Skip to yearly menu bar Skip to main content


Paris Invited Talk Dec 10, 12:00 AM - 1:00 AM AEDT Grand Amphithéâtre

Why informative explanations only exist for simple functions and what this means for the European AI Act

Ulrike Luxburg
Ulrike Luxburg is a full professor of Machine Learning Theory at the University of Tübingen, Germany, where she also founded and coordinates the research cluster "Machine Learning for Science" and the "Institute for AI and Law". She has served as program chair of NeurIPS and COLT, and together with her students, has received seven best student paper awards at NeurIPS, COLT, ALT, IJCAI. In 2024 she received Germany's national award for science communication. She is a member of the German National Academy of Sciences.
Explainable machine learning is often proposed as a tool to increase the transparency of machine learning systems and to satisfy requirements that might come with the European AI Act. However, to satisfy legal requirements, explanations should not merely appear plausible, but actually be reliable. To investigate when this is the case we introduce a mathematical framework that measures whether explanations are informative. Surprisingly, we find that most of the widely used post-hoc explanations are only informative if the underlying decision function is already simple. This questions the use of post-hoc explanations for complex machine learning models in a legal context. We suggest a path forward by augmenting explanations with concrete, valid interpretations. This can make explanations both formally informative and intuitively useful.
View full details
Paris Invited Talk Dec 10, 7:00 PM - 8:00 PM AEDT Grand Amphithéâtre

Is Mathematics Solved?

Timothy Gowers
Timothy Gowers was born in 1963 into a family of musicians. He was a chorister at King's College Cambridge from 1973 to 1976, returning to Cambridge in 1982 where he was an undergraduate, PhD student and research fellow in mathematics at Trinity College. From 1991 to 1995 he was a lecturer at University College London, before returning to Cambridge, initially as a lecturer, then as Rouse Ball Professor of Mathematics, and then as a Royal Society Research Professor. In 2000 he became the holder of the chair in combinatorics at the Collège de France, He retains a part-time position at Cambridge. He has received several awards, including a Fields Medal in 1998. He is the author of Mathematics: A Very Short Introduction and the main editor of The Princeton Companion to Mathematics.
Now that AI has solved a large number of major and long-standing open problems in mathematics, many people have been tempted to say that "mathematics is solved". I shall argue that there is an important sense in which mathematics is far from solved, even if we end up with a black box that is quicker than humans at solving any mathematical problem. I shall also describe some of the work of an automatic theorem proving group I run in Cambridge, and explain what I see as the potential benefits of continuing that work.
View full details
Paris Invited Talk Dec 11, 12:00 AM - 1:00 AM AEDT Grand Amphithéâtre

Claire Monteleoni

Claire Monteleoni
Claire Monteleoni is a senior research scientist at INRIA Paris where she leads the AI Research for Climate Change and Environmental Sustainability (ARCHES) team, and holds a Choose France Chair in AI. She is also a computer science professor at the University of Colorado Boulder (on leave). Her research on machine learning for the study of climate change helped launch the interdisciplinary field of Climate Informatics. She gave an invited tutorial: Climate Change: Challenges for Machine Learning, at NeurIPS 2014. She co-founded the Environmental Data Science journal, with Cambridge University Press, and the International Conference on Climate Informatics, which held its 15th annual event in 2026.
View full details
Paris Invited Talk Dec 11, 7:00 PM - 8:00 PM AEDT Grand Amphithéâtre

What Abundance Hides: Lessons from Building AI Under Constraint

Pelonomi Moiloa
Pelonomi Moiloa is the CEO and founder of Lelapa AI, a company building the world’s most efficient language infrastructure. With a background in electrical and biomedical engineering, she leads Lelapa’s mission to unlock the digital economies of the Global South, starting in Africa, by developing scalable, locally optimised language AI for constrained environments and pioneering more efficient modelling techniques for the world at large. She has been recognised as one of TIME’s 100 Most Influential People in AI, a Bloomberg Catalyst, a Mozilla Rise 25 Award recipient as well as Forbes Africa 30 under 50. She is also a TED speaker, known for her insights on SLMs and human centric AI. Before founding Lelapa AI, Pelonomi worked at the intersection of finance, data science, and machine learning, applying technology to large-scale systems challenges. She also serves as a trustee of a girls’ scholarship fund and a director of a community-based nonprofit. https://womenaibuilders.org/top-100-women-in-ai
Modern Artificial intelligence has progressed under extraordinary abundance: substantial capital, energy and compute, massive model architectures and datasets. These conditions have shaped the systems we have built but have also defined how we measure success. Impressive models that are meant to generalise fail in low resource environments, so what is there to do about it? At Lelapa AI, we develop speech recognition technology for multilingual environments where people routinely mix languages within a single conversation, training data is scarce and compute is constrained. This work exposes two important questions: how well do conventional metrics such as word error rate reflect human judgements of transcription quality? And when can smaller, more efficient models, trained on the right data, outperform larger general-purpose systems on real-world tasks? These questions challenge assumptions about both how we evaluate AI and what it takes to build capable systems. Constraints are not simply obstacles to overcome; they can reveal weaknesses in established methods and open new directions for model development. As AI cannot establish on its own what counts as good. As capabilities and access to compute continue to evolve, the lessons from building under constraint remain valuable: they help us distinguish genuine progress from progress measured by imperfect proxies, and identify what it will truly take to scale.
View full details

No Events Found

Try adjusting your search terms