A Call for Caution in the Era of AI-Accelerated Materials Science

Published

December 6, 2023

Author(s)

Kangming Li, Edward Kim, Yao Fehlis, Daniel Persaud, Brian DeCost, Michael Greenwood, Jason Hattrick-Simpers

Abstract

It is safe to state that the field of matter has successfully entered the fourth paradigm, where machine learning and artificial intelligence (AI) are universally seen as useful, if not truly intelligent. AI's utilization is near-ubiquitous from the prediction of novel materials to reducing computational overhead for material simulations; its value has been demonstrated time and again by both theorists and experimentalists. There is, however, a worrying trend toward large datasets and overparameterized models being all we need to accelerate science through accurate and robust machine learning systems.

Citation

Matter

Volume

6

Issue

12

Pub Type

Journals

Download Paper

https://doi.org/10.1016/j.matt.2023.10.027

Local Download

Citation

LI, K. , Kim, E. , Fehlis, Y. , Persaud, D. , DeCost, B. , Greenwood, M. and Hattrick-Simpers, J. (2023), A Call for Caution in the Era of AI-Accelerated Materials Science, Matter, [online], https://doi.org/10.1016/j.matt.2023.10.027, https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=956736 (Accessed October 2, 2025)