus is on research, but the work should also contribute to good practices, standards and certification, complying with regulations for AI safety and protection of fundamental rights.

Topics to be covered include, but are not limited to:

• Technical robustness and safety, incl. methods for evaluating the resilience of systems, and standardized ways of quantifying robustness of AI
• Privacy preserving techniques and infrastructures
• Human agency and oversight in terms of system security and safety; including explainability in human-readable terms allowing to detect/prevent/mitigate/recover from harm and threats.

Proposals will bring together the best European research teams around AI to join forces and address major technical as well as sector- or societal-driven challenges in strengthening excellence, networking, multidisciplinarity and developing academia-industry synergies, as relevant.

Composition of the Network:

• It should be driven by leading researchers in AI and AI relevant technologies from major excellent AI research centres, and bringing the best scientists across Europe, including also from promising research labs. It will bring on board the necessary level of expertise and variety of disciplines and profiles to achieve their objectives, ensuring a multidisciplinarity and multi-sectorial research approach, while respecting equality and diversity among the attracted talents.

Activities of the Network:

• In order to structure the activities, the proposals will focus on important scientific or technological challenges with industrial and societal relevance where Europe will make a difference, by building on strengths, or strengthening knowledge to fill gaps critical for Europe.
• Based on the identified challenges, the proposals will develop and implement common research agendas. The main vision and roadmap with targets within the projects, as well as methodology to implement and monitor progress will have to be specified in the proposal and can be further developed during the project.
• Scientific progress will have to be demonstrated through testing on application specific datasets or use-cases. By extending the benchmarking of foundational research to application specific areas, the research community will simultaneously address advancements in AI and grand societal and technological challenges.
• The proposals should define mechanisms to foster excellence throughout Europe, to increase efficiency of collaboration, including through networking and exchange programmes, and to develop a vibrant AI network in Europe.
• The network will disseminate the latest and most advanced knowledge to all the academic and industrial AI laboratories in Europe and involving them in collaborative projects/exchange programmes. (This could involve projects defined initially or via financial support to third parties, for maximum 20% of the requested EU contribution, with a maximum of 60k€ per third party^[4]).
• The network will develop, where relevant, interactions with the industry, in view of triggering new scientific questions and fostering take-up of scientific advances
• The network will develop collaboration with the relevant Digital innovation Hubs, World-class reference testing facilities, AI regulatory sandboxes and AI start-up initiatives, to disseminate knowledge and tools, and understand their needs.
• The network should also foster innovation and include mechanisms to exploit new ideas coming out of the network’s work (for instance via incubators, start-up and spinoff initiatives and university tech-transfer funds).
• Overall, each proposal will define mechanisms to become a virtual centre of excellence, offering access to knowledge and serve as a reference in their chosen specific field, including activities to ensure visibility.

The proposals should:

• include mechanisms to spread the latest and most advanced knowledge to all the AI-labs in Europe
• develop synergies and cross-fertilization between industry and academia
• become a common resource and shared facility, as a virtual laboratory offering access to knowledge and expertise and attracting talents
• provide broad access to AI excellence in Europe and also play an important role in increasing visibility
• provide access to the required resources and infrastructure to support the R&D activities of the action, such as cloud and computing capacity, IoT, robotics equipment, support staff and engineers, where relevant, and the capacity to develop prototypes, pilots, demonstrators, etc.
• include a number of major scientific and application challenges which will mobilise the community to join forces in addressing them. Continuous evaluation and demonstration of scientific and technological progress (with qualitative and quantitative KPIs, benchmarking and progress monitoring processes) towards solving the targeted challenges will motivate the entire network and support publications and scientific career developments (providing reference benchmarks to publish comparative results, using the reference data, scenarios, etc.), and also showcase the technology in application contexts, to attract more user industries and eventually foster take up and adoption of the technology.
• include mechanisms to share resources, knowledge, tools, modules, software, results, expertise, and make equipment/infrastructure available to scientists to optimise the scientific and technological progress. To that end, proposals should exploit tools such as the AI-on-demand platform^[5] and further develop and expand the platform, to support the network and sharing of resource, results, tools among the scientific community, maximising re-use of results, and supporting faster progress. Mechanisms to test results and continuously measure and demonstrate progress should be integrated in the platform, which is also important to support the scientific community, allowing also for comparative analysis. Openness and interoperability of components are encouraged to develop synergies and cross-fertilization between different approaches and solutions (e.g. through modularity of components or open interfaces).
• include collaboration mechanisms among the best AI and AI-relevant research teams, but also mechanisms to bring all European AI teams to the highest level of excellence. This is also in view of supporting and encouraging the adoption of AI technologies in all Member States and Associated Countries, with particular emphasis on geographical aspect and elimination of gaps in digital competences and access between Member States and Associated Countries, as well as addressing existing gender disparities.
• exploit and develop technology enablers, such as methodologies, tools and systems and exploit latest hardware development and data spaces, cloud and HPC resources.

Activities are expected to achieve TRL 4-5 by the end of the project.

Proposals are expected to develop synergies:

• With other Networks of excellence centres in AI funded in H2020 or Horizon Europe, with a view of, all together, create vibrant European network of AI excellence centres. To that end, the activities should integrate with and complement the activities of the H2020-ICT-48 projects. The proposals are expected to dedicate tasks to ensure this coherence.
• With relevant activities in AI, Data and Robotics, primarily in destinations 3, 4 and 6, but also in other destinations and clusters (in particular with cluster 3 regarding security-related activities), and share or exploit results where appropriate.

All proposals are expected to allocate tasks to cohesion activities with the PPP on AI, Data and Robotics and funded actions related to this partnership, including the CSA HORIZON-CL4-2021-HUMAN-01-02.

Background

The network of excellence centres on AI safety and security will contribute to the larger objective of the European Commission to establish the European AI lighthouse.

The AI lighthouse is expected to mobilise the AI community to collaborate on key AI research challenges and to progress faster in joined efforts rather than working in silos, leading to fragmented and duplicated efforts. This is essential to reach critical mass and overcome the present fragmentation of AI research in Europe.

The lighthouse will bring together stakeholders from research, innovation and deployment, to become a world reference in AI that can attract investments and the best talents in the field. The lighthouse will build on key pillars, each of them being a network of excellence centres specialising in a given topic where Europe has the potential to become a global champion. The initiative started with the H2020-ICT-48 call establishing the first networks of excellence centres and will further develop in Horizon Europe.

Activities are expected to start at TRL 2-3 and achieve TRL 4-5 by the end of the project – see General Annex B.

Artificial Intelligence
Digital Agenda
Social sciences and humanities

[1]In this section AI is taken in the broad sense and covers AI, Data and Robotics

[2]https://ec.europa.eu/digital-single-market/en/news/communication-artificial-intelligence-europe

[3]https://ec.europa.eu/commission/presscorner/detail/en/IP_18_6689

[4]Maximum amount per third party, received from a given Action, over its entire duration

[5]Initiated under the AI4EU project https://cordis.europa.eu/project/id/825619 and further developed in projects resulting from H2020-ICT-49-2020 call