In Chemistry, Artificial Intelligence (AI) Beats Humans Again
PUBLISHED
2024-01-27
Tireless autonomous work, a week to complete months of doctoral research, “AI chemist” on Science.
An autonomous chemical synthesis AI robot called “RoboChem” not only outperforms human chemists in speed and accuracy, but also shows a high degree of ingenuity.
The research paper, titled “Automated self-optimization, intensification, and scale-up of photocatalysis in flow,” was published in the prestigious scientific journal Science The paper was published in the prestigious scientific journal Science.
RoboChem is described as a precise and reliable “AI chemist” that not only works autonomously around the clock to deliver results quickly and tirelessly, but also rapidly optimizes the process of chemical synthesis and performs a wide range of reactions while generating minimal waste, promising to dramatically accelerate the chemical discovery of molecules used in pharmaceuticals and many other applications. applications, the chemical discovery of molecules.
Timothy Noël, corresponding author of the paper and a professor at the University of Amsterdam, said,
“Within a week, RoboChem can optimize the synthesis of about ten to twenty molecules, which would normally take a Ph.D. student several months.”
Notably, RoboChem not only discovered photocatalytic reactions that require very little light, but also successfully reproduced the findings from four randomly selected papers, producing better results in about 80% of cases.
In response, Noël says, “This leaves me in no doubt that AI-assisted methods will benefit chemical discovery in the broadest sense.”
Why is RoboChem so great?
Traditional chemical synthesis typically takes a long time, with the exact time largely dependent on the complexity of the synthetic target, the number of reaction steps, and the choice of reaction conditions.
In general, synthesizing a new organic molecule can take days or even months. This is because conventional synthesis is usually performed manually, requiring a time-consuming series of steps to continuously optimize reaction conditions, purify the product, and analyze the structure.
In conventional organic synthesis, chemists must design experimental protocols based on experience and literature, and conduct a series of trials to find the optimal conditions. The process requires repeated adjustments and optimization. In addition, the traditional synthesis process carries the risk of human error.
RoboChem, which is controlled by open-source components and simple IoT devices, but equipped with an AI brain, is a good solution to the problem of efficiency in chemical synthesis.
Autonomous Chemical Synthesis AI Robot RoboChem
According to the paper description, RoboChem can continuously learn and optimize the reaction conditions with the help of Bayesian optimization algorithms to ultimately achieve the best synthesis results. This efficient autonomous learning ability enables it to outperform other autonomous synthesis machines in a short period, which not only increases the synthesis speed but also greatly improves the efficiency of the reaction.
In addition, compared with traditional reactors, RoboChem uses a flow chemistry system that not only reduces the size of the reaction system and the generation of waste, but also improves the controllability of the experiments, and allows for more precise control of the reaction conditions while performing multi-step reactions, which in turn improves the purity of the products.
Moreover, RoboChem is not only an experimental equipment, but also a fully automated data collection and analysis system. In each experiment, RoboChem can record a large amount of data and provide timely feedback to the AI brain, enabling it to more accurately assess the advantages and disadvantages of different reaction conditions. This comprehensive approach to data processing makes RoboChem’s optimization process more scientific and reliable.
Surprisingly, RoboChem is also unusually creative. The team notes that in some experiments, RoboChem chose reaction conditions that required only weak light. This not only improved the selectivity of the reaction, but also significantly reduced the energy consumption of the photocatalytic process.
This creative choice of reaction pathways challenges the thinking of traditional chemists and demonstrates the great potential of robots in optimizing chemical reactions.
How does RoboChem work?
RoboChem works based on flow chemistry and artificial intelligence (AI), with specific work on photocatalysis optimization, replication, and scalability, as well as automated sampling and mixing and real-time nuclear magnetic resonance (NMR) analysis. Below:
Flow Chemistry Platform: RoboChem employs a flow chemistry system that replaces traditional step-by-step manual operations. In flow chemistry, chemical reactions take place in micro-continuous flow channels, which helps to control the reaction conditions more precisely and increase the reaction efficiency.
Automated Sampling and Mixing: With an automated liquid handling system, RoboChem can accurately sample different reagents and then mix them in the microreactor. This ensures that the reagents are accurately proportioned and avoids the errors that can occur in manual operations.
Photocatalytic reactions: Photocatalytic technology is used in RoboChem to initiate chemical reactions by exciting a photocatalyst with a powerful LED light source. This method is particularly useful when synthesizing organic molecules, allowing for highly selective and efficient reactions.
Real-time NMR analysis: During the reaction, RoboChem monitors the progress of the reaction using real-time NMR, which provides structural information about the reactants and products, helping to determine how the reaction is proceeding.
AI Algorithms: The “brain” of RoboChem is a computer system driven by AI algorithms. The system continuously optimizes the reaction conditions through machine learning, adjusting parameters based on real-time NMR data to achieve the best synthesis results. This allows RoboChem to learn and adapt to different chemical reactions on its own.
Automated Robotics Platform. a. High-level view of the platform architecture; b. Reaction tracking on the platform by phase sensors, leading to timely triggers, which can be tracked by the sensors as the reaction segment plugs pass through the phase sensors and studied by the algorithms in order to form triggers to be used in the next optimization cycles.
Although RoboChem demonstrated impressive autonomous learning and chemical optimization capabilities, the study still has some limitations:
First, RoboChem used specific flow reactors and NMR equipment during the experiments. The design and capacity of these devices may impose certain limitations on reaction conditions and experimental scale, making certain types of reactions or large-scale production difficult to realize.
Second, since the training of machine learning algorithms is based on pre-existing datasets, RoboChem may not be able to provide optimal reaction conditions in some special cases, especially when complex or rare chemical reactions are involved, and the generalization ability of the machine learning model may be challenged.
Moreover, the performance of RoboChem depends on the quality and purity of the chemicals used in the experiment, such as photocatalysts and reactants. If there are discrepancies in these chemicals, the accuracy and reproducibility of the experimental results may be affected.
It is important to note that AI algorithms are often considered “black boxes,” meaning that their decision-making processes are difficult to interpret. In some cases, it may be difficult for researchers to understand the specific chemical logic behind the optimization conditions provided by RoboChem.
Finally, the results of RoboChem’s research so far have focused on specific types of photocatalytic reactions, and their applicability and generalization to other fields need to be more practiced and verified.
More Than Autonomous Chemical Synthesis
AI-powered robots can not only perform chemical synthesis on their own, but can also assist and even surpass humans in chemistry and the sciences in general in different ways.
For example, Coscientist, which appeared in the journal Nature last December, was able to reproduce a Nobel Prize-winning study in just a few minutes after a single attempt, powered by GPT-4. This research shows that it is possible for humans to effectively utilize AI to increase the speed and volume of scientific discovery and improve the reproducibility and reliability of experimental results. (Click to view original article)
Separately, an AI lab called A-Lab, also featured in Nature, saw artificial intelligence (AI) create 41 new materials on its own in just 17 days. In contrast, human scientists can take months of trial and error to create a new material. (Click for original article )
In addition, automated synthesis flow platforms, intelligent chemical synthesis systems, and high-throughput experimental methods are among the ways to automate chemistry labs.
These researches have pushed the automation and intelligence development in the field of chemical synthesis in different directions, providing new ideas and methods to improve experimental efficiency and discover compounds in new fields.
Taken together, RoboChem realizes a highly automated chemical synthesis process by integrating flow chemistry and AI algorithms. This approach not only improves the efficiency of synthesis, but also reduces human error in experiments, providing a new, faster and more controllable approach to chemical research.
But in Noël’s view, the significance of RoboChem, along with other “computerized” chemical systems, lies in the generation of high-quality data, which will be beneficial to AI applications in the future.
In the future, with the successful application of RoboChem, similar autonomous synthesis machines are expected to be widely used in drug synthesis, new material development and other fields, accelerating the process of scientific research.
At the same time, the strategy of combining machine learning and flow chemistry will also become a new trend in the field of chemistry, providing new possibilities for autonomous optimization of more complex reactions.
