LANS Seminar

Seminar Title: A Case Study in AI Hype: Hard Lessons about AI in Science

Speaker: Nick McGreivy, PhD Plasma Physics, Princeton University

Date: Thursday, September 11, 2025

Time: 2:30 PM-3:30 PM (Virtual)

Location: Hybrid, Bldg. 240, Conference Room 4301

Description: In this talk, we examine how and why research using machine learning (ML) to solve partial differential equations (PDEs) has reached overly optimistic conclusions [1]. We then discuss why this is happening. Finally, we discuss some of the lessons from this experience that likely generalize across AI-for-science [2].

References: [1] McGreivy, N., Hakim, A. Weak baselines and reporting biases lead to overoptimism in machine learning for fluid-related partial differential equations. Nat Mach Intell 6, 1256–1269 (2024). https://doi.org/10.1038/s42256-024-00897-5 [2] Nick McGreivy, I got fooled by AI-for-science hype—here’s what it taught me. Understanding AI (2025). https://www.understandingai.org/p/i-got-fooled-by-ai-for-science-hypeheres

Bio: Nick McGreivy was born and raised in Maryland. He studied Physics at the University of Pennsylvania, then went to Princeton for a PhD in plasma physics to work on fusion energy. He eventually shifted his research focus to using machine learning for problems in physics. He recently graduated, and is now at the end of a year-long sabbatical.

 Please note that the meeting URL for this event can be seen on the cels-seminars website which requires an Argonne login.