ERCIM researchers and EC discuss GenAI: tools or threat?

Last week, ERCIM – the European Research Consortium for Informatics and Mathematics – organized its first visionary event in Brussels for EU policy makers and researchers. The topic was ‘Challenges and opportunities of Foundational Models and Generative AI for science and society’. Foundational Models are, for instance, Large Language Models (LLMs) as used by ChatGPT, Gemini and other systems. With these discussions, a vision document will be formulated that could help European research institutes and the European Commission (EC) with determining research directions. CWI and Inria were co-sponsor of the day.

Broad Visions

Han La Poutré (CWI), was, as Vice President of ERCIM, the main organizer of this day, together with Marie-Hélène Pautrat (Inria) and Joost Geurts (Inria). They say: “ERCIM combines the expertise of 16 renown institutes in computer science and mathematics. Together we have knowledge and strength to envision future scientific challenges in a broad way. This does not just mean scientific roadmaps, but also broader visions about impact and challenges.”

Tools or threats

The organizers add: “Generative AI (GenAI) is in a revolutionary phase of development. This has and will have impact on society as well as science. On the one hand, this impact can be positive, like GenAI being an enabler, on the other hand, it can be negative, like GenAI being a threat to society, or science. This leads to various new research challenges, not only in the research field of GenAI itself, but also in other fields of computer science, exact science, social sciences, and arts: to use GenAI in a positive way or to restrict and control generative AI in order to avoid negative effects.”

“Therefore we invited several prominent European researchers and EU policy makers to discuss this. Among them were Cecile Huet, head of unit AI and Robotics at the DG Connect, Liviu Stirbat, head of the unit AI and Science at the DG Research and Innovation of the European Commission, and Sean Ó hÉigeartaigh, director of the AI: Futures and Responsibility Programme at the University of Cambridge. Discussed topics varied from the impact of Generative AI and LLMs on society to fake realities and challenges for the future.”

La Poutré explains: “Regarding the risks of GenAI, we discussed topics like AI hallucinations and errors, the use of GenAI in propaganda and hate speech, deepfake videos, its impact on jobs and the environment, IP infringements and, for instance military use. On the other hand, GenAI can also be an incredibly powerful tool for scientific and economic progress (one of the speakers showed, for instance, the benefits for giant astronomical telescopes). Apart from the current regulations, there is an urgent need for more international cooperation in AI governance. LLMs should be managed not just as technical tools. Ethical, cultural, and social dimensions should also be encompassed”.

Vision document

La Poutré: “Our main conclusions of this day were that Generative AI and LLMs have a large impact but also generate many challenges for science and society to cope with. After this day, we will write a vision document, not only targeted at the management and researchers of ERCIM institutes and other academia but also at the EC, which can be helpful for future directions in research. This will be available online".

Speakers

The speakers of the ERCIM Visionary Event - “Challenges and Opportunities of Foundational Models and Generative AI for Science and Society” - were:

• Benoît Sagot, Inria
• Cecile Huet, head of unit AI and Robotics, DG Connect, European Commission
• Liviu Stirbat, head of unit AI and Science, DG Research and Innovation, European Commission
• Damien Gratadour, Centre national de la recherche scientifique, Observatoire de Paris
• Dong Nguyen, Utrecht University
• Mathieu Acher, INSA/Inria/CNRS/IRISA (DiverSE team)
• Seán Ó hÉigeartaigh, Centre for the Future of Intelligence, University of Cambridge
• Haris Papageorgiou, Institute for Language and Speech Processing, ILSP/R.C. Athena
• Anastasios (Tassos) Roussos, Institute of Computer Science (ICS), Hellas (FORTH)
• Abdallah El Ali, Centrum Wiskunde & Informatica (CWI)

The event was held in the Maison Irène et Frédéric Joliot Curie in Brussels on on 16 April 2024.

Header picture: Peter Kunz (ERCIM).