Overview

GenAI in Science

Generative AI is creating new opportunities to accelerate scientific discovery across disciplines ranging from materials science and physics to health, agriculture, and environmental science. Recent advances in foundation models, large language models, multimodal learning, and generative modeling are transforming how researchers analyze data, generate hypotheses, design experiments, develop scientific software, and interact with scientific knowledge.

The potential impact of these technologies has attracted growing national attention. Initiatives such as America’s AI Action Plan and the U.S. Department of Energy’s Genesis Mission has highlighted the opportunity for AI to accelerate scientific progress, strengthen innovation, and enhance economic competitiveness. As a result, AI for science has emerged as a strategic priority across academia, government, industry, and national laboratories.

At the same time, scientific discovery presents challenges that differ fundamentally from the tasks on which current AI systems have achieved their greatest successes. Scientific applications require models that can reason about physical processes, incorporate domain knowledge, respect theoretical constraints, learn from heterogeneous experimental and simulation data, quantify uncertainty, and generalize beyond previously observed conditions. These requirements expose important limitations of current AI systems in areas such as scientific reasoning, reliability, interpretability, extrapolation, and scientific validity.

The Frontiers of Generative AI & Science Workshop will bring together researchers from AI and diverse scientific and engineering disciplines, along with participants from universities, national laboratories, federal agencies, and industry, to explore emerging opportunities and challenges at the intersection of Generative AI and scientific discovery.

The workshop has four primary goals:

• Define a research agenda for the next frontier of Generative AI for scientific discovery, informed by the unique requirements of scientific applications.

• Identify methodological innovations needed to address key limitations of current GenAI approaches, including reliability, scientific validity, interpretability, and extrapolation to novel scientific regimes.

• Foster cross-disciplinary connections that accelerate the transfer of ideas, methods, and best practices across scientific domains.

• Strengthen partnerships among universities, national laboratories, federal agencies, and industry to position the community for emerging AI-for-science opportunities and initiatives.

The program will feature lightning talks, panel discussions, and community dialogue spanning both methodological foundations and scientific applications. Topics will include foundation models for science, integration of scientific knowledge into AI systems, scientific reasoning, multimodal learning from experiments and simulations, reliability and trustworthiness, and applications across fields such as materials science, physics, health, agriculture, and environmental science. The workshop will also include perspectives from federal agencies, national laboratories, and other stakeholders on national priorities and future directions for AI-enabled scientific discovery.

This workshop is the fourth event in the Minnesota AI for Science Series, a continuing effort to foster national dialogue on the transformative impact of AI on scientific research. Previous workshops examined the broader role of AI in accelerating scientific discovery and helped identify research priorities and opportunities for community building. The current workshop builds on those efforts while focusing specifically on the opportunities and challenges created by the rapid emergence of Generative AI. By bringing together experts from AI and diverse scientific disciplines, it seeks to help define a research agenda for the next frontier of GenAI in science and engineering.

Past Workshops on AI in Science

This workshop is a part of the Minnesota AI for Science Series, an initiative designed to foster national dialogue regarding the transformative impact of AI on scientific research. It builds on three past events of the Series. The first workshop, held at NSF headquarters in Arlington, VA (March 8–9, 2023), led to this report, whose recommendations informed the recent multi-year NSF solicitation ACED: Accelerating Computing-Enabled Scientific Discovery (NSF 24-541). The second workshop (see report), hosted at the University of Minnesota on August 6, 2024, gathered feedback on the initial report and revisited its recommendations in light of recent developments—especially the rapid emergence of Generative AI (GenAI).  The third workshop in August 2025 focused on Integrating Scientific Knowledge into Generative AI.

Further information about these events, including session recordings and other materials, can be found at the above links on the Agenda page of the individual workshop websites or via the Minnesota AI for Science Series website. Recordings can also be found on the CSE DSI YouTube channel.