The Institut Henri Poincaré is organizing an AI conference on November 16 and 17, 2021, an event around the challenges of artificial intelligence research and the diversity of fields in which it is used. On the program: presentations, exchanges and scientific conversations, from environmental sciences to health, including the next presidential election and justice. Dominique Mouhanna, Deputy Director of the Institut Henri Poincaré, was kind enough to answer our questions.

Don’t miss the AI conference of the Institut Henri Poincaré, in person and online.
Registration on the event website.

Can you present us the content of the two days dedicated to AI organized by the Institut Henri Poincaré and its partners on November 16 and 17?

There will be four half-days. The first three will be devoted to the presentation of fundamental concepts and the current state of the field, and will be aimed at researchers. This will be done from two points of view: the first will focus on the applications of AI, which are now countless, trying to cover a fairly broad field, but obviously not exhaustive over two days. The second point of view will be more a reflection on the intrusion of AI in society and the consequences that follow. It will be a question of taking a more external look at the subject, by trying to tackle various questions, whether ethical, ecological or even philosophical.

The fourth half-day, during which mainly students in thesis or post-doc will intervene, is rather intended for university or even high school students, and will aim at showing in which paths young people who turn to AI are engaged: what are their research subjects, their trainings, their paths. Finally, it will show that AI is a real junction point between fundamental research and the business world. 15 or 20 years ago, the academic world and the business world were still very compartmentalized; today there is an increasingly rapid application of concepts emanating from fundamental research to the business sector. And AI is really one of the key examples. The main lines of these days are: to recall the fundamental principles of AI, to discuss its repercussions on society, and finally to show the variety of jobs and career paths possible in this field.

These conferences will allow us to take stock of the current state of the field. What will be the focus and do you think that the pandemic has changed the AI sector in terms of priority application areas?

You refer to the pandemic. What needs to be stressed is that, globally, medical or bio-medical research is extremely “consuming” AI. This ranges from the identification of very diverse pathologies, whether in the field of traumatology, cardiology, phlebology, but also in cancerology or neurology. AI is used in many medical fields as a tool for analysis and diagnosis, but also as an aid in planning suitable medical protocols.

These fields will be very well represented during these two days. But speakers will also report on advances in the field of robotics, a very dynamic field, in the field of energy, whether it is around the issue of optimal routing of energy, for example, but also around the issue of energy sobriety in our societies. Moreover, we now know very well that AI and modern technologies are themselves extremely energy-intensive. The question of the energy sobriety of AI itself is therefore also raised…

There is also, of course, the field of climatology which, because of its complexity, makes massive use of computers and AI. But more generally, we find it at all levels of society: justice – where AI is used in particular in the context of legal proceedings, which has developed a lot recently in the United States – but also the police, the military sector, online commerce, etc. Moreover, as I said, societal and ethical issues will be raised, such as gender bias or data confidentiality…

So the range of areas covered is very wide. I would also add the field of hard sciences (physics, mathematics) in which more and more work is being done using this tool. However, in mathematics and theoretical physics in particular, it is still a very open question because we are touching on very specific areas where the creative process can intervene very singularly. Where AI works very well is above all in fields where there is a problem of massive data management; medicine or biology are very good examples.

You asked me about the effect of the pandemic. AI has been very interested in COVID 19 both in a “diagnostic” framework, which is more in the realm of infectious diseases, but also in an “epidemiological” framework. But I would like to stress that epidemiology is a field that has long been covered by mathematics, physics – particularly statistics – and computer science. So there has certainly been a focus of AI on health-related fields, but in fact it was already very present.

The IHP is actively collaborating with companies such as Criteo, Aquila data, RTE and even L’Oréal in the context of this conference. How do these partnerships work, what are the major projects and the current dynamics of the IHP with its private and public partners?

First of all, I would like to point out that it is the IHP Endowment Fund, which is celebrating its 5th anniversary in 2021, that is doing a remarkable job in supporting the extension of the IHP and the creation of its Maison Poincaré, by raising funds from companies and individuals. Thanks to this support, many projects have been carried out: exhibitions, films, etc. I would like to mention in particular the “Holo-Math” mixed reality experience, a 3D scientific mediation programme, which is directly linked to advanced technology companies.

The Maison Poincaré, a project to open up to scientific communication and mediation, led by project manager Marion Liewig since 2016 and supervised by the director of the IHP, Sylvie Benzoni, is currently the institute’s major project, in which all its teams are actively involved. This project is supported by public funding, via the CPER (State-Region Plan Contracts) which involves CNRS, the State, the Ile-de-France Region, the City of Paris, and via the institute’s supervisors, CNRS and Sorbonne University.

This conference is a very good example of how we wish to work with the partners of the Endowment Fund: an association in the reflection, within the scientific committees composed of actors and actresses of mediation, communication, research, and in the organization, as in our collaboration with Antonin Braun (Aquila Data Enabler), Julien Mairal (INRIA) and Liva Ralaivola (Criteo AI Lab) We have led this conference with common intentions: to specify the state of current AI research, the excessively broad scope of its applications as well as the perspectives of career paths for the young generation. But, more broadly, there will be, within the framework of the Maison Poincaré, many events around the mediation of mathematics and we count a lot on the business and academic contributors to develop these activities.

This event is aimed at scientists and experts in the field, but also at a wider public, especially students. The number of AI training courses has increased considerably in recent years, what advice would you give to students who wish to embark on a career in AI? What are the main skills to master and the opportunities that may be available to them?

From this point of view, there has been a major evolution in recent years. AI has changed its status: previously it was essentially a research field, very specialized, which required a very solid training in mathematics and computer science. The entry of AI into many areas of society has been accompanied by a diversification of training courses and also of their objectives. Initially, masters’ degrees in research training were the only way to go, and required very high technical skills.

Today there are, in parallel to these advanced courses, integrated courses in which the objective is different, rather turned towards the applications and use of AI. It is important to know that today you can find turnkey AI programs online, and you don’t need to be a high-level computer scientist to use them. The consequence of this evolution is that it is now necessary to have both “minimal” knowledge as an AI user, but also knowledge inherent to the field in which you wish to work: medical, banking, commercial, financial… If you are in finance, you need, for example, to know what a stock portfolio is, what optimization is, etc. But we can also expect partnerships to be created between professionals in a given field and technicians more specifically trained in AI, just as in the medical field, a radiologist is often supported by an X-ray specialist with knowledge of radiation dosimetry.

What is the research project/application/startup that has impressed you the most lately?

There is a bias here related to my own practice, theoretical physics. As I said above, a lot of very spectacular things are happening in the field of medical diagnostics, biology, climatology, I would say the sciences of the “complex”. Here, we find many quantitative problems, i.e. problems related to the management of a gigantic mass of data. Just think, for example, of the sequencing of the human genome, which involves some 3 billion base pairs.

A major question for me is the possibility of making qualitative progress, i.e. of a creative nature: in concrete terms, the possibility of inferring new physical laws or original mathematical theorems. Recently, in the field of physics or mathematics, we have seen proposals of this type circulating. I am attentive to this because we would then have moved from the era of learning to that of “learning about learning”, at a higher level. And this would bring us back to questions that have already been asked, but perhaps from a new perspective, in the context of neuroscience and relating to the functioning of the human being himself: consciousness, free will, creativity. It is this feedback effect that interests me: how technology can refer to the functioning of the human being himself.

Translated from Recherche en IA, formations, domaines d’application : rencontre avec Dominique Mouhanna de l’Institut Henri Poincaré