The CNRS taking up the gauntlet of AI's scientific challenges

You're taking part in the AI Action Summit at the Élysée Palace in Paris from February 6^th to 11^th. What kinds of discussions or collaboration initiatives do you hope to see emerge from the exchanges between the scientists, industrialists and decision-makers attending?

Jalal Fadili : From February 6^th onwards, Paris will be attracting worldwide attention by hosting many events (see box) aimed at bolstering international action in favour of an AI that serves the common good. I'll be taking part in the AI, Science and Society conference on February 6^th and 7^th at the Institut Polytechnique de Paris as a member of the scientific committee and moderator for the session on "The mathematics of machine learning", and also the International Summit scheduled for February 10^th at the Grand Palais. The more scientific events mainly target researchers and aim to showcase the transformations AI has brought about in science and society whereas the February 10^th Summit will be a forum bringing together major global AI stakeholders, including governments, businesses, scientists or members of civil society, thus promoting the cross-fertilisation of ideas. It will involve conferences, round tables and presentations.

The CNRS is co-organising the AI, Science and Society conference at the Institut Polytechnique de Parisbringing together the world's best researchers (Berkeley, MIT, Stanford, Google DeepMind, etc.) in the presence of Clara Chappaz, Minister Delegate for Artificial Intelligence and the Digital Economy, and Philippe Baptiste, Minister for Higher Education and Research. Why was it important to organise these days ahead of the summit? 

J. F.: First, I'd like to highlight the collective work carried out by the scientific committee which I'm a member of on behalf of the CNRS and by the local organising team at the Institut Polytechnique de Paris who made it possible to set up such a large-scale event with a prestigious programme in record time. 

AI is undeniably bringing about a lot of transformations in science and in our societies. So it was essential to organise this kind of event to play our role as scientists and provide essential interdisciplinary expertise to prepare for the high-level discussions between world leaders over the following week in Paris. Society needs science and the contribution of scientific knowledge to respond to the major challenges of today and the future. This is the case in all the key decisive areas today and in the future. Of course AI is one of these fields.

Which subjects will be covered during the discussions?

J. F. This interdisciplinary conference will be an opportunity to understand and explore how AI can contribute to other sciences across a broad scientific spectrum. We hope to promote and facilitate dialogue between experts in AI, natural sciences and social sciences so this conference can enhance our in-depth understanding of the current state of the art in AI and help shape a form of AI that is aligned with humanity's values and priorities by catalysing the input from all disciplines. Researchers from world-renowned public and private organisations have been invited to share their ideas, present their cutting-edge research and innovations, and respond to pressing issues at the interface of AI and its applications in diverse fields like biology, health, physics, mathematics, economics, trust, sustainability, and ethics. We can all work together to create an AI that is both innovative and responsible.

You're the director of the 'AI for Science and Science for AI' Centre (AISSAI) which aims to structure and organise transdisciplinary work at the interfaces with AI. What is your view on the CNRS's position on AI? 
J. F.: The CNRS is at the forefront of this discipline, co-steering or taking part in several national programmes including the AI PEPR  and the Choose France CNRS AI Rising Talents programme . The CNRS runs the Jean Zay supercomputer dedicated to AI which is one of the most powerful in Europe. AI methods are now being rapidly rolled out to major research infrastructures and national and international experimental platforms within the CNRS. This all covers a broad research spectrum - from physics to biology, materials, ecology, climate and so forth. The CNRS knew that our intrinsically multidisciplinary dimension meant we could play a leading role in combining AI with all disciplines so the organisation set up the AISSAI Centre in 2021. This centre implements a range of scientific policies and actions aimed at enabling the CNRS to play a leading role in the interdisciplinary scientific challenges of the future which AI will play a major role in. These interdisciplinary exchanges are crucial as the 2024 Nobel Prizes in Physics and Chemistry  spectacularly illustrated.

The CNRS defined AI as one of its six societal challenges set out in its previous 'Objectives and Performance Contract' (COP) signed with the French State. The 'Generative AI for Science' challenge is also integrated into the CNRS's upcoming 'Objectives, Resources and Performance Contract' (COMP 2024-2028)

J. F.: Yes, AI was already in our strategy as expressed in 2019 as a challenge for science and also for society. As you said, AI is once again present in the CNRS COMP 2024-2028 with the 'Generative AI for science' challenge. A new stage has been reached with generative AI and more generally with foundation models capable of creating multimodal content like text, images, video, 3D models or sequences from a set of examples. This paradigm can help support the process of ideation - the generation of new scientific ideas or hypotheses. Generative AI means researchers can take new approaches and explore novel scenarios and drive and accelerate scientific creativity. The latter could include exploring new materials, structures, components (molecules, drugs, etc.) or models (physics, mathematics, etc.) or creating scenarios in the humanities and social sciences rather than just processing existing data. There are still a lot of challenges when models like these are applied to other disciplines which is why there is a dedicated generative AI session at the AI Action Summit conference to underline the importance of this theme for the future of research.