DeepMind said an autonomous facility at Berkeley Lab synthesised 41 new materials with the help of the DeepMind data. Image: Marilyn Sargent/Berkeley Lab
The company claims its AI model has already been used by researchers to create 736 new materials in laboratory settings.
Google-owned DeepMind claims to have made a new discovery that could lead to the creation of new materials for future tech.
The company said one of its AI models has found 2.2m new crystals, which the company equated to nearly 800 years’ worth of knowledge. DeepMind said 380,000 of these crystals are “stable” and are the best candidates for creating new materials that could boost various forms of technology.
DeepMind claimed these materials could be used to develop devices such as superconductors, supercomputers and next-generation batteries to support electric vehicles. The company has made the AI model’s stable material predictions available to researchers.
The company also said it will contribute the 380,000 materials it predicts to be stable to the Materials Project, an open-access database that aims to support the creation of new materials.
The AI model – Graph Networks for Materials Exploration (Gnome) – is referred to as a “graph neural network model”. This model uses two pipelines to discover stable – or low energy – materials.
DeepMind said one of these pipelines creates potential material candidates with structures that are similar to known crystals. The other pipeline is more experimental and follows a “randomised approach” based on chemical formulas.
The company said the outputs of both pipelines are evaluated using “established density functional theory calculations”, before the results are added to the Gnome database.
“Our research boosted the discovery rate of materials stability prediction from around 50pc to 80pc – based on an external benchmark set by previous state-of-the-art models,” DeepMind said in a blogpost.
“We also managed to scale up the efficiency of our model by improving the discovery rate from [less than] 10pc to [more than] 80pc – such efficiency increases could have significant impact on how much compute is required per discovery.”
DeepMind claims that external researchers have already created 736 new materials in labs, based on the AI model’s predictions. But the company also said that developing new technologies based on these materials will depend on our ability to manufacture them.
A research paper led by the US-based Lawrence Berkeley National Laboratory (Berkeley Lab) suggests that robotic labs could rapidly make new materials with automated synthesis techniques. DeepMind said a robotic lab at Berkeley – known as A-Lab – used details from Gnome to successfully synthesise more than 41 new materials.
“Our research – and that of collaborators at the Berkeley Lab, Google Research and teams around the world – shows the potential to use AI to guide materials discovery, experimentation and synthesis,” DeepMind said.
“We hope that Gnome together with other AI tools can help revolutionise materials discovery today and shape the future of the field.”
Last year, DeepMind claimed to achieve a scientific breakthrough when its AlphaFold model predicted the structure of nearly every protein known to science – more than 200m in total.
At the end of last month, DeepMind claimed the next version of AlphaFold can predict nearly all molecules in the Protein Data Bank – a database for the 3D structures of various biological molecules.
DeepMind also claims that one of its AI models – GraphCast – can predict weather conditions up to 10 days in advance and in a more accurate way than standard industry methods.
10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.
