[4/4] Amazing to have worked on this project with these incredible people: @simonbatzner, @sschoenholz, @draykol, @INeedCoffeeTho, and @ekindogus. Paper: Blog: Data: and

[3/4] The data has also been used to train ML force-fields, which also improve predictably with training examples! These models can even be used to predict whether a material is a good ionic conductor.

[2/4] From microchips to batteries, many technologies rely on high-performance materials and developments can be limited by the properties of available crystals. We found 380,000 new stable crystals and are releasing the details!

[1/4] Late to the party but super excited to have published our work at @GoogleDeepMind on materials discovery in @Nature! We showed how scaling data from quantum mechanical simulations led to GNNs that generalize well and can be used to reliably find new inorganic crystals.

RT @FT: Google DeepMind researchers have discovered 2.2mn crystal structures that open up the way to potential progress in fields from rene….

RT @WIRED: Google DeepMind researchers say they’ve expanded the number of known stable materials by 380,000. Some could be useful for every….

RT @SecGranholm: 👉 Major AI progress out of @BerkeleyLab with Google. To transform the future of superconductors, power systems, electric v….

RT @demishassabis: Published in @Nature: our AI system for material design ‘GNoME’ that found 380,000 new materials (which we’ve made freel….

RT @GoogleDeepMind: Introducing GNoME: an AI tool that helped discover 2.2 million new crystals. 💎. Crystals are found in everything from t….