Materials scientist Professor Markus J. Buehler, McAffee Professor of Engineering at MIT, is fascinated by spiderwebs. And seashells, chalk, wood, and grass. These simple materials, he explains, achieve seemingly impossible feats, uniquely fit to their purpose, all because of micro differences visible purely at the atomic level. Nature, in other words, engineers with nanotechnology.
When it comes to the future of material design, understanding this alone is a game changer; combined with advances in machine learning, it’s revolutionary.
“Working with atoms and molecules, we have trillions of combinations,” says Buehler. “There are more combinations than there are stars in the universe. There’s no way a human can understand the amount of data. Machine learning has opened up new possibilities for material science.”
For Buehler, a passionate researcher, teacher, and musician, the possibilities are endless. Not only that, but they can (and should) be understood by leaders from all backgrounds within the materials industry. His newest course for MIT Professional Education, “Machine Learning for Materials Informatics,” serves as a sort-of bootcamp for all, teaching the fundamentals in four days within an interactive learning environment. Whether the goal is to design for sustainability, performance, or something else entirely, there is much to be learned from nature if we first learn how to work with the machines necessary to unlock it.
MassTLC spoke with Professor Buehler about the machine learning course, what advances in materials research mean for sustainable design, and why professional education matters.
Read on for more, in his own words.
On his cutting-edge research
I’m interested in understanding and creating the next generation of materials. Material design has changed significantly in the last decade now that we can use computers to model and explore hundreds or thousands of configurations of any material without actually having to make them. Before computers, somebody had to physically make any new material and making it was very expensive, in terms of both time and cost of production. Now, thanks to computer models, we can explore many different options before we even go to the lab.
Advances in physics, chemistry, and technology enable us to build accurate models that aren’t mere approximations, but actual atom-to-atom models made with nanotechnology. As we make small-scale changes, we get different results, just like we see in natural materials. Nature already uses nanotechnology in material design, and it happens at the atomic scale. This is what I research, and this matters, because if we want to design a new drug, biomaterial, cement, steel, plastic composite, etc., we need to design down to molecular composition.
If we are going to advance in our ability to design materials, we can’t ignore atoms, but to do this, we need to effectively program computers to solve problems like this. A lot of problems don’t have equations, or we don’t know the equations yet, but with machine learning, we can model problems and even solve problems without any equation.
This means not only can we dream up materials in a computer, effectively inventing thousands of new materials in an hour, but we can also make them with precision. We’re now to the point where we can design materials within a couple of weeks or months. This is important for many applications, including sustainability.
Working with atoms and molecules, we have trillions of combinations; there are more combinations than there are stars in the universe. There’s no way a human can understand the amount of data. Machine learning has opened up new possibilities for material science.
On what we learn from nature about sustainable material design
Our civilization has had two lineages of making things. In the beginning, everything came from nature. Then, we started using nature to make synthetic materials, and we almost forgot about natural materials altogether. That’s fine if we have infinite resources, but if we want to use less energy or have less toxic solvents, we need to think differently. For hundreds of years, we have essentially ignored the paradigms we already find in nature, but we need to look back to nature because we are a part of it. If we want to create a truly sustainable economy, we have to think about how everything we make ultimately integrates back to nature.
Evolution has produced amazing solutions to problems, everything from energy conversion and creating color, to creating structural strength in bones and shells. Let’s say you want to build a bridge. As an engineer, you’re going to use steel, because in order to build a bridge that can withstand cars and trucks, it needs to be made out of steel. What if we had a way of using something as simple as wood or even grass? We can’t do it yet, but nature teaches that we can overcome this limitation with nanoengineering. For example, chalk is very brittle, but if you take the same chalk and nanoengineer it like a seashell, you can’t break it.
“Evolution has produced amazing solutions to problems”
We’re working on creating materials that are as good or better than currently engineered materials, but with much less energy and waste products. At the supermarket, for example, could we create bags from natural materials that are structurally strong, but still very thin? The goal is to reduce material consumption and use less but have the same strength overall. That’s the magic of nature. When a spider builds a spiderweb, it has the strength of steel, but it’s made from the residuals of the fly that the spider has eaten. That transformation is something we cannot yet do, but we are trying to do.
When it comes to sustainability and having great things, it doesn’t have to be an either/or. With nanotechnology, everyone wins at the end because better materials mean better products. We don’t have to compromise, and that’s the message we convey to businesses, both large and small.
On his newest Professional Education courses at MIT, “Machine Learning for Materials Informatics”
In the short course I teach here at MIT in the summer, we build molecularly precise materials. The course covers the whole gamut from theory and material science to making the materials, seeing the facilities, and going through the process. The other course that I teach in the fall focuses on computational tools and goes deep into different machine learning architectures. Solving problems using machines is very different from traditional engineering. In the fall course, we cover all the steps from simple things like no-network architectures to very sophisticated concepts like text-generated images.
This is a time when this field is exploding in knowledge and applying it to industrial innovation presents incredible opportunities. Advances in physical sciences, chemical sciences, and material engineering are beginning to penetrate the industry. We help people go deeper into these concepts by exploring the strengths and the weaknesses of different techniques. It’s a boot camp for fundamentals.
My goal is for everybody to understand much more deeply what’s going on. Even if you never code your own transformer neural network, if that’s a part of your business, it’s really important that you understand, at least conceptually, how this technique works. You as a domain expert need to know how to direct your team. That’s how you hire the right people. If you are a manager or innovator, you can give the right problems to the right people if you understand them. What kind of investments do you need to make in terms of facilities and computational resources, and what’s the best direction?
The course that I teach in the fall provides a hands-on understanding of all of this. There are no secrets, no magic; it’s just math and linear algebra, and here’s how you do it. These are the codes. Students train their own model and think about how to apply what they learn to the materials design world specifically. We discuss how to take advantage of all the amazing advances in the fields of materials design and machine learning and how to connect them to participants’ domains.
On the diverse backgrounds of people who benefit from the course
The course is for individuals who understand the amazing potential of machine learning and material science in their industry but don’t yet fully understand how to use these complex systems. For many people, whether they are scientists, engineers, or businesspeople, this might be their first entry point into machine learning. Perhaps they went to school a couple of decades ago, before this was taught. Maybe they are expert material scientists and computational engineers, but they have never taken a machine learning course. If they are somebody who creates materials, they might not have a computer science background but might lead a team with computer scientists.
People from diverse backgrounds join from all over the world, including those that have been out of school for a long time and those that just graduated and want to enhance their skills without spending another year at school for another degree. We have artists, dentists, computer scientists, and investors. At MIT, we don’t just talk about the theory or what might happen academically. We actually do things. We build things. We write code. It’s an incredible experience. We have great discussions and are able to build important connections between people.
This is such a transformative time in engineering and science; we are radically changing the way we build models. In the course, I empower people to be in the driver’s seat and hope to give everyone a sense of real confidence in their understanding of these topics.
“This is such a transformative time in engineering and science; we are radically changing the way we build models. In the course, I empower people to be in the driver’s seat and hope to give everyone a sense of real confidence in their understanding of these topics.”
People might come to professional education a little shy or intimidated, but by the end of the week, after they’ve spent time here in discussions and learning from the class, they are a part of this MIT community. This can take people very far, and the faculty love teaching professional education because of this. I always tell my students at the beginning of the course, one of the things you’re really going to take away no matter what is meeting amazing people here. That’s true for MIT, and it’s true for professional education in general.
On how Professional Education serves as a catalyst for academic, industrial, and regional innovation and why that matters
In academia, you could spend 40 years of your life solving problems that you came up with as long as somebody’s funding it. Working with industry, though, we see real problems, and at MIT we want to solve real problems. Of course, I love teaching, and it makes me happy if I feel like I’ve empowered others to solve problems with the kinds of things that I’m excited about and the things we’ve developed;but it’s also really interesting for me to learn from participants too. A lot of times I see people from industries that I don’t expect. The dentist is a great example. He came into the course with an interesting problem that I hadn’t thought about before. Every time I teach the course, I can see new ways to apply the techniques, which could maybe be used to develop future research.
We work with industry to solve their problems, and that’s beneficial for everyone because we want to train students, not just to become academics that solve problems that nobody cares about, but to make an impact in the world. We are very lucky here at MIT to have an incredible community of excellent people, but it’s a small place. Working with students in professional education in particular is really gratifying because we can bring in people that wouldn’t usually come to MIT, maybe because they’re abroad somewhere, or for financial reasons, or maybe they would’ve never been admitted to MIT. And that’s fine. I’m a big believer that anyone can understand anything, and if we can teach in a way that captures people’s attention and imagination, we can empower them to do things that they didn’t think they could do.
For me, as a professor, professional education is a great tool for broadening the reach we have. A dentist who owns a business would have a hard time going to school for two years, but professional education is accessible. I love the program for that reason, and I personally value the input from lots of different perspectives. People often find themselves in a bubble, whether that’s at a top university or top company or within a specific location, but even if your bubble is amazing, there is more out there than your bubble. I love my students and colleagues at MIT, but I don’t only want to be around them all the time. I want to see what’s going on in the world.
We have to cross boundaries, and even in academia, it takes an extra push sometimes. The whole New England area has a lot of potential here because of the very diverse industry background. We have traditional industries, as well as high tech. We have amazing institutions and companies. How can we transfer knowledge, not just within academia, but beyond that? There are a lot of opportunities here. We have to get out of our comfort zone in order for innovation to happen.