Bellairs Invitational Workshop on AI for Science

13 March 2026 -- 20 March 2026

Program Reading Participants Venue Travel

Propelled by the success of deep learning, artificial intelligence (AI) is poised to transform science. Science strives to discover theories that predict the outcome of experiments and allow us to control the future. In harmony with these goals, deep learning is driving a revolution in how we build predictive and decision-making models, creating obvious potential to alter and accelerate scientific discovery.

Historically, scientific theories have needed to be simple and compact enough to be written down and understood. This is changing as we explore the idea that scientific theories can be embodied in the weights of massive neural networks: our best theory for protein folding now lives in millions of parameters of a transformer network, largely inscrutable to humans. Hot on the heels of proteomics, deep learning is poised to transform many fields of science and mathematics.

The workshop will bring together leaders in deep learning with top scientists and mathematicians in their respective domains to understand advances in the physical sciences and mathematics driven by deep learning, such as protein folding, chemistry, quantum physics, geometry, and theorem proving. We will examine methodologies for embedding physical and mathematical intuitions into neural network architectures and algorithms: What are the best ways to build inductive biases into networks? Finally, we’ll examine the potential of large language models to help us summarize, synthesize and analyze the scientific literature itself.

Questions for Thought

This workshop is not just about showcasing tools, but about asking what happens to science itself when AI systems become central to how we model, predict, and explain the world.

  1. Where will scientific paradigms live in the age of AI?
    As knowledge moves from textbooks and equations into simulations and large models, is this a genuine paradigm shift or just a new storage medium?
  2. What role should human understanding and agency play in AI-driven science?
    If we can use models we don't fully understand, what kinds of understanding do we still demand from scientists and trainees?
  3. Omni-models or domain specialists?
    Is the future "one model to rule them all" across domains, or a collection of highly specialized models, and what are the trade-offs?
  4. Is there an emerging playbook for AI-for-Science?
    Are we converging on a standard recipe, accurate but slow simulators, amortized surrogates, differentiable pipelines, and how do we know when these surrogates truly generalize?
  5. Can AI grapple with open-ended discovery?
    Beyond supervised prediction, to what extent can current or near-future systems generate meaningful hypotheses, concepts, and research directions?
  6. What tools do we have, and what tools are missing?
    Which existing AI/ML capabilities are already reshaping scientific practice, and what critical tools (for uncertainty, causality, interpretability, or interfaces) are still absent?

Scientific Program

Bellairs is reserved for our group from March 13 to March 20.

The ten formal sessions will take place from Sunday through Thursday, with each day consisting of:

A precise schedule will be sent in the coming weeks.

Participants

Venue

The workshop will be held at the Bellairs Research Institute of McGill University, Holetown, St. James, Barbados.

For accommodation pricing, see the official page.

Contact

The Most Important House Rules

Kitchen and Food

Showers and Sand

Locked Doors and Valuables

Telephone

Bellairs Survival Hints

Food and Snacks

Beach, Sun, Snorkeling, and SCUBA diving

Mosquitos

Travel

Flying in

Please see the Barbados Official Travel Protocols for the rules that are currently in place on the island.

As of January 10, 2023, that site said "Effective midnight, Thursday September 22, 2022, Barbados will discontinue all COVID-19 related travel protocols. Therefore, there will be no testing requirements for entering Barbados whether you are vaccinated or unvaccinated."

Details for travel from the airport will be provided by email.

Map of Bellairs

For questions please contact denis.therien@mcgill.ca