KIST opens unmanned nano materials laboratory

[Courtesy of KIST]

The Korea Institute of Science and Technology (KIST), a multi-disciplinary research institute in South Korea, opened a smart research laboratory that autonomously develops customized nano materials and is paving the way for new paradigms in scientific research.

Robots that synthesize nano particles roam around this smart research laboratory, multiple sources from the science and technology sector said on Sunday.

When researchers input the desired properties for the material, the robots proceed to synthesize nano particles and measure the optical characteristics of the particles to meet the required properties accurately.

“What would take a human around 10,000 attempts to achieve can be accomplished by the smart research laboratory in about 200 attempts,” a KIST researcher said. “Experiments conducted by humans often yield inconsistent results depending on the research environment or researcher, making it difficult to obtain reproducible results. But the smart research laboratory produces consistent high-quality data in large quantities.”

The key to the smart research laboratory’s groundbreaking capabilities lies in Bayesian optimization and early stopping strategies.

Bayesian optimization is a strategy used to optimize results for specific purposes that involves collecting information about the results of a specific objective, using it to select candidate points to try next, building a probabilistic model of previous attempts and then using this model to select the most promising candidate points for the next attempt.

As the name suggests, early stopping involves stopping the training of a machine learning model early. Although machine learning models show high accuracy on training data, they could suffer from decreased performance on new data, known as overfitting, and early stopping prevents this issue.

With the smart research laboratory, KIST researchers have already developed candidate materials for new substances to be applied in catalysts, semiconductors, batteries, and more.

“During the development process, we also acquired new chemical knowledge that was not previously known,” a researcher said. “Smart research laboratories are emerging as a new research and development (R&D) paradigm.”

Unmanned scientific laboratories are gaining attention in the scientific community for several reasons, one of which is its efficiency. The time taken to produce results is drastically reduced compared to when performed by humans, leading to cost savings as all that is required is an electricity supply instead of labor.

Unmanned scientific laboratories can also prevent accidents that occur when humans perform experiments, with Korea also bracing for a decrease in research manpower due to low birth rates. Global efforts to open unmanned scientific laboratories are also gaining momentum with the emergence of AI that can learn scientific knowledge and conduct research without human scientists’ instructions.

A research team at Carnegie Mellon University in the United States announced in December 2023 in the international journal “Nature” that it had developed an AI “coscientist.”

Coscientist, based on large-scale language models (LLMs) such as ChatGPT, searches and learns unknown chemical reactions online and commands robots to produce the desired compounds. When asked to perform experiments on the palladium catalytic reaction widely used in the production of organic light-emitting diodes, it searched various papers, formulated an experimental plan in four minutes, and succeeded in synthesis.

The KIST research team also began business activities based on coscientist. It has equipped several hundred experimental facilities in a facility covering about 4,600 square meters and allowed researchers from universities and companies to use the laboratory facilities for a fee. Although nobody is in the lab, the researchers can access the facilities via the Emerald Cloud Lab system and use them anytime, anywhere.