Letters generated by ChatGPT: Author who?

Abstract

AimChatGPT's role in medical writing is a topic of discussion. I experimented whether ChatGPT almost automatically generates Correspondence or Letter addressed to a “translated” article, and thereby wish to arouse discussion regarding ChatGPT use in medical writing.MethodsI input an English article of mine into ChatGPT, tasking it with generating an English Disagreement Letter (Letter 1). Next, I tasked ChatGPT with translating the manuscript addressed to from English‐French‐Spanish‐German. Then, I once again tasked ChatGPT with generating an English Disagreement Letter addressed to a German manuscript (triplicate translated manuscript) (Letter 2).ResultsLetters 1 and 2 are readable and reasonable, shooting the point that the author (myself) felt as the weakness of the article. Letters addressed to French (single translation) and to Spanish (double translation) and longer Letters (corresponding to Letters 1 and 2) are also readable, and thus stand.ConclusionsSolely based on this experiment, one may be able to write a letter even without understanding the meaning of the paper being addressed, let alone the language of the paper. Although this humble experiment does not conclude anything, I plea for a comprehensive discussion on the implications of these findings.

Response to “Re: ‘Letters generated by ChatGPT: Author who?’”

“…Since the present study 2 was preliminary and was an original article (not an opinion), I could not "conclude" that Letter should be entirely written by humans without ChatGPT's intervention. Please look at my remarks in the last paragraph 2 : "This study strengthened my belief that humans should write personal opinions." I wished to claim this.…”

“…

Dear Editors, Thankfully, Daungsupawong and Wiwanitkit 1 provided insightful comments for my paper. 2 ChatGPT automatically generated a Disagreement Letter addressed to an English paper and also addressed to a German-translated one. 2 I stated, "Although this experiment does not conclude anything, I plead for a comprehensive discussion on the implications of these findings."

…”

“…My intention is not to guide how to improve ChatGPTgenerated letters through human intervention, but to claim letters should be authored by humans rather than relying on ChatGPT. 2,3 I never adopt a strategy: first demanding ChatGPT to generate a "Letter," then, revising it by experienced writers, peers, or myself, and thereby making a better letter "time-efficiently." I believe that "writing," at least writing one's concept (letter, opinion, or perspective), is a kind of "art."…”

“…I further experimented by inputting my original manuscript 2 and Daungsupawong and Wiwanitkit's letter 1 into ChatGPT (version 3.5). I input some different contexts between the two and ordered ChatGPT to write a Reply.…”