Challenges of Using ChatGPT in Nuclear Medicine Academic Writing

Sir,

I read with great interest the recent letter by Dr. Sharma which discussed a contemporary trending topic, large language models with a particular focus on ChatGPT.[1] I would like to share my opinions regarding certain statements made by the authors which I think warrant further consideration and discussion.

I agree with the author that ChatGPT has exceptional capabilities in writing as compared to humans due to its continuous learning process. However, there are some additional limitations that require careful consideration when the tool is used by researchers.

Certainly, a clear benefit lies in the increased speed of research and publication facilitated by artificial intelligence (AI) tools. This allows researchers to dedicate more time to the actual study rather than the writing phase. However, it is crucial to take into account the current academic environment, where factors such as the H-index, based on an author’s citation count, often influence tenure decisions. If scholars were to utilize AI tools solely to enhance their publication output, it could potentially overwhelm journal editors and result in an excess of AI-generated content in vital databases such as PubMed and Scopus. This content would predominantly originate from AI with minimal human involvement. The consequences of this trend on future research outcomes remain uncertain.[2]

Next, most of the time, ChatGPT is unable to correctly reference the sources of the texts it generates. Instead, there is a tendency to generate fake references. Hence, currently, in my own perspective, I only use the tool to generate outlines for articles and to paraphrase text, and if I do use it, I acknowledge its use in the acknowledgments section rather than including it as an author (there was controversy regarding this earlier).[23]

Another aspect is that these new AI tools should make us rethink the purpose of writing a thesis/dissertation and should prompt us to update our education systems, not being limited to medical education, since though there exist tools to detect AI-generated text such as DetectGPT, AI text classifier (developed by OpenAI itself, the creator of ChatGPT), GPTZero, Originality.ai, and Copyleaksno, it is unlikely that they are able to accurately detect/exclude the use of AI tools for scientific writing.[245]

In conclusion, incorporating ChatGPT into academic writing underscores the importance of a thorough examination of the ramifications of AI tools. It is crucial to establish comprehensive guidelines that prioritize academic integrity, transparency, and responsible usage to ensure the ethical integration of these technologies in scholarly endeavors.