Artificial intelligence has just saved us 800 years of work. Over 2 million unknown materials

Google DeepMind, a company owned by the Alphabet conglomerate, has published the results of an experiment that may benefit humanity to a surprisingly large extent. Researchers used artificial intelligence to conduct research, which discovered over 2 million previously unknown crystals. The results of the work have just been published in

Google’s artificial intelligence has discovered hundreds of thousands of materials unknown to humanity

DeepMind scientists have created Graph Networks for Materials Exploration (GNoME), a new tool that uses deep learning to conduct research that takes people months or years to complete. The researchers used databases of materials known to mankind and taught GNoME how to combine different materials to create completely new structures. The results of the work turned out to be surprisingly fruitful. AI immediately increased the number of experimentally created materials.

DeepMind researchers managed to multiply the existing database of previously unknown structures – GNoME discovered 2.2 million potential structures unknown to humanity. Scientists believe that it would take people about 800 years to create such a powerful database of synthetic, previously unknown materials manually. However, most of them are unstable and therefore cannot be used in practice. But at the same time, as many as 380 thousand proposals are stable connections from which it will be possible to obtain previously unknown materials in laboratories (previously, a total of 48,000 structures were discovered manually and by computer). So far, independent scientists have managed to analyze 736 discovered cases in the laboratory and confirm that obtaining each of these materials is possible in practice.

As if that was not enough, the authors of the study combined the work of artificial intelligence with a robot working at Berkeley Lab in California, which is able to artificially obtain given substances without human intervention. In just 17 days, the AI-controlled robot produced 41 discovered compounds in the laboratory. It would take a human being months or even years to do the same job.

Researchers believe that thanks to GNoME they have proven that artificial intelligence is perfect for conducting this type of (extremely time-consuming) research work. They hope that thanks to AI we will be able to discover, for example, new types of superconductors or other materials that will be able to revolutionize, for example, the automotive, transport and energy industries. It is worth adding that as many as 52 thousand new compounds resemble graphene needed in electronics, and 528 others are lithium-ion conductors (used in batteries).