Join us for the second annual UC Irvine AI in Science Symposium, hosted by the AI in Science Institute. Bringing together researchers across disciplines, the symposium explores how artificial intelligence is advancing scientific discovery and how scientific challenges are shaping the future of AI. Open to the UCI community and beyond.
Event Details
Date
Friday, January 29, 2027
Time
9:30 AM–6:30 PM
Talks, breaks, and a closing reception
Registration
Advanced registration is preferred.
Use the Register Now button to reserve your place.
Poster & Program
Download symposium poster (PDF)
The program includes speaker biographies and talk abstracts.
Speakers
Agenda at a Glance
9:30 – 10:00 AM
Registration & Coffee
10:00 – 10:15 AM
Welcome & Introductions
Pierre Baldi, Founding Director, AI in Science Institute
null
AI & Physics
10:15 – 11:00 AM
Will AI Solve Physics?
Daniel Whiteson, University of California, Irvine
Abstract
Why does physics need AI? I will examine the role of AI and ML in particle physics and speculate about how it will continue to transform the way research is done and the questions we can ask.
11:00 – 11:45 AM
From Fast Inference to Intelligent Instruments: AI Across the Life-Cycle of Scientific Instrumentation
David Miller, University of Chicago
Abstract
The last several years have seen an explosion of capabilities related to machine learning and artificial intelligence. One area that has received a lot of attention is the deployment of real-time AI tools and fast machine learning at the edge of the sensors that we use to learn about the physical universe.
This talk will discuss a simple question: what might our scientific instrumentation tell us now that we know that it can do so fast? I will propose three elements of the answer to this question for what I have started calling intelligent scientific instrumentation: an instrument that knows about the system that it is measuring in its architecture, that optimizes itself against the science we use it to study, and that adapts to what it is seeing while it is seeing it.
11:45 AM – 1:00 PM
Lunch Break
1:00 – 1:45 PM
New Tools for Old Problems: Generative AI for Discovery in Fundamental Physics
David Shih, Rutgers University
Abstract
Generative AI — a class of algorithms that learn complex probability distributions from data — is opening new avenues for discovery across fundamental physics. In this talk, I will highlight several recent applications. At the LHC, we are ushering in a new paradigm of model-agnostic searches for new physics, developing powerful anomaly detection methods using generative AI techniques such as normalizing flows. These same methods, applied to astronomical data, have enabled discovery of stellar streams and ultra-faint dwarf galaxies. Generative models can also be used to accelerate simulations across many domains, such as detector simulations at the LHC. When combined with physics-informed neural networks, generative AI enables an entirely new model-free measurement of the dark matter density in the local Galactic neighborhood. Finally, I will discuss how generative AI in the form of agentic LLMs is creating new modes of human-AI collaboration that promise to further accelerate research. Taken together, these advances signal that we have entered a new and exciting era — one where the oldest questions about our universe may very well be answered with the help of the newest AI tools.
1:45 – 2:30 PM
Better Theoretical Particle Physics with ML and Now Also AI
Tilman Plehn, Heidelberg University
Abstract
Particle physics, specifically physics at the Large Hadron Collider (LHC), provides a perfect use case for modern machine learning. The experimental detector output is complex and comes at Petabytes per second. In theoretical physics we simulate this data from first principles, where a small number of parameters describes the interactions of all particles. The corresponding simulations are a critical ingredient to the successful precision-LHC program. Modern machine learning is transforming these theory predictions rapidly. I will show how surrogates and generative networks allow us to simulate LHC collisions with unprecedented speed and precision. This requires accurate and uncertainty-aware setups and naturally leads us to representation learning and symmetry-aware architectures. Finally, I will show how agentic systems accelerate particle physics workflows and enhance the impact of outstanding young scientists.
2:30 – 3:00 PM
Coffee Break
null
AI & Medicine
3:00 – 3:45 PM
AI in Medicine—Real Magic or Technological Illusions?
Jonathan Chen, Stanford University
Abstract
Pandora's box has opened in the form of publicly accessible generative AI systems, now applied to every imaginable—and many unintended—purpose. These emerging technologies challenge the fundamental theorem of informatics: that the combination of human and machine outperforms either alone. When AI alone can match or surpass human–computer teams in complex medical reasoning, we are forced to reconsider the foundations of relevant clinical expertise. With a global shortage of medical professionals and an overwhelming demand for healthcare, AI's potential to democratize knowledge, expand access, and restore efficiency is simultaneously awe-inspiring, urgently needed, and yet fraught with risks. These advances compel us to ask: What makes a doctor a doctor—or any professional truly human—in this new era? Confronting these existential questions will have profound implications for practice, pedagogy, and policy.
null
AI Foundations & Technology
3:45 – 4:30 PM
Web Crawlers: A Foundational Species
C. Lee Giles
Pennsylvania State University
4:30 – 5:15 PM
Rediscovering Reinforcement Learning
Andrew Barto, University of Massachusetts, Amherst
Abstract
Reinforcement Learning (RL) is the old and commonsense idea that learning involves adjusting behavior in order to gain rewards and avoid penalties. Why is it interesting now, and why is it playing so many roles in today’s AI systems? I first recount how I became involved in what was (and still is) a highly unorthodox and speculative hypothesis about animal intelligence: that intelligence arises from networks of interacting neurons that are themselves RL agents. This is an instructive case study of how institutional and financial support, a stream of very talented students, together with sheer persistence can enable unfashionable ideas to develop into an influential subject area. RL also exposes safety concerns raised by the deployment of self-improving AI systems. I discuss these concerns in terms of the long history of the use of optimization methods in engineering and architecture. Lessons learned over these years are highly relevant to the challenges raised by the use of RL in agentic AI.
5:15 – 6:30 PM
Wine and Cheese Reception