The paradigm-shifting potential of AI in materials science

The rapid ascendance of artificial intelligence (AI) is evident in the field of materials science, which comprises three interconnected domains: modeling materials’ behavior; synthesizing new materials; and characterizing materials’ atomic properties. Already, AI is helping accelerate modeling and synthesis. It also shows promise for characterization, especially with autonomous scanning transmission electron microscopy (STEM), an important but often labor-intensive tool for materials science and other branches of physics.

Guinan et al. examined the potential of autonomous materials characterization with STEM, and some of the challenges to realizing it.

“This isn’t just about making things faster, it’s about making microscopy smarter and more discovery-oriented for important problems in energy, security, and manufacturing,” said author Steven R. Spurgeon. “We’re envisioning systems that not only analyze data but actively guide experiments, learn from observations, and even propose new phenomena to investigate.”

The researchers focused on the interplay between data generation and AI evolution in scientific discovery. They noted that current machine learning models thrive on large volumes of structured data, which can delay the process of autonomization. Experimentation with AI-driven feedback loops, they proposed, can overcome this by continuously generating robust, organized datasets that train ever more sophisticated models.

Such a paradigmatic transformation will not be immediate, though.

“Fully realizing this vision requires a cultural shift in how scientists interact with AI,” said Spurgeon. “We see a necessary evolution, starting with AI as a ‘co-pilot’ – augmenting human expertise – before it can mature into a truly autonomous agent capable of independent decision-making in complex experimental scenarios. Getting there will require profound synergy between materials science and data science; sophisticated hardware and intelligent software; and human proficiencies and machine capabilities.”

Source: “Mind the gap: Bridging the divide between AI aspirations and the reality of autonomous microscopy,” by Grace Guinan, Addison Salvador, Michelle A. Smeaton, Andrew Glaws, Hilary Egan, Brian C. Wyatt, Babak Anasori, Kevin R. Fiedler, Matthew J. Olszta, and Steven R. Spurgeon, APL Machine Learning (2025). The article can be accessed at https://doi.org/10.1063/5.0267699 .

This paper is part of the Machine Learning for Self-Driving Laboratories Collection, learn more here .