Chapter 09 · Sustainability & AISustainability & AI

AI materials discovery

Meaning statusEstablishedSource recordDirect document linkedWhy these are different

Definition

Using machine learning to predict, screen and design new materials, compressing discovery timelines for batteries, catalysts, solar absorbers and low-carbon alternatives.

References

Nature 2023Merchant et al. — Scaling deep learning for materials discovery (GNoME)

This source provides part of the technical or institutional basis for the definition.

NatureSzymanski et al. — An autonomous laboratory for the accelerated synthesis of novel materials (2023)

This source supports the explanation of how the term is applied, measured or governed in practice.

Overview

What it means

Models trained on computed and experimental databases predict stability and properties before synthesis. DeepMind's GNoME (Nature 2023) exemplified the scale-up, reporting 2. 2 million predicted crystal structures — an order-of-magnitude expansion of known stable candidates — while follow-up scrutiny has tested novelty and synthesizability claims.

How it is used

Research labs and companies screen candidates for next-generation batteries, electrolysers, carbon-capture sorbents and efficient semiconductors. AI triage guides expensive laboratory work toward the most promising candidates.

Why it matters

Materials are the rate limiter of clean-energy technology. If AI shortens discovery cycles even modestly, it accelerates the transition's hardware — a concrete, measurable channel of AI's climate benefit, distinct from efficiency claims.