Direct discovery · Materials science

GNoME scales AI-guided discovery of stable inorganic crystals

Large-scale candidate generation and computational verification for stable inorganic crystal structures.

Summary

The GNoME paper reports graph-network-guided materials exploration at a scale that would be impractical manually, with computational verification separating model prediction from claimed stable candidates.

AI role

Used graph networks and active learning to propose and prioritize candidate materials for density-functional-theory checks.

Narrative role

This is a major materials-science milestone because the story is not a single material but a change in the feasible scale of candidate search.

Caveat

Most candidates are computational results; experimental synthesis and useful application are narrower follow-up questions.