AI to Transform Science: MIT Professor Rafael Gómez-Bombarelli's Vision for Materials Discovery
MIT Associate Professor Rafael Gómez-Bombarelli has utilized artificial intelligence for over a decade in the creation of new materials. He now holds a tenured position in materials science and engineering and indicates that AI is positioned to transform science. His work focuses on accelerating this future.
Current Advancements and the "Second Inflection Point" in AI
Gómez-Bombarelli describes a "second inflection point" in AI, following an initial wave around 2015 that involved representation learning, generative AI, and high-throughput data in science. The current phase integrates language and multiple modalities into general scientific intelligence, enabling reasoning across language, material structures, and synthesis recipes.
His research combines physics-based simulations with machine learning and generative AI to discover new materials. This work has led to advancements in materials for batteries, catalysts, plastics, and organic light-emitting diodes (OLEDs).
Gómez-Bombarelli has co-founded companies and served on scientific advisory boards for startups applying AI to drug discovery and robotics. His company, Lila Sciences, aims to build a scientific superintelligence platform for life sciences, chemical, and materials science industries. He states that AI for science is a positive application designed to advance the future of scientific research.
Academic Journey and Early AI Innovations
Gómez-Bombarelli's academic path began in Spain, where he studied chemistry and later pursued a PhD investigating DNA-damaging chemicals. During his PhD, he transitioned from experimental work to simulation and computer science.
He completed postdoctoral positions in Scotland, studying quantum effects in biology, and at Harvard University with Professor Alán Aspuru-Guzik. During this time, he was among the first researchers to apply generative AI to chemistry in 2016 and use neural networks to understand molecules in 2015. He also contributed to developing high-throughput experiments by eliminating manual parts of molecular simulations, leading to the discovery of numerous promising materials.
Following his postdoc, Gómez-Bombarelli co-founded a materials computation company that later focused on organic light-emitting diodes, an experience he describes as challenging.
Research at MIT: Merging Physics and AI
In 2018, Gómez-Bombarelli joined MIT's Department of Materials Science and Engineering.
His lab's current research investigates how the composition, structure, and reactivity of atoms influence material performance.
He utilizes high-throughput simulations to create new materials and develops tools that merge deep learning with physics-based modeling. He highlights a virtuous cycle where physics-based simulations improve AI algorithms by providing more data.
His research group is solely computational and collaborates with experimentalists to develop computational tools that assist in triaging AI-generated ideas. The lab also works with industry partners, such as MIT's Industrial Liaison Program, to address material needs and practical commercial development challenges.
Accelerating the Future of Science
Gómez-Bombarelli notes the maturation of AI in science, with major companies now conducting physics-based simulations. He points to initiatives like the U.S. Department of Energy's Genesis Mission, which aims to accelerate scientific discovery using AI.
He emphasizes that large language models, having mastered natural language, have opened new avenues for accelerating science.
He anticipates that scaling principles, which have been effective for simulations and language, will similarly apply to science. Gómez-Bombarelli promotes a collaborative environment within his research group, which includes approximately 25 graduate students and postdocs, focusing on fostering their individual aspirations and strengths.