We were fortunate to be able to contribute to this editorial that explores artificial intelligence's (AI) potential to generate academic papers. We focused here on natural language model-based AI, specifically ChatGPT. ChatGPT is an innovative artificial intelligence (AI) software that can generate content in a conversational flow and has been shown to pass medical licensing exams without additional training.
In this editorial, we highlighted the scientific community's acceptance of language-based AI, with examples of ChatGPT being credited as an author and used to generate an academic paper. We then looked at academic papers, emphasizing the importance of a thesis, persuasive arguments, extensive research, critical thinking, and accurate referencing.
As an exercise and example, we used ChatGPT to create two essays on the pros and cons of using AI in scientific manuscripts and sports medicine to test AI's ability to generate academic papers. While the generated papers were created in minutes, they lacked originality, advanced understanding of the subject matter, and accurate referencing. This raises the issue of plagiarism, ethical concerns about AI-generated content, and questions about AI's utility in generating novel research ideas.
ChatGPT generated essay 1 and 2. Scored on ‘GPT-2 Output Detector’ AI detection, preparaphrasing and post-paraphrasing AI. AI, artificial intelligence.
All of the above raises equity, accuracy, and AI detection issues. It emphasizes the possibility of AI tools becoming prohibitively expensive, potentially causing resource distribution imbalances among researchers. Writing errors and incomplete information are identified as problems, as is the difficulty distinguishing AI-generated abstracts from human-written ones.
New protection methods and enhanced detection tools must address these ethical and accuracy concerns. Journal authorship guidelines should explicitly prohibit using AI-based text generation tools, while alternative approaches could emphasize transparency in reporting the use of AI-based text generation.
Finally, while natural language model-based AI, such as ChatGPT, shows promise in generating conversational text, its use raises significant ethical, accuracy, equity, and detection concerns. To protect scientific integrity in sports medicine, scientific publishers, academic organizations, and journal boards must be vigilant and implement novel protection methods.
The full paper can be found here (Open Access)
Anderson N, Belavy DL, Perle SM, et al. AI did not write this manuscript, or did it? Can we trick the AI text detector into generated texts? The potential future of ChatGPT and AI in Sports & Exercise Medicine manuscript generation. BMJ Open Sport & Exercise Medicine 2023;9:e001568. doi: 10.1136/bmjsem-2023-001568