Close this search box.
Close this search box.

Tailored AI language models could transform research process

Artificial intelligence has the potential to radically improve the research process, according to academics from Duke Kunshan and other universities.

Used in large language models tailored to specific subjects it can accurately sift through vast amounts of scattered information, shortening research time and making swift and accurate insights more likely, according to their paper, published in the science journal Cell Reports, Physical Science.

“Harnessing the power of AI-powered language systems correctly could usher in a new era of efficiency and innovation in many research fields,” said Xin Li, a professor of electrical and computer engineering at DKU, who co-authored the paper.

Professor Xin Li

Titled, “Potential to transfer words to watts with large language models in battery research,” it focuses on the use of AI-powered language systems in the search for improved energy storage technology, used in industries such as electric vehicles and solar power.

General large language models lack the specialized knowledge to answer professional queries about this type of technology, which can lead to potential inaccuracies in query answers. However, tailored models that look for information from specific designated sources can overcome this limitation, they said.

The paper proposes a large language model called BatteryGPT that would source information from a fast-charging database, a repository of research, lectures and other quality information specific to this area of research.

“While large language models excel at addressing general questions, BatteryGPT would be capable of offering expert-level responses, accelerating and enriching the research process,” said Li.

Other researchers involved in producing the paper included: Shuo Zhao and Yang Liu from DKU’s Data Science Research Center; Sihui Chen, Chao Li and Jiayu Wan, from the Global Institute of Future Technology at Shanghai Jiao Tong University; Jiayi Zhou and Tan Tang from the Laboratory of Art and Archaeological Image at Zhejiang University; and Stephen Harris from the Lawrence Berkely National Laboratory at Berkeley.

If you are a journalist looking for information about the University or for an expert to interview for a story, our team can help.

Add our