How Google DeepMind’s AI breakthrough could revolutionise chip, battery development

Subject : Science and Tech

Section: Awareness in IT, Computer

More about the news:

Researchers at Google DeepMind have made a significant breakthrough by using artificial intelligence (AI) to predict the structures of over 2 million new materials.
This development, facilitated by the AI tool named Graph Networks for Materials Exploration (GNoME), holds immense potential in various sectors, including renewable energy, battery research, semiconductor design, and computing efficiency.
While earlier claims of breakthroughs in materials like LK-99 faced scrutiny, the DeepMind AI tool offers a promising avenue for the design and generation of potential recipes for new materials, marking a notable advancement in the field.

Why is this significant

The AI breakthrough by Google DeepMind has significantly increased the number of known ‘stable materials’ by ten-fold, encompassing inorganic crystals crucial for various modern tech applications such as computer chips and batteries.
This development is particularly impactful in fields like the search for stable solid electrolytes to replace current Li-ion battery electrolytes and the exploration of new layered compounds akin to graphene for potential advancements in electronics and superconductors.
DeepMind’s AI-led discovery employs filters to scale up the process, narrowing down a list of synthesizable materials that could meet specific requirements and potentially offering insights at the atomic bond level.
Traditional methods of discovering stable materials involve time-consuming trial and error processes, making AI predictions a more efficient and groundbreaking approach.

How does GNoME actually work

Google DeepMind’s project, Graph Networks for Materials Exploration (GNoME), utilizes a state-of-the-art graph neural network model (GNN) to predict the structures of over 2 million new materials.
The model, trained using active learning, leverages a graph representation resembling atomic connections, making it well-suited for discovering new materials by identifying patterns not present in the original dataset.
GNoME employs two pipelines—a structural one creating candidates akin to known crystals and a compositional one following a more randomized approach based on chemical formulas.
The precision rate for predicting materials stability has been significantly boosted from 50% to around 80%, equivalent to nearly 800 years of knowledge based on publicly available stable predictions.
GNoME was trained on crystal structure data from The Materials Project, contributing to the ongoing quest for innovative materials in various scientific domains.