The Assyr Abdulle Lecture is a public lecture given periodically by a visitor of the Bernoulli Center for Fundamental Studies.

Prof. Assyr Abdulle was a faculty member at the Mathematics Institute at EPFL from 2009 until 2021. He was a world-renowned expert in the field of numerical methods for stochastic differential equations and multiscale problems. He was a passionate researcher engaged in promoting mathematics at EPFL. He was in particular engaged in the development of the Youth at Bernoulli program. This public lecture celebrates his memory.

For the 2026 Assyr Abdulle Lecture, we will be happy to welcome Professor Weinan E  from Peking University.

The lecture will take place on Monday 28 September 2026 from 16:30 to 17:30 at the Rolex Learning Centre, and will be followed by an informal aperitif.

Title: Mathematics, Science and Artificial intelligence

Since Newton established modern science and mathematics, tremendous progress has been made on both fronts by formulating the first principles of science as mathematical objects, developing the mathematical tools needed, and using them to solve scientific problems. One consequence is that we can now label a large part of mechanics as “engineering“, where mathematical and computational tools are very effective.  However, many problems, such as drug discovery, the design of new materials or catalysts, still remain difficult and resist mathematical or computational treatment. The difference between these two kinds of problems is that the former, the ones that we can handle, are “easy”, and the latter, the ones that remain difficult, are “complex”.

This complexity comes in several aspects: high dimensionality, non-locality, large hidden space, etc. Classical mathematical tools, such as polynomials and differential equations, are ineffective in these situations.

Deep learning has brought us new hope. This is a new class of mathematical tools that seem to be able to handle complex problems in artificial intelligence. By extending these methods to science, we have entered the new era of “AI for Science”. This is indeed a paradigm change for science. Tremendous progress has already been made and much more is still to come.

Yet the irony is that AI itself is a singular subject that relies almost entirely on experience, on trial and error. It lacks the guidance from first principles. For this reason, it suffered severe ups and downs in the past, and it has become an extremely costly business nowadays. Why is this so? Is it really the case that the first principles of AI are just too difficult and out of reach at the present time?

In this talk, we will go through these topics with a focus on the issue of complexity: The complexity that we need to deal with in scientific research, AI as an effective tool to deal with complex problems, and the new mathematics that needs to be developed to handle complexity.

Registration details will be shared in due course. 

Stay tuned!