2024 EDITION

AI Factories: Addressing the data challenge

October 2, 2024 - 11:30 - 13:00

Room Gold

AI Factories (AIFs) are open ecosystems formed around European public supercomputers that bring together key material and human resources needed for the development of generative AI models and applications. AI Factories are created with the ultimate objective of leveraging EU supercomputers for the development, training, and deployment of large-scale AI models, to support activities for AI startups, science and innovation ecosystems, and other productive environments.

To achieve this objective, AI Factories must be able to access, acquire and reuse massive amounts of different types of high-quality data, spread out in the cloud and coming from different sources, repositories and data ecosystems. This implies dealing with different types of access, use and re-use terms and conditions, and specific regulations that may apply to different datasets and data ecosystems, particularly important for privately owned data and personal data.

In this regard, EU Data Spaces emerge as ideal instruments that can eventually contribute to cover the mentioned aspects (diversity of sources, data and conditions) but also to address other relevant challenges like data interoperability, security, governance and quality.

BDVA is a private member of the EuroHPC JU, representing the role of user industries and research on data and AI. BDVA is very active in the EU Data Space landscape, partner in the Data Spaces Support Centre, and collaborates closely with other key initiatives on the data spaces domain. BDVA and its members are very active on Industrial AI and Data-driven AI.

Objectives of the session

This session is framed as part of the process that BDVA is running to provide coherence, common knowledge and recommendations to accelerate access of the AI Factories to the EU Data Spaces. The session will also benefit from the discussions taken place to publish the white paper “Generative AI and Data Spaces” and the content therein, led by BDVA and produced in the scope of Data Spaces Support Center (DSSC).

The session provides a unique opportunity to gather policy makers, involved parties and experts on the field physically around the same table, to deep dive on the discussion about existing challenges for this connection, identify potential solutions and agree on next actions.

Based on the above, and the taking the BDVA discussion paper as a reference, the objectives of the session are:

To have the view from policy makers on how AI Factories will foster competitiveness EU AI ecosystems, and the need to connect to EU Data Spaces

To learn firsthand what barriers AI Factories might encounter for this connection, and what requirements are in place

To bring the knowledge and experience from data spaces, about the different concepts, roles, processes, etc. in data spaces that would be relevant for this connection

To discuss potential different scenarios for this connection (from a simple onboarding as data space participant to a fully AIF – DS connection?)

To learn how some data spaces are planning similar connections with other ecosystems

To learn from experts about key aspects as governance, trust, sovereignty, etc. across different ecosystems
To have the view of AI practitioners as potential users

According to the scope of the session and its objectives, it aims to bring the view from representatives from:

European Commission