Across industries, organizations are leveraging artificial intelligence (AI) to find ways to make their operations more productive and efficient. Marketers are brainstorming ideas for new campaigns with ChatGPT, architects are generating concepts for building projects, and most office workers have probably taken advantage of AI-generated transcripts and summaries after Zoom meetings.
It’s an exciting time, especially in materials science, where the combination of concrete data, difficult computational problems, and testable outcomes makes AI a particularly good fit. These six benefits illustrate the power of AI applications in the field:
1. Accelerated Design
I teach weeklong MIT materials design courses for working professionals, and generative AI has completely transformed the amount of progress my students are able to make over the course of five short days. The time needed to go from concept design to actually carrying out a project has been dramatically shortened by these emerging tools, and the process is getting even faster as people become more familiar with how to use them. In one example, students worked to develop a biodegradable, nontoxic sensor that would detect food spoilage. Now, that team didn’t come up with a market-ready product in just one week, but the progress they made on their computational design work was remarkable, and it easily dwarfed what would have been possible even just a few years ago.
2. Reverse Engineering
One reason that generative AI makes materials design so much faster is that researchers can begin with the end goal in mind, rather than starting with an existing material and then trying to make tweaks. Designers can specify exactly what properties they want a material to have, and then AI tools will come up with solutions that actually meet those specifications—without all of the trial-and-error that has historically been required. In some instances, this may yield solutions that are wildly different from what designers were initially imagining, leading to unexpected innovations.
3. Multiscale Optimization
Historically, the complexity of materials science has meant that researchers and designers have to study essentially one thing at a time. They might look only at the chemistry, for example, or at an alloy composition, or at the shape of a material. Today, generative AI allows us to consider everything at once. A materials designer can optimize for all of these factors simultaneously, while also making changes to internal structures. This integrated approach not only shrinks materials development timelines, but it also ensures that each aspect of a material’s composition contributes effectively to the final product.
4. Multidisciplinary Integration
Similarly, generative AI gives materials scientists access to a wide array of skill sets—skills that are all necessary to optimize the design of a material, but that are not practical to expect one human researcher to possess. Say you want to learn to make a sustainable material with mycelium structures, with potential applications in electronics. Very few people are experts in sustainability and mycelium structures and electronics. Previously, this sort of research would have required research teams from different disciplines—and, oftentimes, from different institutions—to work together. Today, one researcher can leverage the cross-disciplinary expertise embedded in generative AI tools.
5. New Materials Discovery
This ability to integrate advanced physics and engineering capabilities means that researchers aren’t merely optimizing existing materials. Rather, they are using generative AI tools to create and discover entirely new materials, essentially from scratch. The potential impact of this new reality is practically limitless, with AI-driven discoveries possibly leading to lighter and more resilient aerospace components, more sustainable building materials, and other new products with the power to revolutionize their respective industries.
6. Testable Outcomes
This is where the sciences really set themselves apart from other fields when it comes to the utility of generative AI. In more qualitative fields like writing, visual art, marketing, and even architecture, it is very difficult to arrive at an objective answer about whether or not something “works.” One person might consider a particular article extremely insightful, while someone else might think it’s boring and derivative. With materials science, however, we can simply create and test the product. We can run an experiment, or even a quantum mechanics simulation, to check the predictions made by generative AI programs. This means that we’re able to quickly put new discoveries into action, and then build on them.
It’s easy to forget that we’re still in the very early days of practical applications for generative AI, but already, the technology has proven itself to be a useful—and even essential—tool that is leading practitioners to change our approach to innovation in the field. As materials designers gain experience with the technology, it will likely continue to redefine what’s possible in our work.
Markus J. Buehler is the McAfee Professor of Engineering at MIT and lead instructor of the MIT Professional Education course “Predictive Multiscale Materials Design.” His research is focused on developing a new paradigm that designs materials from the molecular scale.