top of page

Agora você conta com um assistente de leitura, o chatGPT responde suas perguntas sobre a postagem que está lendo. Somente abra o chat abaixo e faça perguntas!

Writer's pictureJorge Guerra Pires

Papers accepted with responses from ChatGPT: copy and paste, and publish

Updated: Mar 17








It may sound like a joke, but it is not. There are papers being published with outputs from ChatGPT. The first reaction people have is: those are predatory journals, those are third-world journals, those are low-impact journals. Those are the first to be blamed. Predatory journals are from third-world, developed countries are above any possible corruption. Afterall, a main stream journal by a developed country would never do that. Nope, it is a major journal from Elsevier with IF of 6, and a fee of almost 3.000 USD. Elsevier has been showing to profit almost as much as Apple and Google. And they charge according to the IF.


It is hard to guess what happened, two cases below. It is possible someone was trying to make a point (paying 3.000 USD?). It is possible after-acceptance edition (but two papers found?). It is hard to know what happened. Any attempt to guess is speculations. All the papers I have published, they have found is minor typos on the text, now image, a whole paragraph of clear copy and paste from ChatGPT.


It is important to mentioned that SciGen already did that much before ChatGPT: several papers were accepted, even on major journals. SciGen creates random papers, with text and images. It is garbage if you read. They were trying to prove a point: the peer review system is inefficient.


It is possible to do much better than the cases below, and I have a tool here called RefWiz, which I have created and I have used on my papers, and Ideclare that, that can do a much better job at creating paragraphs for papers with references, with citations. One example here. openAI stopped providing a detector since their accuracy was low, even though people are still providing solutions online. If you take a text from AI, and make minor changes, it will undetectable.


There is the possibility that the authors are not English speakers, different from native speakers. Thus, they do not understand what was added to the paper. It is still hard to believe that would pass. Again, I have heard a lot that I am not a native speaker. It could be because of the name; it could be that the reviewers read the papers I have submitted.



I do not believe it is a mistake to use AI on writing papers, it is the future. The mistake is copy and paste. Several journals are already requiring a statement that IA was used, and how much. Taking out the extreme cases as cited (copy and paste), the boundaries are fuzzy. If you make it right, it is impossible to detect you have used AI on your paper.



 


Reading suggestion



 













240 views0 comments

Comments


Suporte
Apoiar
bottom of page