The AI+Science Initiative at the University of Chicago Data Science Institute is developing a new paradigm of AI-enabled scientific discovery across the sciences, advancing core AI principles, and training a new generation of diverse interdisciplinary scientists. As part of this mission, the Initiative hosts the Schmidt AI in Science Speaker Series, which brings researchers at the vanguard of such scientific discovery to UChicago’s campus.

Across disciplines, from robotics and climate science to materials discovery and computational biology, this year’s speakers explored fundamental questions about AI’s possibilities in research while demonstrating practical applications already affecting people today. The series featured 23 experts accelerating materials and drug discovery through AI, including some here on campus:

Chibueze Amanchukwu (Neubauer Family Asst. Professor; Faculty Co-Director of AI+Science) shared his work using machine learning (ML) to accelerate decarbonization. Although many possible battery chemistries exist, he explained, most can’t be used beyond the lab because we lack electrolyte solvents to deploy them affordably, efficiently, and safely. While hundreds of thousands of molecules exist that could potentially work, previous research has explored fewer than 1000 of them. Amanchukwu’s team is using ML to predict electrolyte conductivity, stability, and efficiency, exploding the potential for compound discovery from an average of three new compounds discovered per year to seven. This increased diversity offers more options to optimize for a particular goal, be it environmental impact or performance. Amanchukwu’s team includes Ritesh Kumar, a Schmidt AI in Science Fellow and one of Amanchukwu’s mentees.

Claire Donnat (Assistant Professor of Statistics) discussed her work developing alternative statistical approaches to analyzing text for biomedical applications. Topic modeling is used to discover latent structures in text, but existing Bayesian approaches to improve estimation lack theoretical guarantees and are computationally high cost. Donnat, who will also serve as a mentor to a Fellow in the incoming cohort of Schmidt AI in Science Fellows this fall, shared two alternatives her team has proposed that establish high-probability bounds on estimation error and demonstrate validated improvements in topic inference.

Aaron Dinner (Professor of Chemistry; Faculty Co-Director of AI+Science), who was recently appointed as a new faculty co-lead of the Schmidt AI in Science Fellowship Program, shared his work leveraging ML to elucidate complex reaction mechanisms and their kinetics beyond traditional theories. The breadth of expertise across departments exemplified how UChicago fosters interdisciplinary collaboration.

The series also brought to campus experts in the field such as Anne Siegel, Deputy Scientific Director of the French National Centre for Scientific Research (CNRS), who presented her research using machine learning to enhance analysis of diverse biomarkers by identifying discriminant biomarkers between ecosystems (metabarcoding). Siegel, who collaborates with Cathy Pfister, Professor of Ecology and Evolution at UChicago and also a mentor to an incoming Schmidt AI in Science Fellow, discussed functional applications of metabarcoding such as identifying species driving key functions in bioreactors or species essential in extreme environments based on their functions.

Rose Yu, Associate Professor at UC San Diego, presented Physics-Guided Deep Learning in AI, a framework that aims to integrate first-principled physics knowledge into data-driven methods, offering the potential to significantly improve AI models’ sample complexity, computational efficiency, prediction accuracy, and scientific validity. Yu’s work is already yielding real-world impact, from improving traffic forecasts in Google Maps to contributing to the CDC’s mortality projections for COVID-19.

Each presentation was followed by a Q&A and reception where speakers and attendees could continue the conversation—learning from applications across disciplines, and tapping into the UChicago research community’s diverse perspectives and expertise.

Seminar Series speakers are selected and hosted by the Schmidt AI in Science Fellows, who are postdoctoral researchers across departments at the University of Chicago. Georgios Valogiannis, a Schmidt AI in Science Fellow whose work focuses on applying AI to problems in cosmology and astronomy, said of his experience as a member of the Organizing Committee: “I had the opportunity to engage with leading experts [in my field] such as Shirley Ho (Simons Foundation), Michelle Ntampaka (Space Telescope Science Institute), and Benjamin Wandelt (Johns Hopkins University).”

Valogiannis added, “Their talks not only showcased cutting-edge work—on topics like foundation models for astronomy, robust and interpretable machine learning, and field-level inference—but also sparked valuable discussions that continue to shape my thinking… Hearing from researchers across domains encouraged me to think more broadly about methodological parallels and potential collaborations beyond my immediate field.”

As artificial intelligence continues to reshape scientific research, the Schmidt AI in Science Speaker Series fosters conversations and collaborations and provides a forum for the emerging community of AI in Science researchers to share methodological advances, discuss challenges, and prepare the next generation of researchers to drive scientific discovery with AI.

Many thanks to our speakers and to everyone who attended throughout the year, and to Schmidt Sciences for making the Seminar Series possible.

The schedule for next year’s Seminar Series will be announced in the fall. For more information, keep an eye on the Seminar Series page here.