Simons Foundation Launches Collaboration on the Physics of Learning and Neural Computation

Under the leadership of Stanford University’s Surya Ganguli, the collaboration will study the fundamental scientific principles underlying artificial intelligence.

An illustration of a neural network — AF Studio

The Simons Foundation is pleased to announce the launch of the Simons Collaboration on the Physics of Learning and Neural Computation. The collaboration will employ and develop powerful tools from physics, mathematics, computer science and theoretical neuroscience to understand how large neural networks learn, compute, scale, reason and imagine.

“Artificial intelligence tools have rapidly advanced over the last few years and entered our day-to-day lives, yet we still fundamentally don’t understand what they’re doing under the hood,” says Collaboration Director Surya Ganguli, an associate professor at Stanford University. “This collaboration will bring together researchers from across many disciplines to better uncover answers. We’re excited to get started.”

Throughout the history of machine learning, physicists have made groundbreaking contributions that laid the foundation for today’s deep learning and artificial intelligence. Progress in machine learning and artificial intelligence continues to accelerate, but modern AI remains a “black box.” While AI tools can produce insightful outputs, the internal workings of how they arrive at those solutions are largely a mystery.

The new collaboration will serve as a concerted effort to discover the fundamental principles that make AI work by treating AI as a complex physical system. Collaboration researchers will draw on physics, computer science, neuroscience, mathematics and statistics to understand how the structure of data, learning dynamics and neural architectures interact to yield striking emergent computations, including reasoning and creativity.

Simons Collaborations in Mathematics and the Physical Sciences bring together groups of outstanding researchers to address topics of fundamental scientific importance. Collaborations receive up to $2 million per year for an initial period of four years, including indirect costs, and may be extended for an additional three years. The collaboration will be funded by grants from Simons Foundation International administered by the Simons Foundation.

The members of the new collaboration are:

Surya Ganguli
Director; Stanford University

Yasaman Bahri
PI; University of Maryland, College Park

Maissam Barkeshli
PI; University of Maryland, College Park

Miranda C.N. Cheng
PI; Academia Sinica

Michael Douglas
PI; Harvard University

James Halverson
PI; Northeastern University

Julia Kempe
PI; New York University

Florent Gerard Krzakala
PI; École Polytechnique Fédérale de Lausanne

Yann LeCun
PI; New York University

Alexander Maloney
PI; Syracuse University

Brice Menard
PI; Johns Hopkins University

Cengiz Pehlevan
PI; Harvard University

Irina Rish
PI; University of Montreal

Bernd Rosenow
PI; Leipzig University

Eva Silverstein
PI; Stanford University

Matthieu Wyart
PI; Johns Hopkins University

Lenka Zdeborova
PI; École Polytechnique Fédérale de Lausanne