Sept. 12, 2024 — The Max Planck-Humboldt Research Award, endowed with 1.5 million euros, goes to Geordie Williamson, professor at the University of Sydney, for his use of artificial intelligence for fundamental work in mathematics.

Meanwhile, Max Planck Humboldt Medals go to Laura Waller, professor at the University of California, Berkeley, for her work in computational microscopy, and to Torsten Hoefler, professor at ETH Zurich, for the introduction of AI in high-performance computing, for example in climate research. The awards will be presented on 3 December in Berlin.

Scientists today use artificial intelligence in many areas, especially in the natural sciences, for tasks such as analysing data or images. In theoretical mathematics, on the other hand, AI has barely been used thus far. Now Williamson is aiming to change that. In his previous work he has already used artificial neural networks, which can guide mathematical intuition by drawing attention to previously unrecognised relationships in a large number of mathematical objects. Artificial intelligence can also help to generate examples or counterexamples that prove or disprove mathematical assumptions.

Although artificial neural networks can recognise patterns in large data sets very efficiently and effectively, they know nothing about mathematics. It therefore remains the task of mathematicians to filter out the sensible proposals from AI, to interpret them and, in the case of new assumptions about mathematical relationships, to prove or disprove them. Williamson wants to optimise the possibilities of using AI in theoretical mathematics in the collaboration made possible by the Max Planck-Humboldt Research Award. To this end, he will work closely with researchers from the University of Bonn and the Max Planck Institute for Mathematics in Bonn, where he will also spend two periods of several months each.

Williamson’s previous research work was characterised, among other things, by the fact that he brought together different fields such as combinatorics and geometry. In simple terms, combinatorics can be understood as the branch of mathematics that is dedicated to everything that can be counted; it includes subjects such as graph theory and discrete mathematics. Geometry is about objects in spaces, i.e. straight lines, surfaces, and solids, just like in school maths. Both sub-areas come together in a simple example when the intersection points of a curve and a surface are to be counted. Geordie Williamson has now opened up ways of solving combinatorics problems with geometric tools, for which purpose he first had to develop a kind of common mathematical language for the two fields so that combinatorial problems could be worked on in geometry, but geometry could also be translated into combinatorics. With this approach, Geordie Williamson has proved or disproved various assumptions that mathematicians have been working on intensively, but to no avail, for a long time.

For example, Williamson in collaboration with Ben Elias from the University of Oregon provided a general proof of an important conjecture in mathematics relating to Kazhdan-Lusztig polynomials. The work of David Kazhdan and George Lusztig provided precise recipes for building up certain mathematical objects out of constituent pieces. Imagine a recipe that contains a list of ingredients and instructions on what to do with them, but the recipe does not specify the quantities. Kazhdan and Lusztig hypothesised that there are polynomials in mathematics for such cases, from which the quantities for the recipe can be determined. Polynomials are formulae that are familiar to us in their simple form from the binomial formulae we study in school. Geordie Williamson has proven this assumption, for which evidence had previously been sought in vain for a long time. His methods, borrowed from geometry, also make it much easier to solve the polynomials that provide the unknown data and to analyse them in greater depth.

**Laura Waller – Pioneering Computational Microscopy**

Waller uses algorithms – some of which are based on machine learning – to improve microscopy, particularly of biological samples, as well as the imaging of astronomical objects. This pioneer of computational microscopy is combining computer science and simple instruments to achieve such things as making more details visible and creating three-dimensional images or videos. Among other things, Laura Waller has further developed the phase contrast microscope, which can also image transparent objects. She has formulated algorithms that determine quantitative information about the phase of light waves – in simple terms, this is the displacement of light waves relative to each other– from a few images with illumination from different angles. The resulting images not only better visualise the shape of cells, but also allow better cell tracking. In another invention, the DiffuserCam, Waller places an uneven plastic plate on a light sensor, which scatters the incoming light. Very detailed 3D images can then be reconstructed from a single sensor reading, with applications in microscopy and astronomical imaging. The technology also makes it possible to create high-speed videos with low-speed camera equipment.

**Torsten Hoefler Makes HPC and AI More Efficient**

Hoefler’s computer science work concentrates on increasing the efficiency of algorithms, particularly for applications in high-performance computing and artificial intelligence. Hoefler’s approaches have led to substantial advancements in various fields. For example, his team has found ways to considerably speed up very complex computational problems such as quantum simulations, which are important to the semiconductor industry. He has also developed methods that optimise machine learning algorithms and significantly improve their practical applicability. A particularly remarkable breakthrough for Hoefler and his team was the processing of large amounts of data for climate simulations. Using neural networks, the researchers have compressed this data to a thousandth of its original volume without sacrificing fidelity. By skilfully combining and optimising hardware, software, and algorithms, Hoefler has increased the efficiency of computer systems by a factor of up to one thousand. His work makes a significant contribution to the further development of artificial intelligence and opens up new areas of application in computer science.

The Max Planck Society and the Alexander von Humboldt Foundation present the Max Planck-Humboldt Research Award, along with 1.5 million euros in prize money, to a researcher from abroad. 80,000 euros in personal prize money is also awarded.

The focus here is on personalities whose work is characterised by outstanding potential for the future. The prize is intended to attract particularly innovative scientists working abroad to spend a fixed period of time at a German higher education institution or research facility. The Federal Ministry of Education and Research provides the funding for the award.

The focus of the award alternates each year between natural and engineering sciences, life sciences, humanities and social sciences. In addition, one or two further individuals may be nominated and awarded the Max Planck-Humboldt Medal. This is awarded along with 60,000 euros in prize money.