NeurIPS 2026 Invited Speaker Topics
The NeurIPS 2026 organizing committee is excited to share details about the upcoming invited talks at this year’s conference. New this year, the conference is held in three locations, each with its own speakers. Our distinguished lineup includes six speakers for the main Sydney location, four speakers for Paris and three speakers for Atlanta.
This year’s lineup includes leading researchers who will explore questions at the frontiers of artificial intelligence. They range from the theory of machine learning and reinforcement learning to AI for science, mathematics and climate, and to the ethical, social and environmental questions that come with AI at scale. Though spread across three continents, the lineups share common topics. Each site has a slight accent of its own: science in Sydney, where three talks put machine learning to work in biology, astronomy and industry; mathematics in Paris, with the theory of learning and the future of theorem proving; and society in Atlanta, where two talks look at AI’s footprint on education, communities and the energy grid.
Below is a short preview of each lineup, site by site, in schedule order. Times are given in each site’s local time. Abstracts and full bios are on the invited speaker pages as they are published.
Sydney
The main site hosts six talks over three days, December 8th to 10th:
At 9:00am on Dec 8th, Tie-Yan Liu, founding President of Zhongguancun Academy in Beijing, will describe how to build an industrial learning loop arguing that AI’s next phase will be shaped by learning from real-world experience which in industry is embedded in proprietary workflows, physical experiments, and decisions with real consequences.
At 2:00pm on Dec 8th, Barbara Engelhardt, of Gladstone Institutes and Stanford University, will speak from her work on models for structured biomedical data.
At 9:00am on Dec 9th, Satinder Singh will elaborate on RL as the solution to alignment, making the case of huma-designed reward functions, and will will talk about autonomous agents that can learn to be broadly competent in complex, dynamic, and uncertain environments.
At 2:00pm on Dec 9th, Cheng Soon Ong, science director at CSIRO’s Technology Research Unit, will talk about scoring the performance of machine learning models: he will present the theory and practice of bandit-style adaptive experimental design for scientific discovery, validated in a genomics lab.
At 9:00am on Dec 10th, Shannon Vallor, Baillie Gifford Chair in the Ethics of Data and AI at the University of Edinburgh, will reflect on generative AI as an engine both of knowledge production and of epistemic dispossession and on the wide-scale removal of knowledge as a good created, enacted and held by human beings.
At 2:00pm on Dec 10th, José Hernández-Orallo, Director of Research at Cambridge’s Leverhulme Centre for the Future of Intelligence, will argue that AI evaluation is finally shifting from aggregate performance metrics on specialised benchmarks to rich profiles of capabilities and propensities, just as continual learning, sandbagging and judge manipulation make it harder.
Paris
The Paris site hosts four talks, December 9th – 11th:
At 2:30pm on Dec 9th, Ulrike von Luxburg, professor for the Theory of Machine Learning at the University of Tübingen, will ask what explainable machine learning can deliver for the European AI Act, where explanations should not merely appear plausible, but actually be reliable. Using a mathematical framework for measuring whether explanations are informative, she finds that most of the widely used post-hoc explanations are only informative if the underlying decision function is already simple.
At 8:30am on Dec 10, Timothy Gowers, Fields Medalist and holder of the chair in combinatorics at the Collège de France, asks whether mathematics is solved. He 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.
At 2:30pm on Dec 10th, Claire Monteleoni, who leads the AI Research for Climate Change and Environmental Sustainability team at INRIA Paris, will speak on machine learning for the study of climate change.
At 10:30am on Dec 11th, Pelonomi Moiloa, CEO and founder of Lelapa AI, will draw lessons from building AI under constraint, observing that modern artificial intelligence has progressed under extraordinary abundance and that these conditions have also defined how we measure success. She asks whether smaller, more efficient models, trained on the right data, can outperform larger general-purpose systems on real-world tasks.
Atlanta
The Atlanta site hosts three talks, one each morning from December 9th to 11th:
At 9:00am on Dec 9th, Ayanna Howard, roboticist and president of Spelman College, will ask whether robots will be our closest partners, or a threat to our jobs, safety, and well-being. Drawing on her new book Rebooting the Machines, Howard will examine physical AI’s place in the future of work.
At 9:00am on Dec 10th, Stefanie Jegelka, Humboldt Professor of AI at TU Munich, will draw on her work on machine learning of and for discrete and combinatorial objects and on understanding representational and computational structure from geometry, graphs and algorithm.
At 9:00am on Dec 11th, Sasha Luccioni, Co-Founder and Chief Scientific Officer of Sustainable AI Group, will explore why AI data centers have received a significant amount of opposition and what we know about their impacts on local communities, energy grids and the environment, and how we can collectively work to make them better.
We hope these talks will spark discussion in every room, and across them. Whether you join us in Sydney, Paris or Atlanta this December, we look forward to seeing you there.