Yapay zekâ, erektil disfonksiyon ile ilgili sorulara nasıl yanıtlar veriyor: ChatGPT vs Google AI

Abstract

OBJECTIVE: This study aimed to investigate how accurately Google AI and ChatGPT responded to questions regarding erectile dysfunction (ED). MATERIAL and METHODS: In the study conducted by two andrologists between January and February 2024, 80 specific questions regarding erectile dysfunction (ED) were formulated based on established medical guidelines (EAU guidelines 2023). The questions covered topics including etiology, risk factors, pathophysiology, diagnosis, and treatment. To control potential biases, a single researcher inputted all questions into ChatGPT and Google AI. The responses provided by the artificial intelligence (AI) models were then evaluated by the physicians who formulated the questions. Statistical analysis was performed using IBM Statistical Package for Social Sciences (SPSS) program version 29.0.2.0, employing non-parametric tests due to the small sample size (n <30). Results were presented numerically and graphically, with mean values calculated for each parameter. The Mann-Whitney U test was utilized for comparison between the two groups, with a significance level set at p<0.05. RESULTS: Both language models provided responses with high accuracy to questions related to etiology and risk factors; however, ChatGPT gave slightly lower rates of completely correct answers (3.8±0.41, 3.65±0.58; p=0.56). In terms of pathophysiology, both models provided correct answers without misleading information (3.55±0.51, 3.7±0.47; p=0.43). Regarding diagnosis and treatment-related questions, both artificial intelligence models provided similar rates of correct responses (3.6±0.59, 3.35±0.87; p=0.47) and (3.50±0.68, 3.4±0.94; p=0.70), respectively. However, both Google AI and ChatGPT provided responses containing more incorrect information concerning diagnosis and treatment. CONCLUSION: Current AI models like Google AI and ChatGPT generally offer satisfactory and accurate responses to erectile dysfunction queries, yet performance fluctuates between questions. Keywords: artificial intelligence, ChatGPT, erectile disfunction, Google AI

Ähnliche Arbeiten