"AI could soon create chemical and catalytic reactions named after itself"

Bartosz GRZYBOWSKI is widely known as a group leader at the IBS Center for Soft and Living Matter, as well as a pioneer in the field of nano and micro chemical systems and an expert in computer-based synthesis.

He has published over 300 papers, including more than 20 in top-tier journals like Nature and Science, with his work cited over 37,000 times. Recently, he was appointed as the director of the Center for Algorithmic and Robotized Synthesis.

I had the opportunity to speak with Director Grzybowski at his center, located within the Ulsan National Institute of Science and Technology (UNIST), to hear about his future plans.

Congratulations on your new appointment as the director of the IBS Center for Algorithmic and Robotized Synthesis. Why this specific research center?

Modern science is advancing through the use of data, with AI (artificial intelligence) being a prime example of this trend. I believe that AI is a powerful tool that can help us explore the laws of nature that humans have not yet discovered. It’s a new revolution to advance science, and AI is already being actively used in various fields. For example, AI tools like AlphaFold are used to study protein structures and reactions.

When proposing this new research center to IBS, I thought it would be beneficial to reflect these academic trends. Around the world, efforts to use AI to discover new chemical reactions are already beginning. Research teams in the U.S., Canada, Germany, and the Netherlands are being formed.
As for Asia, our research center might be the first.

When did you first become interested in AI?

Around 20 years ago, I started to believe that AI would become important in chemistry. I was considering ways to make new discoveries in chemistry without relying on serendipity. I thought it would be possible to predict chemical reactions using algorithms. From 1995 to 2003, I was at Harvard University, where I interacted with people who were developing chess AI. I figured that if AI could learn to play chess, it could also learn about chemical molecules. This led to my first paper in 2005.

I’m curious how AI is related to robots.

AI suggests the direction for conducting chemical experiments. Since the number of possible molecules far exceeds the number of atoms in the universe, trying experiments randomly would make it nearly impossible to find interesting results. Before the idea of AI came along, scientists tried automating experiments with simple programs, but this approach didn’t work well because it was essentially random.

The key is for AI to guide the robot on which experiments to conduct. When the robot produces results, AI learns from them and suggests new directions. This creates a cyclical interaction between the two.

What kind of outcomes do you aim to achieve with this process?

Our research center focuses on discovering new chemical reactions and catalysts. We study the characteristics of chosen materials, experimental conditions, and nearly every aspect using AI and robots. When a new reaction or catalyst is found, researchers often name it after the discoverer, like the "Suzuki reaction" (2010 Nobel Prize in Chemistry) or "Grubbs catalyst" (2005 Nobel Prize in Chemistry). With AI, we can discover so many reactions that attaching a person’s name to each one would lose its meaning. We attribute the credit to AI instead. In a recent paper, we named newly discovered reactions MACH-1 and MACH-2, where MACH refers to "machine," representing AI.

In 2024, we can't afford to conduct experiments like in the 19th century, one at a time. Robots help speed up this process—they prepare, conduct, and report on chemical reactions. A postdoctoral researcher can perform about three experiments a day, but with robots, you could do up to 1,000 per day. More experiments allow us to explore fundamental questions about chemical reactions.

Moreover, using AI and robots reduces the resources needed for experiments. A single successful experiment might require hundreds of failed attempts, consuming large amounts of solvents and toxic materials. When AI determines the experiment’s direction, robots can achieve maximum results with minimal waste. This conserves resources and has significant environmental benefits.

If AI plans the experiments and robots carry them out, what role do scientists play?

First, I want to clarify that this hasn’t fully happened yet. And I don't believe that scientists will ever become obsolete, no matter how advanced AI gets. Think back to when steam engines were first invented—people feared they would take away jobs, to the point that they even started destroying machines. But jobs didn’t disappear, they changed. The same will happen in research. With faster ways to develop reactions and catalysts, we’ll be able to ask new questions. AI is a tool that helps us discover more knowledge, but it’s still humans who decide whether or not to follow AI’s suggestions.

Take catalysts, for example. Catalysts are crucial because they reduce energy and resource consumption, and most were discovered by chance over the past 400 years. Should we continue to rely on luck, as in the 19th century, to find new catalysts? As science and technology advance, it’s equally important to make good use of newly developed tools.

How do you develop the robots used at your research center?

The field we’re entering is entirely new, so naturally, no ready-made robots exist for this purpose. Our young researchers collaborate to design and build these robots from scratch. They code the movements and actions required for experiments and assemble the machines themselves. In essence, to study the new field of chemical AI and robotics, we are designing and creating our own AI tools and experimental robots.

It appears you will need people with multiple talents? How are you looking to form your research center?

We’re looking for chemists, but ideally, they should also have an interest in AI and robotics and know how to code. The Center for Algorithmic and Robotized Synthesis is a newly established research group, so there’s a lot to prepare. We're planning to form three groups with about 60–70 people. Our goal is to create the first and best research center for AI- and robot-based chemistry in the world. I want to make it so attractive that even if Korean researchers have the chance to go to MIT in the US or the Max Planck Institute in Germany, they would still choose to come here.

1. The robots used in the research center are programmed and assembled by the researchers themselves.
2. These robots can conduct large volume of experiments at once, paving the way for scientists to explore new ideas.

What kind of personality are you specifically looking for?

We are looking for passionate individuals with an innovative mindset. I hope for people who can create a new research culture with us, regardless of past methods. Currently, our research team consists of people from 9 to 10 different countries, so it's important that everyone approaches each other with respect, given our diverse backgrounds.

I especially want our team to be free from excessive rules. Too many rules create hierarchies and can stifle creative thinking. The only rule in our research center is to publish papers in prestigious journals like Nature or Science and make groundbreaking discoveries.

I hope to find people who can enjoy the beauty of science together with their fellow researchers.

Leading such a large research center must be stressful, how do you deal with that?

When stress levels peak, I exercise. I’ve invested some space to create a workout area. I even bought a ping-pong table with my own money to help the research team relieve stress. I believe it’s been the best investment I’ve made in the past six months. Playing games and chatting with team members has helped build communication and relationships.