Plant blindness and diversity in AI language models.

Journal: Trends in plant science
Published Date:

Abstract

Large language models (LLMs) will benefit science by accelerating task performance. We explored whether answers generated by ChatGPT (generative pretrained transformer) to questions of biology are sufficiently diverse. 'Plant awareness' in ChatGPT answers was found to be highly variable, illustrating the importance of scientists being involved in validating the data and methods used to train artificial intelligence (AI) models.

Authors

  • Anja Geitmann
    Department of Plant Science, McGill University, Ste-Anne-de-Bellevue, QC, Canada. geitmann.aes@mcgill.ca.
  • Amir J Bidhendi
    Department of Plant Science, McGill University, Ste-Anne-de-Bellevue, QC, Canada.