Skip to main content

A European AI On Demand Platform and Ecosystem

Periodic Reporting for period 1 - AI4EU (A European AI On Demand Platform and Ecosystem)

Reporting period: 2019-01-01 to 2020-06-30

AI4EU On-Demand Platform must play a crucial role by lowering the barriers to AI adoption. Europe's main weakness lay in the fragmentation of its ecosystem in a very competitive landscape. By offering a core AI node, the AI4EU On-Demand Platform can contribute to strengthen the community and bridge the gap between research and innovation. AI4EU proposal was built with the following strategic vision. We create an AI On-Demand Platform (WP2 and 3) that will be the AI Factory floor for the European AI Community (WP4 Ecosystem) by bringing excellent AI assets and resources (WP 3 enrichment of the platform, 6 industrial pilots, 7 technology gaps) and by promoting a trustworthy AI (WP5 ethics). We showcase the ability of the platform and the community to bring an ecosystem to solve AI challenges (WP6 industrial pilots and 8 technology transfer program and open calls) and offer a vision of what will be the future(s) for the platform (WP9 Strategic Research Innovation Agenda). At this time, the balance between the platform and the community is an important aspect. The focus is now more on the platform itself as the research community is more structured, both due to the European Commission action and the AI4EU Project. Expectations towards the platform are growing, with different visions on what it should deliver for the stakeholders. The competitive landscape is moving fast with massive investments in developing AI pipeline/machine learning platforms by the GAFAM or new players. Thus, the specificity of the project has to be highlighted. AI4EU is not a standard RIA project, building prototypes, or delivering publications. It’s a project aiming to provide a competitive platform, operational with excellent content and assets, in a moving context with a corresponding limited budget.
All WPs have achieved numerous significant results. The AI4EU platform is public since September 2019. Most of the promised features described in the project are now available, either at TRL 8 (news & events, Community, AI resources) or in a preliminary alpha version (Experiment). The expectations regarding the platform have evolved over time, and new requirements have appeared. Education and Gender issues, or vertical domains (industrial or by area) are, for example, requiring more developments to fulfil the collected requests. New needs, such as a full benchmarking environment or support for challenges, have been envisioned as a powerful incentive for research to adopt the platform.
Ecosystem development and platform notoriety are increasing. The community services are active, registered users (2650 at the end of August), and AI resources (end August 60 assets into the platform) are growing steadily. Registered organizations are reaching one hundred, with numerous groups and discussions. An editorial board has been created to supervise informational content publication. Web Cafés are occurring weekly Twitter, LinkedIn, and Facebook (a recent choice) are feed on an almost daily basis. We have also refined our event strategy for greater efficiency. AI4EU events have been a success.
Work toward platform sustainability has been started with the support of the Industrial board and an ongoing survey to define the AI4EU Platform business model. Work to create the AI4EU Foundation begun and the legal structure will be created by December 2020.
The Ethics observatory is now fully operational, with already more than ninety reference articles published. A dedicated Gender Committee has been created under the supervision of WAI (Women in AI). Regarding WP6, the vision developed for the pilots is that they are instrumental in demonstrating the various functions of the platform, and work is progressing in this way. In WP7, 18 new AI techniques have been developed for specific aspects of human-centred AI. All of them have been demonstrated on test-cases inspired by the AI4EU pilots. More than 40 assets have been published in the AI repository concerning these techniques. 52 scientific papers were published.
A new and more effective open calls strategy has been defined. The first Open Call will be launched in September. An initial vision of the Strategic Research & Innovation Agenda was published in June 2019. A second document was delivered in July 2020, targeting the platform's future.
Progress beyond the state of the art can be summarized in 5 main points.
1) AI4EU On-Demand Platform is existing, and the community now acknowledges the Platform. This was not obvious at the outset of the project. In 18 months, we have created the foundation of “a European AI Factory Floor” where the community will find people, resources, assets, AI experiment functionality. And other features in the future, with the support of the associated projects (ICT 48, 49,…). Multiple limitations can still be pointed out but the consortium is entirely focused on delivering the promised targeted TRL8 platform. It is also very difficult to evaluate this objective globally at M18. For instance, regarding AI4EU experiments, it is worth mentioning that big players (Google, Amazon, H2O.ai...) spent more than four years to reach TRL8. An alpha version of the Acumos subsystem is already functional in the platform. We are now working on an enhanced end-user experience (Single Sign-on) and to populate this environment.
2) We believe the trustworthy AI approach is an asset. The richness of the platform is to connect different worlds, such as ethics, industry, and research. It’s what we have started with WP5,6,7,8 activities and cross-collaboration. By setting-up eight pilots, developing a new open call model, we aim to showcase the ability of the platform to bring an ecosystem and the AI4EU platform functionalities to develop AI solutions.
3) Significant progress has also been achieved for the community. By creating dedicated boards addressing specific themes such as education, gender... we have laid the foundation for lively groups working on strategic issues. The creation of an Architectural Board and a High-Level Advisory board, aiming to gather ICT 48 and 49 winning projects views about the platform development is a first of its kind.
4) We have engaged the next steps for the AI4EU On-Demand Platform by building a sustainability plan, embedded in a set of principles that can make the platform unique. We believe the platform is a common digital good and should be owned by the Community and not by vested interests. A critical stake is to make this vision alive by creating the AI4EU Foundation and/or association, opening the project developments, and finding a compelling business model that can sustain our vision.
5) AI4EU On-Demand Platform is today a two-sided platform. The first dimension is the community and social part. The second is the AI development side, with AI resources and experiments. We are convinced that both are needed and provide originality to the AI4EU On-Demand Platform. With a third dimension, we will strive to build: the openness, trustworthy AI, and the common digital goods the platform should embody
We are at a tipping point. We will reorganize the project to dedicate more resources to the development of the platform, to achieve the features foreseen (AI4EU experiments as a top priority), to connect with other projects, and enhance the user experience. Simultaneously we have to promote the platform widely to the community and the end-users to foster its adoption.
consortium picture