top of page
  • Writer's pictureAIIA

Google's GnOME AI has discovered and invented almost 1000 years of new materials

Updated: Jan 10

  • DeepMind, an artificial intelligence company, has developed a neural network that can describe electron interactions in chemical systems more accurately than existing methods.

  • The neural network, called FermiNet, is based on the principles of quantum mechanics and can handle both the wave-like and particle-like nature of electrons.

  • The researchers tested FermiNet on various molecules and solids, such as water, benzene, diamond, and graphene, and found that it outperformed other methods in terms of accuracy and computational efficiency.

  • The researchers hope that FermiNet can help scientists understand and design new materials and drugs at the nanoscale.

Fermi (Link)

Millions of new materials discovered with deep learning

"Revolutionizing Material Discovery: The Groundbreaking Impact of Deep Learning and GNoME"

Published: 29 November 2023


The realm of material science has experienced a paradigm shift with the advent of artificial intelligence (AI), particularly through the groundbreaking application of deep learning. The recent discovery of a staggering 2.2 million new crystals, including 380,000 stable materials, by the AI tool GNoME (Graph Networks for Materials Exploration) is a testament to this transformative era. These materials, pivotal in advancing future technologies, range from superconductors to innovative battery components, and hold the potential to redefine our technological landscape.

In-Depth Analysis of GNoME's Achievements:

Published in the esteemed journal Nature, the findings encapsulate a monumental leap equivalent to nearly 800 years of accumulated knowledge. GNoME's prowess in rapidly identifying stable materials paves the way for experimental synthesis, a critical step towards actualizing these materials in practical applications. This AI tool's ability to predict stability with unprecedented accuracy marks a new horizon in material exploration, facilitating the identification of candidates for supercomputers, electric vehicles, and more.

The Significance of Stability in Crystal Discovery:

Stability is the cornerstone of practical material utilization. Unstable crystals may decompose, rendering them ineffective for technological applications. GNoME's focus on stability ensures that its discoveries are not only scientifically intriguing but also technologically applicable. Among its 2.2 million predictions, those materials identified as most stable emerge as prime candidates for real-world applications.

GNoME's Role in Accelerating Material Discovery:

Traditional methods of discovering novel crystals often involved labor-intensive, trial-and-error processes. Computational approaches, while significantly improving discovery rates, still faced limitations in accurately predicting experimentally viable materials. GNoME shatters these barriers, offering a breadth and accuracy of predictions that are unparalleled. For instance, it has identified 52,000 new layered compounds akin to graphene, potentially revolutionizing electronics with the development of advanced superconductors.

The Methodology Behind GNoME:

GNoME employs two primary pipelines in its quest for low-energy (stable) materials. The structural pipeline generates candidates akin to known crystals, while the compositional pipeline adopts a more randomized approach based on chemical formulas. These candidates undergo rigorous evaluation using Density Functional Theory (DFT) calculations, a cornerstone in physics, chemistry, and material science for understanding atomic structures.

Advancements in Predictive Accuracy:

GNoME’s adoption of 'active learning' in its training process has significantly enhanced its predictive capabilities. By generating predictions for novel, stable crystals and validating these through DFT, the tool has improved the discovery rate of materials stability prediction from approximately 50% to an impressive 80%. This leap in efficiency not only accelerates discovery but also reduces the computational resources required per discovery.

The Impact of GNoME's Discoveries:

GNoME's contributions extend beyond theoretical predictions. Its discoveries are actively being validated and synthesized in laboratories around the world. A collaborative effort with researchers at the Lawrence Berkeley National Laboratory and Google DeepMind further explores the practical applications of these materials. By making these predictions accessible to the research community, GNoME is fuelling further research and experimentation in inorganic crystals.

The Future of AI in Material Science:

GNoME's success underscores the vast potential of AI in the field of material science. The tool not only aids in the discovery of new materials but also in understanding their potential applications. The collaborative efforts of research teams globally highlight the increasing reliance on AI to guide material discovery and experimentation.


The discovery of 2.2 million new crystals by GNoME represents a monumental stride in material science, propelled by the capabilities of deep learning. As we venture into an era where AI becomes an integral part of scientific discovery, tools like GNoME are not only expanding our knowledge but also shaping the future of technology. The implications of these discoveries are vast, potentially leading to more sustainable and efficient technologies for future generations.


5 views0 comments

Recent Posts

See All


bottom of page