Biodiv Sci ›› 2024, Vol. 32 ›› Issue (9): 24258.  DOI: 10.17520/biods.2024258  cstr: 32101.14.biods.2024258

• Technology and Methodology • Previous Articles     Next Articles

Application of large language models in biodiversity research

Jiqi Gu1,4, Jianping Chen2, Jiangshan Lai3,4,*()   

  1. 1. Ministry of Education of Key Laboratory for Biodiversity Science and Ecological Engineering, Beijing Normal University, Beijing 100875, China
    2. Chenshan Botanical Garden, Shanghai 201602, China
    3. College of Ecology and the Environment, Nanjing Forestry University, Nanjing 210037, China
    4. Research Center of Quantitative Ecology, Nanjing Forestry University, Nanjing 210037, China
  • Received:2024-06-25 Accepted:2024-08-09 Online:2024-09-20 Published:2024-08-09
  • Contact: * E-mail: lai@njfu.edu.cn
  • Supported by:
    National Natural Science Foundation of China(32271551)

Abstract:

Background & Aims: With the development and advancement of artificial intelligence technology, large language models (LLMs), such as Kimi Chat, have begun to play a significant role in biodiversity research. LLMs’s deep learning and natural language processing technologies, augmented by human feedback reinforced learning (RLHF) and proximal policy optimization (PPO), offer new avenues for handling and analyzing large biodiversity data sets.
Progresses: We explore the application of LLMs, taking Kimi Chat as an example, in investigating biodiversity research questions, reviewing literature, designing hypotheses, organizing and analyzing data, and writing research papers, as well as its potential to enhance research efficiency and quality. (1) LLMs can quickly process vast amounts of scientific literature, helping researchers distill key information and swiftly catch up with the latest research trends in specific fields. (2) LLMs can also assist researchers in formulating research hypotheses and designing experimental protocols, thereby providing abundant scientific inspiration, broadening research perspectives, and enhancing the efficiency of the initial stages of research. (3) In terms of research design, LLMs can offer advice on data collection methods, design of experiment, and statistical analyses to ensure the scientific validity and the logic of the research design. (4) LLMs can assist in scientific writing and peer review processes by helping draft scientific papers and providing suggestions for revision and polishing to enhance the quality and readability of the papers, and it also supports researchers in understanding and responding to peer review comments and optimizing the presentation of research findings. We also discuss the challenges and limitations encountered during using LLMs, such as the need for professional judgment, the homogenization of research methods, the accuracy of data and results, and ethical issues. Additionally, we propose strategies for integrating this technology with traditional biodiversity research methods in the future.
Prospects: We demonstrates how LLMs can aid in biodiversity research, thus advancing scientific discovery and ecological conservation strategies.

Key words: large language models, biodiversity research, scientific writing, research design, research methods