Researchers have created mobile robots powered by artificial intelligence (AI) that can conduct chemical synthesis research with remarkable efficiency. A new study illustrates how these AI-enabled robots can perform exploratory chemistry tasks at a speed comparable to that of human researchers, but with significantly quicker results.
Researchers at the University of Liverpool have developed AI-driven mobile robots that can carry out chemical synthesis research with extraordinary efficiency.
In an article published in the journal Nature, the researchers demonstrate that mobile robots utilizing AI decision-making are capable of performing exploratory chemistry tasks at a level similar to human researchers, but at an accelerated pace.
The 1.75-meter-high robots were crafted by the Liverpool team to address three key challenges in exploratory chemistry: executing reactions, analyzing the products, and determining the next steps based on the findings.
Working together, the two robots tackled issues across three distinct areas of chemical synthesis: structural diversification chemistry (important for drug discovery), supramolecular host-guest chemistry, and photochemical synthesis.
The findings indicate that with their AI capabilities, the mobile robots were able to make decisions akin to those made by human researchers, but they did so much more quickly—whereas human decision-making might take hours.
Professor Andrew Cooper from the University of Liverpool’s Department of Chemistry and Materials Innovation Factory, who spearheaded the project, stated:
“Chemical synthesis research is both time-consuming and costly, involving not just the physical experiments but also the decision-making process regarding subsequent experiments. Employing intelligent robots presents a solution to speed up this process.”
“When people visualize robots in chemistry automation, they often think of tasks like mixing solutions and heating reactions. While those are elements of it, the decision-making aspect can be just as demanding. This is especially true in exploratory chemistry, where outcomes are uncertain and decisions about what is interesting depend on multiple datasets. This complexity makes it a challenging task for both researchers and AI.”
Decision-making represents a critical challenge in exploratory chemistry. For instance, a researcher might conduct various trial reactions and choose to scale up only those with favorable yields or intriguing products. AI struggles with this process since determining what is ‘interesting’ can vary based on context, including the novelty of the reaction product and the complexities of the synthesis route.
Dr. Sriram Vijayakrishnan, a former PhD student at the University of Liverpool and a Postdoctoral Researcher involved in the synthesis work, shared: “During my PhD, many chemical reactions were performed manually. Collecting and analyzing analytical data often took as much time as preparing the experiments. This data analysis becomes even more challenging with automation, leading to an overwhelming amount of data.”
“We addressed this by developing an AI logic for the robots that processes analytical datasets to autonomously decide whether to move on to the next step in a reaction. This decision is made almost instantly, so if a robot completes its analysis at 3:00 AM, it will determine which reactions to pursue by 3:01 AM. In contrast, a chemist could spend hours working through the same datasets.”
Professor Cooper noted: “Although the robots lack the extensive contextual understanding of an experienced researcher and won’t have a breakthrough moment, for the tasks assigned to them, the AI logic produced decisions very similar to those of a synthetic chemist across three different chemistry scenarios, making those decisions in the blink of an eye. There’s significant potential to enhance the AI’s contextual awareness by connecting it to relevant scientific literature via large language models.”
Looking ahead, the Liverpool team aims to apply this technology to discover chemical reactions vital for pharmaceutical drug synthesis and to develop new materials for applications like carbon dioxide capture.
While the study involved two mobile robots, the scalability of this approach could potentially accommodate larger teams of robots, making it suitable for extensive industrial laboratories.
