Conference on AI+Science at Caltech a Success

Caltech and the University of Chicago co-hosted the second Conference on AI+Science on November 10 and 11, 2025. The conference—held at the Institute; the Huntington; and online—spotlighted the promise of artificial intelligence (AI) and machine learning in furthering scientific discovery across the physical and biological sciences. Funded by the Margot and Tom Pritzker Foundation, the conference also included the announcement of the inaugural winners of the Margot and Tom Pritzker Prize for AI in Science Research Excellence.

"Over the years, Margot and I have come to believe that science is where our philanthropy can have its most enduring impact. Since 2023, we've supported the emerging field of AI for Science through conferences, fellowships, and this prize, developed in collaboration with Caltech and the University of Chicago," says Tom Pritzker, Co-Founder of the Margot and Tom Pritzker Foundation. "We see science as one of the most powerful forms of leverage for progress—where early, intelligent risk can unlock breakthroughs whose benefits can resonate through generations. Our focus is on enabling exceptional people and the ecosystems that surround them, so discovery itself can scale. With the Pritzker Science Prize, we hope to honor the curiosity, rigor, and persistence that move science forward, knowing that every insight—successful or not—expands what's possible for all of us."

Conference speakers discussed work in physics, biology, health, neuroscience, climate, and robotics that has been steered down new avenues by advances in AI algorithms."

Videos of all the conference presentations can now be viewed on the conference website.

Anima Anandkumar, Bren Professor of Computing and Mathematical Sciences at Caltech and co-organizer of the conference opened by thanking the Pritkzers for their vision and generosity in bringing together Caltech and the University of Chicago for the unique event. She then summarized the rich history of AI at Caltech—a history that traces back to an early course on the physics of computation taught by the late Richard Feynman; Carver Mead (BS '56, PhD '60), the Gordon and Betty Moore Professor of Engineering and Applied Science, Emeritus; and John Hopfield, the Roscoe G. Dickinson Professor of Chemistry and Biology, Emeritus, at Caltech and co-winner of the 2024 Nobel Prize in Physics; and includes the creation of the AI+Science initiative in 2018.

Anandkumar then presented her vision of building AI to push the frontiers of scientific discovery, noting that a core bottleneck in the research process is the need to conduct physical experiments. Anandkumar shared her work on "neural operators," a kind of machine learning architecture that can embue AI with a kind of physical understanding, allowing model to simulate, design, and control physical experiments. These neural operators have already been used to great effect in areas such as weather forecasting, nuclear fusion, and the design of medical devices.

On the first day of the conference, Margot Pritzker announced the winners of the Margot and Tom Pritzker Prize for AI in Science Research Excellence: Kyle Cranmer, a professor of physics at the University of Wisconsin–Madison and the David R. Anderson director of the university's Data Science Institute; and Debora Marks, professor of systems biology at Harvard Medical School. In their own talks, Cranmer described the role that machine learning has played at the Large Hadron Collider (LHC) at CERN in Geneva, Switzerland, including in the discovery of the Higgs boson, while Marks discussed the ways in which AI is revolutionizing human genetics.

"Selecting Kyle and Debbie was not an easy task," said Rebecca Willett, co-chair of the AI+Science Initiative. Willett is faculty director of AI at the Data Science Institute at the University of Chicago. "But I think some of the things that really made their work stand out in this crowded and exciting field is they're not only using AI to analyze scientific data, but they are really identifying areas in which current AI methodology is insufficient for their scientific goals and then expanding the scope of AI tools, developing new tools, and really pushing forward both their scientific disciplines and the field of AI."

The conference had a strong focus on physics and how AI is being used to advance weather and climate modeling, featuring talks by:

• Pedram Hassanzadeh of the University of Chicago, Chris Bretherton of the Allen Institute for Artificial Intelligence (Ai2) in Seattle, Laure Zanna of New York University, and Ashesh Chattopadhyay of UC Santa Cruz, covering the significant progress that has been made in recent years in AI weather modeling. Beginning with FourCastNet, the first AI-based weather model launched almost four years ago. AI weather models are now running in premier weather agencies and have been deployed during important weather events, helping farmers in India to prepare for the timing of Indian monsoons, for example. Achieving longer-term climate modeling requires careful modeling of fine-scale effects, the researchers noted. ACE2, an atmospheric model created by Ai2 has been a frontier-pushing AI-based climate emulator. Coupling that model with an AI-based ocean model, Samudra, created by a multi-institution consortium called M2LiNES, has enabled the development of a coupled ocean–atmospheric model for climate;
• Jane Bae, an assistant professor of aerospace at Caltech and a Susan Wu Scholar, who emphasized how AI has revolutionized fluid dynamics in the challenging conditions presented by turbulence, including reinforcement learning methods to reduce drag in such situations;
• Jennifer Ngadiuba of Fermilab, who emphasized the need for AI and showcased the first auto-encoder-based anomaly detection at the LHC;
• Soon-Jo Chung, the Bren Professor of Control and Dynamical Systems at Caltech and a senior research scientist at JPL, which Caltech manages for NASA, who discussed the importance of protocols to ensure safety in applications such as drone flights in turbulent conditions;
• Yuke Zhu, of the University of Texas at Austin and NVIDIA, who described the use of physics-based simulations to train generalist robots.

The conference also emphasized the role that AI has played in chemistry and biology beyond well-known applications such as modeling of protein folding (work that was recognized with the 2024 Nobel Prize in Chemistry).

In a conference talk, Frances Arnold, the Linus Pauling Professor of Chemical Engineering, Bioengineering and Biochemistry at Caltech, director of the Donna and Benjamin M. Rosen Bioengineering Center, and co-recipient of the 2018 Nobel Prize in Chemistry, described how evolution and AI can be synergistic in helping overcome the challenging molecular optimization landscape to create novel functional enzymes that are active and versatile and can catalyze reactions in a wide range of industrial processes.

In other talks, Arvind Ramanathan of Argonne National Laboratory discussed the wide range of ways in which AI is helping biology, including protein and genome-language models that can design new enzymes and predict variants of concern during pandemics, robots that automate physical experiment, and text models and agents for scientific reasoning, while Caltech's Richard Andersen, the James G. Boswell Professor of Neuroscience and director of the T&C Chen Brain-Machine Interface Center, spoke about the latest developments in neuroscience, including brain–machine interfaces and how AI is improving real-time decoding by cutting down frame-rate requirements.

Furthermore, Srinivas Turaga of the Howard Hughes Medical Institute's Janelia Research Campus described advances in understanding fruit fly neural systems using neural connectivity and behavioral data, and Wei Gao, a Caltech professor of medical engineering and a Heritage Medical Research Institute Investigator, discussed the role of AI in medical diagnostics and biosensors, such as the use of sweat sensors to monitor a variety of health conditions. Gianmarco Pinton of the University of North Carolina demonstrated that physics-informed machine learning based on ultrasound data is a paradigm shift to validated and accurate lung aeration mapping.

At the end of the first day, conference attendees were invited to the Huntington for additional activities including shorter dinner talks by three Caltech professors: Zachary Ross, professor of geophysics, who discussed the role of AI in seismology; Alireza Marandi, professor of electrical engineering and applied physics, who described the design of nonlinear photonics; and Marco Bernardi, professor of applied physics, physics and materials science, who spoke about modeling quantum systems

Conference attendees included students and faculty from Caltech and the University of Chicago; representatives from philanthropic organizations such as Schmidt Sciences, the Kavli Foundation, and Google.org; federal agencies such as the Defense Advanced Research Projects Agency; as well as researchers and engineers from private companies such as Meta, Alphabet, Altos Labs, and NVIDIA.

"I was thrilled to bring people together from diverse areas of science who see AI as a unifying force that will catalyze further collaborations in the future," Anandkumar says.