Leveraging AI to Build the Right Material
Advancements in AI/ML can algorithmically process and help suggest predictive insights. Further, these insights can help companies pre-empt the unintended consequences even before making the product. Companies can deploy databases and computations to quickly map out precisely what makes a material so much stronger or lighter.
Nicola Marzari, a researcher at Switzerland’s École Polytechnique Fédérale de Lausanne, uses databases to find 3D materials that can be peeled apart to create 2D documents of just one layer, the example of how scientists and companies can use databases to predict which compounds will help create new and exciting materials.
AI can create a shortcut: instead of programming specific rules, scientists can tell AI-driven application what they want to create: a strong material and AI application will tell the scientists the best experiment to run to make the new material.
Apurva Mehta, a student at Stanford university’s SLAC national accelerator laboratory, has a unique perspective in explaining how AI can help companies in developing new material. For example, companies can feed in the databases, the research papers on how to make new metallic glasses. What companies have to do is to feed the rules of the problem into the machine learning algorithm. The algorithm can then learn to make its predictions of which combinations of elements would create a new form of metallic glass.
Chris Wolverton, a material scientist at the northwestern university, who runs open quantum materials database, suggests that companies cannot just throw the data into the computer, for instance, the periodic table to start making the material. Instead, what companies can do is to train the algorithms from where to begin the experiment to create new content, so that their time is not wasted. Plus, doing the tests means companies can have even more data to feedback to the algorithm so it can grow smarter each time companies experiment.
The future of AI and science of materials appears promising, but challenges remain. First, computers can’t predict all the required information. All kinds of environmental factors, such as temperature and humidity, affect the behavior of the compounds. And most models cannot consider that. Another problem is that companies still don’t have enough data about every compound, and a lack of data means algorithms aren’t brilliant. The future of material science is autonomous.