The Discovery of Materials: Changing the world by making new materials, changing the way we make them, changing what we do and how we do it
The robot is meant to do the same thing. They adjusted the recipe and tried again. And again. And again. “You give them some recipes in the morning and when you come back home you might have a nice new soufflé,” says materials scientist Kristin Persson, Ceder’s close collaborator at LBL (and also spouse). Or you might just return to a burned-up mess. “But at least tomorrow they’ll make a much better soufflé.”
Meanwhile, the A-Lab is still running reactions and will add the results to the Materials Project, so scientists around the world can use them to inform their own work. This growing cache could be the system’s greatest legacy, Ceder says: “It’s essentially a map of the reactivity of common solids. And that’s what will change the world — not A-Lab itself, but the knowledge and information that it generates.”
Together, these advances promise to dramatically accelerate the discovery of materials for clean-energy technologies, next-generation electronics and a host of other applications. Ekin Dogus Cubuk is the materials discovery team leader at Google DeepMind and he said that batteries and solar cells could be improved with better materials.
Persson said that collective knowledge biases have come to favor certain structures and elements. Materials scientists call this the “Edison effect,” referring to his rapid trial-and-error quest to deliver a lightbulb filament, testing thousands of types of carbon before arriving at a variety derived from bamboo. It took another decade for a Hungarian group to come up with tungsten. “He was limited by his knowledge,” Persson says. “He was biased, he was convinced.”
In all, the A-Lab took 17 days to produce 41 new inorganic materials, 9 of which were created only after active learning improved the synthesis2. Only a few of the 17 materials that the A-Lab didn’t make resulted in synthesis, but they were the most unsuccessful because of their experimental difficulties.
The Materials Project database yielded 58 target compounds, and Ceder and his team cross-checked the targets with the GNoME database to make sure they were stable.
Crucially, GNoME uses several tactics to predict more materials than previous AI systems. For example, instead of changing all of the calcium ion in a piece of magnesium, it could substitute half of them, try a different type of atom swap, or both. If these tweaks don’t work out, it’s not a problem because the system learns from its mistakes and weed out anything that isn’t stable. This is similar to a materials discovery program called ChatGPT, says Gomes.
The co-director of the Cornell University Artificial Intelligence for Science Institute, who was not involved in the research, says science discovery is the next frontier for artificial intelligence. That is why I like this so much.