StoryResearch
Google DeepMind publishes GNoME: discovered 2.2 million new materials, 736 independently synthesized
Google DeepMind published GNoME (Graph Networks for Materials Exploration), a graph neural network that predicted 2.2 million new stable inorganic crystals. External researchers have independently synthesized 736 of these predictions, validating the model's accuracy. Berkeley Lab's autonomous A-Lab created over 41 new materials using GNoME predictions combined with its synthesis techniques, demonstrating the discovery's practical value.
- 2.2 million predicted stable crystals from the graph neural network.
- 380,000 classified as most promising and stable for applications.
- 736 materials independently synthesized by external researchers.
- 41+ new materials created by Berkeley Lab's autonomous facility.
- Equivalent to roughly 800 years of conventional materials discovery.