Nome
Distributed Computing and Digital Infrastructures

Código
FLUID-AI

Entidade Beneficiária

LIP - Laboratório de Instrumentação e Física Experimental de Partículas

Sumário do Projecto

In the current landscape of scientific research, the ability to efficiently reuse, combine, and integrate data and
AI/ML models is a key enabler for advancing knowledge and innovation. The European Commission (EC) has
addressed this challenge by means of the European Open Science Cloud (EOSC), a strategic EC initiative that is
transforming research and innovation in Europe at a significant pace, leaning on the concept of Web of FAIR Data
and Services. The well-known FAIR Data Principles (Findable, Accessible, Interoperable, and Reusable) have been
instrumental in improving data management practices across disciplines, ensuring that data are well-documented,
accessible, and reusable without barriers and have now been applied to multiple research objects in various disciplines.
However, in spite of these substantial efforts, the EOSC and the EU Research Infrastructures (RIs) often face
significant challenges in achieving seamless interoperability and accessibility of their digital assets. While the FAIR
Principles have significantly enhanced data management, they are not yet sufficient in the context of Artificial
Intelligence (AI), both at the data and model levels. AI requires data that are not only FAIR but also AI-ready —i.e.,
data that are structured, annotated, and formatted in a way that facilitates Machine Learning (ML) or other AI
applications and help with AI explainability. AI-ready data should clearly build upon and extend the FAIR Principles
to ensure that data are of high quality, optimized for AI-driven research, and ready to be used without further
processing. Furthermore, and going beyond data, AI/ML models also suffer from technological interoperability and
compatibility challenges in real-world scenarios. This combination, boiled down to the lack of data AI-readiness and
the lack of interoperable solutions, hinders the widespread adoption of AI-based solutions and hampers the impact
of AI in science.
Therefore, in order to further advance the capabilities of the EOSC and related RIs, our project aims to introduce
the concept of Data and Models Liquidity. In this context, Data and Models Liquidity —in an analogy to liquidity
in finance, i.e., the fact for an asset of being quickly and easily available, converted into cash— refers to the ease
with which RIs and their users can utilize, reuse, combine, and integrate AI-related assets (i.e., data, models, and
platforms). High liquidity ensures that data and models are not only FAIR but also optimized for AI applications.
The data liquidity concept, originally defined (only for data) by MIT CSIR in 2021 (Wixom and Piccoli 2021) to
facilitate business data management and monetization, is extended by FLUID-AI to both data and AI/ML models as
a response to the EOSC ecosystem needs. In our project, liquidity is the central concept to overcoming the current
challenges in AI and data interoperability for science, fostering a dynamic and collaborative research environment
where researchers can build upon existing work and drive new discoveries.
In the context of this novel concept of Data and Models Liquidity, current and future research infrastructures
should further advance in several key areas:
• Unified Data and Models Integration: Many RIs operate in isolated environments, with data and models stored
in various registries, marketplaces, formats and systems that lack compatibility and interoperability. This frag-
mentation hinders the ability of researchers to integrate and reuse them, impeding the pace of scientific progress.
• Accessible and Intuitive Platforms: The exploitation of EOSC should not demand significant technical expertise,
posing barriers for researchers who may lack advanced computational skills. Complex technical details should
be hidden from end users and scientists. This technical complexity can discourage potential users, reducing the
overall utilization and impact of the research infrastructure.
• Collaborative Support and Training: There is a clear lack of coordinated support and training for researchers,
particularly in the use of AI/ML tools and data management practices in the EOSC context. Without adequate
support, researchers may struggle to effectively utilize available resources, limiting the potential benefits of the
infrastructure.

---

Suporte sob

Reforçar a investigação, o desenvolvimento tecnológico e a inovação

Região de Intervenção

...

 

Financiamento

---

---

 

Datas

---

---