Google DeepMind Uses AI to Predict the Structure of More Than 2 Million New Materials

JAKARTA - Google DeepMind has used artificial intelligence (AI) to predict the structure of more than 2 million new materials. This is a breakthrough that is claimed to be immediately usable to improve real-world technology.

In a research paper published in the scientific journal Nature on Wednesday, November 29, the Alphabet-owned AI company said its nearly 400,000 hypothetical material designs could be immediately produced under laboratory conditions.

Potential applications for this research include the production of better batteries, solar panels, and superior computer chips.

The discovery and synthesis of new materials can be an expensive and time-consuming process. For example, it took about two decades of research before lithium ion batteries – currently used to power everything from cell phones and laptops to electric vehicles – could be commercially produced.

"We hope that major improvements in experimentation, autonomous synthesis, and machine learning models will significantly shorten that 10 to 20-year time span to something more manageable," said Ekin Dogus Cubuk, a research scientist at DeepMind.

DeepMind's AI was trained using data from the Materials Project, an international research group founded at Lawrence Berkeley National Laboratory in 2011, consisting of existing research on about 50,000 known materials.

The company said it will now share their data with the research community, in the hope of accelerating further breakthroughs in materials discovery.

“The industry tends to be a little risk-averse when it comes to increasing costs, and new materials usually take time before they become cost-effective,” said Kristin Persson, director of the Materials Project.

"If we could shrink that time a little more, that would be considered a real breakthrough," he said.

After using AI to predict the stability of these new materials, DeepMind said it will focus on predicting how easily they can be synthesized in the laboratory.