Scientific journals have begun to change their editorial policy, banning the popular ChatGPT chatbot as a co-author of articles. Such changes, in particular, were announced by the chief editors of Science and Nature. However, some journals will allow the use of ChatGPT when preparing texts of manuscripts, if the authors do not hide information about it, writes The Guardian.
The ChatGPT chatbot is a generative language neural network developed by the American company OpenAI based on the GPT-3.5 language model. ChatGPT was launched on November 30, 2022, after which it quickly attracted the attention of a wide audience due to its ability to perform tasks related to different areas. Among other things, this chatbot can not only answer questions, but also work with code, write scripts, songs and compositions, and give advice.
Recently, an editorial on the Nature website reported that some researchers have begun listing ChatGPT as co-authors of preprints and papers. So, journalists managed to find at least four confirmations of this, for example, on the medRxiv preprint service and in the peer-reviewed journal Oncoscience, which resulted in a discussion as to whether it is acceptable to indicate a neural network as the author of the work, which cannot be held responsible for the content of the written text, says N+1.
Now, the British edition of The Guardian spoke about the attitude of publishers and individual scientific journals. Thus, the newspaper drew attention to a change in the editorial policy of the well-known journal Science, which was announced by its editor-in-chief Holden Thorp. He noted that the journal will not publish articles that list ChatGPT as one of the authors. In addition, Thorp opposed the use of texts generated by the chatbot in scientific manuscripts, since ChatGPT makes numerous errors.
Springer-Nature, which publishes about 3,000 journals, also banned ChatGPT from being credited as the author of the work. However, this tool (or one similar to it) can be used in the preparation of the text if the researchers do not hide information about it. So, the editor-in-chief of Nature Magdalena Skipper believes that the limited use of neural networks can be useful for science. For example, neural networks can help foreigners prepare articles in English. A similar position was taken by the Elsevier publishing house, which, in particular, publishes the well-known magazines Cell and the Lancet, notes NIXsolutions. And eLife Editor-in-Chief Michael Eisen stressed that authors should be honest about how they used different tools and remember that they are responsible for the results.