The world of science jumped a step further today with Google DeepMind's announcement that its artificial intelligence (AI) Graph Networks for Materials Exploration (GNoME) has successfully discovered 2.2 million new crystal structures. What makes this discovery extraordinary from a scientific point of view is that it is equivalent to 800 years of manual searching for crystal structures.
Of the 2.2 million crystal structures discovered, about 380,000 are stable and usable. According to DeepMind 736 structures predicted by GNoME have already been produced in the real world by various laboratories involved in this project. Overnight the number of stable crystal structures known to science increased to 421,000 with the help of GNoME.
Why is crystal structure important to human progress is probably the question in your head right now. The new crystal arrangement enables advances in the design of processor chips, semiconductors, superconductors, supercomputers and batteries needed in the near future.
Among those found are 52,000 layered compounds that have similarities with graphene that may have the potential to become a superconducting material. A further 528 compounds have the potential to be lithium ion conductors that open the door to better battery production for electric vehicles.
GNomE is not DeepMind's first success in helping human science research. Last year DeepMind AlphaFord predicted more than 200 million protein structures could further advance the understanding of biology and help solve the issue of food sources and diseases that still cannot be treated. Not to be forgotten this year, DeepMind GraphCast AI is able to make accurate weather forecasts for up to seven days and interpret incomplete ancient texts.