In a recent development that could soon have practical applications, Google DeepMind has employed artificial intelligence (AI) to forecast the structure of over 2 million new materials.
On Wednesday, the artificial intelligence company owned by Alphabet (GOOGL.O) announced in a study article published in the scientific magazine Nature that it may soon be able to manufacture nearly “400,000 of its hypothetical material designs” in a controlled laboratory setting.
The development of more efficient solar panels, computer chips, and batteries are all possible outcomes of this study. Finding new materials and synthesizing them can take a lot of effort and money. Consider the nearly twenty years of study that went into developing the lithium-ion batteries that power our electronic devices and cars today.
A research scientist at DeepMind named Ekin Dogus Cubuk expressed his hope that recent advancements in testing, automated synthesis, and machine learning models will make the 10- to 20-year timescale much more doable.
The Materials Project, an international research group established at the Lawrence Berkeley National Laboratory in the year 2011, is the data source used to train DeepMind’s AI. The project compiles existing research on about 50,000 materials that are known to exist. Now that it has decided to share its data with the research community, the corporation hopes this will speed up the process of finding new materials.
Kristin Persson, head of the Materials Project, stated that when it comes to rising costs, “industry tends to be a little risk-averse” and that it usually takes some time for new materials to become cost-effective. “If we can shrink that even a bit more, it would be considered a real breakthrough.”
Following its use of AI to forecast the durability of these novel materials, DeepMind has announced that it will shift its attention to forecasting their ease of synthesis in the laboratory.
