ChatGPT: Is this version good for healthcare and research?

Abstract

Background and aims: There have been advancements in artificial intelligence (AI) and deep learning in the past decade. Recently, OpenAI Inc. has launched a new chatbot, called ChatGPT that interacts in a conversational way and its dialogue format makes is user friendly and fast. In this paper we aimed to explore the current position and the accuracy of currently available version of ChatGPT in relation to healthcare and medical research.

Methods: We searched the PubMed, Scopus, and Google databases from 15^th to 25^th February 2023, using the keywords: 'ChatGPT' AND 'medical research, healthcare, and scientific writing'. We found 29 results in PubMed and 9 results in Scopus database., in English language. In addition, we (RV, AM) interacted with ChatGPT multiple times to review accuracy of responses of various medical questions.

Results: Using literature search and interactions with ChatGPT with medical questions, we infer that this version generates answers rapidly but narrates data from existing internet literature in a general manner. However, as emphasised by the company in the landing page of ChatGPT, we found errors in responses to medical questions, Further, narrated data were limited up to September 2021. Positive features include admission of its limitations in medical field, and as it has been designed, learning from previous answers.

Conclusion: Current version of ChatGPT may be useful in a limited manner as a narrative AI chatbot for medical personnel, however, researchers are advised to fact check all statements provided, keeping in mind its limitations.

Keywords: Artificial intelligence; ChatGPT; Healthcare; Research.