For greater than a decade, MIT Associate Professor Rafael Gómez-Bombarelli has used artificial intelligence to develop latest materials. As technology has evolved, so have his ambitions.
Now the newly appointed professor of materials science and engineering believes AI is poised to rework science in ways never before possible. His work at MIT and beyond is devoted to accelerating that future.
“We are at a second turning point” Gómez-Bombarelli says. “The first was around 2015 with the primary wave of representation learning, generative AI, and high-throughput data in some areas of science. These are among the techniques that I first dropped at my lab at MIT. I believe now we’re at a second tipping point where we’re mixing language and merging multiple modalities into general scientific intelligence. We could have all of the model classes and scaling laws we want to reason about language, reason about material structures, and reason about synthesis recipes.”
Gómez Bombarelli's research combines physics-based simulations with approaches equivalent to machine learning and generative AI to find latest materials with promising real-world applications. His work has led to latest materials for batteries, catalysts, plastics and organic light-emitting diodes (OLEDs). He has also co-founded several firms and served on scientific advisory boards for startups using AI in drug discovery, robotics, and more. His newest company, Lila Sciences, is working to construct a scientific superintelligence platform for the life sciences, chemical and materials science industries.
All of this work is meant to be sure that scientific research in the long run runs more easily and productively than research today.
“AI for science is one of the vital exciting and impressive applications of AI,” says Gómez-Bombarelli. “Other applications for AI have more drawbacks and ambiguities. AI for science is about advancing a greater future in a timely manner.”
From experiments to simulations
Gómez-Bombarelli grew up in Spain and was interested by science from an early age. In 2001, he won a Chemistry Olympiad competition, launching an instructional profession in chemistry, which he studied as an undergraduate at his hometown university, the University of Salamanca. Gómez-Bombarelli stayed for his doctoral thesis, by which he examined the function of DNA-damaging chemicals.
“My doctoral work began experimentally, after which about halfway through I used to be bitten by the bug of simulation and computer science,” he says. “I began simulating the identical chemical reactions that I had measured within the lab. I like the best way programming organizes the brain; it felt like a natural approach to organize one's pondering. Programming can be much less limited by what you’ll be able to do together with your hands or with scientific instruments.”
Gómez-Bombarelli then went to Scotland for a postdoctoral position, where he studied quantum effects in biology. Through this work he made contact with Alán Aspuru-Guzik, a chemistry professor at Harvard University, whom he joined for his next postdoc in 2014.
“I used to be one in all the primary people to make use of generative AI for chemistry in 2016, and in 2015 I used to be a part of the primary team to make use of neural networks to know molecules,” says Gómez-Bombarelli. “It was the beginnings of deep learning in science.”
Gómez-Bombarelli also began eliminating manual parts of molecular simulations to conduct more high-throughput experiments. In the tip, he and his collaborators performed lots of of hundreds of calculations across materials and discovered lots of of promising materials to check.
After two years within the lab, Gómez-Bombarelli and Aspuru-Guzik founded a general-purpose materials computing company that eventually focused on producing organic light-emitting diodes. Gómez-Bombarelli joined the corporate full-time and calls it the toughest thing he has ever done in his profession.
“It was great to make something tangible,” he says. “Also, I didn't need to be a professor after seeing Aspuru-Guzik running a lab. My father was a professor of linguistics and I assumed it might be a soothing job. Then I saw Aspuru-Guzik with a gaggle of 40 and he was on the road 120 days a yr. It was crazy. I didn't think I had that sort of energy and creativity in me.”
In 2018, Aspuru-Guzik suggested Gómez-Bombarelli apply for a brand new position in MIT's Department of Materials Science and Engineering. But Gómez-Bombarelli was afraid of being hired as a lecturer and let the deadline pass. Aspuru-Guzik confronted him in his office, slammed his hands on the table and told him: “You should apply for this.” It was enough to get Gómez-Bombarelli to jot down a proper proposal.
Luckily, Gómez-Bombarelli had spent lots of time at his startup occupied with the best way to create value from computational material discovery. During the interview, he says he was drawn to the energy and collaborative spirit at MIT. He also began to understand the research opportunities.
“Everything I did as a postdoc and in the corporate was a part of what I used to be in a position to do at MIT,” he says. “I used to be making products, and I can still do this. Suddenly my work universe was a subset of this latest universe of things I could explore and do.”
It has been nine years since Gómez Bombarelli joined MIT. Today, his lab focuses on how the composition, structure and reactivity of atoms affect material performance. He has also used high-throughput simulations to create latest materials and helped develop tools for combining deep learning with physics-based modeling.
“Through physical simulations, data and AI algorithms turn into higher the more data you make available to them,” says Gómez Bombarelli. “There are all types of virtuous cycles between AI and simulations.”
The research group he arrange works exclusively computationally – it doesn’t perform any physical experiments.
“It’s a blessing because we now have lots of bandwidth and might do lots of things at the identical time,” he says. “We love working with experimenters and attempting to be good partners with them. We also love developing computational tools that help experimenters triage AI's ideas.”
Gómez-Bombarelli also still focuses on the real-world applications of the materials he invented. His lab works closely with firms and organizations equivalent to MIT's Industrial Liaison Program to know the fabric needs of the private sector and the sensible hurdles to business development.
Accelerating science
As enthusiasm for artificial intelligence has exploded, Gómez-Bombarelli has seen the sector mature. Companies like Meta, Microsoft and Google's DeepMind now repeatedly run physics-based simulations paying homage to what he worked on in 2016. In November, the US Department of Energy launched the Genesis mission to speed up scientific discovery, national security and energy dominance using AI.
“AI for simulations has evolved from something which may work to a scientific consensus view,” says Gómez-Bombarelli. “We're at a tipping point. People think in natural language, we write papers in natural language, and it seems that these large language models that do natural language have opened up the potential of accelerating science. We've seen scaling work for simulations. We've seen scaling work for language. Now we're going to see how scaling works for science.”
When he first got here to MIT, Gómez-Bombarelli says he was overwhelmed by how unrivaled things were between the researchers. He tries to bring the identical positive-sum pondering to his research group, which consists of about 25 graduate students and postdocs.
“We’ve naturally turn into a extremely diverse group with different mentalities,” Gomez-Bombarelli says. “Everyone has their very own profession goals and strengths and weaknesses. Figuring out the best way to help people turn into the most effective version of themselves is fun. Now I'm the one insisting that individuals apply for faculty positions after the deadline. I believe I've passed that baton.”
