Evaluating text-based generative artificial intelligence models for patient information regarding cataract surgery.

This article discusses the evaluation of text-based generative artificial intelligence (AI) models for providing patient information about cataract surgery. The study compared the responses of three AI models (ChatGPT3, Microsoft Bing, and Google Bard) to a set of 25 questions about cataract surgery. The study found that all three AI models provided mostly appropriate answers with accurate information, and no model provided dangerously incorrect information. However, the reproducibility of answers was poor, with less than half of the answers being consistent when the same question was posed multiple times. The article suggests that while AI chatbots show promise as a helpful tool for patient information, physician oversight is still important to ensure accuracy and patient safety. [Extracted from the article]