COLOSS B-RAP expert evaluation of beekeeping advice from ChatGPT, part 2

Lotta Fabricius Kristiansen, Linde Morawetz, Robert Chlebo, Alison Gray, Ewa Mazur, Bjørn Dahle, Ana Diéguez-Antón, Georg Schaunitzer, Kristina Gratzer, Hannes Oberreiter, Dirk Louis P. Schorkopf, Johann Fischer, Robert Brodschneider

Research output: Contribution to journalArticle

1 Downloads (Pure)

Abstract

The artificial intelligence (AI) chatbot or advanced language model (LLM) ChatGPT is capable of understanding and generating human-like text. This article is a sequel to our first article on this topic, in which we imagined an AI tool that would serve as a beekeeping advisory tool in the future after having been trained with information stemming from high quality public domain resources on beekeeping and bee diseases (Morawetz et al., Citation2024). In that article, we presented the first part of the COLOSS B-RAP AI challenge, in which we asked ChatGPT-3.5 for advice regarding the serious honey bee disease American Foulbrood, which is a notifiable disease in many countries. We concluded that the tested AI chatbot version is a good teacher for basic or general beekeeping knowledge. It summarizes common knowledge about bee diseases at a level of detail that can be found in general beekeeping textbooks or in a beginners’ course in beekeeping. As a summarization tool, it was, for example, used to write the recently published book on bee diseases by Walker (Citation2024) to provide a comprehensive and structured overview of the topic. Nonetheless, it was concluded by the expert panel that advice from ChatGPT should be avoided for critical situations needing immediate action by the beekeeper, such as notifiable or severe diseases (Morawetz et al., Citation2024). Worryingly, part of the advice given was identified as misleading, wrong, or of limited usefulness in the given context and may easily lead to the spread of notifiable diseases and to severe consequences for the beekeeper and the beekeeping neighbors.
Original languageEnglish
Pages (from-to)1-3
Number of pages3
JournalBee World
Early online date28 Oct 2024
DOIs
Publication statusE-pub ahead of print - 28 Oct 2024

Keywords

  • ChatGPT
  • apiculture
  • beekeeping
  • American Foulbrood

Fingerprint

Dive into the research topics of 'COLOSS B-RAP expert evaluation of beekeeping advice from ChatGPT, part 2'. Together they form a unique fingerprint.
  • COLOSS B-RAP expert evaluation of beekeeping advice from ChatGPT, part 1

    Morawetz, L., Fabricius Kristiansen, L., Brus, J., Dahle, B., Danihlík, J., Diéguez-Antón, A., Fischer, J., Gray, A., Mazur, E., Schaunitzer, G. & Brodschneider, R., 14 May 2024, In: Bee World. 101, 1, p. 2-4 3 p.

    Research output: Contribution to journalArticle

    Open Access
    File
    7 Downloads (Pure)

Cite this