Improving Accuracy in ChatGPT
Makaleler > Tam Makale · 2023
Yazarlar: Alperen Elek
American Journal of Roentgenology, Uluslararası, Wos, 2023, cilt 221, ss 705-705
DOI: 10.2214/ajr.23.29868
Özet
The AJR article titled "Accuracy of Information Provided by ChatGPT Regarding Liver Cancer Surveillance and Diagnosis" highlights the potential value of artificial intelligence (AI), particularly the ChatGPT 3.5 model (OpenAI), in the medical field [1].Recent advancements in AI technology, such as the integration of the GPT-4 model (OpenAI) by the Bing chatbot (Bing Chat, Microsoft), have brought forward newer iterations of ChatGPT (e.g., the plug-in, web-accessing models), broadening the scope for future research in this domain.However, it is important to remember that the default models of ChatGPT 3.5 or ChatGPT-4 do not have Internet access.Their training on biased datasets could potentially limit the accuracy of their responses [2, 3].Therefore, framing the right question or prompt is often not enough to ensure accurate information.The recent introduction of the plug-in model in ChatGPT-4 marks a significant advancement.This feature, which allows these models to access the Internet and thereby to access external databases, can substantially enhance their potential applications in the scientific and medical fields.For instance, with this feature, models could potentially tap into academic databases such as PubMed (National Library of Medicine, NIH) to ensure more accurate responses.It is worth noting that the enhancement of AI capabilities is not limited to OpenAI's developments.Other developers also have the right to create plug-ins, which could foster rapid progress in this area [4].The way a question is phrased to models such as ChatGPT becomes pivotal to increase the accuracy of responses.Enabling web access in ChatGPT or requesting references from the PubMed database after asking questions could potentially enhance the model's accuracy.Adopting such a strategic approach may yield a more realistic and comprehensive understanding of these models' applications in a medical setting.