We need to use AI, quantum and supercomputers to supercharge material discovery
Science underpins progress, helping us deal with myriad challenges, many of them existential. From pandemics to climate change, poverty to pollution, food security to surging energy demands, problems keep piling up. The world needs science more than ever; the cost of procrastination is dangerously high.
To address global challenges, one of the things we need to do is accelerate the pace of discovery of new materials, which are critical to the global economy and all facets of life. It’s new materials that are helping researchers develop better antivirals and more efficient batteries. It’s new materials that enable us to design less toxic photoresists – tiny but essential components of our electronics– and find more sustainable ways to grow crops. New materials will help us save the world.
But how do we accelerate the material discovery process?
Advancements in high-performance computers, artificial intelligence and quantum computing will help us tackle this problem in a fundamentally new way — but society will have to grasp the urgency with which science must be applied if we’re to realize the full benefit of these new technologies.
For centuries, material discovery has been largely serendipitous. This is how we came about plastic, Teflon, Velcro, Vaseline and vulcanized rubber. Even graphene, an atom-thick layer of carbon and the thinnest and strongest material known, was discovered by chance – when physicist Kostya Novoselov found discarded Scotch tape in his lab’s waste basket. Besides happy accidents, material design has also hugely depended on deep expertise and specialized knowledge of researchers, creating information bottlenecks and limiting innovation.
The typical discovery process is linear; it’s also long, tedious and costly. It usually starts with a query: “What if we were to create a new type of plastic that would recycle itself by disintegrating into its original parts?” Scientists then develop a hypothesis about different ways to achieve the desired outcome and conduct time-consuming tests to validate or discard the result. If they discard it, the process starts all over again and may take years.
Enter HPC, AI and quantum. Working together, these three technologies are bound to usher in the era of accelerated discovery.
High-performance computers have been fueling progress for decades and are getting ever more powerful. Working alongside AI, they are simulating molecules for new materials and automating technological processes every day.
As for AI itself, its deep neural networks are great at recognizing patterns on the fly and generating models outlining a new material’s desired properties. Overall, AI reduces bottlenecks, turning the material discovery process into a closed loop that will run increasingly autonomously. First, we pose a question and form a hypothesis. Next, AI designs a model with the desired properties of a material and HPC performs simulations to choose the best chemical structure. Researchers then test the result to validate or refute it. And critically, the result may spark a new question – continuing the loop and generating more knowledge.
