Periodic Reporting for period 2 - AI4EUROPE (AN AI ON-DEMAND PLATFORM TO SUPPORT RESEARCH EXCELLENCE IN EUROPE)
Reporting period: 2023-07-01 to 2024-06-30
OBJECTIVES
AI4Europe's main objective is to support & facilitate a sustainable digital platform and experimentation environment through the creation of open research channels and mechanisms that foster the European AI research ecosystem, academic and industrial, and that maximises the academic, social, and industrial impact while it seamlessly integrates other projects, platforms, and solutions.
The specific objectives are to:
1. Ensure platform support for the research community fostering a cooperative & transparent governance model open to contributions from all stakeholders to address their needs
2. Design and execute a sustainability plan that supports and ensures the long-term viability of technical, financial, & legal features of the AIOD platform
3. Integrate hardware resource provisioning into the platform to provide underlying computational resources, establishing it as a place to carry out and disseminate AI research
4. Advance the AIOD platform as an open distributed tool extendable by the community with new services & tools, and ensure interoperability with other strategic EU platforms
5. Enhance the accessibility & usability of the excellence science of Europe’s AI research community, and to ensure standard, secure, and executable reusable research assets
6. Embed mechanisms to facilitate the development & use of Trustworthy AI aligned with policy and regulatory developments in the EU
7. Foster the uptake of AI in research communities, industry, and society at large by mobilizing & enriching efforts in AI education, upskilling & reskilling
8. Enhance access, sharing, training & mobility for all European nations, ecosystems, & communities, with special attention on young AI scholars, facilitating cooperation & matchmaking
9. Ensure cooperation with the AI, Data and Robotics PPP & other related Digital Europe initiatives, to support them and to maximise the impact of the AIOD platform
IMPACTS
#1 - Scientific
• An AI platform to provide researchers with a comprehensive, sovereign, & sustainable infrastructure following European values
• Mitigate many IPR challenges that restrict the sharing of AI, datasets, & solutions
• An open governance model owned by the community, driven by their needs, based on contribution and fair principles
#2 - Economic/Technological
• An EU-wide platform for AI researchers to translate research into innovation
• By making online educational resources accessible, AI4Europe will promote European values to create a seal of quality
• A sustainable platform that ensures its technical, legal, & financial continuation after the finalisation of the project
• Easy-to-use set of services used by all researcher institutions on AI
• Ease to access heterogeneous hardware
#3 – Societal
• Increased inclusiveness, by supporting a human-centered approach to technology development, aligned with European social & ethical values, as well as sustainability
• Explicit & more transparent connection of the EU’s AI ecosystem with policy priorities
AI4Europe's main objective is to support & facilitate a sustainable digital platform and experimentation environment through the creation of open research channels and mechanisms that foster the European AI research ecosystem, academic and industrial, and that maximises the academic, social, and industrial impact while it seamlessly integrates other projects, platforms, and solutions.
The specific objectives are to:
1. Ensure platform support for the research community fostering a cooperative & transparent governance model open to contributions from all stakeholders to address their needs
2. Design and execute a sustainability plan that supports and ensures the long-term viability of technical, financial, & legal features of the AIOD platform
3. Integrate hardware resource provisioning into the platform to provide underlying computational resources, establishing it as a place to carry out and disseminate AI research
4. Advance the AIOD platform as an open distributed tool extendable by the community with new services & tools, and ensure interoperability with other strategic EU platforms
5. Enhance the accessibility & usability of the excellence science of Europe’s AI research community, and to ensure standard, secure, and executable reusable research assets
6. Embed mechanisms to facilitate the development & use of Trustworthy AI aligned with policy and regulatory developments in the EU
7. Foster the uptake of AI in research communities, industry, and society at large by mobilizing & enriching efforts in AI education, upskilling & reskilling
8. Enhance access, sharing, training & mobility for all European nations, ecosystems, & communities, with special attention on young AI scholars, facilitating cooperation & matchmaking
9. Ensure cooperation with the AI, Data and Robotics PPP & other related Digital Europe initiatives, to support them and to maximise the impact of the AIOD platform
IMPACTS
#1 - Scientific
• An AI platform to provide researchers with a comprehensive, sovereign, & sustainable infrastructure following European values
• Mitigate many IPR challenges that restrict the sharing of AI, datasets, & solutions
• An open governance model owned by the community, driven by their needs, based on contribution and fair principles
#2 - Economic/Technological
• An EU-wide platform for AI researchers to translate research into innovation
• By making online educational resources accessible, AI4Europe will promote European values to create a seal of quality
• A sustainable platform that ensures its technical, legal, & financial continuation after the finalisation of the project
• Easy-to-use set of services used by all researcher institutions on AI
• Ease to access heterogeneous hardware
#3 – Societal
• Increased inclusiveness, by supporting a human-centered approach to technology development, aligned with European social & ethical values, as well as sustainability
• Explicit & more transparent connection of the EU’s AI ecosystem with policy priorities
Key activities undertaken in the technical Work packages (3-4-5) from M1 to M24 are:
WP3
• Decentralized AI-on-demand core architecture design
• Design & implementation of the architectural foundations of the AIoD: the Metadata Catalogue and its connectors, identity management & authentication service
• Development of a Python-based interface to the Metadata Catalogue - Python SDK
• Implementation of MyLibrary, a solution that allows users to browse different types of assets of the Metadata Catalogue and add them to a personal collection of assets that will be available in other AIoD services
• Deployment & maintenance of services at ITI, UNIZAR and EGI as test nodes (production environment at UNIZAR): Metadata Catalogue API, Authentication & Authorization, MyLibrary, RAIL, Marketing Portal
• Refinement of the Metadata model & alignment with user requirements
WP4
• Establishment of development processes for integration of new tools
• Creation of a generative AI experimentation space for the European AI community
• Analysis of the existing reproducibility initiatives in different conferences and journals
• Implementation of two new Hybrid-AI use cases in AI-Builder
• Creation of an open database of Trustworthy Artificial Intelligence tools & services from EU research
• High-fidelity design of the Marketing Portal and subsequent improvements
• Development of a Style sheet for creators of services on AIoD
• Creation of the first version of Practitioner’s Portal (PP), later rebranded to RAIL
WP5
• Representation of physical resources as metadata to (a) allow the publication of physical resources in a catalogue & (b) to provide important information that needs to be considered to deploy datasets, ML experiments on such resources
• High-level definition of the Hardware (HW) agent, a software tool to gather the necessary information from a computational resource in order to add it to the AIoD catalogue
• Initial implementation of the HW Agent using Connector Architecture that enables multiple connectors for several orchestrators, integrating efforts from T5.1 T5.2 T5.3
WP3
• Decentralized AI-on-demand core architecture design
• Design & implementation of the architectural foundations of the AIoD: the Metadata Catalogue and its connectors, identity management & authentication service
• Development of a Python-based interface to the Metadata Catalogue - Python SDK
• Implementation of MyLibrary, a solution that allows users to browse different types of assets of the Metadata Catalogue and add them to a personal collection of assets that will be available in other AIoD services
• Deployment & maintenance of services at ITI, UNIZAR and EGI as test nodes (production environment at UNIZAR): Metadata Catalogue API, Authentication & Authorization, MyLibrary, RAIL, Marketing Portal
• Refinement of the Metadata model & alignment with user requirements
WP4
• Establishment of development processes for integration of new tools
• Creation of a generative AI experimentation space for the European AI community
• Analysis of the existing reproducibility initiatives in different conferences and journals
• Implementation of two new Hybrid-AI use cases in AI-Builder
• Creation of an open database of Trustworthy Artificial Intelligence tools & services from EU research
• High-fidelity design of the Marketing Portal and subsequent improvements
• Development of a Style sheet for creators of services on AIoD
• Creation of the first version of Practitioner’s Portal (PP), later rebranded to RAIL
WP5
• Representation of physical resources as metadata to (a) allow the publication of physical resources in a catalogue & (b) to provide important information that needs to be considered to deploy datasets, ML experiments on such resources
• High-level definition of the Hardware (HW) agent, a software tool to gather the necessary information from a computational resource in order to add it to the AIoD catalogue
• Initial implementation of the HW Agent using Connector Architecture that enables multiple connectors for several orchestrators, integrating efforts from T5.1 T5.2 T5.3
ALL expected results and the status of corresponding Deliverables & Milestones
1. Governance supported by stakeholders - MS2 achieved in RP1; D2.1 submitted in RP2
2. Platform self-sustainable through a legal entity that will ensure technical, legal, and financial continuation - D2.1 submitted in RP2
3. Integration of plugin-like methods to deploy on heterogeneous hardware (Edge, 5G, HPC, Cloud) - D5.1 & D5.2 submitted in RP2; MS6 achieved in RP2
4. A toolkit with specific tools on AI and a set of repositories with contributions from researchers - D3.3 & D7.1 delivered in RP1
5. Dynamic user interfaces easing engagement with content - D4.2 submitted in RP2; D2.2 due in RP3
6. Several repositories EU distributed beyond partner of consortium - D3.1 delivered in RP1
7. Strategy plan and engaged academic community and PhD community - D6.2 submitted in RP2
8. Catalogue of success stories on multiple market domains - D7.2 submitted in RP2
9. Tools that enable Trustworthy AI research by design and validation of AI tools and datasets - D4.2 submitted in RP2
1. Governance supported by stakeholders - MS2 achieved in RP1; D2.1 submitted in RP2
2. Platform self-sustainable through a legal entity that will ensure technical, legal, and financial continuation - D2.1 submitted in RP2
3. Integration of plugin-like methods to deploy on heterogeneous hardware (Edge, 5G, HPC, Cloud) - D5.1 & D5.2 submitted in RP2; MS6 achieved in RP2
4. A toolkit with specific tools on AI and a set of repositories with contributions from researchers - D3.3 & D7.1 delivered in RP1
5. Dynamic user interfaces easing engagement with content - D4.2 submitted in RP2; D2.2 due in RP3
6. Several repositories EU distributed beyond partner of consortium - D3.1 delivered in RP1
7. Strategy plan and engaged academic community and PhD community - D6.2 submitted in RP2
8. Catalogue of success stories on multiple market domains - D7.2 submitted in RP2
9. Tools that enable Trustworthy AI research by design and validation of AI tools and datasets - D4.2 submitted in RP2