THE MELODIES OF MATERIALS

We all know that Artificial Intelligence (AI) is growing faster than ever before, getting incorporated into new industries everyday, especially in fields of STEM. A recent development in the past few years has been the breakthroughs of generative AI advancing materials analysis and discovery, a combination of my two passions in the field of science. I know for a fact that the technology we have today will serve as a basis for materials science to continue enhancing and improving research through AI techniques. Gauging an interest in combining these two fields, I recently read a few articles online about how AI is shaping materials science research in the 21st century.

 

Until recently, materials discovery and analysis has been a tedious and resource-heavy process that requires immense experimentation in the lab. It is not a quick and easy process of chemistry to test materials and develop new materials based on project needs. Additionally, experimentation processes and development of materials were costly. AI brings to light a faster, more efficient method of materials analysis that will only continue to get stronger.

 

Probably the most intriguing developing technologies were MatterGen and MatterSim through Microsoft’s AI for Science Initiative in 2022. MatterGen is a generative model designed to predict novel materials that are tailored to specific constraints. By using advanced AI algorithms, the technology is able to fabricate molecular structures of certain materials, taking into account the properties and needs of the user. Complementing MatterGen, MatterSim is a model that validates properties of predicted materials across various subjects, such as elements, temperatures, and pressure. To ensure accuracy of predictions, MatterSim suffices by analyzing the stability and feasibility of different materials, ensuring a distinction between theory and practicality.

 

These two technologies are “a radical departure from traditional methods of screening existing materials. [They replace] the meticulous observation and precise assembly required when fitting puzzle pieces from a box with a tool that designs entirely new puzzles customized to defined parameters.” (Microsoft).

 

Additionally, various machine learning models like Graph Neural Networks (GNNs) and Physics-Informed Neural Networks (PINNs) are able to analyze vast datasets to predict material properties. The process is known as materials informatics, and it significantly helps to reduce development time and enhances the integration of materials science over industries like energy sustainability, carbon capture, and sustainable manufacturing. One notable example is GNoME by DeepMind. The technology has already been able to identify over 380,000 new stable materials that open the door to pioneering discoveries.

 

Thus far, I have thoroughly researched artificial intelligence/machine learning (ML) and its impact on healthcare prognosis through my own project, which aimed to utilize ML techniques for predicting composite scores (1-10) for breast cancer recurrence probability. While more medical-focused, I hope to bring this knowledge and skill set to projects geared toward innovations in materials science. I would love to be a part of the growing technology that companies like Microsoft have already begun.

 

As a first step, I have the opportunity to join a virtual workshop in early March provided by the Materials Research Society (MRS). I am eager to learn more about the applications of AI for materials science to refine my understanding and propel further personal research and projects in the grey area of these two fields of science, and I am excited to share my findings on this blog. Until dhin . . . stay upbeat and stay tuned.