The artificial intelligence (AI) chatbot ChatGPT that has taken the world by storm has made its formal debut in the scientific literature, pulling up at least four authorship credits on published papers and preprints.
Editors, researchers, and publishers are currently debating whether such AI tools should be included in published literature and whether the bot should be cited as an author. Publishers are thinking of policies to be created for chatbots like chatgpt.
Research publications and preprint servers stated that AIs like ChatGPT don’t meet the criteria for study authors. They cannot assume responsibility for scientific papers. Some publishers claim that an AI’s contributions to writing papers can be recognized in sections other than those on the author list.
ChatGPT is one of 12 authors of a preprint 1, and it was about using the tool in medical education. It was published in the medical repository MedRxiv on December 2022.
Richard Sever, the co-founder of Cold Spring Harbor Laboratory Press in New York, said that the team behind the repository and bioRxiv are discussing whether crediting AI tools like ChatGPT should be used when writing research. He says that conventions could change.
Sever says we should distinguish the formal role of an author of a scientific manuscript from the more general notion of an author as the writer of a paper. He says that authors take on legal responsibility for their work, so only people should be listed. People may try to sneak it in, as has happened at MedRxiv, just like people have listed fictional people, pets, etc. As authors on journal articles in the past, that’s a checking issue instead of a policy issue.
This month’s editorial 2 in Nurse Education in Practice credits AI as a co-author and Siobhan O’Connor, a UK health-technology researcher, in its paper. Roger Watson, editor-in-chief of the journal, claims this credit was wrong and will be rectified soon. He says it was my oversight, as editorials are subject to a different management system than research papers.
Alex Zhavoronkov is the chief executive of Insilico Medicine. Insilico Medicine is an AI-powered drug discovery company in Hong Kong.
Alex credited ChatGPT as a co-author of a perspective paper 3 published in Oncoscience last month. He says that his company has published over 80 papers using generative AI tools. He says, “We aren’t new to this field.”
In the context of Pascal’s wager, the latest paper examines the pros and cons of taking Rapamycin. Zhavoronkov says ChatGPT produced a better article than the previous generation of generative AI tools. After requesting its editor, he said that Oncoscience had peer-reviewed the paper.
Almira Osmanovic Thunstrom (a neurobiologist at Sahlgrenska University Hospital, Gothenburg) said that an article 4, which was co-authored by an older chatbot called GPT-3, was posted on the French preprint server HAL June 2022. It will soon be published in peer-reviewed journals. After reviewing the paper, one journal rejected it. However, a second accepted the article with GPT-3 as the author. She rewrote the article to meet reviewer requests.
Holden Thorp, the chief of Science Family for Journals, stated that using AI-generated texts without proper citations could be considered plagiarised content, and no publication should accept it.
Sabina Alam, the director of publishing ethics & integrity, said that Taylor & Francis in London is currently reviewing its policy. She agreed that authors are responsible for ensuring the integrity and validity of their work and should mention any LLMs used in the acknowledgments section. Taylor & Francis has not yet received submissions that credit ChatGPT co-author.
Steinn Sigurdsson (scientific director at Pennsylvania State University, University Park) says that the board of physical-sciences preprint servers arXiv has been having internal discussions and is starting to agree on a strategy for using generative AIs. He agreed that software tools could not be authors of submissions, partly because they cannot consent to terms and conditions of use. Sigurdsson says that he isn’t aware if any arXiv preprints list ChatGPT among the co-authors. He also states that new guidance is being developed for authors.