Distinguished Lecture Series 2025

Abstract: For over four decades, statistical physics has studied exactly solvable models of artificial neural networks. In this talk, we will explore how these models offer insights into deep learning and large language models. Specifically, we will examine a research strategy that trades distributional assumptions about data for precise control over learning behavior in high-dimensional settings. We will discuss several types of phase transitions that emerge in this limit, particularly as a function of data quantity. In particular, we will highlight how discontinuous phase transitions are linked to algorithmic hardness, impacting the behavior of gradient-based learning algorithms. Finally, we will review recent progress in learning from sequences and advances in understanding generalization in modern architectures, including the role of dot-product attention layers in transformers.

Bio: Lenka Zdeborová is a Professor of Physics and of Computer Science in École Polytechnique Fédérale de Lausanne where she leads the Statistical Physics of Computation Laboratory. She received a PhD in physics from University Paris-Sud and from Charles University in Prague in 2008. She spent two years in the Los Alamos National Laboratory as the Director’s Postdoctoral Fellow. Between 2010 and 2020 she was a researcher at CNRS working in the Institute of Theoretical Physics in CEA Saclay, France. In 2014, she was awarded the CNRS bronze medal, in 2016 Philippe Meyer prize in theoretical physics and an ERC Starting Grant, in 2018 the Irène Joliot-Curie prize, in 2021 the Gibbs lectureship of AMS. She is an editorial board member for Journal of Physics A, Physical Review E, Physical Review X, SIMODS, Machine Learning: Science and Technology, and Information and Inference. Lenka’s expertise is in applications of concepts from statistical physics, such as advanced mean field methods, replica method and related message-passing algorithms, to problems in machine learning, signal processing, inference and optimization. She enjoys erasing the boundaries between theoretical physics, mathematics and computer science.

Abstract: TBA

Bio: Irit Dinur is an Israeli computer scientist. She is professor of computer science at the Weizmann Institute of Science. Irit’s research is in Foundations of Computer Science and in Combinatorics, especially Probabilistically Checkable Proofs, hardness of approximation, and most recently high dimensional expanders.

Abstract: In the last few years, large pretrained models have shown impressive performance in a diverse set of tasks. These models must be trained with large datasets and are, in most cases, opaque on how they process information internally. In this talk I will focus on tools to understand the inner workings of existing pretrained models. Our current line of research aims to build tools that help users understand models, while combining the flexibility of human experimentation with the scalability of automated techniques. We introduce the Multimodal Automated Interpretability Agent (MAIA), which designs experiments to answer user queries about components of AI systems. MAIA iteratively generates hypotheses, runs experiments that test these hypotheses, observes experimental outcomes, and updates hypotheses until it can answer the user query. MAIA equips a pre-trained vision-language model with a set of tools that support iterative experimentation on subcomponents of other models to explain their behavior. These include tools commonly used by human interpretability researchers: for synthesizing and editing inputs, computing maximally activating exemplars from real-world datasets, and summarizing and describing experimental results. Interpretability experiments proposed by MAIA compose these tools to describe and explain system behavior. To conclude, I will talk about the role of data to train large vision models and ask if we can do away with real image datasets entirely when building a computer vision system, instead learning from noise processes.

Bio: Antonio Torralba is the Delta electronics Professor and head of the AI+D faculty at the Department of EECS at MIT. He received the 2010 J. K. Aggarwal Prize, the 2020 PAMI Mark Everingham Prize, the Inaugural Thomas Huang Memorial Prize by the PAMITC in 2021. In 2022, he was named Honoris Causa doctor by the Universitat Politècnica de Catalunya - BarcelonaTech (UPC). He is a AAAI fellow.I am a professor of computer science at the Weizmann Institute of Science.

Abstract: TBA

Bio: Toniann Pitassi was the Bell Canada Chair in Information Systems, in the Department of Computer Science at the University of Toronto, as well as a faculty member of the Vector Institute for AI, and on the research leadership team at the Swartz-Reismann Institute for Technology and Society. She has recently joined Columbia University where her primary research interests are complexity theory, fairness and privacy in machine learning. She currently holds a 6 year visiting professorship at the Institute of Advanced Study in mathematics and is also a recipient of the 2021 EATCS research award.

Abstract: TBA

Bio: Richard S. Zemel is a Canadian-American computer scientist and professor at Columbia University, Department of Computer Science, and a leading figure in the field of Machine Learning and Computer Vision.. He is broadly interested in machine learning, artificial intelligence, statistics, neuroscience, and cognitive science. He is also the Director of the new NSF AI Institute for Artificial and Natural Intelligence.

Abstract: TBA

Bio: Eran Segal is a computational biologist professor at the Weizmann Institute of Science. He leads a multi-disciplinary team of computational biologists and experimental scientists working in the area of Computational and Systems biology. Focuses on Nutrition, Genetics, Microbiome, and Gene Regulation and their effect on health and disease. His lab aims to develop personalized nutrition and personalized medicine using machine learning, computational biology, probabilistic modeling, and analysis of heterogeneous high-throughput genomic and clinical data.